README: index primers/ and us_market/ in the document map

Adds reading-order entries and marks EN-only docs as awaiting
translation.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

README.md

@@ -18,9 +18,41 @@ re-verification before external use.
 
 ## Document map 文档地图
 
-Every document exists in both languages (`_en` / `_zh`), with matching
+Documents exist in both languages (`_en` / `_zh`) with matching section
-section structure so the two versions can be read side by side.
+structure so the two versions can be read side by side; docs marked
-每份文档均有英文与中文两个版本（`_en` / `_zh`），章节结构一一对应，可对照阅读。
+**EN only** are awaiting translation.
+文档以英文与中文两个版本并存（`_en` / `_zh`），章节结构一一对应，可对照阅读；
+标注 **EN only** 的文档中文版待补。
+
+### `primers/` — Background primers 背景入门读物
+
+Market-neutral educational material — no venture-specific analysis.
+与具体业务无关的背景知识读物。
+
+**`primers/grid/` — the grid & markets series 电网与市场系列** (read in order 按序阅读):
+
+| # | EN | 中文 | Contents 内容 |
+|---|---|---|---|
+| 01 | [01_how_the_grid_works_en.md](primers/grid/01_how_the_grid_works_en.md) | [zh](primers/grid/01_how_the_grid_works_zh.md) | Grid physics, the balancing act, blackouts — from first principles. 电网物理原理、实时平衡与停电机理。 |
+| 02 | [02_us_electricity_markets_en.md](primers/grid/02_us_electricity_markets_en.md) | [zh](primers/grid/02_us_electricity_markets_zh.md) | How US wholesale markets are organized. 美国批发电力市场结构。 |
+| 03 | [03_china_electricity_markets_en.md](primers/grid/03_china_electricity_markets_en.md) | [zh](primers/grid/03_china_electricity_markets_zh.md) | China's market structure and reform path. 中国电力市场结构与改革路径。 |
+| 04 | [04_market_mechanics_companion_en.md](primers/grid/04_market_mechanics_companion_en.md) | [zh](primers/grid/04_market_mechanics_companion_zh.md) | Market mechanics companion — clearing, pricing, settlement. 市场机制手册——出清、定价、结算。 |
+| 05 | [05_trading_under_uncertainty_en.md](primers/grid/05_trading_under_uncertainty_en.md) | [zh](primers/grid/05_trading_under_uncertainty_zh.md) | Trading under uncertainty. 不确定性下的交易。 |
+
+**Topical primers 专题读物** (EN only):
+
+| Doc 文档 | File | Contents 内容 |
+|---|---|---|
+| SCUC & SCED | [scuc_sced_en.md](primers/scuc_sced_en.md) | The two optimization problems that clear US wholesale markets; N-1 security, LMP as dual variables. 出清美国批发市场的两大优化问题。 |
+| Newsvendor 报童模型 | [newsvendor_en.md](primers/newsvendor_en.md) | The newsvendor problem applied to electricity bidding — underpins the bid loop in `pjm/design_models_*`. 报童模型在电力竞价中的应用。 |
+| Stochastic optimization 随机优化 | [stochastic_optimization_en.md](primers/stochastic_optimization_en.md) | Stochastic optimization in power systems — scenarios, recourse, hydro, offering under uncertainty. 电力系统中的随机优化。 |
+
+### `us_market/` — US market reference 美国市场综述
+
+| Doc 文档 | EN | 中文 | Contents 内容 |
+|---|---|---|---|
+| Field guide 实务指南 | [field_guide_en.md](us_market/field_guide_en.md) | EN only | Bidding, trading, and the impact of AI — consolidated working discussions, July 2026. 竞价、交易与 AI 影响综合笔记。 |
+| Economists survey 经济学家综述 | [economists_survey_en.md](us_market/economists_survey_en.md) | [economists_survey_zh.md](us_market/economists_survey_zh.md) | How economists shaped US electricity markets — theory, practice, milestones. 经济学家如何塑造美国电力市场——理论、实践与里程碑。 |
 
 ### `pjm/` — US market 美国市场
 
@@ -39,10 +71,11 @@ section structure so the two versions can be read side by side.
 
 ## Suggested reading order 建议阅读顺序
 
-1. **New to PJM 不熟悉 PJM** → `pjm/market_primer_*` → `pjm/design_models_*`
+1. **New to power markets entirely 电力市场零基础** → `primers/grid/01–05` in order 按序阅读
-2. **New to China's market 不熟悉中国市场** → `china/market_vendor_dd_*` §1 → `china/product_design_*`
+2. **New to PJM 不熟悉 PJM** → `pjm/market_primer_*` → `pjm/design_models_*`
-3. **Business/competitive view 商业与竞争视角** → the two vendor DD docs 两份供应商尽调 → synthesis sections 综合研判章节
+3. **New to China's market 不熟悉中国市场** → `china/market_vendor_dd_*` §1 → `china/product_design_*`
-4. **Modeling 建模视角** → `pjm/design_models_*` (A–C + newsvendor) → `china/product_design_*` (Model D math)
+4. **Business/competitive view 商业与竞争视角** → the two vendor DD docs 两份供应商尽调 → synthesis sections 综合研判章节
+5. **Modeling 建模视角** → `primers/newsvendor_en.md` + `primers/stochastic_optimization_en.md` → `pjm/design_models_*` (A–C + newsvendor) → `china/product_design_*` (Model D math)
 
 ## Conventions 体例约定
 

primers/grid/01_how_the_grid_works_en.md

@@ -0,0 +1,247 @@
+# How the Electricity Grid Works: A Primer
+
+*A five-chapter guide from first principles to blackouts — covering the physics, the balancing act, the markets, the renewables transition, and what happens when things break.*
+
+---
+
+## Chapter 1 — The Fundamentals
+
+### 1.1 The machine that can't pause
+
+The electricity grid is often called the largest machine ever built, and the description is literal: every generator, transformer, wire, and appliance on an interconnection forms a single, electrically continuous system. What makes it unlike any other machine is its defining constraint — electricity at grid scale mostly cannot be stored. It must be generated at the same instant it is consumed. Everything in this primer flows from that fact.
+
+### 1.2 Voltage, current, and power
+
+Three quantities govern everything, and a water analogy captures them well. **Voltage** (volts, V) is pressure — how hard electrons are pushed. **Current** (amps, I) is flow rate — how much charge passes per second. **Power** (watts, P) is the rate of useful work, and it is simply their product:
+
+> P = V × I
+
+The same power can be delivered as high voltage with low current, or low voltage with high current. The choice matters because wires waste energy as heat, and those resistive losses scale with the *square* of current:
+
+> P_loss = I² × R
+
+Raise voltage tenfold and current falls tenfold — cutting wire losses by a factor of one hundred. This single relationship dictates the grid's architecture: generate at 10–25 kV, step up to 100–765 kV for long-distance transmission, then step back down through substations to the 120–240 V that arrives at your outlet. A well-run grid loses only about 5–8% of energy between plant and plug.
+
+### 1.3 Why AC won
+
+The grid runs on **alternating current** (AC): voltage swings sinusoidally, completing 50 cycles per second in most of the world and 60 in North America. AC prevailed over Edison's direct current (DC) in the 1890s for one decisive reason: **transformers only work with AC**. A transformer is two coils sharing an iron core; a *changing* current in one coil induces voltage in the other, with the voltage ratio set by the turns ratio. No moving parts, roughly 99% efficient, and it makes changing voltage trivial — which, per section 1.2, is the whole game. Practical DC voltage conversion had to wait a century for modern power electronics (and today, high-voltage DC links do have a niche: very long lines, undersea cables, and connecting grids that aren't synchronized).
+
+The grid actually uses **three-phase** AC: three conductors carrying identical sine waves offset by 120°. Two benefits follow. First, although each phase's power pulses to zero twice per cycle, the sum of all three is perfectly constant — motors run smoothly and generators feel steady torque. Second, three-phase delivers three times the power of a single phase with only about 1.5 times the conductor material. Homes typically tap a single phase; factories take all three.
+
+### 1.4 Power versus energy
+
+The most common confusion in energy discussions is between power and energy.
+
+- **Power** (W, kW, MW, GW) is an instantaneous *rate* — how fast energy flows right now.
+- **Energy** (Wh, kWh, MWh, TWh) is power sustained over time — what meters count and bills charge. A 2 kW heater running 3 hours uses 6 kWh.
+
+A plant therefore has two distinct ratings. Its **capacity** is the maximum power it could produce; its **capacity factor** is the fraction of that maximum it actually delivers over a year. Nuclear plants run near-continuously (~90% capacity factor); onshore wind manages roughly 25–35%; solar 10–25% depending on latitude and climate. A 1 GW solar farm and a 1 GW nuclear plant share a nameplate but produce very different annual energy.
+
+### 1.5 The shape of demand
+
+Demand follows human life. It is low overnight, ramps as people wake, plateaus through the working day, peaks in the evening, and falls again — with weekly rhythms (weekends dip) and seasonal ones (heating or cooling peaks) layered on top. Traditionally, three tiers of generation serve this shape, stacked by cost and flexibility:
+
+1. **Baseload** — nuclear, large hydro, historically coal. Expensive to build, cheap to run, slow to cycle. Runs flat-out serving the demand floor.
+2. **Mid-merit** — typically combined-cycle gas. Ramps up and down to follow the daily swing.
+3. **Peaking** — open-cycle gas turbines. Cheap to build, costly to run; fires up only for the highest-demand hours of the year.
+
+A crucial consequence: the grid is sized for its **peak**, not its average. Wires, transformers, and capacity must all survive the worst winter evening or summer heatwave hour. This is why shaving the peak — via pricing, demand response, or storage — is disproportionately valuable.
+
+### 1.6 Who runs the grid
+
+The grid is a layered ecosystem rather than a single organization:
+
+- **Generators** own and operate power plants, from utility fleets to a single rooftop.
+- The **system operator** (TSO in Europe; ISO/RTO in the US) is the air traffic controller: it owns no plants but schedules them, monitors flows in real time, and bears ultimate responsibility for balance.
+- **Transmission owners** maintain the high-voltage network; **distribution utilities** run the local wires to your street.
+- **Retailers** buy wholesale and sell to consumers.
+- **Regulators** oversee it all, because wires are a natural monopoly — you cannot have competing power lines to a house — so their prices and reliability standards are set publicly even in otherwise competitive markets.
+
+---
+
+## Chapter 2 — Balancing and Stability
+
+### 2.1 Frequency: the grid's heartbeat
+
+Because electricity cannot be meaningfully stored on the grid itself, supply must match demand second by second. The grid's real-time vital sign for this balance is **frequency**. Every conventional generator is a *synchronous machine* — a turbine spinning a rotor phase-locked to the grid — and all of them rotate in lockstep. On a 50 Hz system a two-pole machine spins at exactly 3,000 rpm.
+
+Frequency and balance are linked by physics, not convention. Collectively, the spinning rotors store enormous kinetic energy (E = ½Jω²). When demand exceeds generation, the deficit is drawn from this kinetic reservoir: rotors decelerate and frequency falls. Excess generation accelerates them and frequency rises.
+
+### 2.2 The swing equation and ROCOF
+
+The dynamics are captured by the swing equation. In aggregated form:
+
+> df/dt = ΔP × f₀ / (2 × H × S)
+
+where ΔP is the power imbalance, f₀ the nominal frequency, S the total rated power of connected machines, and H the **inertia constant** — the seconds of full output a machine could supply from stored kinetic energy alone (typically 2–9 s; large steam turbines around 4–7 s).
+
+The left side is the **ROCOF** (rate of change of frequency), the critical number in the first moments after a disturbance. Example: a 50 Hz system loses 5% of its generation with aggregate H = 4 s. ROCOF = 0.05 × 50 / (2 × 4) ≈ 0.31 Hz/s — reaching the first load-shedding threshold (~49 Hz) in roughly three seconds if nothing intervened. Halve the inertia and the same event unfolds twice as fast.
+
+### 2.3 Anatomy of a frequency event
+
+A large plant trip triggers a characteristic sequence:
+
+1. **Inertial response (0–2 s).** Pure physics, no control. The deficit is met by decelerating rotors; ROCOF is set by ΔP and inertia. Inertia does not stop the decline — it slows it, buying time.
+2. **Load damping.** Motor-driven loads draw slightly less as frequency falls (roughly 1–2% of demand per Hz), a natural stabilizer represented as a damping term D·Δf.
+3. **Primary control / governor droop (2–30 s).** Every governor runs a proportional controller with a deliberate "droop" of typically 4–5%: full extra output for a 5% frequency deviation. Droop lets hundreds of machines share the burden in proportion to their size without fighting each other. The frequency **nadir** occurs when governor response plus damping exactly equals the deficit.
+4. **Quasi-steady offset.** Proportional control has no memory of the setpoint, so the system settles *below* nominal frequency.
+5. **Secondary control / AGC (30 s–15 min).** Automatic Generation Control computes an Area Control Error (frequency deviation plus unscheduled interchange with neighbors) and sends setpoint adjustments to regulating plants every few seconds — the integral action that restores exactly 50.00 Hz and re-arms primary reserves.
+6. **Tertiary control (15 min+).** Manual redispatch or unit start-ups restore the secondary reserves themselves.
+
+### 2.4 Protection: load shedding and the cascade risk
+
+If controls cannot arrest a fall, automatic protection intervenes. **Under-frequency load shedding (UFLS)** relays disconnect blocks of demand in staged tranches, typically starting near 49.0 Hz (59.5 Hz on 60 Hz systems) with further stages every ~0.2 Hz. Below roughly 47.5 Hz, generators must disconnect to protect themselves — steam turbine blades have resonant frequencies and sustained off-nominal operation can destroy them. That is the cascade scenario: plants tripping on under-frequency deepen the deficit, tripping more plants. Chapter 5 examines real cases.
+
+### 2.5 The other balance: voltage and reactive power
+
+Frequency is a *system-wide* quantity; voltage is *local*, and it must be held within a few percent of nominal at every node. Voltage control hinges on **reactive power** (measured in VAr) — power that oscillates between source and load without doing net work, but which is required to sustain the magnetic fields of motors and transformers and to hold voltage up along lines.
+
+The practical rule: real power (W) travels well; reactive power does not. It must be supplied close to where it is needed. Tools include generator excitation systems, switched capacitor banks, transformer tap changers, and power-electronic devices (SVCs, STATCOMs). Voltage problems, unlike frequency problems, can be invisible from far away — and voltage collapse, where a heavily loaded region's voltage spirals down as motors stall and draw ever more current, has caused several major blackouts.
+
+### 2.6 Balancing across timescales
+
+Pulling the threads together, balance is maintained by layered mechanisms:
+
+| Timescale | Mechanism |
+|---|---|
+| Milliseconds–seconds | Rotor inertia, load damping, battery fast frequency response |
+| Seconds–minutes | Governor droop (primary reserve) |
+| Minutes | AGC (secondary reserve) |
+| 15 min–hours | Tertiary reserve, redispatch, interconnector trades |
+| Hours–days | Market scheduling from demand and weather forecasts |
+| Months–years | Maintenance planning, capacity procurement, network build |
+
+---
+
+## Chapter 3 — Markets: How Electricity Gets Priced
+
+### 3.1 Why electricity markets are strange
+
+Electricity is a commodity with no storage buffer, no substitute at the moment of use, demand that barely responds to price in real time, and a delivery network shared by all participants. The result is the most volatile commodity market in the world: prices in the same market can range from below zero to several thousand dollars per MWh within a single day.
+
+### 3.2 The merit order and marginal pricing
+
+Most wholesale markets are built on a simple auction. For each delivery hour, generators bid the price at which they are willing to run. The operator stacks bids from cheapest to most expensive — the **merit order** — and accepts them until forecast demand is met. The bid of the *last* (marginal) unit accepted sets the **clearing price paid to every accepted generator**.
+
+Uniform marginal pricing looks odd at first — why pay a cheap nuclear plant the same as an expensive gas peaker? — but it has a clean logic: every generator is paid the system's true marginal cost, giving each one an incentive to bid its actual costs (bid higher and you risk not being dispatched; bid lower and you may run at a loss). Cheap plants earn the spread between price and their costs, which is precisely what repays their high construction costs.
+
+Two consequences follow:
+
+- **Fuel sets the price more than the mix suggests.** If gas plants are marginal most hours, gas prices drive electricity prices even in systems with lots of cheap nuclear or renewables — as Europe discovered painfully in 2022.
+- **Zero-marginal-cost renewables depress prices when they run.** Wind and solar bid near zero (their fuel is free), pushing expensive units out of the stack and lowering the clearing price — the *merit order effect*. This also means renewables cannibalize their own revenue: the more solar on a system, the lower prices are at noon, exactly when solar produces.
+
+### 3.3 The sequence of markets
+
+Trading happens in a sequence of markets that progressively refine the schedule as forecasts sharpen:
+
+1. **Forward/futures markets** (years to days ahead): hedging instruments; most energy is financially contracted here.
+2. **Day-ahead market**: the main physical auction, clearing hourly (or half-hourly) prices for tomorrow based on demand and weather forecasts.
+3. **Intraday market**: continuous trading up to minutes before delivery, letting participants correct for forecast errors (a wind operator selling more as a windy front firms up, or buying back a shortfall).
+4. **Balancing / real-time market**: the operator's domain. Whatever imbalance remains at delivery is settled here, at prices that can spike brutally — the financial incentive to be balanced.
+
+### 3.4 Negative prices
+
+Prices go negative when generation exceeds demand and it costs producers more to switch off than to pay someone to consume. Nuclear plants avoid cycling for technical reasons; some renewables keep producing because subsidies pay per MWh generated; combined-heat-and-power plants must run for their heat customers. Negative prices are a feature, not a bug: they are the market screaming for flexibility — storage to charge, demand to shift, exports to flow.
+
+### 3.5 Locational pricing and congestion
+
+Wires have limits. When the cheapest generation cannot physically reach demand because a line is full, the system splits into price zones. US-style markets price every node on the network (**locational marginal pricing, LMP**): each node's price reflects the marginal cost of serving one more MW *there*, including congestion and losses. European markets mostly use larger zonal prices with separate congestion management ("redispatch"). The design debate is live: nodal prices are economically precise but complex; zonal prices are simple but hide congestion costs and can misplace investment signals.
+
+### 3.6 Beyond energy: capacity and ancillary services
+
+Energy-only revenue may not sustain plants needed only rarely — the peaker that runs 100 hours a year, or firm capacity backing up renewables. Many systems therefore add:
+
+- **Capacity markets**: payments for being *available*, procured through auctions years ahead (e.g., GB, PJM in the US).
+- **Ancillary service markets**: separate products for the reserves of Chapter 2 — frequency response, spinning reserve, voltage support, black-start capability. Batteries have transformed these markets; fast frequency response was their first killer app.
+- **Scarcity pricing**: letting energy prices rise very high (e.g., to a cap of $5,000+/MWh) in shortages, so rare hours can fund peaking capacity — the Texas ERCOT philosophy.
+
+---
+
+## Chapter 4 — The Renewables Transition
+
+### 4.1 What actually changes
+
+Wind and solar differ from conventional plants in three fundamental ways, each with its own consequences:
+
+1. **Variable**: output follows weather, not demand.
+2. **Uncertain**: forecastable, but with errors that grow with horizon.
+3. **Inverter-based**: connected via power electronics, not spinning synchronous machines — so they contribute no natural inertia (Chapter 2) and no short-circuit strength.
+
+None of these is a showstopper, but each dismantles an assumption the grid was built on.
+
+### 4.2 The duck curve
+
+Solar reshapes the daily *net load* — demand minus renewable output, which is what conventional plants and storage must serve. With heavy solar, net load sags deeply at midday, then ramps ferociously as the sun sets just when evening demand rises. Plotted over the day, the curve resembles a duck (deep belly at noon, steep neck at sunset) — hence California's famous **duck curve**. The steep evening ramp, often tens of GW within three hours, is the operational crux: a fleet of flexible resources must swing from minimum to maximum in step with the sunset.
+
+### 4.3 Curtailment and the value of flexibility
+
+When renewable output exceeds what the system can absorb — demand is low, exports are full, must-run plants can't go lower — the surplus is **curtailed**: wind and solar are told to spill free energy. Some curtailment is economically optimal (building a grid to absorb every last windy hour would cost more than the energy is worth), but rising curtailment is the signal that flexibility, not generation, is the binding constraint. The remedies, roughly in order of maturity:
+
+- **Storage.** Lithium-ion batteries dominate intra-day shifting (charge at noon, discharge at the evening ramp), with typical durations of 1–4 hours. Pumped hydro remains the giant of bulk storage. Multi-day and seasonal storage (hydrogen, thermal, other chemistries) is the unsolved frontier.
+- **Demand flexibility.** EV charging, heat pumps with thermal buffering, industrial load shifting — demand becoming an active grid resource rather than a passive given.
+- **Transmission.** Bigger and smarter interconnection smooths weather over geography: it is usually windy or sunny *somewhere*.
+- **Sector coupling.** Turning surplus power into other vectors — hydrogen via electrolysis, heat into district networks.
+
+### 4.4 Keeping the lights on with less iron
+
+As inverter-based resources displace synchronous machines, system inertia falls, ROCOF rises, and frequency events unfold faster (Chapter 2). Short-circuit strength — the grid's ability to hold voltage steady through faults, which protection relays also rely on to detect faults — falls too. The toolkit answering this:
+
+- **Fast frequency response** from batteries: full injection in well under a second, faster than any governor.
+- **Synthetic inertia**: inverters programmed to inject power proportional to df/dt, mimicking a rotor.
+- **Grid-forming inverters**: the deeper fix. Conventional ("grid-following") inverters measure the grid's voltage waveform and inject current in step with it — they need a grid to follow. Grid-forming inverters create their own voltage waveform, behaving like a voltage source that other resources can synchronize to. They can, in principle, run a system with no synchronous machines at all.
+- **Synchronous condensers**: generators without turbines, spun purely to provide inertia, short-circuit strength, and reactive power — old technology in a new starring role.
+
+### 4.5 The economics of "cheap but not free"
+
+Levelized cost per MWh of wind and solar has fallen below almost everything else, but system value is not the same as levelized cost. Each additional MW of solar produces at the same hours as the existing solar, depressing its own capture price (section 3.2) while shifting costs to whoever provides the evening ramp, firm winter capacity, and grid reinforcement. Honest accounting compares *system* costs across scenarios rather than plant-level costs — which is why storage economics, capacity mechanisms, and flexible demand are inseparable from the renewables build-out itself.
+
+---
+
+## Chapter 5 — When Things Break
+
+### 5.1 Designing for failure: the N-1 rule
+
+Grid planning starts from the assumption that things fail. The core reliability standard is **N-1**: the system must survive the loss of any single element — the largest generator, any one transmission line, a transformer — without cascading, without load loss, and while remaining within limits. Operators continuously run contingency analysis: software that simulates hundreds of possible failures against the *current* system state, every few minutes, flagging any contingency that would breach a limit so operators can reposition the system preemptively. Critical corridors are often held to N-2.
+
+The reserve sizing of Chapter 2 follows the same logic: primary reserve on an interconnection is dimensioned to cover the *reference incident* — typically the two largest units tripping simultaneously (3,000 MW in continental Europe).
+
+### 5.2 How cascades happen
+
+Blackouts are almost never one failure; they are chains in which each link makes the next more likely:
+
+1. An initiating event — a line fault, a plant trip, often during stress (heat, storm, high load).
+2. Power reroutes over remaining paths (electricity follows physics, not contracts), overloading them.
+3. Overloaded lines sag into vegetation or trip on protection; each trip reroutes more flow onto fewer paths.
+4. Voltage and frequency excursions begin tripping generators and loads.
+5. The system splits into islands, most of which are imbalanced and collapse.
+
+The recurring accelerants across post-mortems: hidden failures in protection relays that trip when they shouldn't, operators lacking visibility into the true system state, and inadequate coordination across control-area boundaries.
+
+### 5.3 Case studies in one paragraph each
+
+**US/Canada Northeast, August 2003 (~50 million people).** A hot afternoon in Ohio: a generation trip, then transmission lines sagging into untrimmed trees — while an alarm-system software failure left operators blind to the deteriorating state. Overloads cascaded across the Midwest and Northeast in minutes. Lessons: situational awareness tools, mandatory (not voluntary) reliability standards, and vegetation management became enforceable rules.
+
+**South Australia, September 2016 (state-wide).** Tornado-strength storms downed transmission lines, causing voltage disturbances; the ride-through protection settings on multiple wind farms reduced their output after repeated voltage dips, overloading the interconnector to Victoria, which tripped — islanding the state with a massive deficit and near-record ROCOF. Lessons: inverter ride-through settings matter enormously; the event catalyzed the world's first big grid battery (Hornsdale) and stricter performance standards.
+
+**Great Britain, August 2019 (~1 million customers).** A lightning strike — normally routine — coincided with the near-simultaneous loss of an offshore wind farm and a gas plant, plus embedded generation tripping on the frequency dip. The combined loss exceeded the reserve held, frequency fell below 48.8 Hz, and load shedding operated as designed but disconnected some critical infrastructure (notably rail). Lessons: reserves sized to single contingencies can be beaten by correlated failures; know what's actually connected to load-shedding blocks.
+
+**Texas, February 2021 (Winter Storm Uri).** Not a cascade but a supply catastrophe: extreme cold froze gas wells, pipelines, wind turbines, and plant instrumentation simultaneously, while heating demand hit records. ERCOT shed up to 20 GW for days to avoid total collapse — operators later estimated the system was minutes from a months-long black start. Lessons: weatherization of both power *and* fuel supply chains; the danger of correlated failures across an entire fleet; the limits of an energy-only market in valuing extreme-event resilience.
+
+### 5.4 Restoration: the black start
+
+Restarting a dead grid is a delicate bootstrapping problem, because most power plants need external power to start — pumps, controls, ignition. Designated **black-start units** (hydro and gas turbines with on-site batteries or diesels) can self-start, energize a path to a larger plant, bring it online, and gradually re-energize "islands" of the grid that are then carefully resynchronized — matching frequency, voltage, and phase before closing the breaker between them. Load is reconnected in small blocks so demand never outruns the recovering generation. Full restoration takes hours to days; system operators rehearse it regularly.
+
+### 5.5 The reliability outlook
+
+The threat list is evolving: extreme weather is now the dominant cause of major outages and stresses correlated-failure assumptions; cyber attacks on control systems are a demonstrated reality (Ukraine 2015/2016); and the transition itself changes the physics (Chapter 4) faster than standards traditionally update. The countervailing trend is that the grid is becoming more observable and more controllable than ever — synchrophasors streaming precise measurements many times a second, batteries responding in milliseconds, and grid-forming inverters promising stability services from the very resources that once only consumed them.
+
+---
+
+## Where to go deeper
+
+- **Balancing and stability**: look up "swing equation", "governor droop", and your local system operator's frequency-response procurement documents — they're surprisingly readable.
+- **Markets**: day-ahead price data is public for most markets (e.g., European power exchanges, US ISOs) and exploring a single day's merit order teaches more than any textbook chapter.
+- **The transition**: system operator studies on operating low-inertia grids (Ireland's EirGrid and Australia's AEMO publish the most candid ones).
+- **Failures**: official blackout post-mortem reports (US-Canada 2003 Task Force report, AEMO's South Australia report) are gripping engineering literature.
+
+*End of primer.*

primers/grid/01_how_the_grid_works_zh.md

@@ -0,0 +1,247 @@
+# 电网是如何运转的：入门读本
+
+*一份从第一性原理讲到大停电的五章指南——涵盖物理基础、供需平衡、电力市场、可再生能源转型，以及系统故障时会发生什么。*
+
+---
+
+## 第一章 基础知识
+
+### 1.1 一台不能暂停的机器
+
+电网常被称为人类建造过的最大机器，这个说法是字面意义上的：一个同步电网中的每一台发电机、每一台变压器、每一根导线和每一件电器，共同构成一个电气上连续的整体系统。它与其他任何机器的根本不同在于一个决定性的约束——电网尺度的电能基本无法储存，发电与用电必须在同一瞬间完成。本书的一切内容都源于这个事实。
+
+### 1.2 电压、电流与功率
+
+三个物理量支配着一切，用水流类比最为直观。**电压**（伏特，V）是压力——推动电子的力度；**电流**（安培，I）是流量——每秒通过的电荷量；**功率**（瓦特，P）是做有用功的速率，它就是二者的乘积：
+
+> P = V × I
+
+同样的功率，既可以用高电压配小电流输送，也可以用低电压配大电流输送。这个选择至关重要，因为导线会把能量耗散为热，而电阻损耗与电流的*平方*成正比：
+
+> P_loss = I² × R
+
+电压提高十倍，电流就降为十分之一——线路损耗随之降为百分之一。正是这一条关系决定了电网的整体架构：发电机在 10–25 kV 发电，升压到 100–765 kV 进行远距离输电，再经变电站逐级降压，最终以 120–240 V 到达你的插座。一个运行良好的电网，从电厂到插头的总损耗只有约 5–8%。
+
+### 1.3 交流电为什么赢了
+
+电网使用**交流电**（AC）：电压按正弦规律摆动，世界上大部分地区每秒 50 个周期，北美为 60 个。交流电之所以在 1890 年代的"电流之战"中战胜爱迪生的直流电（DC），有一个决定性原因：**变压器只对交流电起作用**。变压器不过是共用一个铁芯的两组线圈；一侧线圈中*变化的*电流会在另一侧感应出电压，电压比等于匝数比。没有运动部件，效率约 99%，让电压变换变得轻而易举——而按 1.2 节所述，这正是整个游戏的关键。实用的直流电压变换技术直到一个世纪后的现代电力电子出现才成为可能（如今高压直流输电确有自己的用武之地：超长距离线路、海底电缆，以及连接互不同步的电网）。
+
+电网实际使用的是**三相**交流电：三根导线承载完全相同、但彼此错开 120° 的正弦波。这带来两大好处。其一，虽然每一相的功率每个周期两次降到零，但三相之和是完全恒定的——电动机运转平稳，发电机承受的转矩均匀。其二，三相输电用约 1.5 倍的导线材料就能输送单相 3 倍的功率。普通住宅通常只接入其中一相；工厂则使用全部三相。
+
+### 1.4 功率与电量
+
+能源讨论中最常见的混淆，就是功率与电量（能量）之分。
+
+- **功率**（W、kW、MW、GW）是瞬时*速率*——此刻能量流动有多快。
+- **电量**（Wh、kWh、MWh、TWh）是功率在时间上的累积——电表计量、账单收费的对象。一台 2 kW 的电暖器运行 3 小时，消耗 6 kWh。
+
+因此一座电厂有两个截然不同的指标。**装机容量**是它所能输出的最大功率；**容量因子**是它全年实际发电量占最大可能发电量的比例。核电站近乎连续运行（容量因子约 90%）；陆上风电大约 25–35%；光伏则视纬度和气候为 10–25%。一座 1 GW 的光伏电站和一座 1 GW 的核电站铭牌容量相同，年发电量却相去甚远。
+
+### 1.5 需求的形状
+
+用电需求跟随人类生活的节奏：夜间低谷，清晨爬坡，工作时段维持平台，傍晚达到高峰，随后回落——其上还叠加着周节律（周末下降）和季节节律（取暖或制冷高峰）。传统上，三个层次的电源按成本和灵活性堆叠起来，共同满足这一形状：
+
+1. **基荷电源**——核电、大型水电，历史上还有煤电。建设成本高、运行成本低、不宜频繁启停，全天满发，覆盖永远存在的需求底部。
+2. **腰荷电源**——典型为联合循环燃气机组，随日内需求起伏上下调节。
+3. **调峰电源**——开式循环燃气轮机，建设便宜、运行昂贵，只在全年需求最高的少数小时启动。
+
+一个关键推论：电网是按**峰值**而非平均值建设的。线路、变压器和发电容量都必须扛得住最严酷的冬夜或酷暑时刻。这也是削峰——通过电价、需求响应或储能——具有超比例价值的原因。
+
+### 1.6 谁在运营电网
+
+电网不是一家机构，而是一个层层分工的生态系统：
+
+- **发电企业**拥有并运营电厂，从大型电力集团到一块屋顶光伏。
+- **系统运营机构**（欧洲称 TSO，美国称 ISO/RTO）是空中交通管制员：它不拥有电厂，却负责调度电厂、实时监控潮流，并对供需平衡承担最终责任。
+- **输电公司**维护高压骨干网；**配电公司**运营通到你家街道的本地网络。
+- **售电公司**在批发市场购电，再连同账单和客服一起卖给用户。
+- **监管机构**统辖全局。因为电网线路是天然垄断——不可能有两家公司竞争着往你家拉电线——所以即便在竞争性市场中，网络的价格和可靠性标准也由公共监管确定。
+
+---
+
+## 第二章 平衡与稳定
+
+### 2.1 频率：电网的心跳
+
+由于电能无法在电网上大规模储存，供给与需求必须逐秒匹配。电网衡量这一平衡的实时"生命体征"就是**频率**。每一台常规发电机都是*同步电机*——汽轮机带动转子旋转，其电气输出与电网锁相——所有同步电机步调完全一致地旋转。在 50 Hz 系统中，一台两极电机的转速恰好是每分钟 3000 转。
+
+频率与平衡之间的联系是物理规律而非人为约定。所有旋转转子合计储存着巨大的动能（E = ½Jω²）。当需求超过发电时，缺口就从这个动能库中提取：转子减速，频率下降；发电过剩则转子加速，频率上升。
+
+### 2.2 摇摆方程与频率变化率（ROCOF）
+
+这一动态过程由摇摆方程刻画，其聚合形式为：
+
+> df/dt = ΔP × f₀ / (2 × H × S)
+
+其中 ΔP 是功率不平衡量，f₀ 是额定频率，S 是并网机组的总额定容量，H 是**惯量常数**——仅靠储存的动能，机组能以额定功率输出多少秒（典型值 2–9 秒；大型汽轮机组约 4–7 秒）。
+
+方程左侧 df/dt 就是**频率变化率（ROCOF）**，是扰动发生后最初时刻里最关键的数字。举例：一个 50 Hz 系统损失 5% 的发电出力，系统聚合惯量常数 H = 4 秒，则 ROCOF = 0.05 × 50 / (2 × 4) ≈ 0.31 Hz/s——若无任何干预，大约三秒即触及第一级切负荷门槛（约 49 Hz）。惯量减半，同样的事件就以两倍的速度上演。
+
+### 2.3 一次频率事件的解剖
+
+大型机组跳闸会触发一个特征性的过程序列：
+
+1. **惯量响应（0–2 秒）**。 纯物理过程，无任何控制介入。缺口由转子减速填补；ROCOF 由 ΔP 和系统惯量决定。惯量并不能阻止频率下跌——它只是延缓下跌，为控制系统争取时间。
+2. **负荷阻尼**。 电动机类负荷在频率下降时耗电略微减少（约为每赫兹 1–2% 的需求量），这是一种天然的稳定因素，在方程中表现为阻尼项 D·Δf。
+3. **一次调频 / 调速器下垂（2–30 秒）**。 每台机组的调速器都是一个带有预设"下垂率"的比例控制器，典型下垂率 4–5%：频率偏差达 5% 时输出全部备用出力。下垂特性使成百上千台机组按各自容量比例分担缺口，而不会相互争抢。当调速器响应加负荷阻尼恰好等于缺口时，频率到达**最低点（nadir）**。
+4. **准稳态偏差**。 比例控制不记忆设定值，因此系统会稳定在*低于*额定频率的某个水平。
+5. **二次调频 / AGC（30 秒–15 分钟）**。 自动发电控制计算区域控制偏差（频率偏差加上与邻网的非计划交换功率），每隔几秒向调频电厂下发出力调整指令——正是这一积分作用把频率精确恢复到 50.00 Hz，并释放一次备用以应对下一次扰动。
+6. **三次调频（15 分钟以后）**。 通过人工再调度或启动机组，恢复二次备用本身。
+
+### 2.4 保护措施：切负荷与连锁风险
+
+如果控制手段无法止住频率下跌，自动保护装置就会介入。**低频减载**（UFLS）继电器按分级方案切除成块负荷，通常从 49.0 Hz 附近开始（60 Hz 系统约为 59.5 Hz），此后大约每降 0.2 Hz 再切一级。频率低于约 47.5 Hz 时，发电机组必须解列自保——汽轮机叶片存在共振频率，长时间偏离额定频率运行可能造成机械损毁。这正是连锁崩溃的剧本：机组因低频跳闸使缺口进一步扩大，导致更多机组跳闸。第五章将剖析真实案例。
+
+### 2.5 另一种平衡：电压与无功功率
+
+频率是*全网一致*的量，电压却是*局部*的，必须在每一个节点都保持在额定值的百分之几以内。电压控制的核心是**无功功率**（单位：乏，VAr）——它在电源与负荷之间来回振荡而不做净功，却是维持电动机和变压器磁场、支撑线路电压所必需的。
+
+实践法则是：有功功率（W）可以远距离输送，无功功率不行——它必须在需要之处就近提供。手段包括发电机励磁系统、投切电容器组、变压器有载调压分接头，以及电力电子装置（SVC、STATCOM）。与频率问题不同，电压问题在远处可能毫无征兆——而电压崩溃（重负荷地区电压不断下滑，电动机失速后电流反而越抽越大，形成恶性循环）已经酿成过多次大停电。
+
+### 2.6 跨越各时间尺度的平衡
+
+把线索拢在一起，供需平衡由层层机制共同维持：
+
+| 时间尺度 | 机制 |
+|---|---|
+| 毫秒–秒 | 转子惯量、负荷阻尼、电池快速频率响应 |
+| 秒–分钟 | 调速器下垂（一次备用） |
+| 分钟 | AGC（二次备用） |
+| 15 分钟–小时 | 三次备用、再调度、联络线交易 |
+| 小时–天 | 基于负荷和气象预测的市场排产 |
+| 月–年 | 检修计划、容量采购、电网建设 |
+
+---
+
+## 第三章 电力市场：电价是如何形成的
+
+### 3.1 电力市场为何如此奇特
+
+电力这种商品没有库存缓冲，在用电的那一刻没有替代品，实时需求对价格几乎无弹性，而且所有参与者共用同一张输送网络。结果是全世界波动最剧烈的大宗商品市场：同一个市场的价格可以在一天之内从负值飙到每兆瓦时数千美元。
+
+### 3.2 边际成本排序与边际定价
+
+大多数批发市场建立在一个简单的拍卖机制之上。针对每个交割时段，发电商申报自己愿意开机的价格。运营机构把报价从低到高排成一列——即按**边际成本排序**（merit order）——依次接受，直到满足预测需求为止。最后一台（边际）中标机组的报价，就是**支付给所有中标机组的统一出清价格**。
+
+统一边际定价乍看有些奇怪——凭什么给便宜的核电和昂贵的燃气调峰机组付同样的价格？——但其逻辑十分清晰：每台机组都按系统真实的边际成本获得支付，因而每家都有动力如实申报成本（报高了有落选风险，报低了可能亏本运行）。低成本机组赚取电价与自身成本之间的价差，而这正是回收其高昂建设投资的来源。
+
+由此引出两个推论：
+
+- **决定电价的是边际燃料，而非电源结构本身**。 如果大多数小时里燃气机组是边际机组，那么即使系统中有大量廉价核电或可再生能源，天然气价格依然主导电价——2022 年的欧洲对此有切肤之痛。
+- **零边际成本的可再生能源在出力时压低电价**。 风电和光伏以接近零的价格申报（它们的"燃料"免费），把昂贵机组挤出排序，从而拉低出清价——这就是*优序效应*。这也意味着可再生能源会蚕食自己的收入：系统里光伏越多，正午电价越低，而那恰恰是光伏出力的时候。
+
+### 3.3 市场的时序
+
+电力交易在一连串市场中依次展开，随着预测精度提高而逐步校准计划：
+
+1. **远期/期货市场**（提前数年至数天）：套期保值工具，大部分电量在此完成金融合约。
+2. **日前市场**：最主要的实物拍卖，基于负荷和气象预测，出清次日每小时（或每半小时）的价格。
+3. **日内市场**：连续交易直至交割前数分钟，让参与者修正预测误差（风电场在大风前锋确认后加卖电量，或在出力不足时买回缺口）。
+4. **平衡/实时市场**：系统运营机构的领地。交割时刻残余的一切不平衡在此结算，价格可能剧烈飙升——这正是促使各方保持平衡的经济鞭策。
+
+### 3.4 负电价
+
+当发电超过需求、且发电商停机的代价比付钱请人用电更高时，电价就会转负。核电机组出于技术原因避免启停；部分可再生能源因补贴按发电量支付而继续发电；热电联产机组必须为供热用户运行。负电价是特性而非缺陷：它是市场在高声呼唤灵活性——储能该充电了，负荷该转移了，电力该外送了。
+
+### 3.5 分区电价与阻塞
+
+导线有容量极限。当最便宜的电力因线路满载而无法送达需求侧时，系统就会分裂为不同价区。美式市场对电网的每个节点定价（**节点边际电价，LMP**）：每个节点的价格反映在*该处*多供应 1 MW 的边际成本，包含阻塞与网损。欧洲市场大多采用较大的分区价格，辅以独立的阻塞管理（"再调度"）。这场设计之争至今未决：节点电价经济上精确但复杂，分区电价简单却掩盖阻塞成本，可能误导投资选址。
+
+### 3.6 电能之外：容量与辅助服务
+
+仅靠电能收入，可能养不活那些只在少数时刻被需要的电源——一年只运行 100 小时的调峰机组，或为可再生能源兜底的可靠容量。因此许多系统增设了：
+
+- **容量市场**：为"随时可用"本身付费，通过提前数年的拍卖采购（如英国、美国 PJM）。
+- **辅助服务市场**：为第二章所述的各类备用设立独立品种——频率响应、旋转备用、电压支撑、黑启动能力。电池储能彻底改变了这些市场；快速频率响应是它的第一个杀手级应用。
+- **稀缺定价**：允许电价在短缺时冲得极高（例如上限 5000 美元/MWh 以上），让极少数稀缺小时养活调峰容量——这是德州 ERCOT 的市场哲学。
+
+---
+
+## 第四章 可再生能源转型
+
+### 4.1 到底什么变了
+
+风电和光伏与常规电源有三个根本差异，各有其后果：
+
+1. **波动性**：出力跟随天气，而非跟随需求。
+2. **不确定性**：可以预测，但误差随预测时长增大。
+3. **基于逆变器**：通过电力电子而非旋转同步电机并网——因此不提供天然惯量（见第二章），也不提供短路容量。
+
+三者都不是不可逾越的障碍，但每一条都拆掉了电网赖以建立的一个前提假设。
+
+### 4.2 鸭子曲线
+
+光伏重塑了日内的*净负荷*——总需求减去可再生出力，即常规电源和储能真正需要满足的部分。在光伏比例高的系统中，净负荷在正午深深下凹，日落时分又随着晚高峰来临而陡峭攀升。把这条曲线画出来，形似一只鸭子（正午的深腹、日落的陡颈）——这就是加州著名的**鸭子曲线**。傍晚的陡坡（常常是三小时内爬升数十吉瓦）是运行上的症结所在：必须有一支灵活电源舰队随着日落从最小出力一路拉到最大。
+
+### 4.3 弃电与灵活性的价值
+
+当可再生出力超过系统的消纳能力——需求低迷、外送通道已满、必开机组无法再压——多余部分只能被**弃掉**（弃风弃光）：免费的能量被白白倾泻。一定程度的弃电在经济上是最优的（为吸纳每一个大风小时而扩建电网，成本会超过那些电的价值），但弃电率持续上升是一个信号：约束系统的已不是发电，而是灵活性。对策按成熟度大致排列：
+
+- **储能**。 锂电池主导日内平移（正午充电、晚高峰放电），典型时长 1–4 小时；抽水蓄能仍是大容量储能的巨人；跨多日和跨季节储能（氢、储热及其他化学体系）是尚未攻克的前沿。
+- **需求灵活性**。 电动车充电、带蓄热缓冲的热泵、工业负荷转移——需求从被动给定变成主动的电网资源。
+- **输电**。 更大、更智能的互联互通用地理广度平滑天气：*总有某个地方*在刮风或出太阳。
+- **部门耦合**。 把过剩电力转化为其他能源载体——电解水制氢、余电供热入区域热网。
+
+### 4.4 钢铁少了，灯还要亮
+
+随着逆变器型电源取代同步电机，系统惯量下降、ROCOF 上升，频率事件的演化更快（第二章）。短路容量——电网在故障期间维持电压、同时也是保护继电器识别故障所依赖的能力——同样在下降。应对的工具箱包括：
+
+- 电池提供的**快速频率响应**：不到一秒即可满出力注入，比任何调速器都快。
+- **虚拟惯量**：将逆变器编程为按 df/dt 成比例地注入功率，模拟转子行为。
+- **构网型逆变器**：更根本的解法。常规（"跟网型"）逆变器测量电网电压波形、随之注入电流——它们需要一个可跟随的电网。构网型逆变器则自己建立电压波形，表现为一个可供其他电源同步的电压源，原则上可以支撑一个完全没有同步电机的系统。
+- **同步调相机**：没有汽轮机的发电机，空转只为提供惯量、短路容量和无功——老技术担纲新主角。
+
+### 4.5 "便宜但不免费"的经济学
+
+风电和光伏的平准化度电成本已降到几乎低于一切电源，但系统价值不等于平准化成本。每新增一兆瓦光伏，都在与既有光伏相同的时段发电，压低自己的实际售电价格（见 3.2 节），同时把成本转嫁给提供晚高峰爬坡、冬季可靠容量和电网加固的一方。诚实的核算应比较不同情景下的*系统*总成本，而非电厂层面的成本——这正是储能经济性、容量机制与需求灵活性同可再生能源建设本身密不可分的原因。
+
+---
+
+## 第五章 系统故障时会发生什么
+
+### 5.1 为失效而设计：N-1 准则
+
+电网规划的出发点就是假定设备会坏。核心可靠性标准是 **N-1**：系统必须经受得住任何单一元件的损失——最大的机组、任意一条输电线、一台变压器——不发生连锁反应、不损失负荷、且各项指标保持在限值以内。运营机构持续运行预想事故分析：软件每隔几分钟就用*当前*系统状态模拟数百种可能故障，一旦发现某个预想事故会越限就发出警示，让调度员提前调整运行方式。关键通道往往执行 N-2 标准。
+
+第二章所述的备用容量整定遵循同样的逻辑：一个同步电网的一次备用按*基准事故*配置——通常是两台最大机组同时跳闸（欧洲大陆电网为 3000 MW）。
+
+### 5.2 连锁崩溃是如何发生的
+
+大停电几乎从来不是单一故障，而是一条环环相扣的链条，每一环都让下一环更容易断裂：
+
+1. 初始事件——线路故障或机组跳闸，往往发生在系统承压时（高温、风暴、重负荷）。
+2. 潮流沿剩余通道重新分布（电流遵循物理定律，不遵循合同），使这些通道过载。
+3. 过载线路下垂触树或被保护切除；每断开一条，就有更多潮流涌向更少的通道。
+4. 电压和频率的偏移开始导致机组和负荷相继跳开。
+5. 系统解列成若干孤岛，其中多数供需失衡而相继崩溃。
+
+历次事故报告中反复出现的加速因素：保护继电器中的隐性缺陷（在不该动作时动作）、调度员对系统真实状态缺乏可见性，以及跨调度区边界的协调不足。
+
+### 5.3 案例研究，每例一段
+
+**美加东北部，2003 年 8 月（约 5000 万人受影响）**。 俄亥俄州一个炎热的下午：先是机组跳闸，随后输电线路下垂触碰未修剪的树木——与此同时，告警系统的软件故障使调度员对不断恶化的局面浑然不觉。过载在数分钟内连锁席卷中西部和东北部。教训：态势感知工具、强制性（而非自愿性）可靠性标准，以及线路走廊植被管理自此成为可执行的硬性规定。
+
+**南澳大利亚，2016 年 9 月（全州停电）**。 龙卷风级别的风暴刮倒输电线路并引发电压扰动；多座风电场的穿越保护整定在电压反复跌落后削减了出力，导致通往维多利亚州的联络线过载跳闸——全州瞬间孤岛运行，面对巨大功率缺口和接近纪录的 ROCOF。教训：逆变器故障穿越整定至关重要；这次事件催生了世界第一个大型电网储能电站（Hornsdale）和更严格的并网性能标准。
+
+**英国，2019 年 8 月（约 100 万用户）**。 一次本属寻常的雷击，恰与一座海上风电场和一座燃气电厂的近乎同时脱网叠加，加上分布式电源在频率跌落中相继跳开。总损失出力超过了系统持有的备用，频率跌破 48.8 Hz，低频减载按设计动作——却切掉了部分关键基础设施（尤以铁路为甚）。教训：按单一事故整定的备用可能被相关性故障击穿；必须清楚切负荷轮次里到底接着什么。
+
+**德克萨斯州，2021 年 2 月（冬季风暴 Uri）**。 这不是连锁跳闸，而是一场供给端的浩劫：极寒同时冻住了气井、管道、风机和电厂仪表，而取暖需求刷新纪录。ERCOT 连续数日切除高达 20 GW 的负荷以避免全网崩溃——运营方事后估计，系统距离一场需要数月才能恢复的黑启动只有几分钟之遥。教训：电力系统*和*燃料供应链都必须防寒加固；警惕席卷整个机组群的相关性故障；纯电能量市场在为极端事件韧性定价上存在局限。
+
+### 5.4 恢复供电：黑启动
+
+重启一张彻底停电的电网是一个精巧的"自举"难题，因为大多数电厂启动时本身就需要外部电源——驱动水泵、控制系统、点火装置。指定的**黑启动机组**（配有站用电池或柴油机的水电和燃气轮机）能够自行启动，为通往更大电厂的路径充电送电，将其带入运行，然后逐步为电网中一个个"孤岛"恢复供电，再小心翼翼地将孤岛彼此并列——频率、电压、相位三者对齐后才能合上联络断路器。负荷以小块逐步恢复，确保需求永远不超过正在恢复的发电能力。全面恢复需要数小时到数天；系统运营机构会定期演练。
+
+### 5.5 可靠性展望
+
+威胁清单在不断演变：极端天气已成为重大停电的头号原因，不断冲击关于"故障相互独立"的传统假设；针对控制系统的网络攻击已是被证实的现实（乌克兰 2015/2016）；而能源转型本身正在以超过标准更新速度的节奏改变系统物理特性（第四章）。与之抗衡的趋势是，电网正变得前所未有地可观测、可控制——同步相量测量装置每秒数十次地传回精确量测，电池在毫秒级响应，构网型逆变器则有望让那些曾经只会"消耗"稳定性的资源反过来提供稳定服务。
+
+---
+
+## 延伸阅读建议
+
+- **平衡与稳定**：查阅"摇摆方程""调速器下垂"，以及你所在地区系统运营机构的频率响应采购文件——它们的可读性出人意料地好。
+- **电力市场**：大多数市场的日前电价数据是公开的（如欧洲电力交易所、美国各 ISO），研究某一天的边际成本排序，胜过教科书里的任何一章。
+- **能源转型**：各系统运营机构关于低惯量电网运行的研究报告（爱尔兰 EirGrid 和澳大利亚 AEMO 发布的最为坦率）。
+- **系统故障**：官方大停电调查报告（美加 2003 年联合工作组报告、AEMO 的南澳事故报告）是引人入胜的工程文献。
+
+*全文完。*

primers/grid/02_us_electricity_markets_en.md

@@ -0,0 +1,222 @@
+# US Electricity Markets: A Primer
+
+*How power gets priced, traded, and paid for in the world's most intricate electricity market landscape — from the history and the players to the market mechanics, the money flows, and the stresses testing the system today.*
+
+---
+
+## Introduction: why US markets are like nothing else
+
+The United States does not have *an* electricity market. It has seven organized wholesale markets run by regional operators, layered over three separate synchronous grids, coexisting with large regions that never restructured at all, overseen by one federal regulator, one reliability corporation, and fifty state commissions. This complexity is not an accident; it is the residue of a century of institutional evolution plus a half-finished revolution in the 1990s. Understanding US electricity markets means understanding both the elegant optimization machinery at their core and the jurisdictional patchwork wrapped around it.
+
+Two facts orient everything. First, electricity is the most unforgiving commodity: no meaningful storage at grid scale, demand that barely responds to price in real time, and a shared delivery network governed by physics rather than contracts. Second, in the US, *wholesale* power (sales between companies, and interstate transmission) is federal turf, while *retail* power (sales to consumers, local distribution, generation siting) belongs to the states. Nearly every controversy in the industry lives on that seam.
+
+---
+
+## Chapter 1 — How we got here
+
+### 1.1 The regulated century
+
+From the early 1900s until the 1980s, the model was uniform: a vertically integrated utility owned the generation, transmission, and distribution in its exclusive territory. In exchange for its monopoly, the utility submitted to rate regulation — state commissions set retail prices to cover prudent costs plus a fair return on invested capital. The model built the grid, electrified the country, and worked passably while technology kept driving costs down. Its flaw emerged in the 1970s: cost-plus regulation rewards capital spending regardless of need, and ratepayers bore the overruns of the nuclear construction era.
+
+### 1.2 Cracks, then restructuring
+
+Three legislative and regulatory waves dismantled the monopoly on generation:
+
+- **PURPA (1978)** forced utilities to buy power from qualifying independent cogenerators and small renewables — accidentally proving that non-utility companies could build and run power plants.
+- **The Energy Policy Act (1992)** created a broad class of exempt wholesale generators and authorized FERC to open the transmission network.
+- **FERC Order 888 (1996)** delivered the opening: utilities had to offer transmission service to competitors on the same terms they gave themselves. Order 2000 (1999) then encouraged utilities to hand grid operations to independent regional bodies — the RTOs.
+
+In parallel, states chose their own paths. Restructuring states (the Northeast, mid-Atlantic, Illinois, Ohio, Texas, California) forced or induced utilities to divest generation and opened retail choice. Most of the Southeast and the non-coastal West declined, keeping vertically integrated utilities under traditional regulation.
+
+### 1.3 California, and the map freezes
+
+California's restructured market collapsed in 2000–2001 under a badly designed framework: retail prices frozen while wholesale prices were free to spike, utilities forbidden from long-term hedging, and traders (most infamously Enron) gaming transmission and scheduling rules. Rolling blackouts and a utility bankruptcy followed. The political lesson absorbed nationwide was caution: no state has restructured since. The result is today's frozen map — roughly two-thirds of US load served through organized RTO markets, one-third under traditional regulation, with hybrid arrangements in between (notably, vertically integrated utilities in MISO and SPP that participate in RTO dispatch while remaining state-regulated monopolies).
+
+### 1.4 The modern reform era
+
+FERC's recent landmark orders sketch where the system is heading: Order 745 (demand response paid full LMP), Order 841 (storage must be allowed to participate in all markets), Order 2222 (aggregated distributed resources may participate), Order 2023 (interconnection queue reform: cluster studies, readiness requirements, penalties), and Order 1920 (long-term regional transmission planning). Each order is FERC using its wholesale jurisdiction to pull new technologies and processes into markets designed for a fossil, one-way grid.
+
+---
+
+## Chapter 2 — The players
+
+### 2.1 The referees
+
+**FERC** (Federal Energy Regulatory Commission) is the apex wholesale regulator: it approves every RTO tariff and market rule change, sets transmission rates of return, polices market manipulation with penalties up to seven figures per violation per day, and reviews mergers. Five commissioners, appointed to staggered terms, no more than three from one party.
+
+**NERC** (North American Electric Reliability Corporation) writes the mandatory reliability standards — over 100 of them covering operations, planning, cyber security (the CIP standards), and personnel — enforced through six regional entities with FERC-backed penalties. Standards became mandatory only after the 2003 blackout; before that, compliance was voluntary.
+
+**State public utility commissions** control the retail side: rates for distribution and (in non-restructured states) generation, integrated resource plans, siting of plants and lines, and consumer protection. In restructured states their generation role shrank; in vertically integrated states they effectively decide what gets built.
+
+### 2.2 The operators: profiles of the seven
+
+- **PJM Interconnection** — 13 mid-Atlantic and Midwest states plus DC, ~67 million people, the largest market and the intellectual origin of LMP and capacity markets. Currently the epicenter of data-center-driven stress.
+- **MISO** — a vast north-south footprint from Manitoba to Louisiana; mostly vertically integrated utility members; known for its seams with SPP and PJM and its large wind fleet.
+- **ERCOT** — most of Texas; energy-only design, no capacity market, its own interconnection (avoiding FERC jurisdiction over its wholesale market design); the fastest-growing fleet of solar, storage, and gas in the country.
+- **CAISO** — California plus bits of Nevada; the renewables laboratory (duck curve, storage fleet); operates the Western Energy Imbalance Market and its day-ahead extension (EDAM) for utilities across the West.
+- **NYISO** — New York alone; sharp zonal price separation between upstate (cheap, clean) and New York City (constrained, expensive).
+- **ISO-NE** — six New England states; winter gas scarcity is the defining reliability problem.
+- **SPP** — the wind-rich central plains; like MISO, mostly integrated utilities using the market for dispatch efficiency.
+
+An RTO is a nonprofit membership organization. It owns no wires and no plants; it operates the system, runs the markets, performs regional planning, and administers the interconnection queue. Its rules are made through a stakeholder process (committees of generators, utilities, consumers, states) and filed with FERC.
+
+### 2.3 The commercial cast
+
+**Generators**: increasingly independent power producers (IPPs) and large competitive fleets (Constellation, Vistra, NextEra, Talen) rather than utilities. In vertically integrated regions, the utility still owns most plants.
+
+**Load-serving entities (LSEs)**: whoever sells to end users — the incumbent utility, or in retail-choice states, competitive retail suppliers. LSEs buy at wholesale (spot or hedged) and sell at retail.
+
+**Financial participants**: banks, hedge funds, and prop traders with no physical assets, trading virtual bids, FTRs, and futures. They supply liquidity and price discovery — and occasionally manipulation cases.
+
+**Demand response and aggregators**: companies that enroll factories, buildings, and now households and batteries, selling their flexibility into energy, capacity, and reserve markets.
+
+**Independent market monitors**: each RTO retains an external watchdog (e.g., Monitoring Analytics for PJM, Potomac Economics for several others) that screens every offer for market power, mitigates offers in constrained areas, publishes state-of-the-market reports, and refers suspected manipulation to FERC.
+
+---
+
+## Chapter 3 — The energy markets
+
+### 3.1 The two-settlement design
+
+Every RTO runs the same core: a **day-ahead market** and a **real-time market**, settled sequentially.
+
+The **day-ahead market** clears each afternoon for every hour of the next day. Generators submit multi-part offers (energy price curves, start-up and no-load costs, ramp rates, minimum run times); LSEs and traders submit demand bids. The RTO solves **security-constrained unit commitment (SCUC)**: a mixed-integer optimization choosing which units to turn on and at what level, minimizing total as-offered cost subject to meeting demand, transmission limits, reserve requirements, and each unit's physical constraints — while also ensuring the system would survive any single contingency (the N-1 rule embedded directly in the market). The outputs are binding hourly schedules and day-ahead LMPs at every node.
+
+The **real-time market** then handles reality. Every five minutes, **security-constrained economic dispatch (SCED)** re-optimizes output levels of committed units against actual load, wind, and outages, producing five-minute LMPs. Deviations settle at real-time prices: a generator that produced less than its day-ahead schedule buys back the shortfall at real-time; one that produced more sells the surplus.
+
+The two-settlement structure is a hedging machine: the day-ahead market is where positions are locked (roughly 95%+ of energy settles there), and real-time is the balancing market whose volatility everyone tries to avoid — or exploit.
+
+### 3.2 Locational marginal pricing in depth
+
+Every node's price decomposes into three components:
+
+**Energy** — the marginal cost of serving load at a reference bus, identical across all nodes; the pure merit-order price, set by the offer of the marginal unit.
+
+**Congestion** — each node's share of the cost of binding transmission constraints. For every constraint at its limit, the optimization computes a *shadow price* (the system savings if the limit were 1 MW higher). A node's congestion component is the sum over binding constraints of shadow price times the node's *shift factor* (the fraction of an injection at that node that physically flows across that line). Nodes whose injections aggravate a constraint see negative components; import-side nodes see positive ones. Congestion is zero most hours, dominant during stress, the driver of negative prices in trapped-renewables pockets, and the only component hedgeable with FTRs.
+
+**Losses** — pricing of marginal transmission losses. Because losses grow with the square of flow, the marginal loss of delivery to a distant node is roughly twice the average loss; components run a few percent of the energy price, positive far from generation and occasionally negative where injection reduces net flows. The over-collection inherent in marginal-loss pricing is rebated.
+
+Prices are settled at three granularities: individual **nodes** (where generators are paid), **zones** (aggregations, typically utility territories, where load pays), and **hubs** (published averages of many nodes, stable enough to serve as reference prices for bilateral contracts and futures).
+
+### 3.3 Price formation extremes
+
+- **Negative prices** occur where must-run generation (nuclear avoiding shutdown cycles, subsidized renewables earning per-MWh credits, hydro under environmental flow rules) exceeds what the grid can absorb locally. They are a functioning signal: charge storage, shift demand, build transmission.
+- **Scarcity pricing** operates at the other end. When reserves run short, administrative adders push prices far above any unit's marginal cost — in ERCOT via the Operating Reserve Demand Curve (ORDC), which adds a premium that grows as reserves shrink, up to a cap (raised and restructured after Winter Storm Uri, when prices pinned at the then-cap of $9,000/MWh for days). Scarcity pricing is deliberate: those rare hours are meant to fund the fixed costs of resources that exist only for reliability — the energy-only market's substitute for capacity payments.
+- **Price caps and offer mitigation** bound the system: offer caps limit what generators may bid (with cost verification above thresholds), and market monitors mitigate offers from units with local market power inside constrained pockets.
+
+---
+
+## Chapter 4 — Beyond energy: the other markets
+
+### 4.1 Ancillary services
+
+Energy is only one product the grid needs. RTOs run co-optimized markets for reliability services, meaning the same optimization allocates each unit's capability between energy and services according to value:
+
+- **Regulation** — capacity that follows the AGC signal second by second to hold frequency and area interchange. Paid for both capacity held and "mileage" (movement). Batteries dominate here: millisecond response and precise tracking.
+- **Spinning (synchronized) reserve** — online headroom deliverable within 10 minutes, the buffer against a large unit trip.
+- **Non-spinning / supplemental reserve** — offline but fast-start capacity (10–30 minutes).
+- **Newer products** — several markets have added ramping products (capacity held to manage steep net-load ramps) and are designing fast-frequency-response products as synchronous inertia declines.
+- **Voltage support and black start** — typically compensated through cost-based tariffs rather than auctions; reactive power is too local for a liquid market.
+
+Co-optimization matters: a plant offering both energy and reserves is cleared to whichever mix maximizes system welfare, and the reserve price automatically reflects the energy profit the unit forgoes by holding back — an elegant piece of market design that eliminates the need for separate opportunity-cost payments.
+
+### 4.2 Capacity markets: paying for existence
+
+The "missing money" problem: if energy prices are capped and scarcity hours are rare, revenues may not cover the fixed costs of resources needed only occasionally — yet reliability requires them to exist. Two rival answers dominate.
+
+**The capacity-market answer (PJM, NYISO, ISO-NE).** A mandatory forward auction procures commitments to be available during the delivery year. PJM's Reliability Pricing Model (RPM) is the archetype:
+
+- The demand side is an administratively drawn downward-sloping curve (the Variable Resource Requirement curve) anchored to the cost of new entry and the target reserve margin (a ~20% cushion above forecast peak, calibrated to the "one outage event in ten years" standard).
+- The supply side is capacity offers from generators, demand response, and storage — measured in **UCAP** (unforced capacity: nameplate derated for forced-outage history) with variable resources credited via **ELCC** (effective load-carrying capability: the marginal reliability contribution of another MW of that resource class, which declines as penetration rises — solar's evening-peak value erodes as more solar shifts the risk hours to sunset).
+- Clearing sets a $/MW-day price paid to all cleared capacity; in exchange, resources take on availability obligations with severe non-performance penalties during emergencies (post-2014 "capacity performance" reforms, born of plants freezing during the 2014 polar vortex).
+- Locational constraints split the auction into zones when transmission limits imports — constrained zones clear higher, mirroring LMP logic at annual scale.
+
+**The energy-only answer (ERCOT).** No capacity payments; instead, let real-time prices rise to a very high cap during scarcity so that rare hours fund reliability. Sharper incentives and no administrative demand curves, but revenue arrives in unpredictable lumps and the political tolerance for price spikes is finite. Uri pushed ERCOT to supplement the design (higher ORDC floors at lower reserve levels, a lower cap, new reliability services) — a drift toward hybrid.
+
+**The obligation answer (CAISO).** California imposes resource adequacy requirements directly on LSEs, who contract bilaterally for qualifying capacity; the ISO runs no central capacity auction.
+
+The debate is one of the liveliest in energy economics: capacity markets are criticized for administrative complexity, over-procurement, and mispricing new technologies; energy-only markets for volatility and political fragility. The current demand shock is stress-testing both.
+
+### 4.3 Financial instruments
+
+**Financial transmission rights (FTRs)** entitle the holder to the day-ahead congestion price difference between two nodes. Sold by RTOs in auctions (funded by congestion revenue the RTO collects), they let anyone hedge congestion — a generator locking the spread from its plant to the hub, a load hedging its zone, or a fund speculating on which constraints will bind. Some markets issue them as obligations (negative payouts possible), others as options. Auction revenue rights (ARRs) allocate the auction proceeds back to the transmission customers who funded the grid.
+
+**Virtual transactions** — increment offers (INCs, virtual supply) and decrement bids (DECs, virtual demand) — are purely financial positions in the day-ahead market that automatically liquidate at real-time prices. Their function is convergence: if day-ahead prices run systematically above real-time, virtual sellers profit until the gap closes. Convergence makes the day-ahead market an honest forecast, improving unit commitment. The dark side: virtuals have featured in manipulation cases where traders used them to move prices that other positions (like FTRs) profited from.
+
+**Futures and swaps** trade on exchanges (ICE, Nymex) settled against hub prices, extending the hedging horizon years out.
+
+---
+
+## Chapter 5 — Contracts, settlement, and the money
+
+### 5.1 The contract stack
+
+Spot markets set reference prices, but most energy is hedged long before delivery:
+
+- **Utility full-requirements and default-service auctions** — restructured-state utilities procure their default retail supply through periodic competitive solicitations.
+- **Power purchase agreements (PPAs)** — long-term (10–25 year) contracts, the financing backbone of renewables: a fixed price per MWh delivered gives the project bankable revenue and gives the buyer (utility or, increasingly, corporate — tech companies are now the largest PPA buyers) a hedge and the renewable attributes.
+- **Virtual PPAs** — purely financial contracts-for-differences settled against a hub price, no physical delivery; the dominant corporate structure.
+- **Tolling agreements** — the buyer supplies fuel and dispatch rights, paying the plant owner a capacity fee: renting the plant rather than buying its output.
+- **Hedges and heat-rate options** — the standard toolbox of any commodity market, denominated at hubs.
+
+### 5.2 How settlement actually works
+
+Each participant's monthly RTO invoice nets: day-ahead energy positions at day-ahead LMPs; real-time deviations at real-time LMPs; ancillary obligations and credits; capacity charges/credits; FTR payouts; transmission charges; and **uplift** — out-of-market payments that make units whole when dispatch instructions leave them unable to recover as-offered costs (a persistent, deliberately uncomfortable residual: high uplift signals the market design is missing something, and RTOs work to price it into LMPs instead).
+
+### 5.3 To the retail bill
+
+A US retail bill stacks: wholesale energy (the LMP-weighted cost of the LSE's supply portfolio), capacity charges, transmission (FERC-regulated), distribution (state-regulated), plus riders for public programs. Wholesale energy is typically only a third to a half of the total — which is why even dramatic wholesale swings reach consumers diluted, and why the current capacity-price surge in PJM is politically explosive: it flows to bills with little dilution.
+
+---
+
+## Chapter 6 — Oversight, gaming, and enforcement
+
+Markets create incentives to cheat, and the system's immune response has three layers. **Ex-ante mitigation**: automated screens flag offers from units with local market power (the "three pivotal supplier" test in PJM), capping them at cost-based references inside constrained areas. **Monitoring**: the independent market monitors screen conduct continuously and publish annual State of the Market reports that are the best public analytics on each market. **Enforcement**: FERC's Office of Enforcement investigates manipulation under its post-2005 anti-manipulation authority, with penalties that have reached hundreds of millions (JP Morgan's $410M settlement over bidding strategies in California; multiple cases involving virtual trades used to benefit FTR positions; classic wash-trading and false-scheduling cases from the Enron era that shaped the rules).
+
+The recurring manipulation patterns are worth knowing because they map to design seams: physical withholding (declaring false outages), economic withholding (offering far above cost in a pocket you dominate), cross-product manipulation (moving a price you profit from elsewhere), and information games in the interconnection or outage-scheduling processes.
+
+---
+
+## Chapter 7 — The stress tests of the 2020s
+
+### 7.1 The demand shock
+
+After fifteen flat years, US electricity demand is growing again — roughly 2% annually in recent years — led by data centers, manufacturing reshoring, and electrification. Forecasts of data-center growth dominate planning: PJM projects tens of GW of peak-load growth by 2030, nearly all of it data centers. Load forecasting, long a sleepy actuarial exercise, is now the most contested number in the industry, because speculative or duplicated data-center requests can trigger billions in procurement.
+
+### 7.2 PJM's capacity crunch: the live case study
+
+PJM's capacity auctions tell the story in three data points: clearing prices rose from about $29/MW-day (2024/25) to $270 (2025/26) to $329 (2026/27) to a capped $333 (2027/28) — roughly tenfold in three years, with the last auction procuring several GW less than the reliability target for the first time and total costs around $16 billion per year. The market monitor attributes the bulk of the increase to data-center load forecasts, much of it for facilities not yet built. Without a negotiated price cap, prices would have cleared far higher. Meanwhile almost no new capacity cleared, because supply is stuck behind the interconnection queue, permitting, financing, and turbine supply chains. Governors have publicly threatened to pull states from PJM; states are creating data-center rate classes and re-examining incentives.
+
+This is the textbook collision: the market is producing exactly the scarcity signal it was designed to produce, while the physical and administrative machinery prevents supply from answering it — converting a price signal into a political crisis.
+
+### 7.3 The supply bottleneck
+
+Over 2,000 GW of generation and storage — more than the entire existing US fleet — waits in interconnection queues; historically only a small minority of queued capacity is ever built, and median queue-to-operation time exceeds five years (eight in PJM). FERC Order 2023's cluster studies, readiness deposits, and deadlines are raising throughput; several RTOs added fast-track lanes for reliability-critical resources. But the deeper constraint is transmission: interconnection costs explode where the network is weak, and the US builds little interregional transmission. Order 1920's long-term planning mandate is the structural answer, on a decade timescale.
+
+### 7.4 Design frontiers
+
+The open questions defining the next decade: how to credit storage and hybrids in capacity constructs as ELCC values shift; whether large flexible loads (data centers that can curtail) should be planned and priced as resources; co-location of load at power plants and the federal-state fight over who governs it; day-ahead market expansion in the West (EDAM vs. competing proposals) potentially finally organizing the last non-RTO region; and whether capacity markets, energy-only designs, or state contracting will prove most adaptable to a fleet dominated by zero-marginal-cost resources.
+
+---
+
+## Quick-reference glossary
+
+| Term | Meaning |
+|---|---|
+| RTO / ISO | Regional operator running the grid and wholesale markets |
+| LMP | Locational marginal price = energy + congestion + losses |
+| SCUC / SCED | The day-ahead commitment and real-time dispatch optimizations |
+| Node / zone / hub | Pricing points: individual bus / load aggregation / trading reference |
+| FTR / ARR | Congestion hedge / entitlement to FTR auction revenue |
+| INC / DEC | Virtual supply offer / virtual demand bid (financial, day-ahead) |
+| UCAP / ICAP | Outage-derated capacity value / nameplate capacity |
+| ELCC | Reliability credit for variable resources |
+| RPM / BRA / VRR | PJM's capacity market / its auction / its demand curve |
+| ORDC | ERCOT's scarcity-pricing reserve demand curve |
+| RA | California's bilateral resource adequacy obligation |
+| Uplift | Out-of-market make-whole payments |
+| PPA / VPPA | Long-term power contract / its financial (contract-for-differences) form |
+| IMM | Independent market monitor |
+| PURPA / Order 888 / 2000 / 841 / 2222 / 2023 / 1920 | The landmark statutes and FERC orders of restructuring and reform |
+
+*End of primer.*

primers/grid/02_us_electricity_markets_zh.md

@@ -0,0 +1,222 @@
+# 美国电力市场入门读本
+
+*电力在这个全球最错综复杂的电力市场版图中如何定价、交易与结算——从历史沿革与参与主体，到市场机制、资金流转，以及当下正在考验这套体系的种种压力。*
+
+---
+
+## 引言：美国电力市场为何独一无二
+
+美国并没有"一个"电力市场。它有七个由区域运营机构运营的批发市场，叠加在三个彼此独立的同步电网之上，与大片从未进行市场化改革的地区并存，其上是一个联邦监管机构、一个可靠性组织和五十个州级监管委员会。这种复杂性并非偶然，而是一个世纪制度演化的沉积，再加上 1990 年代一场进行到一半的改革革命。理解美国电力市场，既要理解其核心那套优雅的优化机器，也要理解包裹在外的管辖权拼图。
+
+两个基本事实可以为一切定向。第一，电力是最不宽容的大宗商品：电网尺度上没有真正的储存手段，实时需求对价格几乎无弹性，共享的输电网络遵循物理定律而非商业合同。第二，在美国，*批发*电力（企业间购售、州际输电）属于联邦事权，而*零售*电力（面向终端用户的销售、本地配电、电源选址）归各州管辖。这个行业几乎所有的争议都发生在这条缝隙上。
+
+---
+
+## 第一章 历史沿革
+
+### 1.1 管制的世纪
+
+从 1900 年代初到 1980 年代，模式高度统一：一家垂直一体化的公用事业公司在其专营区域内拥有发电、输电和配电资产。作为垄断经营权的对价，公司接受价格管制——州监管委员会核定零售电价，覆盖审慎发生的成本并给予合理的资本回报。这一模式建成了电网、实现了全国电气化，在技术持续降低成本的年代运转尚可。它的缺陷在 1970 年代显现：成本加成式管制奖励资本开支而不论其必要性，核电建设时代的巨额超支最终由电费用户承担。
+
+### 1.2 裂缝出现，市场化启动
+
+三波立法与监管浪潮拆除了发电环节的垄断：
+
+- **《公用事业管制政策法》**（PURPA，1978）强制公用事业公司收购符合条件的独立热电联产和小型可再生能源项目的电力——无意中证明了非公用事业公司同样能够建设和运营电厂。
+- **《能源政策法》**（1992）创设了豁免批发发电商这一宽泛类别，并授权 FERC 开放输电网络。
+- **FERC 第 888 号令**（1996）完成了开放：公用事业公司必须以与自用完全相同的条件向竞争者提供输电服务。随后的第 2000 号令（1999)鼓励公用事业公司将电网运营移交给独立的区域机构——即 RTO。
+
+与此同时，各州各行其道。推行市场化改革的州（东北部、大西洋中部各州、伊利诺伊、俄亥俄、得克萨斯、加利福尼亚）强制或引导公用事业公司剥离发电资产并开放零售选择权。东南部大部和内陆西部则拒绝改革，保留了传统管制下的垂直一体化模式。
+
+### 1.3 加州危机，版图冻结
+
+2000–2001 年，加州市场在一套设计糟糕的框架下崩溃：零售电价被冻结而批发电价可以自由飙升，公用事业公司被禁止签订长期套保合约，交易商（最臭名昭著的是安然）则在输电和调度规则上上下其手。轮流停电和一家公用事业公司的破产接踵而至。全国吸取的政治教训是审慎：此后再无一州推行市场化改革。结果就是今天这幅冻结的版图——约三分之二的美国负荷通过 RTO 市场供应，三分之一仍处于传统管制之下，其间还有混合形态（最典型的是 MISO 和 SPP 中的垂直一体化公用事业公司：参与 RTO 调度，但仍是州管制下的垄断企业）。
+
+### 1.4 现代改革时期
+
+FERC 近年的里程碑式法令勾勒出体系的走向：第 745 号令（需求响应按全额 LMP 获得支付）、第 841 号令（储能必须被允许参与所有市场）、第 2222 号令（聚合的分布式资源可以参与市场）、第 2023 号令（并网排队改革：集群研究、就绪要求、违约罚则）、第 1920 号令（长期区域输电规划）。每一道法令，都是 FERC 运用其批发管辖权，把新技术、新流程拉进这些为化石能源、单向潮流时代设计的市场。
+
+---
+
+## 第二章 参与主体
+
+### 2.1 裁判员
+
+**FERC**（联邦能源监管委员会）是批发层面的最高监管者：审批每一份 RTO 收费准则和市场规则修改，核定输电投资回报率，以每项违规每天最高逾百万美元的罚则查处市场操纵，并审查并购。委员会由五名委员组成，任期交错，同一党派不得超过三人。
+
+**NERC**（北美电力可靠性公司）制定强制性可靠性标准——一百余项，覆盖运行、规划、网络安全（CIP 标准）与人员管理——通过六个区域实体执行，罚则由 FERC 背书。标准在 2003 年大停电之后才成为强制性要求；此前遵守全凭自愿。
+
+**州公用事业监管委员会**掌管零售侧：配电（在未市场化的州还包括发电）价格、综合资源规划、电厂和线路选址，以及消费者保护。在市场化改革州，其发电职能已经收缩；在垂直一体化州，它们实际上决定着建什么电源。
+
+### 2.2 运营机构：七大市场速写
+
+- **PJM**——大西洋中部及中西部 13 州加华盛顿特区，约 6700 万人口，最大的市场，也是 LMP 和容量市场的思想发源地。当前是数据中心冲击的震中。
+- **MISO**——从马尼托巴延伸到路易斯安那的巨大南北走廊；成员以垂直一体化公用事业公司为主；以其与 SPP、PJM 的接缝问题和庞大的风电机群著称。
+- **ERCOT**——得克萨斯州大部；纯电能量市场设计，没有容量市场，拥有独立的同步电网（从而使其批发市场设计避开 FERC 管辖）；全美扩张最快的光伏、储能和燃气机群。
+- **CAISO**——加州及内华达一角；可再生能源实验室（鸭子曲线、储能机群）；运营西部能量不平衡市场（EIM）及其日前市场延伸（EDAM），把非 RTO 的西部各公用事业逐步编织起来。
+- **NYISO**——仅纽约州；上州（便宜、清洁）与纽约市（受限、昂贵）之间的分区价差十分尖锐。
+- **ISO-NE**——新英格兰六州；冬季天然气短缺是其标志性的可靠性难题。
+- **SPP**——风资源富集的中部平原；与 MISO 类似，以一体化公用事业公司为主，借市场实现调度效率。
+
+RTO 是非营利的会员制组织。它不拥有线路也不拥有电厂；它负责运行系统、运营市场、开展区域规划并管理并网排队。其规则通过利益相关方程序（由发电商、公用事业、用户、州政府组成的各类委员会）形成，并报 FERC 备案。
+
+### 2.3 商业阵容
+
+**发电商**：越来越多是独立发电商（IPP）和大型竞争性发电集团（Constellation、Vistra、NextEra、Talen），而非公用事业公司。在垂直一体化地区，公用事业公司仍持有大部分电厂。
+
+**售电主体（LSE）**：一切向终端用户售电的实体——在位公用事业公司，或在零售选择州的竞争性售电公司。LSE 在批发侧购电（现货或套保），在零售侧售出。
+
+**金融参与者**：不持有实物资产的银行、对冲基金和自营交易商，交易虚拟申报、FTR 和期货。它们提供流动性和价格发现——偶尔也制造操纵案件。
+
+**需求响应与聚合商**：把工厂、楼宇乃至家庭和电池组织起来，将其灵活性卖入电能量、容量和备用市场的公司。
+
+**独立市场监测机构**：每个 RTO 都聘有外部监督者（如 PJM 的 Monitoring Analytics、多家市场共用的 Potomac Economics），逐笔筛查报价中的市场力，对受限区域内的报价实施缓解措施，发布市场运行状况报告，并将涉嫌操纵的行为移送 FERC。
+
+---
+
+## 第三章 电能量市场
+
+### 3.1 两级结算设计
+
+每个 RTO 都运行同一核心结构：**日前市场**与**实时市场**，依次结算。
+
+**日前市场**每天下午对次日各小时出清。发电商提交多段式报价（电能价格曲线、启动费用与空载费用、爬坡速率、最小运行时间）；LSE 和交易商提交需求申报。RTO 求解**安全约束机组组合（SCUC）**：一个混合整数优化问题——在满足需求、输电限额、备用要求和每台机组物理约束的前提下，选择开哪些机组、开到什么水平，使按报价计算的总成本最小，同时确保系统能够承受任一单一预想事故（N-1 准则直接内嵌于市场之中）。产出是具有约束力的逐小时计划和每个节点的日前 LMP。
+
+**实时市场**随后处理现实。每五分钟，**安全约束经济调度**（SCED）根据实际负荷、风电出力和设备停运情况重新优化已开机组的出力水平，生成五分钟级 LMP。偏差按实时价格结算：实际出力低于日前计划的发电商按实时价格买回缺口；超发部分按实时价格卖出。
+
+两级结算结构是一部套保机器：日前市场是锁定头寸之处（约 95% 以上的电量在此结算），实时市场则是人人都想避开——或有人专门利用——的波动性平衡市场。
+
+### 3.2 节点边际电价（LMP）详解
+
+每个节点的价格分解为三个分量：
+
+**电能分量**——在参考节点供应新增一兆瓦的边际成本，全网所有节点相同；这是纯粹的边际成本排序价格，由边际机组的报价决定。
+
+**阻塞分量**——每个节点分摊的、有约束力的输电约束的成本。对每条达到限额的约束，优化会算出一个*影子价格*（若该限额放宽 1 MW，系统总成本能节省多少）。节点的阻塞分量等于对所有起作用约束求和：影子价格乘以该节点的*转移分布因子*（在该节点注入的功率中实际流经该线路的比例）。注入会加剧约束的节点获得负分量；受端节点获得正分量。阻塞在大多数小时为零，在系统紧张时占据主导，是可再生能源富集"孤岛"中负电价的推手，也是唯一可以用 FTR 对冲的分量。
+
+**网损分量**——对边际输电损耗的定价。由于损耗与潮流的平方成正比，向远端节点送电的边际损耗约为平均损耗的两倍；该分量通常只占电能价格的百分之几，在远离电源处为正，在注入反而能降低净潮流之处偶尔为负。边际损耗定价固有的超收部分会返还给市场成员。
+
+价格按三个颗粒度结算：单个**节点**（发电商按此获付）、**分区**（通常按公用事业公司辖区聚合，负荷按此付费）和**枢纽**（多个节点的公开均价，足够稳定，可作为双边合约和期货的参考价格）。
+
+### 3.3 价格形成的两个极端
+
+- **负电价**出现在必开机组（避免启停循环的核电、按发电量领取补贴的可再生能源、受环境流量规则约束的水电）超过当地电网消纳能力之处。它是一个正常运作的信号：给储能充电、转移需求、修建输电线路。
+- **稀缺定价**作用于另一端。当备用不足时，行政性加价把电价推到远高于任何机组的边际成本——在 ERCOT 通过运行备用需求曲线（ORDC）实现：备用越少，加价越高，直至价格上限（在冬季风暴 Uri 之后上限被下调并重构；那次风暴期间电价曾连续数日钉在当时 9000 美元/MWh 的上限）。稀缺定价是有意为之：这些稀有小时就是用来回收那些只为可靠性而存在的资源的固定成本——这是纯电能量市场替代容量付费的方案。
+- **价格上限与报价缓解**为系统设定边界：报价上限约束发电商的申报（超过门槛需成本核证），市场监测机构对受限区域内拥有局部市场力的机组实施报价缓解。
+
+---
+
+## 第四章 电能量之外：其他市场
+
+### 4.1 辅助服务
+
+电能量只是电网需要的产品之一。RTO 为各类可靠性服务运行联合优化的市场——即同一个优化引擎按价值高低，把每台机组的能力在电能量与各项服务之间分配：
+
+- **调频（Regulation）**——逐秒跟踪 AGC 信号、维持频率和区域交换功率的容量。按持有容量和"里程"（实际调节量）双重付费。电池在此占据统治地位：毫秒级响应、精准跟踪。
+- **旋转（同步）备用**——在线机组的富余出力，10 分钟内可释放，是应对大机组跳闸的缓冲。
+- **非旋转/补充备用**——离线但可快速启动的容量（10–30 分钟）。
+- **新产品**——多个市场已增设爬坡产品（为应对陡峭的净负荷爬坡而持有的容量），并在设计快速频率响应产品，以应对同步惯量的下降。
+- **电压支撑与黑启动**——通常通过基于成本的收费准则补偿而非拍卖；无功过于本地化，难以形成有流动性的市场。
+
+联合优化至关重要：同时申报电能量和备用的机组，会被出清到使系统福利最大化的组合上，而备用价格自动反映机组因压出力而放弃的电能利润——这是市场设计中一个优雅的细节，省去了单独的机会成本补偿。
+
+### 4.2 容量市场：为"存在"付费
+
+"缺失货币"问题：如果电价有上限、稀缺小时又很稀少，收入可能覆盖不了那些只在偶尔时刻被需要的资源的固定成本——但可靠性又要求它们存在。两种针锋相对的答案占据主流。
+
+**容量市场方案（PJM、NYISO、ISO-NE）**。通过强制性远期拍卖采购交割年度内的可用性承诺。PJM 的可靠性定价模型（RPM）是原型：
+
+- 需求侧是一条行政绘制的向下倾斜曲线（可变资源需求曲线，VRR），锚定于新建机组成本和目标备用率（高出预测峰荷约 20% 的缓冲，按"十年一次停电事件"标准校准）。
+- 供给侧是发电、需求响应和储能的容量报价——以 **UCAP**（无强迫停运容量：铭牌容量按历史强迫停运率折减）计量；波动性资源按 **ELCC**（等效负荷承载能力）授信：即该类资源再增加一兆瓦的边际可靠性贡献，且随渗透率上升而递减——光伏的晚高峰价值会随着更多光伏把风险时段推向日落而不断缩水。
+- 出清形成按美元/兆瓦·日支付给全部中标容量的价格；作为交换，资源承担可用性义务，紧急时段不履约将面临严厉罚则（2014 年后的"容量履约"改革，源于 2014 年极地涡旋期间大批机组被冻停）。
+- 当输电限制外来电时，拍卖按位置拆分为多个分区——受限分区以更高价格出清，是 LMP 逻辑在年度尺度上的镜像。
+
+**纯电能量方案（ERCOT）**。不设容量付费；转而允许实时电价在稀缺时冲到极高的上限，让稀有小时供养可靠性。激励更锐利、没有行政需求曲线，但收入呈难以预测的脉冲式到账，而公众对价格尖峰的政治容忍度是有限的。Uri 风暴促使 ERCOT 为这一设计打上补丁（在更低备用水平即触发更高的 ORDC 加价、下调价格上限、新增可靠性服务）——正在向混合模式漂移。
+
+**义务方案（CAISO）**。加州把资源充裕性（RA）义务直接压给各 LSE，由其双边签约采购合格容量；ISO 本身不组织集中容量拍卖。
+
+这是能源经济学中最热闹的辩论之一：容量市场被批评为行政复杂、过度采购、对新技术定价失真；纯电能量市场则被批评为波动剧烈、政治脆弱。当前的需求冲击正在同时对两者进行压力测试。
+
+### 4.3 金融工具
+
+**金融输电权**（FTR）赋予持有者两个节点间日前阻塞价差的收益权。由 RTO 通过拍卖出售（资金来源是 RTO 收取的阻塞盈余），让任何人都能对冲阻塞——发电商锁定电厂到枢纽的价差，负荷对冲自己所在分区，基金则押注哪些约束会起作用。有的市场以义务形式发行（可能出现负收益），有的以期权形式。拍卖收入权（ARR）则把拍卖所得返还给为电网建设出资的输电用户。
+
+**虚拟交易**——增量报价（INC，虚拟供给）和减量申报（DEC，虚拟需求）——是日前市场中的纯金融头寸，自动按实时价格平仓。其功能在于收敛：若日前价格系统性高于实时，虚拟卖方就会获利，直到价差弥合。收敛让日前市场成为诚实的预测，从而改善机组组合。阴暗面：多起操纵案中，交易商利用虚拟交易推动价格，使其在其他头寸（如 FTR）上获利。
+
+**期货与掉期**在交易所（ICE、Nymex）挂牌，按枢纽价格结算，把套保期限延展到数年之外。
+
+---
+
+## 第五章 合约、结算与资金
+
+### 5.1 合约体系
+
+现货市场设定参考价格，但大部分电量早在交割之前就已完成套保：
+
+- **公用事业全额需求与默认供电拍卖**——市场化改革州的公用事业公司通过定期竞争性招标采购其默认零售供电。
+- **购电协议（PPA）**——10–25 年的长期合约，可再生能源融资的支柱：按交付兆瓦时支付固定价格，给项目以可融资的收入流，给买方（公用事业公司，或日益增多的企业买家——科技公司已是最大的 PPA 买家）以套保和绿色属性。
+- **虚拟 PPA**——按枢纽价格结算的纯金融差价合约，无实物交割；企业采购的主流结构。
+- **代加工协议（tolling）**——买方提供燃料并掌握调度权，向电厂业主支付容量费：租厂而非购电。
+- **套保与热耗率期权**——任何大宗商品市场的标准工具箱，以枢纽为计价基准。
+
+### 5.2 结算如何运转
+
+每个市场成员的月度 RTO 账单是净额结算：日前电能头寸按日前 LMP；实时偏差按实时 LMP；辅助服务义务与收入；容量费用/收入；FTR 收益；输电费用；以及 **uplift（市场外补足付款）**——当调度指令使机组无法按报价回收成本时的补齐款项（一个持续存在、且刻意令人不适的残差项：uplift 居高不下，说明市场设计漏掉了某些东西，各 RTO 都在努力把它定价进 LMP）。
+
+### 5.3 传导到零售账单
+
+一张美国零售电费单叠加着：批发电能（LSE 供电组合按 LMP 加权的成本）、容量费用、输电费（FERC 管制）、配电费（州管制），再加公共政策附加。批发电能通常只占总额的三分之一到一半——这就是为什么批发价格的剧烈波动传导到用户账单时往往被稀释；也是为什么当前 PJM 容量价格的飙升在政治上如此爆炸：容量成本几乎不打折扣地直达账单。
+
+---
+
+## 第六章 监督、博弈与执法
+
+市场创造了作弊的激励，这套体系的免疫系统有三层。**事前缓解**：自动筛查标记出拥有局部市场力的机组报价（如 PJM 的"三家关键供应商"测试），在受限区域内将其报价压至基于成本的参考值。**持续监测**：独立市场监测机构不间断筛查市场行为，发布年度《市场运行状况报告》——这是各市场最好的公开分析资料。**执法**：FERC 执法办公室依据 2005 年后获得的反操纵授权开展调查，罚金已达数亿美元量级（摩根大通就其在加州的申报策略支付 4.1 亿美元和解金；多起利用虚拟交易为 FTR 头寸牟利的案件；以及安然时代塑造了今日规则的对倒交易和虚假调度经典案例）。
+
+值得记住那些反复出现的操纵模式，因为它们正好映射着设计的接缝：实物截留（谎报停运）、经济截留（在自己主导的受限区域内远高于成本报价）、跨产品操纵（推动一个价格，在别处的头寸上获利），以及并网排队和检修计划流程中的信息博弈。
+
+---
+
+## 第七章 2020 年代的压力测试
+
+### 7.1 需求冲击
+
+在十五年的平台期之后，美国电力需求重新增长——近年约为每年 2%，由数据中心、制造业回流和电气化牵引。数据中心增长预测主导着系统规划：PJM 预计到 2030 年新增数十吉瓦峰荷，几乎全部来自数据中心。负荷预测这门昔日沉闷的精算功课，如今成了行业里争议最大的数字——因为投机性或重复申报的数据中心接入请求，可能触发数十亿美元量级的采购。
+
+### 7.2 PJM 容量危机：进行中的活案例
+
+PJM 的容量拍卖用三个数据点讲完了这个故事：出清价格从约 29 美元/兆瓦·日（2024/25）涨到 270（2025/26）、329（2026/27），再到触及上限的 333（2027/28）——三年约十倍，且最近一次拍卖的采购量历史上首次低于可靠性目标数吉瓦，年度总成本约 160 亿美元。市场监测机构把涨幅的大头归因于数据中心负荷预测，其中很大部分对应尚未建成的设施。若无一项经谈判达成的价格上限，出清价还会高得多。与此同时，几乎没有新增容量中标——供给被卡在并网排队、许可审批、融资环境和燃机供应链之后。多州州长公开威胁退出 PJM；一些州开始设立数据中心专属费率类别并重新审视税收优惠。
+
+这是教科书式的对撞：市场正在产生它被设计来产生的稀缺信号，而物理与行政机器却阻止供给作出回应——把一个价格信号变成了一场政治危机。
+
+### 7.3 供给瓶颈
+
+超过 2000 吉瓦的发电和储能项目——超过美国现有全部机组容量——正在并网队列中等待；历史上真正建成的排队容量只是少数，从申请到投运的中位时长已超过五年（PJM 为八年）。FERC 第 2023 号令的集群研究、就绪保证金和硬性期限正在提升处理速度；数个 RTO 为可靠性关键资源开辟了快速通道。但更深层的约束是输电：电网薄弱之处的并网成本会爆炸式上升，而美国几乎不建设跨区域输电。第 1920 号令的长期规划要求是结构性答案——以十年为尺度。
+
+### 7.4 设计前沿
+
+定义下一个十年的开放问题：随着 ELCC 数值不断漂移，储能和混合电站在容量机制中如何授信；可灵活调节的大负荷（可削减用电的数据中心）是否应被当作一种资源来规划和定价；负荷与电厂同址接入及其引发的联邦—州管辖权之争；西部日前市场扩张（EDAM 与竞争方案）能否最终把最后一片非 RTO 地区组织起来；以及在零边际成本资源主导的机组结构下，容量市场、纯电能量设计与州政府签约模式究竟谁更具适应力。
+
+---
+
+## 速查术语表
+
+| 术语 | 含义 |
+|---|---|
+| RTO / ISO | 运营电网和批发市场的区域机构 |
+| LMP | 节点边际电价 = 电能 + 阻塞 + 网损 |
+| SCUC / SCED | 日前安全约束机组组合 / 实时安全约束经济调度 |
+| 节点 / 分区 / 枢纽 | 三级定价颗粒度：单一母线 / 负荷聚合区 / 交易参考点 |
+| FTR / ARR | 阻塞对冲工具 / FTR 拍卖收入的分配权 |
+| INC / DEC | 虚拟供给报价 / 虚拟需求申报（日前、纯金融） |
+| UCAP / ICAP | 按停运率折减的容量值 / 铭牌容量 |
+| ELCC | 波动性资源的可靠性授信（等效负荷承载能力） |
+| RPM / BRA / VRR | PJM 容量市场 / 其基础拍卖 / 其行政需求曲线 |
+| ORDC | ERCOT 的稀缺定价备用需求曲线 |
+| RA | 加州的双边资源充裕性义务 |
+| Uplift | 市场外补足付款 |
+| PPA / VPPA | 长期购电协议 / 其金融差价合约形式（虚拟 PPA） |
+| IMM | 独立市场监测机构 |
+| PURPA / 888 / 2000 / 841 / 2222 / 2023 / 1920 号令 | 市场化改革与现代改革的里程碑法律和 FERC 法令 |
+
+*全文完。*

primers/grid/03_china_electricity_markets_en.md

@@ -0,0 +1,188 @@
+# China's Electricity Markets: A Primer
+
+*How the world's largest power system is building markets — the history from plan to price, the players, the layered market architecture, the 2025 reform wave, and the road to a unified national market.*
+
+---
+
+## Introduction: the largest system, the youngest market
+
+China runs the world's largest power system by every measure — capacity approaching 4 TW, consumption near 10,000 TWh — yet its electricity *markets* are among the youngest of any major economy. For most of its history, Chinese electricity was not priced at all in any economic sense: it was allocated. Plants received administratively set tariffs and administratively allocated operating hours; users paid catalog prices set by the state. The past decade has been a deliberate, cautious, and now rapidly accelerating effort to replace that machinery with markets — while never letting go of the state's grip on reliability, affordability, and industrial strategy.
+
+Understanding China's markets requires holding two ideas at once. First, the direction is unmistakably toward markets: by 2025, spot markets were mandated to cover essentially the whole country, and all new wind and solar must sell through the market. Second, the *philosophy* differs from the Anglo-American one: prices in China have traditionally been treated as cost-recovery instruments to be stabilized, not scarcity signals to be unleashed. Every reform is a negotiation between those instincts. The result is a hybrid unlike anything in the US or Europe — and given that China's grid is three times the size of America's, this hybrid is quietly becoming the most consequential electricity market design experiment in the world.
+
+---
+
+## Chapter 1 — From plan to market: the reform arc
+
+### 1.1 The monopoly era and the 2002 separation
+
+Until 2002, a single entity — the State Power Corporation, heir to the Ministry of Electric Power — owned nearly everything. The foundational reform of 2002 (State Council Document No. 5) broke it apart along one axis: **separating plants from the grid** (厂网分开). Generation assets were distributed among five major state-owned generation groups (Huaneng, Datang, Huadian, Guodian, and China Power Investment — later consolidated and joined by others), while the wires went to two grid companies: **State Grid Corporation of China** covering ~88% of the territory, and **China Southern Power Grid** covering five southern provinces. Crucially, the 2002 reform stopped there: the grid companies remained the sole buyer and sole seller of electricity, purchasing from plants at regulated on-grid tariffs and selling to users at regulated catalog prices, with dispatch run administratively from inside the grid companies themselves.
+
+### 1.2 The administrative market (2002–2015)
+
+The next decade ran on two administrative pillars. **Benchmark tariffs**: each province had a regulated on-grid price for coal power (the "coal benchmark"), with renewables later receiving feed-in tariffs pegged above it. **Fair dispatch** (三公调度): rather than merit order, plants of the same class were allocated roughly equal annual operating hours — equitable among state-owned peers, but economically blind, keeping inefficient plants alive and giving no reward to flexibility. Pilot "direct trading" between large users and generators appeared, but the system remained fundamentally a plan.
+
+### 1.3 Document No. 9 (2015): the modern reform charter
+
+The March 2015 "Opinions on Further Deepening Power Sector Reform" (中发9号文) launched the current era with a memorable formula: "control the middle, liberalize the two ends" (管住中间、放开两头). Transmission and distribution would remain regulated monopolies with cost-audited tariffs; generation prices and retail would be progressively opened to competition. The document spawned the institutions of today's market: provincial power exchanges, independent(ish) trading platforms, thousands of registered retail companies, and — from 2017 — the first batch of provincial **spot market pilots** (Guangdong, Shanxi, Shandong, Zhejiang, Gansu, western Inner Mongolia, Sichuan, Fujian).
+
+### 1.4 The 2021 shock and the great liberalization
+
+In late 2021, surging coal prices collided with capped power tariffs; generators lost money on every MWh and power rationing spread across provinces. The crisis forced the boldest pricing move yet: in October 2021, all coal-fired generation was pushed into market trading, with prices allowed to float ±20% around the benchmark (no upward limit for energy-intensive users), and regulated catalog tariffs for all commercial and industrial users were abolished — pushing effectively all C&I consumption into the market. Residents and agriculture kept protected prices. In one stroke, market-traded volume jumped to well over 60% of consumption.
+
+### 1.5 The construction decade: 2022–2026
+
+The subsequent policy cadence has been relentless:
+
+- **2022, Document No. 118 territory**: the NDRC/NEA blueprint for a **unified national power market system** — a multi-layer design of national, regional, and provincial markets, initially targeted to be preliminarily established by 2025 and basically complete by 2030.
+- **2023**: national trial **Basic Rules for the Electricity Spot Market** standardized the spot design; Shanxi and Guangdong graduated from years of trial settlement to formal spot operation; and a **capacity payment for coal power** was introduced from 2024 — explicitly recasting coal as a backup and balancing resource paid for availability rather than volume.
+- **2024**: the overarching **Basic Rules for Power Market Operation** took effect, the first regulation covering the whole market architecture.
+- **2025 — the reform's biggest year**: February's **Document No. 136** pushed *all* new wind and solar into market trading with a CfD-style settlement mechanism (Chapter 4); April's spot-market notice ordered essentially **nationwide spot market coverage by the end of 2025**, with named deadlines for each province; the first national **Basic Rules for the Ancillary Services Market** standardized products and introduced beneficiary-pays cost allocation; and mandatory storage-pairing requirements for renewables were abolished.
+- **2026**: the State Council set the goal of a unified national market **basically established by 2030 and fully established by 2035**, confirming both the direction and the realism about pace.
+
+---
+
+## Chapter 2 — The players
+
+### 2.1 The state
+
+**NDRC (National Development and Reform Commission)** is the super-ministry: it owns pricing policy (through its price bureau), approves major reforms, and issues the numbered documents that structure the entire sector. **NEA (National Energy Administration)**, nominally under NDRC, handles energy planning, market rules, licensing, and supervision, with regional bureaus overseeing market conduct. There is no independent regulator on the FERC model: the referee is also the planner and part-owner of the teams.
+
+**Provincial governments** are arguably the most underrated players. Electricity reform in China is implemented province by province; provinces set local market rules within national frameworks, run their own exchanges, own stakes in generators, protect local industry through electricity pricing, and historically resisted imports that would idle local plants. The "provincial fortress" problem — balkanized markets that trap cheap power inside borders — is the central obstacle the unified-market program exists to break.
+
+### 2.2 The grid companies
+
+**State Grid** and **Southern Grid** are unlike anything in Western markets: state-owned giants that own the wires, run the dispatch centers (there are no independent ISOs — dispatch sits inside the grid companies), operate the two national-level trading platforms through their affiliates (the **Beijing Power Exchange Center** for the State Grid area, the **Guangzhou Power Exchange Center** for the south, handling inter-provincial trade), and historically earned the spread between regulated purchase and sale prices. Reform has converted their regulated revenue to audited transmission-distribution tariffs and forced trading functions into somewhat more independent subsidiaries, but the structural conflict — the market operator, system operator, and largest commercial actor sharing one balance sheet — remains the deepest governance difference from Western designs.
+
+### 2.3 The commercial cast
+
+Generation is dominated by large state-owned groups — the central-government "Big Five" (now reorganized around Huaneng, Datang, Huadian, SPIC, and CHN Energy from the Guodian–Shenhua merger) plus nuclear specialists (CGN, CNNC), the Three Gorges group in hydro, provincial energy groups, and an enormous, genuinely diverse renewables sector including many private developers. **Retail companies** number in the thousands since 2015, though most are thin intermediaries reselling MLT contracts; grid companies still serve residents and act as default supplier. **Users**: all commercial and industrial consumers are in the market; large industrials increasingly trade directly and sign green PPAs. And uniquely, the **carbon market (national ETS)** sits alongside: covering the power sector since 2021 with intensity-based free allocation, it prices coal plants' emissions and interacts (still weakly) with power prices.
+
+---
+
+## Chapter 3 — The market architecture
+
+China's wholesale market is a layered stack, dominated by contracts, with spot markets as the new foundation being slid underneath.
+
+### 3.1 Medium- and long-term contracts: the backbone
+
+Unlike US markets where day-ahead auctions anchor everything, China's system is built on **medium- and long-term (MLT) contracts** — annual, quarterly, monthly, and now intra-month bilateral and centrally-listed trades that historically covered around 80% of market volume, and which policy still requires to cover the large majority of most participants' positions (a stability mandate: the MLT book is the shock absorber that keeps average prices anchored). Reform is making MLT progressively more spot-compatible: contracts are decomposing from flat monthly blocks into time-of-use and even hourly-shaped positions that settle as differences against spot prices — converging on the international model of financial contracts layered over a physical spot market.
+
+### 3.2 Provincial spot markets: the new foundation
+
+The spot design that emerged from the pilots is broadly uniform: **centralized day-ahead and real-time markets at 15-minute granularity**, security-constrained dispatch, with a characteristic pricing asymmetry — **generators are paid location-based prices while load pays a uniform provincial price**. In effect: nodal or zonal on the supply side, single-zone on the demand side — a design closer to Europe's zonal markets than to US nodal LMP, but with locational granularity on generation that Europe lacks.
+
+Shanxi and Guangdong were first into formal continuous operation (December 2023), followed by Shandong and a lengthening list; the April 2025 acceleration notice set province-by-province deadlines — a group including Hubei and Zhejiang to reach formal operation by end-2025, sixteen more provinces to enter continuous trial settlement, and the southern region to push cross-provincial mechanisms — amounting to a mandate for effectively nationwide spot coverage by the end of 2025. Spot prices are bounded by provincial caps and floors (Jiangsu, for example, caps offers at ¥1.5/kWh with a floor of ¥0 for most trading, while several provinces allow negative prices — Zhejiang's pilot saw −¥0.18/kWh during peak solar hours, and Shandong has recorded extended negative-price episodes during holiday solar floods).
+
+### 3.3 Inter-provincial and regional markets
+
+Cross-border trade runs on a separate track: government-framework agreements and grid-to-grid contracts underpin the big west-to-east UHV flows, supplemented by the Beijing exchange's **inter-provincial spot market for surplus renewables** — a mechanism that lets otherwise-curtailed wind and solar find buyers across provincial lines — and by the **southern regional market**, which extended Guangdong's spot trading toward a five-province regional design and stands as the prototype for multi-province market operation. Regional spot markets launched trial operations in late 2024. This layer is where the unified national market will be won or lost: over 80% of traded volume remains intra-provincial, transmission rights and pricing between provinces are still rigid, and exporting provinces retain strong incentives to serve local industry first.
+
+### 3.4 Green power, certificates, and carbon
+
+Three environmental-value markets coexist. **Green electricity trading** (since 2021) lets users buy renewable MWh bundled with their green attributes at a premium, mostly via MLT contracts — the Chinese green PPA. **Green Electricity Certificates (GECs)** are the unbundled instrument, now issued for essentially all renewable generation and increasingly demanded by exporters facing carbon border adjustments and by sectors under new renewable-consumption mandates (steel, cement, polysilicon, and some data centers must cover set shares of consumption with renewables). The **national ETS** prices carbon intensity for the power sector. Coordination among the three — avoiding double-counting, aligning incentives — is an active reform frontier.
+
+### 3.5 Ancillary services and capacity
+
+Ancillary services markets (frequency regulation, reserves, ramping) developed province by province, with a peculiarity: costs were historically shared among *generators* rather than passed to load. The 2025 national basic rules standardize product definitions and shift toward **beneficiary-pays** allocation, folding ancillary costs into user tariffs as spot markets mature — and opening participation to storage, virtual power plants, and demand response. On capacity: rather than auctions, China uses **administered capacity payments** — two-part tariffs for pumped hydro (around ¥500/kW-year), the 2024 coal capacity payment (a set fraction of a reference fixed cost, stepping up over time), and, functionally, the Document 136 mechanism price for renewables — an emerging suite of availability-based payments that substitutes for a capacity market, with the trade-off that prices are set by regulators rather than discovered by auction.
+
+---
+
+## Chapter 4 — Pricing mechanics
+
+### 4.1 The coal benchmark and the band
+
+The anchor of Chinese power pricing remains the provincial **coal benchmark tariff**, a legacy administered price now serving as the reference point for the market band: coal power trades within **±20% of benchmark** (uncapped upward for energy-intensive industries). MLT prices, spot caps, and even renewable mechanism prices are all expressed and negotiated relative to this anchor. It is a deliberately bounded market — enough room for meaningful price formation, not enough for politically intolerable outcomes.
+
+### 4.2 Time-of-use and the retail link
+
+Since C&I catalog tariffs were abolished, business users pay market-based energy costs plus regulated transmission-distribution tariffs, system charges, and surcharges. Provincial **time-of-use schedules** — with peak/valley ratios often 3:1 or wider, and deep-valley periods added around midday solar — remain administratively set but are progressively being aligned with (and in spot provinces, superseded by) actual spot price shapes. Residential and agricultural users stay on protected catalog prices, sustained by cross-subsidy from C&I users — one of the system's most guarded political settlements and a growing tension as industrial users gain the option to procure efficiently.
+
+### 4.3 Document 136: renewables meet the market
+
+The February 2025 reform (formally the "Notice on Deepening Market-Oriented Reform of On-Grid Tariffs for New Energy") is the pivotal renewables policy of the decade. Its content:
+
+- **All new wind and solar output enters market trading** — ending the era of guaranteed offtake at fixed prices, which still covered roughly half of renewable generation when the policy landed.
+- In its place, each province operates a **sustainable pricing settlement mechanism** (机制电价): a two-way contract-for-difference. Projects receive a strike ("mechanism") price for a covered share of their output; when market prices settle below it, they are topped up; when above, they refund the difference.
+- **Existing projects** (pre-June 2025) transition at prices capped by the coal benchmark, preserving legacy economics. **New projects** must win their mechanism price in annual **competitive auctions**, with volumes linked to provincial renewable-consumption quotas and prices capped at the coal benchmark.
+- Unlike European CfDs that typically cover all output, Chinese provinces underwrite only a portion — commonly 40–80% of generation — leaving the remainder fully exposed to market prices.
+- The same document abolished mandatory storage-pairing for new renewable projects, replacing an administrative burden with market incentives (widened peak-valley spreads plus provincial capacity compensation have made standalone storage economic in several provinces).
+
+Early auctions delivered a cold shower: Shandong's first round cleared wind at ¥0.319/kWh but solar at just ¥0.225/kWh — below what many believe is financeable — while Yunnan set ¥0.33 for both. The policy triggered a historic installation rush before the June 2025 cutoff, followed by a market-digesting slowdown. The design's known gaps: annual price-setting versus multi-decade project economics, and no protection against curtailment (the CfD pays only for power actually delivered).
+
+### 4.4 What spot prices are revealing
+
+Where spot markets run, they are doing exactly what theory predicts: midday prices collapse (and go negative) under solar floods; evening peaks widen; peak-valley spreads have expanded sharply in high-renewables provinces; flexible coal, storage, and shiftable demand are being paid for what they actually provide. Average full-load hours of large plants fell notably in 2025 — the statistical signature of a system pivoting from volume to flexibility.
+
+---
+
+## Chapter 5 — The system's distinctive logic
+
+### 5.1 The dual track
+
+China runs market and plan simultaneously and deliberately. Priority-purchase categories (residential, agricultural, some public-interest and legacy contracts) sit outside the market; provincial planners still issue annual generation guidance; inter-provincial framework agreements predate and constrain market trade. The dual-track is shrinking but is a feature, not a transition artifact: it is how reformers ring-fence the politically untouchable while marketizing the rest.
+
+### 5.2 Provincial fortresses
+
+The signature pathology. Provinces are fiscal and political units whose governments answer for local growth, employment, and energy security; a governor gains little from importing cheaper power that idles local plants and taxable industry. The resulting barriers — rigid annual export/import plans, preferential local dispatch, resistance to transmission-capacity release — are why central documents now use pointedly muscular language about "breaking provincial barriers," why regional markets are being pushed hard, and why the unified national market timeline has repeatedly stretched.
+
+### 5.3 Consumption mandates and curtailment
+
+The renewable portfolio standard analog is the **renewable consumption responsibility weight** (消纳责任权重): annual provincial quotas for the share of consumption met by renewables, cascaded to grid companies and users, now extended to specific industries. Quotas, GECs, mechanism-price volumes, and green trading all interlock through this instrument. It exists because of curtailment: with capacity growing faster than local demand and export capability in the northwest, curtailment rates in the worst pockets (parts of Gansu and neighboring regions) have climbed above 30% for solar — the system's clearest signal that integration, not installation, is the binding constraint.
+
+### 5.4 Dispatch and the philosophy of price
+
+The deepest transformation is conceptual. The old system dispatched administratively and treated price as a cost-recovery formula; the new one lets security-constrained economic dispatch replace fair-hours allocation wherever spot markets formally operate, and asks prices to coordinate behavior — while regulators still visibly flinch at volatility, bounding every market with caps, floors, bands, and mechanism prices. China is adopting the machinery of price signals while retaining the instinct of price control; how far the second yields to the first is the single best question to ask of every future reform document.
+
+---
+
+## Chapter 6 — China vs. the US: the same physics, opposite architectures
+
+| Dimension | United States | China |
+|---|---|---|
+| Market coverage | ~2/3 of load in RTO markets; rest regulated | All C&I in market; residents/agriculture protected |
+| Anchor market | Day-ahead auction (spot-centric) | MLT contracts (~80%), spot layering in beneath |
+| Spot design | Nodal LMP both sides | Location-based for generation, uniform provincial price for load; 15-min |
+| System operator | Independent ISOs/RTOs | Dispatch inside the two grid companies |
+| Regulator | Independent (FERC) + states | NDRC/NEA (planner-regulator) + provinces |
+| Capacity adequacy | Auctions (PJM), scarcity pricing (ERCOT), obligations (CAISO) | Administered capacity payments + mechanism prices |
+| Renewables support | Tax credits + markets + state policies | FiT era → Document 136 CfD auctions + consumption quotas |
+| Congestion between regions | Priced (LMP), hedged (FTRs) | Largely administered; framework agreements + emerging inter-provincial spot |
+| Core constraint | Building supply (queues, permitting) | Absorbing supply (integration, provincial barriers) |
+| Price philosophy | Scarcity signal | Cost-recovery instrument, cautiously liberalizing |
+
+The two systems are converging from opposite directions: the US is re-learning planning (transmission mandates, fast-track procurement, capacity interventions) while China is learning prices (spot markets, CfDs, beneficiary-pays). Each is adopting precisely the tool the other's crisis proved necessary.
+
+---
+
+## Chapter 7 — Trajectory and open questions
+
+The official roadmap now reads: nationwide provincial spot coverage (essentially achieved on paper at end-2025), regional markets deepening from the southern prototype, a unified national market basically established by 2030 and fully established by 2035, with inter-provincial trade and renewable allocation as its stated core purpose. The open questions that will define whether the blueprint becomes reality:
+
+1. **Will provinces yield?** Everything depends on whether transmission access, dispatch, and settlement rules genuinely erode provincial fortresses, against entrenched local interests.
+2. **Can renewables finance on auctioned CfDs?** If mechanism prices keep clearing below financeable levels (as early solar auctions suggest), either designs adjust, costs fall further, or the buildout slows toward the ~200 GW/year pace implied by the 2035 target of 3,600 GW.
+3. **Who pays for flexibility?** The suite of administered capacity payments (coal, pumped hydro, storage compensation, mechanism prices) must eventually cohere — or be replaced by scarcity pricing regulators have so far resisted.
+4. **Will an independent operator emerge?** A genuinely unified market strains the model of dispatch and trading inside the grid companies; whether China creates something ISO-like is among the most consequential institutional questions.
+5. **Demand as a resource**: spot price signals are only now reaching commercial users; virtual power plants, storage, and flexible industrial and computing loads are the designated growth area of the next Five-Year Plan.
+
+---
+
+## Quick-reference glossary
+
+| Term | Meaning |
+|---|---|
+| Document No. 9 (2015) | The charter of the current reform round: "control the middle, liberalize the two ends" |
+| Document No. 136 (2025) | All new renewables into the market; CfD-style mechanism price via provincial auctions |
+| Mechanism price (机制电价) | The CfD strike price covering a share of a renewable project's output |
+| MLT | Medium- and long-term contracts, the dominant trading layer |
+| Coal benchmark | Provincial reference tariff; market prices float ±20% around it |
+| Consumption responsibility weight | Provincial renewable-consumption quota (China's RPS analog) |
+| GEC | Green Electricity Certificate, the unbundled renewable attribute |
+| Fair dispatch (三公调度) | The old equal-hours administrative dispatch, being replaced by economic dispatch |
+| Provincial fortress | Provincial protectionism fragmenting the national market |
+| Beijing / Guangzhou PX | The two inter-provincial power exchange centers (State Grid / Southern Grid areas) |
+| Capacity payment | Administered availability payments (coal from 2024, pumped hydro two-part tariff) |
+| Unified national market | The 2030 (basic) / 2035 (full) goal: national-regional-provincial multi-layer market |
+| ETS | The national carbon market, covering power since 2021 |
+
+*End of primer.*

primers/grid/03_china_electricity_markets_zh.md

@@ -0,0 +1,188 @@
+# 中国电力市场入门读本
+
+*全球最大的电力系统如何构建市场——从计划到价格的历史、参与主体、多层市场架构、2025 年改革浪潮，以及通往全国统一电力市场之路。*
+
+---
+
+## 引言：最大的系统，最年轻的市场
+
+无论以何种口径衡量，中国都运行着全球最大的电力系统——装机容量逼近 40 亿千瓦，用电量接近 10 万亿千瓦时——但它的电力*市场*，却是主要经济体中最年轻的之一。在其历史的大部分时间里，中国的电力在经济学意义上根本没有"定价"，而是被"分配"：电厂拿到行政核定的电价和行政分配的利用小时，用户按国家制定的目录电价付费。过去十年，是一场审慎推进、如今急剧加速的努力——用市场替换这套机器，同时始终不放松国家对可靠性、可负担性和产业战略的掌控。
+
+理解中国电力市场需要同时把握两点。第一，方向毫无疑义地指向市场：到 2025 年，现货市场已被要求基本覆盖全国，所有新建风电光伏都必须通过市场售电。第二，其*哲学*不同于英美路径：在中国的传统里，电价是需要稳定的成本回收工具，而非需要释放的稀缺信号。每一项改革都是这两种本能之间的谈判。其结果是一个不同于美国也不同于欧洲的混合体——而鉴于中国电网的体量是美国的三倍，这个混合体正悄然成为全球最具分量的电力市场设计实验。
+
+---
+
+## 第一章 从计划到市场：改革的弧线
+
+### 1.1 垄断时代与 2002 年厂网分开
+
+2002 年之前，几乎一切都归一家机构所有——脱胎于电力工业部的国家电力公司。2002 年的奠基性改革（国务院 5 号文）沿一条轴线将其拆分：**厂网分开**。发电资产分给五大国有发电集团（华能、大唐、华电、国电、中电投——后来经过整合并有新成员加入），电网则归两家电网公司：覆盖约 88% 国土的**国家电网公司**，和覆盖南方五省区的**中国南方电网公司**。关键在于，2002 年改革到此为止：电网公司仍是电力的唯一买方和唯一卖方，按核定上网电价向电厂购电，按目录电价向用户售电，调度则由电网公司内部行政运作。
+
+### 1.2 行政化市场（2002–2015）
+
+此后十年靠两根行政支柱运转。**标杆电价**：每省设定燃煤发电的核定上网价格（"煤电标杆"），可再生能源随后获得高于标杆的固定上网电价（FiT）。**三公调度**：不按边际成本排序，而是给同类机组分配大致相等的年利用小时——在国有同侪之间显得公平，经济上却是盲目的，既让低效机组得以存续，也不给灵活性任何回报。大用户与发电企业之间的"直接交易"试点开始出现，但系统本质上仍是一份计划。
+
+### 1.3 9 号文（2015）：现代改革的纲领
+
+2015 年 3 月《关于进一步深化电力体制改革的若干意见》（中发 9 号文）以一个令人过目难忘的公式开启了当前时代："管住中间、放开两头"。输配电继续作为受管制的垄断环节实行成本核定电价；发电价格和售电侧则渐进开放竞争。这份文件催生了今日市场的各项制度：省级电力交易中心、相对独立的交易平台、数以千计注册的售电公司，以及从 2017 年起的第一批省级**现货市场试点**（广东、山西、山东、浙江、甘肃、蒙西、四川、福建）。
+
+### 1.4 2021 年冲击与大放开
+
+2021 年下半年，飙升的煤价撞上被管制的电价；发电企业每发一度电都在亏钱，多省出现有序用电。危机倒逼出迄今最大胆的价格改革：2021 年 10 月，全部燃煤发电电量进入市场交易，价格允许围绕标杆上下浮动 20%（高耗能用户上浮不设限），同时取消全部工商业用户目录电价——实际上把所有工商业用电推入市场。居民和农业用电保留保护性价格。一步之间，市场化交易电量跃升至用电量的 60% 以上。
+
+### 1.5 建设的十年：2022–2026
+
+此后的政策节奏密不透风：
+
+- **2022 年，118 号文所处的脉络**：发改委、能源局印发**全国统一电力市场体系**蓝图——国家、区域、省级市场多层协同的设计，最初目标是 2025 年初步建成、2030 年基本建成。
+- **2023 年**：《电力现货市场基本规则（试行）》统一了现货设计；山西、广东在多年试结算之后转入正式运行；并自 2024 年起对煤电实行**容量电价**——明确将煤电重新定位为按可用性而非电量获得补偿的兜底与调节资源。
+- **2024 年**：《电力市场运行基本规则》生效，这是覆盖整个市场架构的第一部总纲性规章。
+- **2025 年——改革力度最大的一年**：2 月的 **136 号文**将*所有*新建风电光伏推入市场交易，配套类差价合约的结算机制（见第四章）；4 月的现货市场通知要求 **2025 年底基本实现现货市场全国覆盖**，并为各省点名设定期限；第一部全国性《电力辅助服务市场基本规则》统一了产品定义并引入"谁受益、谁承担"的费用分摊；新能源强制配储要求被废止。
+- **2026 年**：国务院明确全国统一电力市场 **2030 年基本建成、2035 年全面建成**的目标——既确认了方向，也体现了对节奏的现实态度。
+
+---
+
+## 第二章 参与主体
+
+### 2.1 国家
+
+**国家发展改革委**（NDRC）是超级部委：掌管价格政策（通过价格司），审批重大改革，发布为整个行业定调的编号文件。名义上隶属发改委的**国家能源局**（NEA）负责能源规划、市场规则、许可和监管，其派出机构监督市场行为。中国没有 FERC 式的独立监管机构：裁判同时也是规划者，还是场上球队的部分所有者。
+
+**省级政府**堪称最被低估的玩家。中国的电力改革是逐省落地的；各省在国家框架内制定本地市场规则、运营自己的交易中心、持有发电企业股权、通过电价保护本地产业，并且历来抵制会让本省机组闲置的外来电。"省间壁垒"问题——把廉价电力困在省界之内的市场碎片化——正是统一市场工程要攻克的核心障碍。
+
+### 2.2 电网公司
+
+**国家电网**与**南方电网**在西方市场中找不到对应物：它们是拥有线路、内设调度机构（中国没有独立 ISO——调度就在电网公司内部）、通过下属机构运营两大国家级交易平台（国网区域的**北京电力交易中心**、南方区域的**广州电力交易中心**，承担跨省跨区交易）、并曾长期赚取购销价差的国有巨头。改革已将其受管制收入改为核定输配电价，并把交易职能装入相对独立的子机构，但结构性矛盾仍在——市场运营者、系统运营者和最大商业主体共用一张资产负债表，这是与西方设计最深层的治理差异。
+
+### 2.3 商业阵容
+
+发电侧由大型国有集团主导——中央层面的"五大"（几经重组后为华能、大唐、华电、国家电投，以及国电与神华合并而成的国家能源集团），加上核电专业公司（中广核、中核）、水电的三峡集团、各省能源集团，以及一个规模巨大、成分真正多元、包含大量民营开发商的可再生能源板块。**售电公司**自 2015 年以来数以千计，但多数只是转售中长期合约的轻资产中间商；电网公司仍为居民供电并承担保底供电职责。**用户**：全部工商业用户已入市；大型工业用户越来越多地直接交易并签订绿电购电协议。此外还有一个独特的邻居——**全国碳市场（ETS）**：2021 年起覆盖电力行业，采用基于强度的免费配额分配，为煤电的排放定价，并与电价（目前仍较弱地）相互作用。
+
+---
+
+## 第三章 市场架构
+
+中国的批发市场是一个分层堆叠的体系：合约占主导，现货市场作为新地基正被垫入其下。
+
+### 3.1 中长期合约：主心骨
+
+与以日前拍卖为锚的美国市场不同，中国的体系建立在**中长期（MLT）合约**之上——年度、季度、月度乃至月内的双边协商与集中挂牌交易，历史上约占市场电量的 80%，且政策至今仍要求大多数主体的头寸以中长期合约覆盖大头（这是一项稳定性要求：中长期合约簿是把平均电价锚住的减震器）。改革正让中长期合约日益与现货兼容：合约正从平段月度大块分解为分时段乃至逐小时的曲线头寸，并作为差价对现货结算——正在向"金融合约叠加实物现货"的国际模式收敛。
+
+### 3.2 省级现货市场：新地基
+
+试点中成型的现货设计大体统一：**日前与实时市场集中出清、15 分钟颗粒度**、安全约束调度，外加一个标志性的定价不对称——**发电侧按分位置价格结算，用户侧按全省统一价格付费**。也就是说：供给侧是节点/分区定价，需求侧是单一价区——这个设计更接近欧洲的分区市场而非美国的全节点 LMP，但在发电侧拥有欧洲所没有的位置颗粒度。
+
+山西和广东最早转入正式连续运行（2023 年 12 月），山东及一串越来越长的名单紧随其后；2025 年 4 月的加速通知逐省设定了期限——湖北、浙江等一组省份须在 2025 年底前正式运行，另有十六个省份进入连续结算试运行，南方区域同步探索跨省机制——实质上是要求 2025 年底现货市场基本覆盖全国。现货价格受各省上下限约束（例如江苏对多数交易设定 1.5 元/千瓦时的报价上限与 0 元的下限，而多个省份允许负电价——浙江试点曾在光伏大发时段出现 −0.18 元/千瓦时，山东在节假日光伏洪峰期间多次录得较长时间的负价）。
+
+### 3.3 省间与区域市场
+
+跨省交易走的是另一条轨道：政府间框架协议和网对网合同支撑着西电东送的特高压大动脉，辅之以北京交易中心的**省间现货（可再生能源余量）市场**——让本会被弃掉的风光电量跨省找到买家——以及**南方区域市场**：它把广东的现货交易向五省区域设计延伸，是多省市场化运行的原型。区域现货市场于 2024 年底启动试运行。这一层是全国统一市场成败之所系：交易电量的 80% 以上仍在省内完成，省间输电权与价格机制依然僵硬，送端省份始终有优先保障本地产业的强烈动机。
+
+### 3.4 绿电、绿证与碳
+
+三个环境价值市场并存。**绿电交易**（2021 年起）让用户以溢价购买捆绑绿色属性的可再生电量，多以中长期合约成交——这是中国版的绿色 PPA。**绿证**（GEC）是解绑后的凭证工具，如今已对几乎全部可再生能源发电核发，需求端日益扩大：面临碳边境调节的出口企业，以及新纳入可再生能源消费强制要求的行业（钢铁、水泥、多晶硅及部分数据中心须以可再生能源覆盖规定比例的用电）。**全国碳市场**则为电力行业的碳强度定价。三者之间的协调——避免重复计算、理顺激励——是当前活跃的改革前沿。
+
+### 3.5 辅助服务与容量
+
+辅助服务市场（调频、备用、爬坡）逐省发展，且带有一个特色：费用历来在*发电侧*内部分摊，而非传导给用户。2025 年的全国基本规则统一了产品定义，转向"**谁受益、谁承担**"的分摊原则，随着现货市场成熟将辅助服务费用纳入用户电价——并向储能、虚拟电厂和需求响应开放参与。容量方面：中国不搞拍卖，而采用**行政核定的容量补偿**——抽水蓄能两部制电价（约 500 元/千瓦·年）、2024 年起的煤电容量电价（按参考固定成本的一定比例支付并逐步提高），以及在功能上属于同类的 136 号文新能源机制电价——一套以可用性为基础的付费组合，替代了容量市场，其代价是价格由监管者核定而非拍卖发现。
+
+---
+
+## 第四章 价格机制详解
+
+### 4.1 煤电标杆与浮动区间
+
+中国电价体系的锚仍是各省的**煤电标杆电价**——一个遗留的行政核定价格，如今充当市场浮动区间的基准：煤电在**标杆上下 20%** 的范围内交易（高耗能行业上浮不设限）。中长期价格、现货上下限乃至新能源机制电价，都以这个锚为参照来表达和谈判。这是一个刻意设界的市场——留出足以形成有效价格的空间，却不给政治上无法容忍的结果留下余地。
+
+### 4.2 分时电价与零售传导
+
+工商业目录电价取消后，企业用户支付的是市场化的电能成本，加上核定输配电价、系统运行费用和政府性基金附加。各省的**分时电价**——峰谷比常达 3:1 甚至更宽，并在正午光伏大发时段增设深谷——仍由行政制定，但正逐步向真实的现货价格形态靠拢（在现货省份则逐步被其取代）。居民和农业用户继续执行受保护的目录电价，由工商业用户交叉补贴维系——这是这套体系中被守护得最紧的政治安排之一，而随着工业用户获得高效采购的选择权，其张力正在加大。
+
+### 4.3 136 号文：新能源入市
+
+2025 年 2 月的改革（全称《关于深化新能源上网电价市场化改革 促进新能源高质量发展的通知》）是这十年最关键的新能源政策。其内容：
+
+- **所有新建风电光伏电量全部进入市场交易**——终结了固定价格保障性收购的时代；政策落地时，仍有约一半的可再生能源发电依靠保障性收购。
+- 取而代之的是各省运行的**新能源可持续发展价格结算机制（机制电价）**：一个双向差价合约。项目就其纳入机制的电量获得一个执行（"机制"）价格；市场价低于机制价时补齐差额，高于时退回超额部分。
+- **存量项目**（2025 年 6 月前投产）按不超过煤电标杆的价格平稳衔接，保全既有经济性。**新增项目**必须通过每年一度的**竞争性招标**赢得机制电价，纳入机制的电量规模与各省可再生能源消纳权重挂钩，价格上限为煤电标杆。
+- 与通常覆盖全部电量的欧洲 CfD 不同，中国各省只对一部分电量提供托底——普遍在发电量的 40%–80% 之间，其余部分完全暴露于市场价格。
+- 同一文件废止了新建新能源项目的强制配储要求，用市场激励取代行政摊派（峰谷价差的拉大加上各省容量补偿政策，已使独立储能在多个省份具备了经济性）。
+
+首批拍卖泼来一盆冷水：山东第一轮风电以 0.319 元/千瓦时出清，光伏却只有 0.225 元/千瓦时——低于多数人认为可融资的水平；云南则为风光统一定在 0.33 元。政策在 2025 年 6 月截止日前引发了历史性的抢装潮，随后进入市场消化式的放缓。设计上已知的缺口：逐年定价与数十年项目经济性的错配，以及对弃电毫无约束（差价合约只对实际上网的电量付费）。
+
+### 4.4 现货价格正在揭示什么
+
+在现货市场运行之处，它们的表现与理论预测如出一辙：正午价格在光伏洪峰下坍塌（乃至转负）；晚高峰拉高；高比例新能源省份的峰谷价差显著扩大；灵活煤电、储能和可转移负荷开始按其真实贡献获得回报。2025 年大型电厂平均利用小时数的明显下降——正是一个系统从"拼电量"转向"拼灵活性"的统计学签名。
+
+---
+
+## 第五章 这套体系的独特逻辑
+
+### 5.1 双轨制
+
+中国是有意地让市场与计划并行。优先购电类别（居民、农业、部分公益性和存量合同）置于市场之外；省级规划部门仍下达年度发电指导；省间框架协议先于并约束着市场化交易。双轨在收窄，但它是特性而非过渡的残迹：改革者正是用它把政治上不可触碰的部分圈护起来，再把其余部分推向市场。
+
+### 5.2 省间壁垒
+
+标志性的顽疾。省是财政与政治单元，省政府要为本地增长、就业和能源安全负责；进口更便宜的外来电、让本地电厂和纳税产业闲置，对一位省级主官几乎无利可图。由此产生的壁垒——僵硬的年度送受电计划、本地机组优先调度、对释放输电通道能力的抵制——正是中央文件如今用"破除省间壁垒"这类强硬措辞的原因，是区域市场被大力推进的原因，也是全国统一市场时间表一再拉长的原因。
+
+### 5.3 消纳责任与弃电
+
+中国版的可再生能源配额制是**可再生能源消纳责任权重**：逐年下达的省级可再生能源消费占比指标，层层分解到电网企业和用户，如今又延伸到特定行业。权重、绿证、机制电价规模和绿电交易，都通过这一工具相互咬合。它的存在源于弃电：在西北，装机增长快于本地需求和外送能力，最严重地区（甘肃部分区域及邻近地带）的光伏弃电率已升破 30%——这是整个系统最清晰的信号：约束在于消纳，而不在于安装。
+
+### 5.4 调度与价格哲学
+
+最深层的转变是观念性的。旧体制以行政方式调度，把价格当作成本回收公式；新体制在现货市场正式运行之处，让安全约束经济调度取代三公分配，并要求价格去协调各方行为——与此同时，监管者面对波动仍会本能地退缩，用上限、下限、浮动区间和机制电价为每一个市场设界。中国正在采纳价格信号的机器，却保留着价格管制的本能；后者向前者让渡到什么程度，是审视未来每一份改革文件时最值得追问的问题。
+
+---
+
+## 第六章 中美对照：同样的物理，相反的架构
+
+| 维度 | 美国 | 中国 |
+|---|---|---|
+| 市场覆盖 | 约 2/3 负荷在 RTO 市场，其余受管制 | 全部工商业入市；居民、农业受保护 |
+| 锚定市场 | 日前拍卖（现货中心） | 中长期合约（约 80%），现货自下垫入 |
+| 现货设计 | 双侧全节点 LMP | 发电侧分位置定价、用户侧全省统一价；15 分钟 |
+| 系统运营者 | 独立的 ISO/RTO | 调度设在两大电网公司内部 |
+| 监管者 | 独立监管（FERC）+ 各州 | 发改委/能源局（规划者兼监管者）+ 各省 |
+| 容量充裕性 | 拍卖（PJM）、稀缺定价（ERCOT）、义务（CAISO） | 行政核定容量电价 + 机制电价 |
+| 可再生能源支持 | 税收抵免 + 市场 + 州政策 | FiT 时代 → 136 号文 CfD 竞标 + 消纳权重 |
+| 区域间阻塞 | 定价（LMP）、可对冲（FTR） | 以行政安排为主；框架协议 + 新兴省间现货 |
+| 核心约束 | 供给建不出来（排队、许可） | 供给消纳不掉（整合、省间壁垒） |
+| 价格哲学 | 稀缺信号 | 成本回收工具，审慎放开中 |
+
+两套体系正从相反的方向彼此靠近：美国在重新学习规划（输电规划强制令、快速采购通道、容量干预），中国在学习价格（现货市场、差价合约、受益者付费）。双方各自采纳的，恰恰是对方的危机所证明必需的那件工具。
+
+---
+
+## 第七章 走向与悬而未决的问题
+
+官方路线图如今写作：省级现货全国覆盖（2025 年底在纸面上基本实现），区域市场以南方为原型继续深化，全国统一电力市场 2030 年基本建成、2035 年全面建成，其明确的核心使命是省间交易与新能源资源配置。决定蓝图能否成真的开放问题：
+
+1. **省会让步吗**？一切取决于输电通道开放、调度和结算规则能否真正侵蚀省间壁垒，对抗根深蒂固的地方利益。
+2. **拍出来的 CfD 撑得起新能源融资吗**？若机制电价持续出清在可融资水平之下（如早期光伏竞标所示），要么设计调整、要么成本继续下降、要么建设节奏放缓到 2035 年 36 亿千瓦目标所隐含的每年约 2 亿千瓦。
+3. **灵活性由谁买单**？煤电容量电价、抽蓄两部制、储能补偿、机制电价这套行政付费组合终须自洽——或被监管者迄今抗拒的稀缺定价所取代。
+4. **会出现独立运营机构吗**？真正统一的市场将使"调度与交易内嵌于电网公司"的模式承压；中国是否会催生某种类 ISO 的机构，是最具分量的制度问题之一。
+5. **需求成为资源**：现货价格信号才刚刚触达工商业用户；虚拟电厂、储能、可灵活调节的工业与算力负荷，是下一个五年规划点名的增长领域。
+
+---
+
+## 速查术语表
+
+| 术语 | 含义 |
+|---|---|
+| 9 号文（2015） | 本轮改革的纲领："管住中间、放开两头" |
+| 136 号文（2025） | 新能源全面入市；通过省级竞标形成类差价合约的机制电价 |
+| 机制电价 | 覆盖新能源项目部分电量的差价合约执行价格 |
+| 中长期合约（MLT） | 占主导地位的交易层 |
+| 煤电标杆电价 | 各省参考电价；市场价格围绕其上下浮动 20% |
+| 消纳责任权重 | 省级可再生能源消费占比指标（中国版 RPS） |
+| 绿证（GEC） | 解绑的可再生能源环境属性凭证 |
+| 三公调度 | 旧的等利用小时行政调度，正被经济调度取代 |
+| 省间壁垒 | 使全国市场碎片化的省级保护主义 |
+| 北京/广州电力交易中心 | 两大跨省跨区交易平台（国网区域/南网区域） |
+| 容量电价 | 行政核定的可用性付费（煤电自 2024 年起；抽蓄两部制） |
+| 全国统一电力市场 | 2030 年基本建成、2035 年全面建成的国家—区域—省级多层市场目标 |
+| 全国碳市场（ETS） | 2021 年起覆盖电力行业的碳排放权交易体系 |
+
+*全文完。*

primers/grid/04_market_mechanics_companion_en.md

@@ -0,0 +1,171 @@
+# Power Market Mechanics: A Deep-Dive Companion
+
+*Bidding, price formation, and settlement in the US and China — a structured capture of the questions the primers raise once you start pulling threads: how offers actually work, where prices mathematically come from, how contracts and spot markets interlock, and who does what inside China's market machinery.*
+
+*This companion assumes the three primers (the grid, US markets, China markets) as background.*
+
+---
+
+## Part I — Bidding
+
+### 1.1 What a US generator actually submits
+
+Plants bid with both price and quantity — and considerably more. A day-ahead submission is a **multi-part offer**:
+
+- **An energy offer curve**: up to ~10 price-quantity segments. A 500 MW gas plant might offer 200 MW at $32/MWh, the next 150 MW at $38, the next 100 MW at $45, the last 50 MW at $60. The upward slope reflects real physics (efficiency falls off away from the design point) plus margin. Segmentation allows *partial* clearing: at an LMP of $40, this unit is dispatched to 350 MW — its in-the-money segments only.
+- **Start-up cost**: the fuel and wear of firing up — tens of thousands of dollars for a large steam unit.
+- **No-load cost**: the hourly cost of being synchronized before producing the first useful MW.
+- **Operating parameters**: min/max output, ramp rate, minimum run time, minimum down time.
+
+The commitment optimization weighs all parts jointly — paying a start-up cost may be worthwhile for a six-hour evening run but not a one-hour peak. If a committed unit's energy revenues fail to cover its as-offered costs (it was committed for reliability but prices stayed low), it receives a **make-whole payment** — the "uplift" of the settlement chapter. The demand side mirrors the structure with price-quantity bids, though much load still bids as fixed quantity regardless of price.
+
+### 1.2 China: the same engine, three differences
+
+Chinese provincial spot markets imported the clearing architecture — centralized day-ahead and real-time markets, security-constrained dispatch, segmented price-quantity offers, 15-minute granularity. Three differences distinguish the bidding process:
+
+1. **Who bids.** US markets are two-sided: generation offers meet demand bids. Most Chinese markets launched **single-sided** — generators bid, while demand enters as the dispatch center's load forecast; users pay the resulting price without participating in forming it (changing under the 2025 reforms, but demand-side bidding remains the immature half). On the supply side there is a formal tiering: conventional units 报量报价 ("report quantity and price" — full offers), while renewables and typically nuclear participate 报量不报价 ("report quantity, not price") — declaring expected output and clearing as price takers at the front of the stack.
+2. **What's in an offer.** The Chinese offer is closer to an energy-only curve; start-up costs, no-load costs, and detailed parameters play less of a role in the market optimization, with commitment leaning more on the dispatch center's engineering and administrative processes, and start-up/must-run compensation often handled outside the market.
+3. **What the bid is for.** In the US, the day-ahead market sets most physical positions. In China, the MLT contract book covers the large majority of volume, and spot bidding largely determines the settlement of *deviations* — the commercial outcome is mostly decided in annual and monthly contract negotiations, with spot trimming the edges. As spot volumes grow and contracts become hourly-shaped financial instruments, the gap closes.
+
+### 1.3 The sequence: prices are outputs, not inputs
+
+A common misconception is that generators see the day-ahead price and then bid quantities against it. The order is the reverse — **the day-ahead price does not exist until after everyone has bid**:
+
+1. Morning of D−1: offers submitted for every interval of tomorrow (full curves from conventional units; quantity-only from price takers). Load forecast and network model finalized.
+2. Midday of D−1: the clearing optimization runs. Day-ahead prices come into existence as its solution.
+3. Afternoon of D−1: schedules and prices for all 96 intervals are published.
+
+Nobody bids against an observed outcome; bidding is the mechanism by which the price is discovered. Two features feed the contrary intuition: price takers do submit "quantity only," but they are committing to accept whatever price emerges, not reacting to one; and bidders see abundant price *information* beforehand (yesterday's results, historical patterns, published indicative forecasts) — expectations, not the actual number. The one place where participants genuinely act against known prices is the operating day: by then day-ahead prices are published and real-time prices stream continuously, and deviation decisions are made against visible numbers. The two-tempo structure — blind auction to set the reference, visible prices to steer the physics — is the same in Shanxi as in PJM.
+
+---
+
+## Part II — The financial layer (US)
+
+### 2.1 How big a role pure financial players play
+
+Larger than commonly assumed — in some corners, financial players *are* the market:
+
+- **Day-ahead energy**: virtual transactions (INCs/DECs) routinely run at volumes on the order of 10–20% of physical load in markets like PJM and MISO, dominated by specialized proprietary trading firms, and concentrated exactly where day-ahead and real-time prices are likely to diverge — the marginal arbitrageur at the seams.
+- **FTR markets**: the majority of positions in most RTOs are held by banks, funds, and specialist congestion traders rather than by the physical parties the instrument nominally hedges. The perennial controversy — whether FTR auctions systematically underprice congestion, transferring ratepayer-funded value to traders — flared with the **GreenHat default** (PJM, 2018): a three-person firm amassed one of the market's largest FTR portfolios on minimal collateral, defaulted, and socialized ~$180M of losses, forcing a rework of credit rules.
+- **Futures/OTC**: financial institutions supply most liquidity in hub-settled power futures; the population shifted from banks toward hedge funds and commodity specialists after Dodd-Frank and FERC enforcement actions.
+
+The verdict is genuinely double-sided: measurable benefits (better day-ahead/real-time convergence, deeper FTR auction competition) alongside real costs (centrality in marquee manipulation cases — JP Morgan's $410M settlement, the classic scheme of virtual bids at illiquid nodes moving prices that FTR positions profit from — plus uplift contribution and credit risk). The US response has been to tighten plumbing, not remove players. Note the design stance: "pure financial player" is a *designed-in* role with dedicated instruments; China currently has no equivalent — no virtuals, no FTR analog, physical participants only.
+
+### 2.2 FTR vs. the day-ahead market
+
+Different layers entirely. The day-ahead market trades energy — MWh for delivery tomorrow at full LMP. An FTR is a financial contract on the **congestion-component spread** between a source and sink node, acquired in RTO auctions months or years ahead, paying out mechanically against each day's day-ahead results for its whole term. No energy, no losses component, no electrons.
+
+Purpose: FTRs solve a problem locational pricing creates — a generator in cheap-land selling to a buyer in expensive-land bleeds the congestion spread every congested hour; the FTR pays exactly that spread back, converting unpredictable congestion exposure into a known upfront auction cost. Funding closes elegantly: congestion splits prices so load pays more than generators receive; that surplus (congestion revenue) accrues to the RTO and is precisely the pot that funds FTR payouts, with auction proceeds returned via ARRs to transmission customers.
+
+Worked example: buy a June FTR, 100 MW, node A→B, at $2/MWh. In a given June hour, A's congestion component is −$5, B's +$15: the FTR pays (15−(−5))×100 = $2,000. If June congestion averages above $2/MWh, the position profits; if the constraint barely binds, it expires worthless. A physical A→B seller holding the FTR is simply made whole — for a hedger the point is cancellation, not profit. Mental model: the day-ahead market is the casino floor where prices are made daily; the FTR is a side contract whose payout is computed from what happens on that floor — which is why FTR trading is really congestion *forecasting*.
+
+---
+
+## Part III — How US prices are generated
+
+### 3.1 The pipeline
+
+Both day-ahead and real-time prices come from optimization engines, not order books:
+
+1. **Inputs** (morning D−1): multi-part offers, demand bids, virtuals; load/renewable forecasts; network model with line ratings and outages.
+2. **Day-ahead SCUC**: security-constrained unit commitment — a mixed-integer program with millions of variables choosing which units run and at what level, minimizing total as-offered cost subject to nodal energy balance, all transmission limits under normal and post-contingency (N-1) conditions, reserve requirements, and unit physical envelopes. A companion **pricing run** (the dispatch LP with commitments fixed) produces hourly day-ahead LMPs — financially binding, published in the afternoon.
+3. **Reliability check (RAC/RUC)**: if the RTO's own forecast says the financial market cleared short of expected real load, additional units are committed out-of-market (a source of uplift).
+4. **Operating day**: the **state estimator** converts thousands of live measurements into the actual system state; every five minutes **SCED** — now a pure linear program, commitments fixed — re-solves dispatch against actual load, renewables, and flows, issuing setpoints and fresh real-time LMPs.
+
+### 3.2 Where the price mathematically comes from
+
+An LMP is not set by anyone: it is the **shadow price (Lagrange multiplier) of the energy-balance constraint at that node** — the marginal change in total system cost if demand there rose by 1 MW — computed as a byproduct of the optimization's dual solution. All structure falls out automatically: unconstrained, every node equals the marginal unit's offer; when a line binds, its shadow price propagates through shift factors into nodal differences — the congestion component. Nobody decides prices at 10,000 nodes; the solver's duals *are* the prices.
+
+Three refinements: **co-optimization** (energy and reserves clear jointly, so reserve prices embed foregone energy profit); **scarcity adders** (reserve demand curves inject administrative scarcity value into the duals when reserves run short — how $1,000+ prices form without anyone offering that high); **fast-start pricing** (block-loaded peakers aren't "marginal" under strict duals, so most RTOs relax constraints in the pricing run to let them set price — theoretical purity traded for investment signals).
+
+### 3.3 Optimization vs. equation-solving, precisely
+
+Both markets are optimizations; the distinction is the *kind*. Day-ahead SCUC is a **mixed-integer program** — on/off commitment decisions are discrete, making it fundamentally hard. Real-time SCED is a **linear program**: by the operating day the discrete layer is fixed (steam plants don't materialize in five minutes), leaving only continuous dials, solvable in seconds. And because MIPs don't yield clean duals, day-ahead prices actually come from an LP pricing run with commitments fixed — so, precisely, *both* price sets come from linear programs; the day-ahead just requires a monster MIP first to decide which plants exist in that LP. Pure equation-solving does appear in the pipeline — the state estimator and power-flow calculations that reconstruct the grid's physical state — but as the *input* to the optimization, not the source of prices. Summary: optimization generates every price; equations tell the optimizer where the grid is.
+
+---
+
+## Part IV — How China's prices are generated
+
+### 4.1 Price generation is plural: four layers
+
+Where the US answer is "everything comes from SCUC/SCED duals," China's depends on which price you ask about:
+
+1. **Spot provinces — the same engine, modified.** Centralized day-ahead and real-time clearing, security-constrained optimization, 15-minute prices from the solution. Modifications: mostly **single-sided** (demand = the dispatch center's forecast, entered as fixed quantity); the **pricing asymmetry** (locational prices for generators, one uniform provincial price for load); renewables and nuclear clearing as **price takers**; tight administrative **caps and floors**; and the whole computation running inside the grid company's dispatch center rather than an independent ISO, with limited public visibility into the model.
+2. **Contract prices — negotiated and auctioned around the benchmark.** The majority of settled volume takes its price from MLT negotiations and exchange listing/matching sessions, within the mandated ±20% band around the provincial coal benchmark. In spot provinces these contracts settle as differences against spot, so spot increasingly *disciplines* contract pricing without *generating* it.
+3. **Administered prices — set, not discovered.** Transmission-distribution tariffs from cost audits; residential/agricultural catalog prices; time-of-use ratios from provincial price bureaus; the coal capacity payment; pumped-hydro two-part tariffs.
+4. **Mechanism prices — annual auctions.** Document 136's provincial reverse auctions discover the CfD strike price for new renewables — genuine competitive price discovery, but of an annual administered contract, not an operational price.
+
+Non-spot regions have no operational price generation yet: contracts plus regulated tariffs, with administrative balancing. The reform's trajectory is for layer 1 to become the reference that disciplines all others — the role day-ahead LMP plays in America.
+
+### 4.2 The uniform provincial load price, in depth
+
+The design: the optimization computes locational prices; generators settle at them; but consumers pay a single provincial purchase price — conceptually a load-weighted average of the locational results — uniform across the province though fully varying by time (users see the 15-minute peaks and valleys; they don't see place).
+
+Why: intra-provincial locational retail prices are politically near-untouchable (price-as-fairness is the inherited philosophy); the demand side is administratively immature (most users entered the market only in 2021, via thin retailers, without bidding); and it descends structurally from the single-buyer era's uniform catalog prices.
+
+Money consequence: the congestion surplus — which in PJM funds FTR payouts — is collected in provincial settlement and socialized through the uniform price. With no locational basis risk on the demand side, an intra-provincial FTR would have nothing to hedge; hence no FTR analog.
+
+Cost: half the market receives no locational signal. A data center siting next to stranded northern wind vs. in the congested southern load pocket sees identical energy prices; storage earns locational prices discharging but faces the flat price charging; demand response can answer time signals but not place signals — a real cost in a country whose central problem is renewables in the wrong place relative to demand.
+
+Fair comparison: the US isn't symmetric either (generators nodal, most load zonal) — but US zones are smaller, sophisticated users can opt into nodal settlement, and zonal averaging is a settlement convenience atop a nodal market rather than a policy commitment. Europe is symmetric the other way (one price both sides, zone-wide, congestion managed by redispatch). China's asymmetric design is a middle path: locational where investment responds most directly (generation), suppressed where politics are hardest (consumption). Whether demand ever gets locational granularity — even two or three load zones per province — is a quiet design question to watch.
+
+---
+
+## Part V — Contracts and the settlement chain
+
+### 5.1 MLT contracts as contracts-for-differences
+
+Once a spot market exists, dispatch follows offers, not contracts — so a contract cannot be a literal delivery obligation. It becomes purely financial, settled in two layers: (1) everything physical settles at spot — every injected MWh paid the locational spot price, every consumed MWh charged the provincial purchase price; (2) for the contracted volume in each interval, the parties exchange the difference between contract price and the spot reference. Spot below contract: buyer tops the seller up; spot above: seller refunds the excess. Net result: each party achieves exactly the contract price on contracted volume regardless of spot.
+
+Worked interval: contract 10 MWh at ¥420/MWh; spot ¥360; generator produces 12 MWh, factory consumes 9. Spot layer: generator receives 12×360 = ¥4,320; factory pays 9×360 = ¥3,240. Difference layer: factory pays generator 10×(420−360) = ¥600. Net: generator earned contract price on 10 MWh plus spot on its extra 2; factory paid contract price on 10 minus spot value of the 1 MWh unconsumed. Contracted positions perfectly hedged; only deviations face spot.
+
+The discipline mechanism: once contracts settle against spot, signing at ¥0.42 when spot is expected to average ¥0.36 is a ¥0.06 gift — so every MLT negotiation becomes a negotiation about expected spot plus a risk premium, exactly as gas and oil forwards relate to their spot markets. Before spot, MLT prices were anchored to benchmark and bargaining power with no operational truth to check them; after, a contract is a forward on spot. Contract *shape* therefore matters: a flat block is a poor hedge for a profiled consumer — hence the reform push decomposing MLT from flat monthly blocks into time-of-use and hourly-shaped curves, converging on the hub-settled power swap. Caveats: trial-settlement and non-spot provinces remain messier than pure CfD logic; and mandated hedge ratios plus the ±20% band mean the discipline operates within a corridor.
+
+### 5.2 The three-layer settlement chain: where day-ahead fits
+
+The precise chain in the mainstream design:
+
+**MLT contract → settles differences against the day-ahead price. Day-ahead position → settles differences against the real-time price. Real-time → settles against metered physics.**
+
+Worked example across all layers: contract 10 MWh at ¥420; day-ahead cleared 12 MWh at ¥360; actual production 11 MWh; real-time price ¥300. Settlement: day-ahead 12×360 = ¥4,320; real-time buys back the 1 MWh shortfall: (11−12)×300 = −¥300; contract difference: 10×(420−360) = +¥600. Total ¥4,620 for 11 MWh — decomposing exactly into: contract price on 10 contracted MWh (¥4,200), day-ahead price on 2 surplus scheduled MWh (¥720), real-time charge for the 1 MWh delivery miss (−¥300).
+
+The day-ahead price is therefore the system's most important single price, playing three roles: the **hedging reference** (the entire MLT book is economically a strip of forwards on the day-ahead price); the **scheduling instrument** (positions are managed a day early, converting physical risk into priced financial risk); and the top of the **risk gradient** (contracts absorb slow price risk; day-ahead absorbs forecastable daily structure; real-time carries only residual surprises — each layer more volatile but smaller in volume; a well-hedged participant might sit 90% contract / 8% day-ahead / 2% real-time). Design footnotes: provinces vary on the CfD reference (day-ahead, real-time, or blends; day-ahead is mainstream because settling contracts against real-time weakens scheduling incentives), and since demand doesn't bid, the day-ahead price this edifice settles against is generated with one side of the market a spectator — tightening that loop is a central reform goal.
+
+### 5.3 What factors go into the day-ahead price
+
+Four buckets: **Supply curve** — fuel prices (offers ≈ heat rate × fuel + VOM; gas-heavy US regions make day-ahead power nearly a gas derivative; coal plays the role in China; carbon where it binds); the available fleet (outages, maintenance, deratings; hydro conditions with opportunity-cost offering); renewable forecasts (near-zero-priced volume at the front of the stack — the biggest day-to-day driver, with forecast shape setting the midday trough and evening ramp); start-up/no-load lumpiness; bidding behavior (opportunity costs, risk premia, margin where mitigation permits). **Demand** — weather above all; calendar (weekends, holidays — Chinese New Year craters industrial load for weeks); in the US, price-responsive bids and virtuals embedding real-time expectations; in China, the dispatch center's forecast with no expectations channel yet. **Network** — transmission limits and outages (a line out can split a market for a week); interchange (US: expected flows between markets; China: rigid inter-provincial schedules acting as pre-set supply shifters). **Rulebook** — reserve requirements and scarcity mechanisms; caps, floors, bands truncating the distribution; in China the coal-benchmark band plus price-taking renewable volume compressing how much of the supply curve can express itself.
+
+Compression: day-ahead price ≈ marginal fuel cost of the expected marginal unit ± congestion at your location + scarcity adders if tight — where the marginal unit is determined by load forecast × renewable forecast × available fleet. Day-to-day these factors set level and shape; across years, the fleet itself (investment, retirement, fuel mix) moves the whole curve.
+
+---
+
+## Part VI — Institutions and the real-time layer in China
+
+### 6.1 Dispatch center vs. power exchange
+
+Formally separate entities; practically, separate desks of the same family. The **dispatch center** (调度中心) is an internal department of the grid company (national/regional/provincial/local hierarchy) running physical operations *and the spot clearing engines* — the closest analog to an RTO operations floor, but on the grid company's org chart. The **power exchange** (电力交易中心) is a distinct legal entity from the post-2015 reform — Beijing and Guangzhou centers for inter-provincial trade, one exchange per province — running the commercial layer: registration, MLT platforms, green power trading, settlement calculation. Shareholding reform has diluted grid-company ownership of the exchanges ("relatively independent," 相对独立 — the word "relatively" carrying great weight), but the grid company typically remains dominant, and the dispatch/clearing function has not been separated at all. The governance issue: the grid company is simultaneously transmission monopolist, dispatch operator, dominant exchange shareholder, settlement counterparty, default supplier, and (through trading arms) a market participant — the standing question being whether a unified national market eventually requires something ISO-like, independent of both grid companies.
+
+### 6.2 How real-time prices are generated, and what drives them
+
+Mechanics: a rolling SCED-style clearing every 15 minutes inside the provincial dispatch center, fed by the actual system state (SCADA/state estimation), ultra-short-term load and renewable forecasts, current topology — and generator offers **carried over from day-ahead**, with limited or no intraday re-declaration. Prices emerge as the solution's marginal values (locational for generation, uniform provincial for load), bounded by caps and floors, with AGC executing between runs.
+
+Factors, in rough order: **renewable forecast error** (the headline driver — price-taking renewables put the entire miss onto the priced stack; over-delivery pushes real-time toward the floor or negative, under-delivery up); **load forecast error** (doubly important in a single-sided market, where the forecast *is* the demand side); **contingencies** (unit trips, UHV/line outages — repriced within minutes off pre-submitted curves); **inter-provincial schedule adjustments** (lumpy, administratively adjusted flows acting as sudden supply shifts — a bigger real-time factor than inter-market flows in the US); **flexibility scarcity** (ramp and range, not energy: sunset ramps spike prices; holiday-noon solar floods against inflexible coal at technical minimums drive floors and negative prices — Shandong's Labor Day negative-price stretches being the canonical case); **reserve coupling and administrative bounds** (tails clipped by design). Exposure is structurally small — contracts and day-ahead positions insulate most volume — but marginal incentives are large: real-time is where forecast accuracy becomes a revenue skill, storage earns intra-day arbitrage, and flexible units capture ramp scarcity. The pattern of real-time prices is the reformers' clearest empirical readout of what the physical system needs.
+
+### 6.3 "Real-time" is minutes-ahead — and in China, nothing is bid at that stage
+
+No market prices the literal present. The real-time clearing runs shortly before each interval, using ultra-short-term forecasts, issuing setpoints and the price *for the upcoming interval* — ex-ante. The price for 14:00–14:15 was computed around 13:45–14:00. Inside the interval, the second-by-second residual is absorbed by AGC regulation, compensated as an ancillary service (capacity + mileage), not priced as energy. The temporal ladder: contracts (months–years) → day-ahead (12–36 h) → real-time (minutes ahead, 15-min resolution; 5-min in the US) → AGC (continuous, a service not a price). Most designs, China's included, settle on ex-ante prices — forecast noise accepted as the cost of actionable signals.
+
+And in China's mainstream design, the 15-minute price involves **no live bidding at all**: conventional offers are frozen from the day-ahead submission; renewables are represented by output forecasts; demand is a forecast. The engine re-runs the dispatch each interval against updated conditions, walking yesterday's frozen curves — an auction mechanism with no live auction participation, closer to a computed scarcity index than a traded price. The US contrast is partial, not total: US offers are also largely set day-ahead, but intraday updates are allowed under rules, storage re-optimizes continuously, and border transactions are effectively bid every interval — live edges China mostly lacks. The consequence: China's real-time price can reveal stress but almost nothing can respond within the day — the market screams, and only the dispatch engine is listening. Opening intraday re-declaration, admitting storage and VPPs as active real-time bidders, and giving C&I users independent declaration and settlement (the 2025 push) is precisely the program of growing live participants around the clearing engine, so the 15-minute-ahead price graduates from diagnostic readout to genuinely traded price.
+
+---
+
+## One-page synthesis
+
+- A US offer is multi-part (curve + start-up + no-load + parameters); a Chinese offer is mostly an energy curve, with renewables as formal price takers and demand as a forecast.
+- Prices are never bid or set — they are the dual solution of dispatch optimizations: a MIP-then-LP at day-ahead, an LP every 5/15 minutes in real time; equations (state estimation) locate the grid, optimization prices it.
+- The US has one price-generation mechanism feeding everything; China has four layers (spot optimization, banded negotiation, administration, annual CfD auctions), with spot designed to gradually discipline the rest.
+- China's signature asymmetry: locational prices for generators, one provincial price for load — deleting the demand side's locational signal and, with it, any need for FTRs.
+- Contracts in spot provinces are CfDs settling against day-ahead; day-ahead positions settle against real-time; real-time settles against meters — a risk gradient from slow/large to fast/small exposure.
+- "Real-time" everywhere means minutes-ahead ex-ante; in China it is additionally bid-frozen — the near-term reform frontier is precisely adding live participants (re-declaration, storage, VPPs, bidding demand) around the clearing engine.
+
+*End of companion.*

primers/grid/04_market_mechanics_companion_zh.md

@@ -0,0 +1,171 @@
+# 电力市场机制深潜手册
+
+*美国与中国的报价、价格形成与结算——把三本入门读本读完后一旦开始追问就绕不开的问题，做一次结构化的沉淀：报价究竟如何运作、价格在数学上从何而来、合约与现货市场如何咬合，以及中国市场机器内部各方各司何职。*
+
+*本手册以三本入门读本（电网、美国电力市场、中国电力市场）为前置背景。*
+
+---
+
+## 第一部分 报价
+
+### 1.1 美国发电商实际提交的是什么
+
+电厂的申报既有价格也有数量——而且远不止这两样。一份日前申报是一个**多段式报价（multi-part offer）**：
+
+- **电能量报价曲线**：最多约 10 个价格—数量段。一台 50 万千瓦的燃气机组可能这样报：前 200 MW 报 32 美元/MWh，接下来 150 MW 报 38 美元，再 100 MW 报 45 美元，最后 50 MW 报 60 美元。曲线上翘反映真实的物理规律（偏离设计工况效率下降）加上利润空间。分段使**部分出清**成为可能：若 LMP 出清在 40 美元，该机组被调度到 350 MW——只有价内的段被调用。
+- **启动费用**：点火开机的燃料与损耗成本——大型汽轮机组可达数万美元。
+- **空载费用**：并网待发、尚未输出第一度有用电之前的每小时成本。
+- **运行参数**：最小/最大出力、爬坡速率、最小运行时间、最小停机时间。
+
+机组组合优化对所有部分统筹权衡——为一段六小时的晚间运行支付启动费可能划算，为一小时的尖峰则未必。若已开机机组的电能收入不足以覆盖其按报价计算的成本（比如因可靠性需要被开机而价格持续低迷），它将获得**补足付款**——即结算章节里的 uplift。需求侧结构对称，提交价格—数量申报，尽管大量负荷仍以不问价格的固定数量申报。
+
+### 1.2 中国：同一引擎，三点不同
+
+中国省级现货市场引进了这套出清架构——集中式日前与实时市场、安全约束调度、分段价格—数量报价、15 分钟颗粒度。报价流程有三点不同：
+
+1. **谁在报价**。美国市场是双边的：发电报价与需求申报相遇。中国多数市场以**单边**起步——发电侧报价，需求以调度中心的负荷预测形式进入；用户承受出清结果却不参与价格形成（2025 年改革正在改变这一点，但需求侧报价仍是市场中不成熟的那一半）。供给侧还有一个正式分层：常规机组**报量报价**（完整报价曲线），而新能源和通常核电**报量不报价**——申报预计出力，作为价格接受者排在供给堆栈最前端出清。
+2. **报价里有什么**。中国的报价更接近纯电能量曲线；启动费、空载费和精细运行参数在市场优化中的作用较小，机组组合更多依赖调度中心的工程判断和行政流程，启动与必开机组补偿也常在市场之外另行处理。
+3. **报价为了什么**。在美国，日前市场确定大部分实物头寸。在中国，中长期合约簿覆盖了绝大部分电量，现货报价主要决定*偏差*的结算——商业结果大体在年度和月度合约谈判中就已定局，现货只是修边。随着现货电量增长、合约演化为分时曲线化的金融工具，这个差距正在收窄。
+
+### 1.3 时序：价格是输出，不是输入
+
+一个常见误解是：发电商先看到日前价格，再据此申报数量。顺序恰恰相反——**在所有人报价完成之前，日前价格根本不存在**：
+
+1. D−1 上午：提交次日所有时段的报价（常规机组报完整曲线；价格接受者只报数量）。负荷预测与网络模型定稿。
+2. D−1 午间：出清优化运行。日前价格此刻才作为优化解诞生。
+3. D−1 下午：公布次日全部 96 个时段的出清计划与价格。
+
+没有人能对着已观察到的结果报价；报价本身就是价格被*发现*的机制。有两个真实特征容易引出相反的直觉：价格接受者确实"只报数量"——但他们不是在对某个已见的价格作反应，而是承诺*接受任何出清结果*；报价者事前也确实看得到大量价格*信息*（昨日出清结果、同类日的历史规律、部分省份交易机构发布的参考预测）——那是预期，不是实际数字。参与者真正对着已知价格行动的环节是运行日：那时日前价格已公布、实时价格持续滚动发布，偏差决策是对着看得见的数字做出的。这种双节拍结构——盲拍确定基准、明价引导物理——山西与 PJM 别无二致。
+
+---
+
+## 第二部分 金融层（美国）
+
+### 2.1 纯金融玩家的分量有多大
+
+比一般想象的大——在市场的某些角落，金融玩家*就是*市场：
+
+- **日前电能量**：虚拟交易（INC/DEC）在 PJM、MISO 等市场的规模常达实物负荷的 10–20% 量级，由少数专业自营交易公司主导，并且恰恰集中在日前与实时价格最可能分叉之处——缝隙上的边际套利者。
+- **FTR 市场**：多数 RTO 中，FTR 头寸的大头由银行、基金和专业阻塞交易公司持有，而非这一工具名义上要对冲的实物主体。长期争议——FTR 拍卖是否系统性低估阻塞价值，把由电费用户出资形成的价值转移给交易商——在 2018 年 PJM 的 **GreenHat 违约事件**中彻底引爆：一家三人公司以极低的保证金积累了市场上数一数二的 FTR 组合，随后违约，约 1.8 亿美元损失由全体 PJM 成员分摊，倒逼信用规则全面重写。
+- **期货/场外**：枢纽结算的电力期货流动性大半由金融机构提供；《多德—弗兰克法案》和 FERC 的一系列执法之后，主力从银行转向对冲基金和商品交易专业机构。
+
+评价是货真价实的两面性：可度量的收益（日前—实时价格收敛改善、FTR 拍卖竞争加深）与同样真切的代价（在标志性操纵案中占据核心位置——摩根大通 4.1 亿美元和解、在流动性稀薄节点下虚拟单推动价格以令 FTR 头寸获利的经典跨产品手法——外加对 uplift 的贡献和无实物抵押的信用风险）。美国的应对是拧紧管道而非清除玩家。值得注意的设计立场：在美国 RTO 中，"纯金融玩家"是一个*被设计进来*的角色，配有专属工具（虚拟交易、FTR）；中国目前没有对应物——没有虚拟报价、没有 FTR 类似物，参与者仅限实物主体。
+
+### 2.2 FTR 与日前市场的区别
+
+完全是两个层面。日前市场交易的是电能——按节点全额 LMP 交割的次日兆瓦时。FTR 则是关于源节点与汇节点之间**阻塞分量价差**的金融合约，提前数月乃至数年在 RTO 拍卖中购得，在整个持有期内逐日、逐小时按日前结果机械兑付。不含电能分量、不含网损分量，更没有电子流动。
+
+用途：FTR 解决的是分位置定价自身制造的问题——身处低价区、卖给高价区买家的发电商，在每个阻塞小时都在流失阻塞价差；沿该路径持有 FTR 恰好把这笔价差补回，把不可预测的阻塞暴露转化为已知的前期成本（FTR 的拍卖价格）。资金闭环十分优雅：阻塞使价格分裂后，负荷在高价节点支付的总额高于发电在低价节点收到的总额；这笔盈余（阻塞盈余）归集于 RTO，正是兑付 FTR 的资金池，拍卖收入再经 ARR 返还给出资建设电网的输电用户。
+
+算例：3 月以 2 美元/MWh 的拍卖价买入 6 月 A→B 路径 100 MW 的 FTR。6 月某小时，A 的阻塞分量为 −5 美元、B 为 +15 美元：该小时 FTR 兑付 (15−(−5))×100 = 2,000 美元。若 6 月平均阻塞高于所付的 2 美元/MWh，头寸盈利；若约束几乎未起作用，则近乎归零。一个实际从 A 向 B 售电的发电商若持有该 FTR，恰好被补齐——对套保者而言，意义不在盈利而在对消。心智模型：日前市场是每天真实形成价格的赌场大厅；FTR 是场外提前签订的边注合约，其兑付由大厅里发生的一切机械算出——这也是为什么 FTR 交易的本质是阻塞*预测*。
+
+---
+
+## 第三部分 美国价格如何生成
+
+### 3.1 流水线
+
+日前与实时价格都出自优化引擎，而非撮合订单簿：
+
+1. **输入**（D−1 上午）：多段式报价、需求申报、虚拟交易；负荷与新能源预测；含线路限额与停运计划的网络模型。
+2. **日前 SCUC**：安全约束机组组合——数百万变量的混合整数规划，决定开哪些机组、开到什么水平，在满足各节点电力平衡、正常及 N-1 预想事故下的全部输电约束、备用需求和机组物理包络的前提下，使按报价计的总成本最小。随后的**定价运行**（固定开机决策后的调度线性规划）产出每小时日前 LMP——具有财务约束力，午后发布。
+3. **可靠性校核（RAC/RUC）**：若 RTO 自己的预测显示金融市场出清量低于实际预期负荷，则在市场外追加开机（uplift 的来源之一）。
+4. **运行日**：**状态估计器**将数千个实时量测还原为系统真实状态；每 5 分钟运行一次 **SCED**——此时是纯线性规划（开机已定）——对实际负荷、新能源和潮流重新求解调度，下发出力设定值并生成新的实时 LMP。
+
+### 3.2 价格在数学上从何而来
+
+LMP 不由任何人制定：它是**该节点电力平衡约束的影子价格（拉格朗日乘子）**——该节点需求增加 1 MW 时系统总成本的边际变化——作为优化对偶解的副产品被计算出来。所有结构自动涌现：无约束时，各节点价格都等于边际机组报价；某线路达限时，其影子价格经转移分布因子传播为各节点差异——即阻塞分量。没有人在一万个节点上决定价格；求解器的对偶解*就是*价格。
+
+三个细化：**联合优化**（电能与备用同时出清，备用价格自动内含压出力所放弃的电能利润）；**稀缺加价**（备用不足时，备用需求曲线把行政性稀缺价值直接注入影子价格——这就是没人报到 1,000 美元、价格却能到 1,000 美元以上的原因）；**快速启动定价**（整块带载的调峰机组在严格对偶意义下并非"边际"，多数 RTO 在定价运行中放松其约束使之可以定价——以理论纯度换取投资信号）。
+
+### 3.3 优化与解方程，说精确些
+
+两个市场都是优化；区别在于优化的*种类*。日前 SCUC 是**混合整数规划**——开/停机是离散决策，这正是问题的难点所在。实时 SCED 是**线性规划**：到了运行日，离散层已被固定（汽轮机组不可能五分钟内凭空出现），只剩连续旋钮可调，数秒内可解。又因为整数规划给不出干净的对偶价格，日前价格实际上来自固定开机后的 LP 定价运行——所以严格地说，*两套*价格都出自线性规划；日前只是需要先解一个庞大的 MIP，来决定这个 LP 里有哪些电厂。流水线中确实存在纯粹的解方程环节——重建电网物理状态的状态估计与潮流计算——但那是优化的*输入*，不是价格的来源。一句话总结：优化生成系统中的每一个价格；方程只是告诉优化器电网此刻在哪里。
+
+---
+
+## 第四部分 中国价格如何生成
+
+### 4.1 价格生成是多元的：四个层次
+
+美国的答案是"一切出自 SCUC/SCED 的对偶解"，中国的答案则取决于你问的是哪一种价格：
+
+1. **现货省份——同一引擎，加以改造**。集中式日前与实时出清、安全约束优化、15 分钟价格出自优化解。改造之处：多为**单边**（需求 = 调度中心预测，以固定数量进入）；**定价不对称**（发电侧分位置价格，用户侧全省统一价）；新能源和核电作为**价格接受者**出清；较紧的行政**上下限**；且整套计算在电网公司调度中心内部运行而非独立 ISO，模型对外可见度有限。
+2. **合约价格——围绕标杆的谈判与集中交易**。大部分结算电量的价格来自中长期谈判和交易中心的挂牌、集中撮合，被限定在煤电标杆上下 20% 的区间内。在现货省份，这些合约按现货差价结算，因此现货日益*约束*着合约定价，却并不*生成*它。
+3. **行政定价——核定而非发现**。成本审核形成的输配电价；居民、农业目录电价；省价格主管部门制定的分时电价比例；煤电容量电价；抽蓄两部制电价。
+4. **机制电价——年度竞标**。136 号文的省级反向竞标为新建新能源发现差价合约执行价——货真价实的竞争性价格发现，但发现的是一年一定的行政性合约价格，而非运行层面的价格。
+
+非现货地区目前尚无运行层面的价格生成：合约加管制电价，辅以行政平衡。改革的方向，是让第 1 层逐步成为约束其余各层的基准——正是日前 LMP 在美国扮演的角色。
+
+### 4.2 用户侧全省统一价，深入拆解
+
+设计本身：优化算出分位置价格；发电按其结算；但用户支付单一的全省购电价——概念上是各位置价格按负荷加权的平均值——全省空间上统一，时间上则完全随行就市（用户能看到 15 分钟级的峰谷，只是看不到地点差异）。
+
+为什么这样设计：省内分地点的销售电价在政治上近乎不可触碰（价格即公平，是这套体系继承的哲学）；需求侧在行政上尚不成熟（多数用户 2021 年才入市，经由轻资产售电公司参与，且并不报价）；结构上它是单一购买者时代统一目录电价的市场化后裔。
+
+对资金流的影响：注意阻塞盈余的去向——在 PJM 用于兑付 FTR 的那笔钱，在中国被归集于省级结算，通过统一价和系统费用摊回全体用户。需求侧既然没有地点性基差风险，省内 FTR 就无物可保——这正是中国没有 FTR 类似物的关键原因。
+
+代价：市场的整整一半收不到任何位置信号。一座数据中心无论选址在省北的滞销风电旁，还是省南的阻塞负荷中心里，看到的电价一模一样——节点化负荷定价本应创造的选址激励并不存在；储能放电时按位置价格获酬，充电时却面对全省平价，削弱了"专往富余被困之处充电"的动力；需求响应能响应时间信号，却响应不了地点信号。对一个核心难题是"新能源长错了地方"的国家来说，主动删除价格信号的位置维度是实打实的成本——批评者的矛头正指向此处。
+
+公平的比较：美国也并非完全对称（发电按节点、多数负荷按分区均价结算）——但美国的分区比省小，成熟的大用户可选择节点结算，且分区平均只是搭在全节点市场之上的结算便利，而非一项政策承诺。欧洲则朝另一个方向对称（整个价区两侧同价，靠行政再调度管理阻塞）。中国的不对称设计其实是一条独特的中间道路：在投资响应最直接之处（电源怎么建、怎么运行）保留位置信号，在政治最艰难之处（用户付多少钱）将其抹平。需求侧未来是否获得位置颗粒度——哪怕只是把一个省拆成两三个负荷区——是值得长期观察的一个安静的设计问题。
+
+---
+
+## 第五部分 合约与结算链
+
+### 5.1 中长期合约作为差价合约
+
+现货市场一旦存在，调度就跟着报价走而不跟着合约走——合约因此不可能是字面意义上的实物交付义务，而变成纯金融安排，分两层结算：（1）一切实物按现货结算——每一度上网电量按发电侧位置现货价获酬，每一度用电按全省购电价付费；（2）对每个时段内的合约电量，双方互相找补合约价与现货基准价之差。现货低于合约：买方补足卖方；现货高于合约：卖方退回超额。净效果：无论现货如何波动，双方在合约电量上实现的都恰好是合约价格。
+
+时段算例：合约该小时 10 MWh、价格 420 元/MWh；现货 360 元；发电商实发 12 MWh，工厂实用 9 MWh。现货层：发电商收 12×360 = 4,320 元；工厂付 9×360 = 3,240 元。差价层：工厂向发电商支付 10×(420−360) = 600 元。净结果：发电商在 10 MWh 合约电量上拿到合约价，多发的 2 MWh 按现货卖出；工厂在 10 MWh 上按合约价付费，少用的 1 MWh 按现货价值冲回。合约头寸被完美对冲；只有偏差面对现货。
+
+约束机制的原理：合约一旦按现货差价结算，在市场普遍预期现货均价 0.36 元时以 0.42 元签约，就等于买方白送 0.06 元——于是每一场中长期谈判都变成关于现货预期加风险溢价的谈判，与天然气、原油远期同现货的关系如出一辙。现货出现之前，中长期价格锚定标杆和议价能力，没有运行层面的真值可校验；现货出现之后，年度合约就是现货的远期，其价格向市场的现货预期收敛。这就是"约束"：现货不制定合约价格，但它使任何远离现货预期的合约价格对其中一方而言变得非理性。合约的*形状*因此变得关键：一份平段合约对晚高峰用电的光伏买家是糟糕的对冲——正午被保护、傍晚在裸奔——这正是改革推动中长期从月度平段大块分解为分时乃至逐小时曲线的原因，终点是与枢纽结算电力掉期的功能等价。两点诚实的保留：试结算与非现货省份的实际操作比纯粹 CfD 逻辑更混杂；中长期签约比例要求与 ±20% 标杆区间意味着这种约束在一条走廊之内运行。
+
+### 5.2 三层结算链：日前价格的位置
+
+主流设计中的精确链条：
+
+**中长期合约 → 按日前价格结算差价。日前头寸 → 按实时价格结算差价。实时 → 按计量表计的物理量结算**。
+
+跨全部三层的算例：合约 10 MWh、420 元；日前出清 12 MWh、价格 360 元；实际发电 11 MWh；实时价格 300 元。结算：日前层 12×360 = 4,320 元；实时层回购 1 MWh 缺口：(11−12)×300 = −300 元；合约差价层：10×(420−360) = +600 元。合计 4,620 元对应 11 MWh——可精确分解为：10 MWh 合约电量按合约价（4,200 元）、2 MWh 计划盈余按日前价（720 元）、1 MWh 交付缺口按实时价扣款（−300 元）。
+
+日前价格因此是全系统最重要的单一价格，一身三任：**套保基准**（整个中长期合约簿在经济上就是一串以日前价格为标的的远期）；**排产工具**（提前一天管理头寸，把实物风险转化为已定价的金融风险）；**风险梯度的顶端**（合约吸收缓慢的大级别价格风险；日前吸收可预测的日内结构；实时只承载残余的意外——每一层比上一层更波动、但对应的电量更小；一个套保良好的主体可能是 90% 合约 / 8% 日前 / 2% 实时的结构）。设计脚注：各省对差价合约基准的选择确有差异（日前、实时或加权混合；日前基准是主流，因为按实时结算合约会削弱日前市场的地位和准确排产的激励）；而由于需求不报价，这整座大厦所结算的日前价格，仍出自一侧市场缺席的优化——收紧这一环，正是当前改革阶段的核心目标之一。
+
+### 5.3 日前价格由哪些因素决定
+
+四个筐：**供给曲线**——燃料价格（报价 ≈ 热耗率 × 燃料价 + 可变运维；美国气电为主的区域里日前电价近乎天然气衍生品；在中国由煤价扮演此角色；碳价在起作用之处同理）；可用机组（检修、非停、降额；水库水位及水电的机会成本报价——今天省下的水可以明天卖）；新能源预测（近零价电量排在堆栈最前端——高比例新能源系统中日际价格波动的头号驱动，且预测*形状*与水平同等重要：光伏预测决定正午深谷的深度和傍晚爬坡的陡度）；启动/空载成本的块状性；报价行为（机会成本、风险溢价、缓解规则允许范围内的加成）。**需求**——气温高于一切；日历效应（工作日/周末、节假日——春节能把工业负荷压低数周）；美国还有内嵌实时预期的价格响应型申报和虚拟交易，把日前拉向预期实时；中国则是调度中心的预测，暂无来自买方的预期通道。**网络**——输电限额与停运（一条线路检修可让市场分区一周）；联络交换（美国：市场间预期潮流；中国：僵硬的省间计划，等效于出清前就位的供给曲线平移）。**规则**——备用需求与稀缺机制（紧张日即便无故障也抬升电价）；上限、下限与区间截断价格分布；在中国，煤电标杆区间加上价格接受型新能源电量，共同压缩了供给曲线能在价格中表达自己的幅度。
+
+压缩成一行：某小时日前价格 ≈ 预期边际机组的边际燃料成本 ± 所在位置的阻塞 + 紧张时的稀缺加价——而哪台机组是边际机组，由负荷预测 × 新能源预测 × 可用机组共同决定。日际因素决定水平与形状；跨年尺度上，机组结构本身（投资、退役、燃料结构）移动整条曲线——"能源转型"最终就是以对远期曲线上每一个小时的重新定价的方式呈现。
+
+---
+
+## 第六部分 中国的机构与实时层
+
+### 6.1 调度中心与电力交易中心
+
+法律上是两个实体；实际上是同一家族的两张办公桌。**调度中心**是电网公司的内设部门（国、区域、省、地四级），负责物理运行*兼现货出清引擎*——最接近 RTO 运行大厅的存在，只是挂在电网公司的组织架构上。**电力交易中心**是 2015 年后改革创设的独立法人——跨省跨区层面的北京、广州两大中心，加上每省一家——运营商业层：市场注册、中长期交易平台、绿电交易、结算计算。股份制改革不断稀释电网公司对交易中心的持股（"相对独立"——"相对"二字承载千钧），但电网公司通常仍是控股方，而调度/出清职能则完全没有从电网公司剥离。治理症结：电网公司同时是输电垄断者、调度运营者、交易中心控股股东、结算对手方、保底供电商，以及（通过交易机构）市场参与者——悬而未决的问题是，真正统一的全国市场是否终究需要某种独立于两大电网公司之外的类 ISO 机构。
+
+### 6.2 实时价格如何生成，由什么驱动
+
+机制：省级调度中心内部每 15 分钟滚动运行一次 SCED 式出清，输入为系统真实状态（SCADA/状态估计）、超短期负荷与新能源预测、当前网络拓扑——以及**沿用日前申报的发电报价**，日内允许的重新申报很少或没有。价格作为解的边际值产生（发电分位置、用户全省统一），受上下限约束，AGC 在两次出清之间执行设定值。
+
+驱动因素，按大致重要性排序：**新能源预测误差**（高比例新能源省份的头号驱动——价格接受型新能源使整个预测误差全部落在有报价的堆栈上；超发把实时压向下限乃至负值，欠发则推高）；**负荷预测误差**（在单边市场中双倍重要：预测不只是输入，它实际上*就是*需求侧，其误差是需求侧唯一的"动作"方式）；**预想外事故**（机组跳闸、特高压或线路停运、降额——数分钟内沿预先提交的曲线重新定价）；**省间计划调整**（块状的、行政调整的外来电变动，等效于突发的供给曲线平移——在中国是比美国市场间潮流更大的实时因素）；**灵活性稀缺**（稀缺的是爬坡与调节范围而非电量：日落光伏塌方考验爬坡速度；节假日正午的光伏洪峰撞上压至技术出力下限的煤电，正是价格触底与转负的推手——山东劳动节假期的连续负价即经典案例）；**备用耦合与行政边界**（尾部被设计性削平）。结构性暴露很小——合约与日前头寸保护了绝大部分电量——但边际激励巨大：实时市场是预测精度变成收入能力之处，是储能赚取日内价差之处，是灵活机组捕获傍晚爬坡稀缺溢价之处。实时价格的形态——正午触底的频率、日落尖峰的高度——是改革者手中关于物理系统真实需求的最清晰的经验读数。
+
+### 6.3 "实时"其实是提前若干分钟——而在中国，这一环节无人报价
+
+没有任何电力市场为字面意义上的"此刻"定价。实时出清在每个时段开始前不久滚动运行，使用超短期预测，为*即将到来的*时段下发设定值和价格——事前（ex-ante）定价。14:00–14:15 的价格约在 13:45–14:00 之间算出。时段*内部*逐秒的残余不平衡从不由电能市场定价，而由 AGC 调频资源连续吸收，按辅助服务（容量+里程）获得补偿。完整的时间阶梯：合约（数月—数年）→ 日前（12–36 小时）→ 实时（提前数分钟，15 分钟颗粒度；美国为 5 分钟）→ AGC（连续，作为服务补偿而非作为价格）。包括中国在内的多数现代设计以事前价格作为结算基准——接受少量预测噪声，换取可据以行动的信号。
+
+而在中国的主流设计中，这个 15 分钟价格**完全没有现场报价**：常规机组的报价自日前申报后冻结；新能源以出力预测的形式被代表；需求是一份预测。引擎每个时段对着更新后的条件重跑调度，在昨天冻结的报价曲线上上下行走——一个有拍卖机制、却无现场竞拍者的拍卖，与其说是被交易出的价格，不如说是被计算出的稀缺指数。与美国的对照是程度之别而非有无之别：美国的报价也大体在日前确定，但日内允许按规则更新，储能持续再优化，市场边界的交易实际上每个时段都在报价——这些"活的边缘"中国大多还没有。后果：中国的实时价格能*揭示*压力，日内却几乎无人能*响应*压力——市场在呐喊，听见的只有调度引擎。放开日内重新申报、让储能和虚拟电厂成为活跃的实时报价者、要求工商业用户具备独立申报与结算能力（2025 年的推进重点），正是在出清引擎周围培育现场参与者的工程——让这个提前 15 分钟的价格，从诊断读数毕业为真正有人竞价的市场价格。
+
+---
+
+## 一页综述
+
+- 美国的报价是多段式的（曲线+启动+空载+参数）；中国的报价基本是电能曲线，新能源为正式的价格接受者，需求则是一份预测。
+- 价格从不被报出或制定——它们是调度优化的对偶解：日前是"先 MIP 后 LP"，实时是每 5/15 分钟一次的 LP；方程（状态估计）定位电网，优化为其定价。
+- 美国用一套价格生成机制喂养一切；中国有四个层次（现货优化、区间内谈判、行政核定、年度 CfD 竞标），并以现货逐步约束其余各层为设计意图。
+- 中国的标志性不对称：发电分位置定价、用户全省一价——删除了需求侧的位置信号，也随之删除了对 FTR 的任何需求。
+- 现货省份的合约是按日前结算的差价合约；日前头寸按实时结算差价；实时按表计结算——一条从"慢而大"到"快而小"的风险暴露梯度。
+- 各国的"实时"都是提前数分钟的事前价格；在中国它还额外处于报价冻结状态——近期改革的前沿恰恰是在出清引擎周围增添现场参与者（重新申报、储能、虚拟电厂、报价的需求侧）。
+
+*全文完。*

primers/grid/05_trading_under_uncertainty_en.md

@@ -0,0 +1,91 @@
+# Trading Renewables Under Uncertainty: Newsvendor Logic, Quantile Bidding, and the Economics of Forecast Errors
+
+*The mathematical foundation beneath declaration optimization — where it comes from, the canonical results, the literature that established them, and how China's market design changes the classical answers. Companion primer to the venture documents' Pool 1 Economics; self-contained for readers with basic probability.*
+
+---
+
+## 1. The problem, stated once
+
+A renewable producer must commit, before delivery, to a quantity q for each future interval — a day-ahead offer, a declaration, a schedule. Actual production G is random at commitment time. The market pays the committed quantity at one price and settles the deviation (G − q) at another. The producer's question: **what q maximizes expected revenue (or expected utility), given a probabilistic forecast of G and beliefs about the prices?**
+
+This is one of the oldest problems in operations research wearing an electricity costume, and recognizing the costume is half the education.
+
+## 2. Two-settlement revenue and imbalance pricing schemes
+
+Per interval, with day-ahead price P_DA and imbalance settlement prices, revenue is:
+
+$$R(q) = q \cdot P_{DA} + \psi^{+}\,(G-q)^{+} - \psi^{-}\,(q-G)^{+}$$
+
+where (x)⁺ = max(x, 0), ψ⁺ is the price *received* for surplus (over-delivery), and ψ⁻ the price *paid* to cover shortfall (under-delivery). Market designs differ in how ψ± relate to P_DA and to the real-time/balancing price P_RT:
+
+- **Dual pricing (two-price imbalance)** — the classic European balancing design: deviations that *aggravate* system imbalance settle at the (unfavorable) balancing price, deviations that *help* settle at P_DA. Consequence: ψ⁺ ≤ P_DA ≤ ψ⁻ always; deviating is never profitable in expectation, only costly. This is the setting of the canonical results.
+- **Single pricing (one-price imbalance)** — both deviation directions settle at the same real-time price: ψ⁺ = ψ⁻ = P_RT. Deviations can be profitable (over-deliver into a high P_RT and you gain). US real-time markets and China's provincial DA/RT settlement are of this type: the deviation term is (G − q)(P_RT), i.e., R = q·P_DA + (G−q)·P_RT — exactly the identity used throughout the venture documents.
+- **Hybrids** — single pricing plus regulatory penalty envelopes (bands, assessments) that reintroduce dual-pricing-like asymmetry administratively. China's 两个细则 regime makes the Chinese design effectively a hybrid: financial single pricing inside a penalty corridor.
+
+Keep this taxonomy in hand — the optimal-offer answer changes qualitatively across it.
+
+## 3. The newsvendor problem and its classical solution
+
+The 19th-century original: a news vendor stocks Q papers before knowing demand D. Each unsold paper loses the overage cost c_o; each unmet customer loses the underage cost c_u. Expected cost is minimized not at E[D] but at the famous **critical fractile**:
+
+$$Q^{*} = F_D^{-1}\!\left(\frac{c_u}{c_u + c_o}\right)$$
+
+where F_D is demand's CDF. The intuition is a marginal argument: at the optimum, the probability-weighted cost of stocking one more unit (c_o with probability F(Q)) equals the probability-weighted cost of stocking one less (c_u with probability 1 − F(Q)). Asymmetric costs ⇒ deliberately biased decision. The decision-relevant object is the *distribution* F, not the point forecast — a 150-year-old argument for probabilistic forecasting.
+
+## 4. The mapping to renewable offering (dual pricing)
+
+Under dual pricing, define the producer's two unit opportunity costs: offering too *little* forgoes c_u = P_DA − ψ⁺ per surplus MWh (sold at ψ⁺ instead of P_DA); offering too *much* costs c_o = ψ⁻ − P_DA per shortfall MWh (bought back above P_DA). Maximize E[R(q)]:
+
+$$\frac{d\,\mathbb{E}[R]}{dq} = P_{DA} - \psi^{+}\big(1 - F_G(q)\big) - \psi^{-} F_G(q) = 0$$
+
+$$\boxed{\;q^{*} = F_G^{-1}\!\left(\frac{P_{DA} - \psi^{+}}{\psi^{-} - \psi^{+}}\right) = F_G^{-1}\!\left(\frac{c_u}{c_u + c_o}\right)\;}$$
+
+The optimal offer is a **quantile of the production distribution at the newsvendor critical ratio**. If buying back shortfalls is twice as painful as discounting surpluses (c_o = 2c_u), the ratio is 1/3: offer the 33rd percentile of your production forecast — deliberately underbid, because the market punishes shortfalls harder. This is the foundational result of renewable trading: **"offer a quantile, not the mean."** In expectation-maximizing form it needs only the marginal prices' expectations (replace ψ± by E[ψ±] when prices are independent of G — an assumption we are about to break).
+
+Worked micro-example: P_DA = ¥400/MWh, E[ψ⁺] = ¥320, E[ψ⁻] = ¥520. c_u = 80, c_o = 120, ratio = 80/200 = 0.40. If the plant's hourly production forecast has quantiles Q30 = 41 MWh, Q40 = 44, Q50 = 47 — declare 44, three MWh below the median, and that shading is *optimal*, not conservative.
+
+## 5. What single pricing does — and why correlation becomes the whole story
+
+Set ψ⁺ = ψ⁻ = P_RT and expected revenue becomes E[R] = E[G·P_RT] + q·E[P_DA − P_RT]. If prices were independent of the producer's output, q's coefficient is a constant: the risk-neutral optimum is a **corner** (declare the maximum if E[S] > 0, the minimum if E[S] < 0) and the "forecast" is irrelevant to the offer — an old and initially shocking observation about one-price imbalance settlement. Reality restores an interior optimum through three channels, and they are precisely the venture documents' structure:
+
+1. **Error–price correlation.** The producer's error shares weather with the fleet, so ε and S are dependent; expected deviation cost gains the term **Cov(ε, S)** — the "structural bleed." For solar the covariance is adverse on both tails (surplus sells into crashed P_RT; shortfall buys back at spikes), so the effective cost function is again asymmetric and forecast-dependent even under single pricing. This is the literature's "price-taker in volume, price-maker in error" regime, studied for wind under system-imbalance-correlated balancing prices.
+2. **Penalty envelopes.** Regulatory assessments on |G − q| (China's 两个细则, tolerance bands) act as an administrative dual-pricing layer, restoring newsvendor curvature with rule-defined c_u, c_o that vary by band.
+3. **Risk aversion.** Even absent both, variance matters to a leveraged plant; CVaR-constrained bidding (a standard extension) pulls the offer off the corner toward distribution-aware interior solutions.
+
+The general problem — maximize E[q·S] − Cov-adjusted deviation cost − penalty(q) over the *joint* distribution of (G, P_DA, P_RT) — no longer has a closed-form fractile, but the newsvendor logic survives as its skeleton: the optimum still balances marginal over- against under-commitment costs, now computed from a joint, regime-dependent distribution. That is exactly what a modern declaration optimizer does numerically.
+
+## 6. The pinball connection: why forecast evaluation and trading are the same mathematics
+
+The **pinball (quantile) loss** for a quantile forecast \(\hat{q}_\alpha\) is L_α(G, \hat q) = α(G − \hat q)⁺ + (1−α)(\hat q − G)⁺. Compare with the newsvendor cost c_u(G−q)⁺ + c_o(q−G)⁺: they are the same function with α = c_u/(c_u+c_o). Minimizing pinball loss at level α *is* solving a newsvendor with that critical ratio — which yields two practically important corollaries. First, a forecaster trained with pinball loss at the market's critical ratio is directly producing optimal offers; forecasting and bidding are one optimization, not a pipeline. Second, the standard probabilistic-forecast skill score (CRPS = pinball loss integrated over all α) measures exactly the economic quantity a trader cares about, averaged over all possible cost asymmetries — the rare case where the statistician's loss function and the trader's P&L are the same object.
+
+## 7. A literature map
+
+- **Foundations of probabilistic wind power forecasting**: Bremnes (2004) on local quantile regression for wind power; the subsequent decade of ensemble-based predictive densities (Pinson and co-authors; Gneiting's proper-scoring framework giving CRPS/pinball their theoretical footing).
+- **The canonical trading results**: Pinson, Chevallier & Kariniotakis (2007), *Trading wind generation from short-term probabilistic forecasts of wind power* — the quantile-offer result in operational form; earlier strands on imbalance-cost minimization (e.g., Bathurst, Weatherill & Strbac 2002; Skytte's 1999 regression analysis of imbalance prices).
+- **The textbook**: Morales, Conejo, Madsen, Pinson & Zugno, *Integrating Renewables in Electricity Markets* (Springer, 2014) — the standard graduate treatment: newsvendor offering, stochastic programming formulations, dual vs. single pricing, risk measures.
+- **Correlated/endogenous prices**: Zugno, Morales, Pinson & Madsen on price-maker bidding (bilevel/MPEC formulations); the strand analyzing balancing prices as functions of system imbalance, which formalizes Cov(ε, S).
+- **Extensions**: CVaR/risk-averse offering (Conejo et al.'s stochastic programming school); multi-market sequencing (day-ahead → intraday → balancing) and the value of trading closer to delivery; storage-renewable co-optimization as multistage stochastic programs; reinforcement-learning bidding (recent, promising, validation-hungry).
+- **Practice**: European renewable balancing-responsible parties (Statkraft, Axpo, Danske Commodities and peers) have run quantile-based imbalance optimization commercially since roughly the early 2010s — the standing existence proof that the theory monetizes.
+
+## 8. The China mapping — what transfers, what mutates, what is new
+
+| Classical element | Chinese provincial reality | Consequence |
+|---|---|---|
+| Dual-price imbalance | Single-price DA/RT settlement + 两个细则 penalty envelope | Hybrid: newsvendor curvature comes from penalties and from Cov(ε,S), not from ψ± spreads |
+| Price-taking offer, exogenous prices | Fleet-correlated errors in solar-saturated provinces | Cov(ε, S) is first-order; joint (ε, S) modeling is the core asset |
+| Quantity-only offer | 报量报价 where provided: the pair (q, p) | The price bid adds a floor-regime self-curtailment option absent from the classical problem |
+| Continuous liquid intraday for position repair | Limited/no re-declaration | The D−1 offer carries more weight; nowcasting informs operations, not re-trading (yet) |
+| CfD/PPA hedges on own output | Mechanism price vs. a fleet-average reference | Relative-performance economics: beating the fleet average is retained alpha on covered volume — no classical analog |
+| Stationary market microstructure | Frozen offers, non-bidding demand | Spreads unusually physically determined; regime models outperform behavioral ones (a transient, valuable state) |
+
+The honest summary for practitioners: the skeleton (newsvendor logic, quantile thinking, pinball-trained forecasting, stochastic-program offering) transfers intact and is fifteen-plus years validated; the flesh (which distributions, which penalties, which options exist) is province-specific, months old, and partially unwritten — the part where operating in China means contributing to the literature rather than consuming it.
+
+## 9. Five takeaways
+
+1. The optimal offer is a quantile at a critical ratio — asymmetric costs make deliberate bias optimal; "accuracy" is not the objective.
+2. Under single pricing, the game is the joint distribution: Cov(ε, S) plus penalties replace the ψ± spread as the source of curvature.
+3. Forecast evaluation and trading share one loss function (pinball ↔ newsvendor) — build the forecaster for the decision, not for RMSE.
+4. Everything above is established science with a commercial track record; the defensible assets are calibrated local distributions, rule encodings, and data — never the algebra.
+5. Where Chinese rules diverge (price bids, fleet-average CfDs, frozen-offer microstructure), the classical literature goes silent — those are the open problems, and the opportunity.
+
+*End of primer.*

primers/grid/05_trading_under_uncertainty_zh.md

@@ -0,0 +1,91 @@
+# 不确定性下的新能源交易：报童逻辑、分位数报价与预测误差的经济学
+
+*申报优化之下的数学地基——它从哪里来、经典结论是什么、由哪些文献确立，以及中国的市场设计如何改变经典答案。作为创业文档《价值池一经济学》的配套读本；具备基础概率知识即可自足阅读。*
+
+---
+
+## 1. 问题，只说一遍
+
+新能源生产者必须在交付之前，对每个未来时段承诺一个电量 q——一份日前报价、一份申报、一份计划。承诺时刻，实际出力 G 是随机的。市场按一个价格支付承诺电量，按另一个价格结算偏差 (G − q)。生产者的问题是：**给定 G 的概率预测与对价格的判断，什么样的 q 使期望收入（或期望效用）最大？**
+
+这是运筹学最古老的问题之一穿上了电力的外衣——认出这件外衣，就完成了一半的学习。
+
+## 2. 两级结算收入与不平衡定价体制
+
+每时段，日前价格 P_DA 与不平衡结算价格下，收入为：
+
+$$R(q) = q \cdot P_{DA} + \psi^{+}\,(G-q)^{+} - \psi^{-}\,(q-G)^{+}$$
+
+其中 (x)⁺ = max(x, 0)，ψ⁺ 是盈余（超发）*获得*的价格，ψ⁻ 是弥补缺口（欠发）*支付*的价格。市场设计的差别在于 ψ± 与 P_DA、实时/平衡价格 P_RT 的关系：
+
+- **双价制（两价不平衡结算）**——欧洲平衡市场的经典设计：*加剧*系统不平衡的偏差按（不利的）平衡价格结算，*帮助*系统的偏差按 P_DA 结算。推论：恒有 ψ⁺ ≤ P_DA ≤ ψ⁻；偏差在期望上永不盈利、只有成本。经典结论正是在这一设定下得出。
+- **单价制（一价不平衡结算）**——两个方向的偏差按同一实时价格结算：ψ⁺ = ψ⁻ = P_RT。偏差可以盈利（在高 P_RT 时超发即获利）。美国实时市场与中国省级日前/实时结算属于此类：偏差项为 (G − q)·P_RT，即 R = q·P_DA + (G−q)·P_RT——正是创业文档通篇使用的恒等式。
+- **混合制**——单价制之上叠加监管性罚则包络（容差区间、考核），以行政方式重新引入双价式的不对称。两个细则使中国设计实际上是混合制：罚则走廊内的金融单价制。
+
+请握住这套分类——最优报价的答案在其间发生质变。
+
+## 3. 报童问题及其经典解
+
+十九世纪的原型：报童在知道需求 D 之前进货 Q 份报纸。每份卖不掉的报纸损失过量成本 c_o；每个没被满足的顾客损失缺量成本 c_u。期望成本的最小值不在 E[D]，而在著名的**临界分位**：
+
+$$Q^{*} = F_D^{-1}\!\left(\frac{c_u}{c_u + c_o}\right)$$
+
+其中 F_D 是需求的累积分布函数。直觉是一个边际论证：最优点上，多进一份的概率加权成本（概率 F(Q) 下的 c_o）恰等于少进一份的概率加权成本（概率 1 − F(Q) 下的 c_u）。成本不对称 ⇒ 刻意有偏的决策才是最优。决策所需的对象是*分布* F 而非点预测——这是一个有一百五十年历史的、支持概率预测的论证。
+
+## 4. 到新能源报价的映射（双价制）
+
+双价制下，定义生产者的两个单位机会成本：报得太*少*，每兆瓦时盈余放弃 c_u = P_DA − ψ⁺（本可按 P_DA 卖、只按 ψ⁺ 卖）；报得太*多*，每兆瓦时缺口付出 c_o = ψ⁻ − P_DA（高于 P_DA 买回）。最大化 E[R(q)]：
+
+$$\frac{d\,\mathbb{E}[R]}{dq} = P_{DA} - \psi^{+}\big(1 - F_G(q)\big) - \psi^{-} F_G(q) = 0$$
+
+$$\boxed{\;q^{*} = F_G^{-1}\!\left(\frac{P_{DA} - \psi^{+}}{\psi^{-} - \psi^{+}}\right) = F_G^{-1}\!\left(\frac{c_u}{c_u + c_o}\right)\;}$$
+
+最优报价是**出力分布在报童临界比处的分位数**。若买回缺口的痛苦是折价卖出盈余的两倍（c_o = 2c_u），临界比为 1/3：申报出力预测的第 33 百分位——刻意低报，因为市场对缺口的惩罚更重。这就是新能源交易的奠基结论：**"报一个分位数，不报均值。"**在期望最大化形式下只需边际价格的期望（当价格独立于 G 时以 E[ψ±] 代入——这个假设我们马上要打破）。
+
+微型算例：P_DA = 400 元/兆瓦时，E[ψ⁺] = 320 元，E[ψ⁻] = 520 元。c_u = 80，c_o = 120，临界比 = 80/200 = 0.40。若电站该小时出力预测的分位数为 Q30 = 41 MWh、Q40 = 44、Q50 = 47——申报 44，比中位数低 3 MWh，而这份削减是*最优*，不是保守。
+
+## 5. 单价制之下——为何相关性成了整个故事
+
+令 ψ⁺ = ψ⁻ = P_RT，期望收入变为 E[R] = E[G·P_RT] + q·E[P_DA − P_RT]。若价格与生产者出力无关，q 的系数是常数：风险中性最优解是一个**角点**（E[S] > 0 报最大、E[S] < 0 报最小），"预测"与报价无关——这是关于一价不平衡结算的一个古老且初看惊人的观察。现实通过三条通道恢复内点最优，而它们恰是创业文档的结构：
+
+1. **误差—价格相关**。生产者的误差与机群共享天气，故 ε 与 S 相关；期望偏差成本多出 **Cov(ε, S)** 一项——"结构性出血"。对光伏，协方差在两条尾巴上都不利（盈余卖进暴跌的 P_RT；缺口在尖峰时买回），于是即便在单价制下，有效成本函数依然不对称且依赖预测。这就是文献中"电量上是价格接受者、误差上是价格制造者"的情形，风电在与系统不平衡相关的平衡价格下已被系统研究。
+2. **罚则包络**。对 |G − q| 的监管考核（两个细则、容差区间）充当行政性双价层，以规则定义的、分区间的 c_u、c_o 恢复报童式曲率。
+3. **风险厌恶**。即便二者皆无，方差对杠杆化的电站也有意义；CVaR 约束下的报价（标准扩展）把报价从角点拉回依赖分布的内点解。
+
+一般问题——在 (G, P_DA, P_RT) 的*联合*分布上最大化 E[q·S] − 协方差调整的偏差成本 − 罚项(q)——不再有闭式分位解，但报童逻辑作为骨架存活：最优点仍在边际的多报成本与少报成本之间取平衡，只是二者现在要从一个联合的、依状态的分布中算出。这正是现代申报优化器在数值上做的事。
+
+## 6. Pinball 联结：为何预测评估与交易是同一套数学
+
+分位数预测 \(\hat{q}_\alpha\) 的 **pinball（分位数）损失**为 L_α(G, \hat q) = α(G − \hat q)⁺ + (1−α)(\hat q − G)⁺。与报童成本 c_u(G−q)⁺ + c_o(q−G)⁺ 对照：二者是同一个函数，α = c_u/(c_u+c_o)。在水平 α 上最小化 pinball 损失*就是*在该临界比下解报童问题——由此得到两个极具实践意义的推论。其一，以市场临界比对应的 pinball 损失训练的预测器，直接产出最优报价；预测与报价是一个优化，不是一条流水线。其二，标准的概率预测技巧评分（CRPS = pinball 损失对全部 α 的积分）度量的恰是交易者关心的经济量在所有成本不对称度上的平均——统计学家的损失函数与交易员的损益在此罕见地是同一个对象。
+
+## 7. 文献地图
+
+- **概率化风电功率预测的奠基**：Bremnes（2004）用局部分位数回归做风电功率预测；随后十年基于集合的预测密度（Pinson 及合作者；Gneiting 的proper scoring 框架为 CRPS/pinball 提供理论立足点）。
+- **经典交易结论**：Pinson、Chevallier & Kariniotakis（2007）《Trading wind generation from short-term probabilistic forecasts of wind power》——运营形态的分位数报价结论；更早的不平衡成本最小化脉络（如 Bathurst、Weatherill & Strbac 2002；Skytte 1999 对不平衡价格的回归分析）。
+- **教科书**：Morales、Conejo、Madsen、Pinson & Zugno《Integrating Renewables in Electricity Markets》（Springer，2014）——标准研究生教材：报童式报价、随机规划形式化、双价对单价、风险度量。
+- **相关/内生价格**：Zugno、Morales、Pinson & Madsen 的价格制造者报价（双层/MPEC 形式化）；把平衡价格作为系统不平衡函数来分析的脉络——Cov(ε, S) 的形式化出处。
+- **扩展**：CVaR/风险厌恶报价（Conejo 等的随机规划学派）；多市场时序（日前 → 日内 → 平衡）与临近交割交易的价值；储能—新能源联合优化的多阶段随机规划；强化学习报价（新近、有前景、亟需验证）。
+- **实践**：欧洲新能源平衡责任方（Statkraft、Axpo、Danske Commodities 等）自 2010 年代初起商业化运行基于分位数的不平衡优化——理论可变现的常设存在性证明。
+
+## 8. 中国映射——什么原样迁移、什么发生变异、什么是全新的
+
+| 经典要素 | 中国省级现实 | 推论 |
+|---|---|---|
+| 双价不平衡结算 | 单价日前/实时结算 + 两个细则罚则包络 | 混合制：报童曲率来自罚则与 Cov(ε,S)，不来自 ψ± 价差 |
+| 价格接受式报价、外生价格 | 光伏饱和省份中与机群相关的误差 | Cov(ε, S) 是一阶项；(ε, S) 联合建模是核心资产 |
+| 只报数量 | 有规定之处报量报价：二元组 (q, p) | 报价新增经典问题中不存在的触底状态自弃电期权 |
+| 连续流动的日内市场供修复头寸 | 有限/无重新申报 | D−1 报价分量更重；临近预报服务于运营而非再交易（暂时） |
+| 对自身出力的 CfD/PPA 套保 | 机制电价对全场均价参考 | 相对绩效经济学：覆盖电量上跑赢全场平均即留存阿尔法——无经典对应物 |
+| 平稳的市场微观结构 | 报价冻结、需求不报价 | 价差异常地由物理决定；状态模型胜过行为模型（一个短暂而宝贵的状态） |
+
+给从业者的诚实总结：骨架（报童逻辑、分位数思维、pinball 训练的预测、随机规划式报价）原样迁移且经十五年以上验证；血肉（哪些分布、哪些罚则、哪些期权存在）是分省的、只有几个月历史的、部分尚未成文的——在中国运营意味着为这部文献添砖加瓦，而不只是消费它。
+
+## 9. 五条要点
+
+1. 最优报价是临界比处的一个分位数——成本不对称使刻意有偏成为最优；"精度"不是目标函数。
+2. 单价制下，博弈在联合分布上：Cov(ε, S) 加罚则取代 ψ± 价差成为曲率来源。
+3. 预测评估与交易共享同一个损失函数（pinball ↔ 报童）——预测器要为决策而建，不为 RMSE 而建。
+4. 以上一切都是有商业业绩记录的成熟科学；可防御的资产是经校准的本地分布、规则编码与数据——从来不是代数本身。
+5. 在中国规则偏离经典之处（报价、全场均价差价合约、报价冻结的微观结构），经典文献陷入沉默——那里是开放问题，也是机会所在。
+
+*读本全文完。*

primers/newsvendor_en.md

@@ -0,0 +1,59 @@
+# The Newsvendor Problem: A Compact Primer for Electricity Markets
+
+## 1. The Model in One Paragraph
+
+A vendor must decide how many newspapers `q` to stock before knowing the day's demand `D`, a random variable with distribution `F`. Each unit of unmet demand costs `c_u` (the underage cost: lost margin, penalty, or replacement at a higher price); each unit of unsold stock costs `c_o` (the overage cost: purchase price minus salvage). The vendor minimizes expected cost:
+
+```
+minimize over q :   E[ c_u · (D − q)⁺ + c_o · (q − D)⁺ ]
+```
+
+This is the smallest interesting stochastic optimization problem: one decision, one random variable, one period, and — crucially — a closed-form solution that carries the central lesson of the entire field.
+
+## 2. The Critical Ratio: Why the Answer Is a Quantile, Not a Forecast
+
+Take the derivative of expected cost in `q` (or argue marginally: the last unit stocked is worth it if the expected underage it avoids exceeds the expected overage it risks). At the optimum, the marginal benefit `c_u · P(D > q)` equals the marginal cost `c_o · P(D ≤ q)`, giving the **critical ratio** (also called the critical fractile):
+
+```
+F(q*) = c_u / (c_u + c_o)        ⟹        q* = F⁻¹( c_u / (c_u + c_o) )
+```
+
+Three things to notice. First, the optimal decision is a **quantile** of the demand distribution — the mean appears nowhere. If underage is three times as painful as overage, stock at the 75th percentile; if overage dominates, stock below the median. Second, the deterministic habit of "plan against the forecast" is optimal only in the knife-edge case `c_u = c_o` *and* a symmetric distribution — asymmetric costs or a skewed distribution make the point forecast systematically the wrong answer. Third, the solution needs the whole distribution (or at least the right quantile of it), not a point estimate: this is the precise sense in which uncertainty quantification has decision value.
+
+The expected cost at the optimum also has a clean interpretation: it grows with the *spread* of `D` around the chosen quantile. Forecast improvement (variance reduction) and decision improvement (choosing the right quantile) are complementary but distinct levers — a lesson organizations routinely blur.
+
+## 3. The Newsvendor in Electricity, Usually Unnamed
+
+Electricity markets are full of one-shot quantity decisions made before uncertainty resolves and settled asymmetrically afterward. Each is a newsvendor once you identify `q`, `D`, `c_u`, and `c_o`.
+
+**Renewable day-ahead offering.** A wind or solar producer offers quantity `q` into the day-ahead market at price `π_DA`, then settles the deviation between actual production `D` and the offer at imbalance prices. In a dual-price balancing market, under-delivery is bought back at a price above `π_DA` (unit underage cost `c_u` = buy-back premium) and over-delivery is sold at a price below it (`c_o` = sell-down discount). The optimal offer is exactly the critical-ratio quantile of the production distribution — a producer facing harsh shortfall penalties should offer conservatively (a low quantile), and vice versa. This result, due to the trading literature of the 2000s (Bremnes; Pinson, Chevallier, and Kariniotakis), is why quantile forecasts, not point forecasts, are the deliverable that matters for renewable traders, and it is the entry point for the richer multi-market offering models in the stochastic optimization primer.
+
+**Reserve sizing.** How many MW of operating reserve should the system hold against net-load forecast error and outages? Holding an unused MW costs the reserve procurement price (overage); being a MW short costs expected load shedding at the value of lost load, or expensive emergency actions (underage). With `c_u` in the thousands of $/MWh (VOLL) and `c_o` in the tens, the critical ratio sits far in the right tail — which is exactly why reserve requirements correspond to high quantiles (e.g., 95th–99.9th percentile) of the net-load error distribution. Probabilistic reserve methodologies and ramping-product demand curves in modern ISOs are institutionalized newsvendor logic.
+
+**Operating reserve demand curves (ORDCs).** Run the logic in reverse: instead of fixing a quantity, price each increment of reserve at its expected marginal value, `VOLL × LOLP(r)` — the value of lost load times the probability of losing load given reserve level `r`. ERCOT's ORDC is the canonical implementation. The demand curve *is* the newsvendor first-order condition traced out over quantities.
+
+**Retail and fuel procurement.** A retailer contracting energy forward for uncertain customer load faces spot-price exposure on the shortfall and resale losses on the surplus; a gas-fired generator nominating day-ahead pipeline capacity against uncertain dispatch faces imbalance penalties both ways. Both reduce to critical fractiles of load or dispatch distributions, complicated in practice by the correlation between quantity and price (you tend to be short exactly when prices are high), which shifts the effective quantile — a first taste of why joint distributions matter.
+
+## 4. The Forecasting Connection: Pinball Loss Is Newsvendor Loss
+
+The link between this decision problem and modern probabilistic forecasting is exact, not analogical. The **pinball (quantile) loss** used to train and evaluate quantile forecasts,
+
+```
+L_τ(q, d) = τ·(d − q)⁺ + (1 − τ)·(q − d)⁺ ,
+```
+
+is the newsvendor cost function with `c_u = τ` and `c_o = 1 − τ`; minimizing its expectation yields the τ-quantile. So a forecaster minimizing pinball loss at `τ = c_u/(c_u + c_o)` is *directly* solving the trader's or operator's decision problem — forecasting and optimization collapse into one step. This observation underlies the current "decision-focused" or "predict-then-optimize" research agenda: rather than fitting a distribution and optimizing against it, train the forecast on the loss the decision actually incurs. The newsvendor is the one case where this program has a complete, closed-form answer.
+
+## 5. Risk Aversion and Other Extensions
+
+The classical model is risk-neutral. Real participants care about tails, and the standard fix is to replace expected cost with a **CVaR** or mean–CVaR objective. The elegant result: for the newsvendor, CVaR optimization again yields a quantile decision, just at a *shifted* critical ratio — risk aversion tilts the offer toward whichever side has the worse tail. Under the Rockafellar–Uryasev formulation the problem remains a small LP given scenarios, previewing how risk enters the larger models in the stochastic primer without breaking tractability.
+
+Other standard extensions each add one ingredient and point toward a larger literature: **price-dependent demand** (the price-setting newsvendor) toward joint pricing/quantity decisions; **quantity–price correlation** toward portfolio and hedging models; **multi-period inventory with carryover** toward base-stock policies and, in electricity, storage arbitrage (a battery's charge decision against uncertain future prices is a newsvendor chained through time); **unknown distribution** toward the distributionally robust newsvendor — Scarf's 1958 closed-form solution for the worst distribution matching a given mean and variance is one of the earliest DRO results and still a touchstone.
+
+## 6. The Bridge to Two-Stage Stochastic Programming
+
+Formally, the newsvendor *is* a two-stage stochastic program: `q` is the here-and-now decision, and the recourse problem is trivial — buy the shortfall at `c_u` or salvage the surplus at `c_o`, with the recourse cost `Q(q, D) = c_u(D − q)⁺ + c_o(q − D)⁺` available in closed form. Everything in the full stochastic optimization toolkit is this structure with the trivialities removed: replace the scalar `q` with a vector of unit commitments, replace the closed-form recourse with a dispatch LP per scenario, and the closed-form quantile solution disappears — but the *economics* survives intact. Stochastic unit commitment is, at heart, the system operator asking the newsvendor question in a thousand coupled dimensions: how much slow, cheap capacity to commit now, given asymmetric costs of being long (startup and no-load costs wasted) versus short (peakers, scarcity prices, load shedding) against the net-load distribution. Whenever a stochastic model's output seems opaque, the newsvendor is the right sanity check: identify the implicit `c_u/c_o` ratio, and ask whether the model is choosing a sensible quantile.
+
+## 7. Suggested Reading
+
+The model traces to Edgeworth's 1888 bank-reserve analysis and Arrow, Harris, and Marschak (1951) in its modern form; Porteus's *Foundations of Stochastic Inventory Theory* and the survey by Qin et al. (2011) cover the classical theory and extensions. Scarf (1958) is the distribution-free origin. For electricity: Bremnes (2004) and Pinson, Chevallier, and Kariniotakis (2007) establish the quantile-offering result for wind; Morales, Conejo, Madsen, Pinson, and Zugno, *Integrating Renewables in Electricity Markets*, chapter 7, gives the full trading treatment; Hogan's ORDC papers develop the scarcity-pricing incarnation; Rockafellar and Uryasev (2000) supply the CVaR machinery reused throughout.

primers/scuc_sced_en.md

@@ -0,0 +1,88 @@
+# A Primer on SCUC and SCED: The Optimization Engines of Wholesale Electricity Markets
+
+## 1. What These Acronyms Actually Are
+
+Security-Constrained Unit Commitment (SCUC) and Security-Constrained Economic Dispatch (SCED) are the two optimization problems that clear organized wholesale electricity markets in the United States and, in closely related forms, around the world. Every locational price you see published by PJM, MISO, CAISO, ERCOT, SPP, NYISO, or ISO-NE is a byproduct — literally a dual variable or its mixed-integer analogue — of one of these two problems. Understanding them means understanding how roughly two-thirds of US electricity is scheduled and priced.
+
+The two problems answer different questions on different timescales. **SCUC answers "which units should be on?"** — a commitment decision involving startup costs, minimum run times, and hours of lead time, solved primarily in the day-ahead market. **SCED answers "given the units that are on, how much should each produce right now?"** — a dispatch decision solved every five minutes in real time. The "security-constrained" prefix, common to both, means the solution must respect the transmission network not only under normal conditions but also under a defined list of contingencies: the system must survive the sudden loss of any single major element. This is the engineering criterion known as N-1 security, and embedding it in the market optimization is what distinguishes modern LMP markets from simple single-node auctions.
+
+## 2. The Physical and Institutional Setting
+
+Three physical facts shape everything. First, supply and demand must balance continuously — frequency deviates within seconds of an imbalance, and large deviations cascade into blackouts. Second, power flows distribute themselves across the network according to Kirchhoff's laws, not according to contracts; you cannot route power, so congestion on one line is determined by injections everywhere. Third, generators are lumpy and inflexible: a large thermal unit may need 8–24 hours notice to start, cost tens of thousands of dollars per startup, be unable to run below 40–50% of capacity, and be limited in how fast it can ramp.
+
+Institutionally, an Independent System Operator (ISO) or Regional Transmission Organization (RTO) runs a two-settlement market. The **day-ahead market** closes in the morning of the day before the operating day: participants submit offers (generators) and bids (loads and financial traders), and the ISO runs SCUC to produce hourly commitments, schedules, and day-ahead LMPs at which those schedules are financially settled. In the **real-time market**, SCED runs every five minutes against actual conditions, producing real-time LMPs; deviations between real-time delivery and day-ahead schedules settle at real-time prices. Between the two, a **reliability unit commitment (RUC)** pass re-runs a SCUC-like model against the ISO's own load forecast to commit any additional capacity needed for reliability that the financial day-ahead market did not procure.
+
+## 3. The SCED Formulation
+
+SCED is the simpler problem and the right place to start. In its DC-linearized form it is a linear program:
+
+```
+minimize    Σ_g C_g(p_g)                        (offer-based cost)
+subject to  Σ_g p_g = Σ_n d_n                   (system power balance)  → dual λ
+            P_g^min ≤ p_g ≤ P_g^max             (dispatch limits of committed units)
+            |p_g − p_g^prev| ≤ R_g · Δt          (ramp limits over the interval)
+            −F_k ≤ Σ_n PTDF_{k,n} · inj_n ≤ F_k  (base-case line flow limits) → duals μ_k
+            (same flow constraints under each contingency c in the monitored set)
+            reserve constraints (co-optimized regulation, spin, non-spin)
+```
+
+The cost functions C_g are piecewise-linear, built from each generator's offer curve (up to ten price-quantity segments in most ISOs). The network is represented through Power Transfer Distribution Factors (PTDFs) — linear sensitivities mapping nodal injections to line flows under the DC approximation, which neglects losses and reactive power (losses are typically added back through marginal loss factors). Contingency security enters by adding flow constraints computed with post-contingency sensitivities (LODFs, line outage distribution factors): for each monitored line k and each contingency c, the post-outage flow must remain within the line's emergency rating. In practice only a screened "active set" of binding or near-binding constraint pairs is included, because the full set (thousands of lines × thousands of contingencies) is enormous and almost entirely slack.
+
+Modern SCED co-optimizes energy with **operating reserves and regulation**: a unit's headroom can be sold as reserve instead of energy, and the LP chooses the split that minimizes total cost against reserve demand curves. It typically also runs in a short **look-ahead** mode (several 5–15 minute intervals) so that ramp constraints are managed proactively rather than myopically.
+
+## 4. Locational Marginal Prices
+
+The elegance of SCED is that its dual solution *is* the market price. The LMP at node n decomposes as:
+
+```
+LMP_n = λ  +  loss component  −  Σ_k μ_k · PTDF_{k,n}
+```
+
+that is, the system marginal energy price, plus the marginal cost of losses at that location, plus a congestion component built from the shadow prices of binding transmission constraints. When nothing is congested, all nodes see (nearly) the same price. When a constraint binds, prices separate: nodes whose injections would relieve the constraint are paid more, nodes that would aggravate it are paid less — the price system internalizes Kirchhoff's laws. Because SCED is a convex LP, these prices support the dispatch: every generator dispatched between its limits is exactly indifferent at the margin, and no committed generator loses money on its incremental energy at the LMP. Congestion rents collected by the ISO (load pays more than generation receives, in aggregate, when lines bind) fund Financial Transmission Rights, the hedging instruments for congestion risk.
+
+## 5. The SCUC Formulation
+
+SCUC wraps the dispatch problem in binary commitment decisions over a 24–48 hour horizon:
+
+```
+minimize    Σ_t Σ_g [ SU_g · v_{g,t} + NL_g · u_{g,t} + C_g(p_{g,t}) ]
+subject to  u_{g,t} ∈ {0,1}                         (on/off status)
+            v_{g,t} ≥ u_{g,t} − u_{g,t−1}            (startup logic)
+            minimum up and down times                (once on, stay on ≥ UT_g hours, etc.)
+            P_g^min · u_{g,t} ≤ p_{g,t} ≤ P_g^max · u_{g,t}
+            ramp limits, startup/shutdown ramp rates
+            hourly power balance, reserve requirements
+            base-case and contingency transmission limits
+```
+
+The objective now includes **startup costs** (SU, often dependent on how long the unit has been off — hot, warm, cold starts) and **no-load costs** (NL, the cost of running at zero net output), alongside incremental energy. The binary variables and minimum up/down time logic make SCUC a large mixed-integer linear program: a large RTO's day-ahead SCUC spans roughly 1,000+ generators, 36–48 hourly periods, tens of thousands of network constraints after screening, and on the order of 10⁵ binary variables. ISOs solve it within strict market timelines (roughly 1–4 hours of wall-clock budget) to an optimality gap tolerance, typically around 0.1%.
+
+Several modeling refinements matter enormously in practice. **Combined-cycle units** are modeled with configuration-based transitions (1x1, 2x1, etc.), each configuration its own pseudo-unit with transition logic. **Tight MIP formulations** of the up/down-time and ramping polytopes (e.g., the Rajan–Takriti facets and their successors) dramatically reduce solve times — formulation strength has been as important as solver progress in making large SCUC tractable. And **demand bids, virtual bids (INCs/DECs), and price-sensitive load** enter the day-ahead problem symmetrically with generation, which is what makes the day-ahead market a genuinely financial, two-sided auction rather than a pure engineering schedule.
+
+## 6. The Pricing Problem SCUC Creates: Non-Convexity and Uplift
+
+SCUC breaks the clean duality story of SCED. With binary variables, the problem is non-convex, and no uniform set of prices generally supports the solution: a unit committed for reliability may, at the resulting LMPs, fail to recover its startup and no-load costs; conversely a unit might see prices at which it would prefer a different schedule. Markets patch this with **make-whole payments (uplift)**: side payments guaranteeing that any resource following ISO instructions recovers its as-offered costs over the day. Uplift is discriminatory (different resources receive different effective prices) and non-transparent (it is not hedgeable the way LMPs are), so market designers work to minimize it.
+
+Two pricing responses are in active use. **Fast-start pricing** relaxes the binary constraints of online fast-start units so their startup and no-load costs can set the LMP when they are marginal — several ISOs adopted variants after FERC pushed the issue in the late 2010s. More ambitiously, **Extended LMP / convex-hull pricing** seeks prices that minimize total uplift by pricing off the convex hull of the non-convex cost functions; MISO runs an approximation of this. The theoretically clean object — the convex-hull price — is the Lagrangian dual optimum of the commitment problem, but computing it exactly at scale remains hard, and the debate over how far prices should deviate from marginal-cost logic to reflect commitment costs is one of the liveliest in market design.
+
+## 7. How the Pieces Fit Together Across the Operating Timeline
+
+A useful mental model is a cascade of the same core problem re-solved with shrinking horizons and hardening decisions. Day-ahead SCUC (financial, bid-in demand) sets hourly commitments and day-ahead settlements. RUC (reliability pass, forecast demand) tops up physical commitment. Intraday or hour-ahead SCUC processes re-commit fast and intermediate units as forecasts update (e.g., ERCOT's HRUC, CAISO's short-term unit commitment). Real-time SCED, every five minutes, dispatches the committed fleet and sets real-time prices, with a look-ahead of an hour or so; fast-start units can still be started by a real-time commitment layer. Each stage inherits the previous stage's decisions as fixed and re-optimizes what remains flexible — the cascade itself is an implicit, rolling form of recourse against uncertainty, even though each individual solve is deterministic.
+
+This is also where the connection to stochastic optimization sits: production SCUC/SCED are deterministic in their treatment of load and renewable uncertainty (a single forecast, exogenous reserve requirements), while handling *equipment* uncertainty robustly through the N-1 contingency set. Stochastic SCUC — scenarios for net load, commitment as here-and-now, dispatch as recourse — is the research extension of exactly this problem, used today in advisory look-ahead tools and in setting reserve and ramp-product requirements rather than in the clearing engine itself, chiefly because of the pricing, cost-recovery, and transparency questions that scenario-based clearing raises.
+
+## 8. Computation
+
+Day-ahead SCUC at RTO scale is among the largest MIPs solved on a daily industrial cadence. The practical toolkit: state-of-the-art commercial MIP solvers (branch-and-cut with presolve, cutting planes, heuristics); tight formulations as noted above; **lazy-constraint handling of transmission security** — solve with a small active set of network constraints, check the full N-1 set against the solution, add violated constraints, and iterate, which typically converges in a handful of passes because so few constraints ever bind; decomposition legacies (Lagrangian relaxation was the pre-MIP workhorse and still informs pricing theory; Benders variants appear in extensions); and warm-starting from the previous day's solution. SCED, being an LP (or nearly so) with an active-set network model, solves in seconds — its challenge is data pipeline latency and robustness at a five-minute cadence, not optimization difficulty. A notable research thread applies machine learning to accelerate SCUC by predicting active constraint sets, good commitment patterns, or screening contingencies.
+
+## 9. Variations Outside the US
+
+The SCUC/SCED architecture is characteristic of US-style centralized pool markets with locational pricing. European markets differ structurally: day-ahead clearing (EUPHEMIA algorithm) is a zonal, portfolio-based auction in which unit commitment is left to the participants and non-convexities enter through block orders; congestion within zones is managed afterward by TSO redispatch. Australia's NEM runs a five-minute regional dispatch resembling SCED without a day-ahead commitment market. The trade-offs between nodal/centralized-commitment and zonal/self-commitment designs — pricing accuracy and congestion management versus simplicity and participant autonomy — are a standing debate, with Europe gradually adding flow-based coupling and discussions of finer granularity, and the US designs serving as the reference point for locational pricing.
+
+## 10. Current Pressure Points
+
+Several forces are reshaping both problems. **Storage** requires state-of-charge dynamics inside SCUC/SCED and raises subtle questions about opportunity-cost-based offers and whether the ISO or the owner should manage the state of charge. **Renewables and net-load volatility** stress the deterministic-forecast assumption, driving ramping products, more frequent intraday commitment, and interest in stochastic and robust look-ahead tools. **Distributed energy resources and demand flexibility** (FERC Order 2222 in the US) push aggregations of small resources into the same optimization, multiplying model size. **Co-optimization scope** keeps expanding — energy, multiple reserve classes, ramping products, and in some designs day-ahead ancillary demand curves — enlarging the already-large MIP. And **pricing reform** (convex-hull approximations, fast-start pricing, scarcity pricing via operating reserve demand curves) continues to evolve the mapping from these optimizations to the prices that drive investment.
+
+## 11. Suggested Reading
+
+For the market-design foundations: Schweppe, Caramanis, Tabors, and Bohn, *Spot Pricing of Electricity* (1988), the origin of LMP theory; and Hogan's papers on contract networks and FTRs. For formulations: Knueven, Ostrowski, and Watson, "On Mixed-Integer Programming Formulations for the Unit Commitment Problem" (2020) is the modern survey; O'Neill et al., "Efficient market-clearing prices in markets with nonconvexities" (2005) and Gribik, Hogan, and Pope's 2007 paper on ELMP frame the pricing problem; Rajan and Takriti's 2005 report gives the classic tight up/down-time polytope. For practice, the ISOs' own business practice manuals (e.g., PJM Manual 11, MISO's energy and operating reserve market documentation) are unusually informative primary sources on how these models are actually configured and solved.

primers/stochastic_optimization_en.md

@@ -0,0 +1,104 @@
+# A Primer on Stochastic Optimization in Electricity Markets
+
+## 1. Why Uncertainty Is the Central Problem in Power Systems
+
+Electricity is unusual among commodities: it must be produced at nearly the same instant it is consumed, storage remains expensive relative to demand, and the physics of the network constrain how power flows. Market operators must therefore commit resources — deciding which power plants to turn on, how much reserve to hold, how to schedule hydro reservoirs — *before* they know exactly what conditions will materialize. The gap between when decisions are made and when uncertainty resolves is what makes electricity markets a natural home for stochastic optimization.
+
+The principal sources of uncertainty are load (demand fluctuates with weather, behavior, and increasingly with distributed resources behind the meter), renewable generation (wind and solar output can swing by large fractions of installed capacity within hours), equipment availability (generators and transmission lines fail unexpectedly), fuel and electricity prices (relevant for market participants making offering and hedging decisions), and inflows (for hydro-dominated systems, reservoir inflows over weeks to years). Historically, load uncertainty was modest and manageable with deterministic forecasts plus reserve margins. The rapid growth of wind and solar changed that calculus: net load (demand minus renewable output) is far more volatile and harder to forecast, which has pushed both system operators and market participants toward decision-making frameworks that represent uncertainty explicitly rather than papering over it with heuristic margins.
+
+## 2. From Deterministic to Stochastic Formulations
+
+The workhorse problems of electricity markets — economic dispatch, unit commitment, market clearing — are traditionally posed as deterministic optimization problems. Economic dispatch minimizes generation cost subject to a power balance constraint and generator limits, typically as a linear or quadratic program. Unit commitment adds binary on/off decisions, minimum up/down times, startup costs, and ramping limits, making it a mixed-integer program (MIP). Both take a single forecast of demand and renewable output as given.
+
+A stochastic formulation replaces the single forecast with a probability distribution, usually approximated by a finite set of scenarios, and restructures the problem around the timing of decisions relative to the resolution of uncertainty. The key conceptual device is the distinction between *here-and-now* decisions, which must be made before uncertainty resolves and are therefore identical across all scenarios, and *wait-and-see* (recourse) decisions, which can adapt once the outcome is observed. In unit commitment, the commitment of slow-start thermal units is here-and-now: a coal or combined-cycle plant needing several hours to start must be committed the day before. Dispatch levels, fast-start unit commitments, and reserve deployment are recourse: they adjust in real time to the realized net load.
+
+## 3. Two-Stage Stochastic Programming and Stochastic Unit Commitment
+
+The canonical model is the two-stage stochastic program with recourse:
+
+```
+minimize    c'x + E_ω[ Q(x, ω) ]
+subject to  x ∈ X
+
+where  Q(x, ω) = min { q(ω)'y : W y = h(ω) − T(ω) x,  y ≥ 0 }
+```
+
+Here `x` collects first-stage decisions (unit commitments), `Q(x, ω)` is the optimal recourse cost in scenario `ω` (the dispatch cost given the commitment), and the objective minimizes first-stage cost plus *expected* second-stage cost. With a finite scenario set {ω₁, …, ω_S} with probabilities p_s, the expectation becomes a weighted sum and the whole problem is one large "extensive form" MIP: a copy of the dispatch problem for each scenario, all coupled through the shared commitment variables and *nonanticipativity* constraints (first-stage decisions cannot depend on which scenario occurs).
+
+**Stochastic unit commitment (SUC)** applies exactly this template. The day-ahead commitment of slow units is chosen to minimize expected total cost across scenarios of net load, where each scenario's dispatch respects transmission limits, ramping, and reserves. The value of doing this rather than committing against a point forecast is quantified by two standard metrics. The **Value of the Stochastic Solution (VSS)** compares the expected cost of the stochastic solution against the expected cost of implementing the deterministic (expected-value) solution — it measures what representing uncertainty is worth. The **Expected Value of Perfect Information (EVPI)** compares the stochastic solution against a clairvoyant who solves each scenario separately — it bounds what better forecasting could be worth. Studies on realistic systems typically find VSS on the order of a fraction of a percent to a few percent of production cost, which sounds small until multiplied by the billions of dollars flowing through a large market annually.
+
+A closely related operational use is **reserve determination**: instead of fixing reserve requirements exogenously (e.g., "3% of load plus the largest contingency"), a stochastic model determines reserves endogenously — the model holds back exactly as much flexible capacity as the scenario set makes economical, and it holds it in the right locations relative to transmission constraints.
+
+## 4. Multi-Stage Problems and Stochastic Dual Dynamic Programming
+
+Many electricity problems are inherently multi-stage: uncertainty unfolds gradually and decisions interleave with observations. The paradigm example is **hydrothermal scheduling** in reservoir-dominated systems such as Brazil, Norway, Colombia, and New Zealand. Water stored today can displace expensive thermal generation later, but inflows over the coming months are uncertain. Releasing too much risks scarcity if inflows disappoint; hoarding wastes free energy through spill if inflows are strong. The problem spans months to years, with a decision each week or month after observing that period's inflow.
+
+A scenario tree for such a problem explodes combinatorially — with even ten inflow outcomes per stage and 60 monthly stages, the tree is astronomically large. **Stochastic Dual Dynamic Programming (SDDP)**, developed by Pereira and Pinto in 1991 precisely for the Brazilian system, tames this by exploiting stage-wise independence (or Markovian structure) of the randomness. SDDP approximates the future cost function of dynamic programming — the expected cost-to-go as a function of reservoir storage — with an outer envelope of linear cuts (Benders cuts), refined iteratively through forward simulation passes and backward cut-generation passes. Because the cost-to-go function is convex in storage under linear dispatch models, the cuts converge to the true function. SDDP and its descendants are in production use for national hydro scheduling in several countries, and the marginal water values it produces feed directly into market prices in hydro systems.
+
+## 5. Scenario Generation and Reduction
+
+A stochastic program is only as good as its scenarios. Scenario generation is the modeling step where statistical craft matters most, and it has several requirements: scenarios must capture the marginal distributions of each uncertain quantity, the temporal autocorrelation (a windy hour tends to be followed by a windy hour), the spatial correlation across sites (weather systems are large), and the dependence between variables (cold snaps raise load and can affect both wind output and thermal plant availability).
+
+Common approaches include time-series models (ARMA/GARCH on forecast errors), Gaussian copula methods that impose realistic dependence structure on marginal forecast distributions, sampling from ensemble numerical weather predictions (each ensemble member is a physically consistent scenario), and, increasingly, machine-learning generative models. For discrete events such as generator outages, Monte Carlo draws from Markov availability models are standard.
+
+Because solver effort grows with the number of scenarios, practitioners generate a large scenario set and then apply **scenario reduction** — selecting a small representative subset and reweighting it to stay close (in a probability metric such as the Wasserstein distance) to the original set. The Heitsch–Römisch forward-selection and backward-reduction algorithms are the classical tools. A practical warning: aggressive reduction tends to discard exactly the extreme scenarios that drive reserve and commitment decisions, so reduction criteria are often modified to preserve tails.
+
+## 6. Alternatives and Cousins: Robust, Chance-Constrained, and Distributionally Robust Optimization
+
+Stochastic programming optimizes an expectation, which requires a credible probability distribution and accepts poor outcomes in bad scenarios if they are improbable. Several alternative frameworks address different attitudes toward uncertainty, and all have found electricity applications.
+
+**Robust optimization (RO)** dispenses with probabilities entirely: uncertainty lives in a set (e.g., net load within a polyhedral band with a "budget of uncertainty" limiting how many hours can be simultaneously extreme), and the model minimizes cost under the *worst case* in that set. Two-stage adaptive robust unit commitment became prominent in the early 2010s, with ISO New England among the first system operators to study it seriously. RO's appeal is that it needs no distribution and yields strong reliability guarantees; its drawback is conservatism — the worst case may be wildly expensive to hedge. The budget parameter gives the operator a dial between economy and security.
+
+**Chance-constrained optimization** requires constraints to hold with at least a specified probability — for example, "transmission line flows remain within limits with probability 0.99 given wind forecast errors." Under Gaussian error assumptions, many chance constraints reformulate into tractable second-order cone constraints, which made chance-constrained DC optimal power flow a popular research vehicle. It offers a transparent reliability knob but inherits sensitivity to the assumed distribution.
+
+**Distributionally robust optimization (DRO)** hedges between the two: it optimizes against the worst distribution within an ambiguity set (distributions matching observed moments, or within a Wasserstein ball around the empirical distribution). DRO has become an active research direction for reserve procurement and OPF because it acknowledges that the distribution itself is estimated from limited data.
+
+A rough practical taxonomy: use stochastic programming when you trust your scenarios and care about expected cost; robust optimization when failure is unacceptable and distributions are unreliable; chance constraints when reliability targets are naturally probabilistic; DRO when you have data but not enough to trust it fully.
+
+## 7. Stochastic Market Clearing and the Pricing Problem
+
+Using stochastic optimization inside the *market clearing* itself — not merely as an operator's advisory tool — raises deep economic questions. In a stochastic market clearing, the day-ahead and real-time (balancing) stages are co-optimized: day-ahead schedules are chosen to minimize expected total cost across scenarios, anticipating balancing actions. Research consistently shows this improves expected system cost relative to sequential, deterministic clearing, especially at high renewable penetration, because day-ahead schedules pre-position flexibility where scenarios say it will be needed.
+
+The difficulties are institutional and economic rather than computational:
+
+**Pricing and cost recovery.** In a deterministic convex market, locational marginal prices (LMPs) from the dual variables support the dispatch — every scheduled resource at least breaks even at the price. In stochastic market clearing, no single set of day-ahead prices generally supports the solution scenario-by-scenario: prices that guarantee cost recovery *in expectation* may leave participants losing money in particular realized outcomes, which strains acceptability. Research on "equilibrium-supporting" pricing schemes and revenue-adequacy properties of stochastic clearing remains active, but no major market has adopted full stochastic clearing, in part for this reason.
+
+**Fairness and transparency.** Scenario choices directly shape who gets scheduled and paid. A market operator would have to defend its scenario-generation methodology as a quasi-regulatory matter.
+
+**What markets actually do.** Real markets have adopted uncertainty-aware elements incrementally rather than wholesale: probabilistic reserve requirements and new ramping products (CAISO's flexible ramping product, MISO's ramp capability product) that are sized using distributions of net-load forecast error; robust look-ahead commitment tools running alongside deterministic market engines; and out-of-market operator actions informed by probabilistic forecasts. Stochastic optimization thus influences markets today mostly through *requirement-setting* and *advisory tools* rather than through the clearing engine itself.
+
+## 8. The Participant's Side: Offering, Trading, and Asset Optimization
+
+Market participants face their own stochastic problems, and here adoption has been faster because no institutional consensus is required — a firm simply uses better models to make more money.
+
+**Renewable offering.** A wind producer selling into a day-ahead market must offer a quantity before knowing its output, then settle deviations at imbalance prices. This is a newsvendor-type problem: the optimal day-ahead offer is a quantile of the production distribution determined by the ratio of penalties for under- and over-delivery. Richer versions co-optimize offers across day-ahead, intraday, and balancing markets as multi-stage stochastic programs.
+
+**Storage and hydro valuation.** A battery or pumped-hydro operator arbitraging uncertain prices solves a stochastic dynamic program over the state of charge; the resulting marginal value of stored energy ("opportunity cost curve") is exactly what the asset should bid. For seasonal hydro, SDDP-derived water values serve the same role.
+
+**Virtual bidding and financial trading.** Traders placing virtual (purely financial) bids between day-ahead and real-time markets are effectively trading the distribution of the DA–RT spread, and portfolio construction over such positions is a classic stochastic optimization with risk constraints.
+
+**Risk measures.** Because market participants are rarely risk-neutral, expected cost is usually replaced or augmented by risk measures, most commonly **Conditional Value-at-Risk (CVaR)** — the expected loss in the worst α-fraction of scenarios. CVaR is coherent and, crucially, preserves linearity: it adds only linear constraints and variables to a scenario-based LP/MIP (Rockafellar–Uryasev formulation), so risk-averse stochastic programs remain tractable. A weighted combination of expectation and CVaR is the de facto standard objective in trading and asset-optimization models.
+
+## 9. Computation: Decomposition and Practical Tactics
+
+Extensive-form stochastic MIPs get large fast: a unit commitment with 1,000 units, 36 hours, and 50 scenarios has tens of millions of variables. Three decomposition families dominate:
+
+**Benders decomposition (the L-shaped method)** splits the problem into a master over first-stage variables and scenario subproblems whose duals generate optimality and feasibility cuts. It is natural when the second stage is a continuous LP (dispatch given commitment) and parallelizes trivially across scenarios. SDDP is its multi-stage, sampled cousin.
+
+**Lagrangian relaxation** dualizes coupling constraints — either the nonanticipativity constraints or the system-wide balance constraints — so the problem separates by scenario or by unit. It was the workhorse of unit commitment for decades and provides strong bounds, though recovering primal-feasible solutions requires heuristics.
+
+**Progressive Hedging (PH)** (Rockafellar and Wets) relaxes nonanticipativity with an augmented-Lagrangian penalty: each scenario subproblem is solved independently with a penalty pulling its first-stage decisions toward the scenario-weighted average, iterating to consensus. PH is a heuristic without convergence guarantees in the mixed-integer case but works well in practice on SUC and is embarrassingly parallel.
+
+Beyond decomposition, practitioners rely on **sample average approximation (SAA)** with out-of-sample validation to choose scenario counts, warm-starting from deterministic solutions, tightened MIP formulations of unit constraints (which matter as much as decomposition), and generous parallel hardware. Open-source tooling has lowered the barrier considerably: PySP/mpi-sppy and StochasticPrograms.jl for general stochastic programming, SDDP.jl for multi-stage problems, and power-system-specific platforms such as PowerSimulations.jl, PLEXOS, and PSR's SDDP implementation.
+
+## 10. A Worked Miniature: Two-Scenario Commitment Intuition
+
+To make the recourse logic concrete, consider one hour, two units, and two equally likely net-load scenarios of 400 MW and 700 MW. Unit A is a slow unit: 500 MW capacity, $40/MWh, $10,000 startup, must be committed now. Unit B is a fast peaker: 300 MW, $120/MWh, committable in real time. The deterministic model sees the expected load of 550 MW, commits A, and plans 550 MW from it — but in the high scenario A alone cannot serve 700 MW, so the peaker runs 200 MW at high cost, and in the low scenario A runs part-loaded. The stochastic model evaluates the *actual* expected cost of each commitment plan against both scenarios, correctly weighing A's startup cost against the probability-weighted peaker usage, and would also detect cases where the deterministic plan is infeasible in the tail (e.g., if the high scenario were 900 MW, expected-value commitment would leave the system short). The general lesson generalizes: deterministic models are systematically blind to asymmetries between scenarios, and the cost of that blindness grows with net-load variance — which is exactly what renewable growth delivers.
+
+## 11. Open Frontiers
+
+Several directions define current research and near-term practice. First, **learning-augmented optimization**: using machine learning to warm-start commitments, predict active constraints, or learn cost-to-go functions, cutting solve times enough to make stochastic clearing computationally routine. Second, **distributionally robust and data-driven formulations** that connect the scenario-generation step and the optimization step into one statistically principled pipeline. Third, **pricing theory for non-convex, stochastic markets** — extending convex-hull and extended-LMP ideas to settings with scenarios, so that stochastic clearing could produce defensible settlement prices. Fourth, **multi-stage market designs** (continuous intraday trading, shorter gate closures) that reduce the here-and-now burden by letting more decisions become recourse. And fifth, **demand-side and storage recourse**: as flexible demand and batteries proliferate, the cheapest hedge against uncertainty increasingly sits on the consumption side, and models are evolving to co-optimize it.
+
+## 12. Suggested Reading
+
+For foundations, Birge and Louveaux, *Introduction to Stochastic Programming*, and Shapiro, Dentcheva, and Ruszczyński, *Lectures on Stochastic Programming*. For electricity applications, Conejo, Carrión, and Morales, *Decision Making Under Uncertainty in Electricity Markets*, is the standard reference; Morales, Conejo, Madsen, Pinson, and Zugno, *Integrating Renewables in Electricity Markets*, covers stochastic clearing and renewable offering in depth. Pereira and Pinto's 1991 paper "Multi-stage stochastic optimization applied to energy planning" introduces SDDP; Bertsimas et al.'s "Adaptive robust optimization for the security constrained unit commitment problem" (2013) is the touchstone for the robust alternative; and Rockafellar and Uryasev's "Optimization of conditional value-at-risk" (2000) underpins the risk-averse formulations used throughout the industry.

us_market/economists_survey_en.md

@@ -0,0 +1,272 @@
+# How Economists Have Shaped the US Electricity Markets: Theory, Practice, and Key Milestones
+
+**A Survey**
+
+---
+
+## Abstract
+
+Few industries have been as thoroughly re-engineered by academic economics as the US electric power sector. Over roughly four decades, economists supplied the intellectual critique that dismantled cost-of-service regulation of generation, developed the pricing theory that underpins every organized wholesale market in the country, diagnosed the failures of first-generation market designs, and continue to shape the rules governing capacity, demand response, storage, and the integration of renewable energy. This survey traces that influence across four dimensions: (i) the intellectual origins of restructuring in the economics of regulation; (ii) the theoretical foundations of modern market design, most notably peak-load pricing, spot pricing, and locational marginal pricing; (iii) the institutional milestones—from PURPA (1978) through FERC Orders 888, 2000, 745, 841, and 2222—where economic ideas were translated into regulation; and (iv) the empirical and applied literature through which economists diagnosed crises (California 2000–01, Texas 2021), built market-monitoring institutions, and evaluated whether restructuring delivered its promised efficiencies. We argue that US electricity restructuring is best understood not as "deregulation" but as a sustained exercise in applied mechanism design, in which economists moved from critics of regulation to architects of institutions, and we conclude with the open design questions—resource adequacy under deep decarbonization, non-convex pricing, hybrid markets—where economic theory is once again being pressed into service.
+
+---
+
+## 1. Introduction
+
+When the Federal Energy Regulatory Commission (FERC) issued Order No. 888 in April 1996, requiring open, non-discriminatory access to the interstate transmission grid, it set in motion the most consequential reorganization of the American electric power industry since the 1930s. Within a decade, two-thirds of US electricity demand was served through organized wholesale markets operated by Independent System Operators (ISOs) and Regional Transmission Organizations (RTOs), in which the price of energy is computed every five minutes at thousands of distinct network locations by solving a large-scale optimization problem whose objective function—maximization of social surplus subject to the physics of power flow—comes straight out of welfare economics.
+
+That last fact is the central theme of this survey. In most deregulated industries—airlines, trucking, telecommunications, natural gas—economists supplied the *case* for competition, and markets then evolved through decentralized entrepreneurial activity. Electricity was different. Because power flows obey Kirchhoff's laws rather than contracts, because electricity was (until recently) not economically storable, and because supply and demand must balance second by second to avoid cascading blackouts, a workable electricity "market" could not simply emerge; it had to be *designed*. The result is that economists did not merely advocate for electricity competition—they wrote its constitution. The pricing algorithm at the heart of every US organized market (locational marginal pricing), the financial instruments that hedge congestion (financial transmission rights), the auctions that procure future capacity, the scarcity-pricing mechanisms that replace them in Texas, and the market-power mitigation machinery that polices all of the above are, to an unusual degree, the direct product of academic economic research.
+
+This survey synthesizes that history for a general economics audience. Section 2 traces the intellectual origins of restructuring in the postwar economics of regulation and natural monopoly. Section 3 develops the theoretical core: peak-load pricing, spot pricing, and Hogan's locational marginal pricing framework. Section 4 walks through the legislative and regulatory milestones through which these ideas became law. Section 5 examines the California crisis of 2000–01, the episode in which empirical economists demonstrated their value as market diagnosticians and which permanently institutionalized economic market monitoring. Section 6 covers the resource-adequacy problem—"missing money," capacity markets, and energy-only alternatives. Section 7 surveys mechanism-design contributions: auctions, non-convex pricing, and virtual bidding. Section 8 turns to the demand side: dynamic retail pricing and the demand-response wars. Section 9 addresses the modern frontier—renewables, storage, distributed resources, and the Texas crisis of 2021. Section 10 reviews the empirical literature evaluating restructuring's effects. Section 11 discusses critiques and the emerging debate over "hybrid" markets, and Section 12 concludes.
+
+Three caveats. First, we focus on the United States; the UK privatization of 1990, though enormously influential on American thinking (and itself heavily shaped by economists such as Stephen Littlechild), enters only as background. Second, "economists" is interpreted broadly to include the engineering-economics tradition at MIT and elsewhere—Fred Schweppe was an electrical engineer, but *Spot Pricing of Electricity* is a work of economics. Third, no survey of this length can be exhaustive; our aim is an accurate map of the main intellectual lineages and institutional turning points.
+
+---
+
+## 2. Intellectual Origins: The Economics of Regulation and Its Discontents
+
+### 2.1 The old regime and its theoretical rationale
+
+From the 1920s until the 1980s, the US electricity industry was organized around vertically integrated, investor-owned utilities holding exclusive retail franchises, with prices set by state commissions on a cost-of-service basis and wholesale/interstate matters governed by the Federal Power Act of 1935. The intellectual rationale was the theory of natural monopoly: with enormous fixed costs and (it was believed) pervasive economies of scale in generation, transmission, and distribution alike, competition was thought to be wasteful or unsustainable, and regulated monopoly the least-bad institutional response.
+
+Economists were present at the creation of this regime—the "fair rate of return on fair value" apparatus descends from institutionalist economics and rate-base litigation of the early twentieth century—but the profession's most important contribution came later, in the form of critique.
+
+### 2.2 The postwar critique
+
+Four strands of postwar economics converged to undermine confidence in the regulated-monopoly model.
+
+**Incentive distortions.** Averch and Johnson (1962) showed formally that rate-of-return regulation gives utilities an incentive to over-invest in capital—the "gold-plating" or A–J effect. Whatever the empirical magnitude of the effect (still debated), the paper reframed regulation as an incentive problem rather than an accounting exercise, and seeded the later literature on incentive regulation (Laffont and Tirole 1993) that would inform performance-based ratemaking.
+
+**Regulatory capture.** Stigler (1971) and Peltzman (1976) argued that regulation is supplied to politically effective interest groups rather than in the public interest, while Demsetz (1968) asked why natural monopoly required regulation at all when the *right to serve* could be auctioned—franchise bidding as competition *for* the market. These Chicago contributions eroded the presumption that regulation reliably corrected market failure.
+
+**Evidence on scale economies.** Empirical work—most influentially Christensen and Greene (1976)—found that by the 1970s most large US generating firms had exhausted economies of scale at the plant and firm level. If generation was no longer a natural monopoly, the case for regulating it collapsed, even if wires remained naturally monopolistic. This "unbundling" insight—competition where feasible (generation, retail), regulation where necessary (transmission, distribution)—became the master blueprint of restructuring.
+
+**The deregulation movement in practice.** Alfred Kahn's two-volume *The Economics of Regulation* (1970–71) supplied the synthesis, and Kahn himself, as chairman of the Civil Aeronautics Board under President Carter, demonstrated that an academic economist could dismantle a regulatory regime from within. Airline (1978), trucking (1980), railroad (1980), and natural gas (1978–92) deregulation created both the political template and the professional confidence for tackling electricity—by common consent the hardest case.
+
+### 2.3 The proximate crisis and the pivotal book
+
+The 1970s destroyed the postwar equilibrium of the utility industry: oil shocks, the collapse of nuclear construction economics (with celebrated cost overruns and the WPPSS default), stagnating demand growth, and double-digit rate increases turned regulators and consumers alike against the incumbent model. The Public Utility Regulatory Policies Act of 1978 (PURPA), discussed in Section 4, inadvertently ran a natural experiment demonstrating that non-utility generators could build and operate power plants.
+
+Into this environment came Paul Joskow and Richard Schmalensee's *Markets for Power: An Analysis of Electric Utility Deregulation* (1983), the single most influential economic analysis of electricity restructuring ever written. Joskow and Schmalensee did two things. First, they systematically dismantled the claim that the status quo was efficient, documenting the incentive problems of cost-of-service regulation. Second—and this is often forgotten—they were deeply *cautious* about wholesale competition, cataloguing the technical obstacles (loop flow, reliability externalities, market power in transmission-constrained regions) that naïve deregulation would encounter. The book's lasting contribution was to define the research agenda: every obstacle they identified became, over the following two decades, a design problem that economists set out to solve. Joskow's subsequent forty-year output—on transmission investment, contract structures, capacity markets, and lately decarbonization—constitutes the connective tissue of the entire field, and his role as advisor, board member, and expert in countless proceedings made him arguably the most influential single economist in the sector's history.
+
+---
+
+## 3. Theoretical Foundations: From Peak-Load Pricing to Locational Marginal Prices
+
+### 3.1 Marginal-cost pricing and the peak-load problem
+
+The deep theory begins in postwar France. Marcel Boiteux, an economist-engineer at Électricité de France (and student of Maurice Allais), worked out in the late 1940s and 1950s how to price electricity at marginal cost when demand varies over time and capacity is costly: off-peak users should pay energy cost only, while peak users pay energy cost plus the marginal cost of capacity (Boiteux 1949, 1956; translated and introduced to Anglophone readers largely via Nelson 1964). Steiner (1957) independently developed the peak-load pricing model in the American literature, and Williamson (1966) refined the welfare analysis. Boiteux's companion work on second-best pricing under a budget constraint (Ramsey–Boiteux pricing) became the standard framework for recovering fixed costs with minimal distortion. EDF's *tarif vert*, implemented in 1956, proved these were not blackboard curiosities.
+
+Two further ingredients completed the classical toolkit. Turvey (1968) and others connected peak-load pricing to investment planning. And the theory of pricing under uncertainty with rationing—what is the efficient price when capacity is short?—pointed toward the concept that would later dominate resource-adequacy debates: the value of lost load (VoLL), the willingness of consumers to pay to avoid interruption. Efficient scarcity prices should rise toward VoLL as the system approaches involuntary curtailment; this single idea underlies everything from ERCOT's price caps to capacity-market demand curves.
+
+### 3.2 Spot pricing: Schweppe and the MIT school
+
+The decisive step from tariff theory to market design was taken at MIT in the late 1970s and 1980s by Fred Schweppe—an electrical engineer—working with economists and engineers Michael Caramanis, Richard Tabors, and Roger Bohn. Their program, culminating in *Spot Pricing of Electricity* (Schweppe, Caramanis, Tabors, and Bohn 1988), posed the question: what is the *true* marginal cost of electricity, recognizing that it varies not just by time of day but continuously, stochastically, and—crucially—by location on the network?
+
+Their answer: the efficient spot price at each moment and each bus of the network equals the system lambda (the marginal cost of energy at the reference bus) adjusted for marginal transmission losses and the shadow prices of binding transmission constraints. Prices differ across locations whenever the network is congested, because Kirchhoff's laws prevent cheap power from reaching expensive locations. Schweppe et al. envisioned these prices as the basis for a decentralized "homeostatic" control of the power system: generators and (in principle) consumers responding to real-time prices would replicate the least-cost dispatch that central engineers had always computed—but through markets. The book was written before any institution existed to implement it; it became the technical bible of every subsequent market design.
+
+### 3.3 Hogan: locational marginal pricing and financial transmission rights
+
+The person who transformed spot-pricing theory into implementable market architecture was William Hogan of Harvard's Kennedy School. Beginning with "Contract Networks for Electric Power Transmission" (Hogan 1992), Hogan solved the problem that had stymied decentralized approaches: how to define tradable transmission rights on an AC network where power flows cannot be directed along contract paths.
+
+Hogan's insight had two parts. First, embrace rather than fight central dispatch: let an independent system operator run a *bid-based, security-constrained economic dispatch*—the same optimization utilities had always solved, but with offer curves submitted by competing generators—and let the dual variables (shadow prices) of that optimization define the market-clearing **locational marginal prices (LMPs)** at every node. The LMP at a node decomposes exactly as Schweppe prescribed: energy plus losses plus congestion. Second, define transmission rights *financially* rather than physically: a **financial transmission right (FTR)** between nodes A and B entitles its holder to the congestion-rent difference LMP(B) − LMP(A) on a specified quantity. FTRs are simultaneously feasible whenever the underlying injections are (the "revenue adequacy" theorem), they hedge congestion risk perfectly for matching physical schedules, and they can be auctioned or allocated to historical grid users—thereby solving the political problem of transition as well as the economic problem of hedging.
+
+This "pool-plus-FTR" or "integrated" architecture—centralized dispatch, nodal prices, financial rights, two settlements (day-ahead and real-time), co-optimized energy and ancillary services—was fought over ferociously in the 1990s against a rival "decentralized/bilateral" vision (associated with, among others, some California designers and the original UK Pool's critics), which favored self-scheduling, zonal prices, and physical rights. The verdict of experience was unambiguous. PJM abandoned zonal for nodal pricing in 1998 after gaming of zonal prices; New York launched with LMP (and its FTR variant, Transmission Congestion Contracts) in 1999; New England converted in 2003; the Midwest ISO launched nodal in 2005; California, after its crisis, reorganized around full nodal pricing in 2009; Texas converted from zones to nodes in 2010; and the Southwest Power Pool followed in 2014. Every organized US market now runs on Hogan's architecture. It is difficult to name another instance in which a single economist's design is executed, every five minutes, across most of a continental economy.
+
+Hogan's subsequent work extended the framework to virtually every contested design margin: multi-settlement systems, FTR options and obligations, scarcity pricing via operating reserve demand curves (Section 6.4), uplift and non-convex pricing (with Gribik and Pope), transmission investment incentives, and—continuing today—pricing for decarbonized grids. Harvey and Hogan's running debate with the Berkeley empiricists over California (Section 5) also set the methodological standards for market-power measurement.
+
+### 3.4 The supporting cast of theory
+
+Three further theoretical literatures fed the design canon:
+
+- **Auction and mechanism design.** Vickrey's (1961) foundations; the Wilson–Milgrom research program on auctions with common values and multi-unit demand; and Robert Wilson's *Architecture of Power Markets* (Econometrica, 2002), which framed electricity design as a problem in "market architecture"—the decomposition of an economic problem into linked auction sub-markets. Wilson himself advised on power-market designs internationally and in California, and Paul Milgrom's consulting practice extended auction design into electricity procurement. The 2020 Nobel to Milgrom and Wilson cited work whose applications explicitly include electricity.
+- **Supply function equilibrium.** Klemperer and Meyer (1989) modeled competition in supply schedules under uncertainty—precisely what generators submit to ISOs—and Green and Newbery (1992) applied SFE to the British Pool, providing the workhorse oligopoly model for electricity markets and an early, influential warning that a duopolistic pool would price far above marginal cost. This transatlantic result shaped American thinking about market structure and divestiture requirements.
+- **Contracts and vertical arrangements.** Joskow's earlier empirical work on long-term coal contracts and asset specificity (in the Williamsonian tradition), and the Allaz–Vila (1993) result that forward contracting can mitigate spot market power, informed both the design of transition contracts and the enduring policy preference for high contract cover—one of the clearest lessons economists drew from comparing California (uncontracted, catastrophic) with other markets.
+
+---
+
+## 4. Institutional Milestones: Ideas Become Law
+
+### 4.1 PURPA (1978): the accidental experiment
+
+The Public Utility Regulatory Policies Act of 1978, passed as energy-crisis legislation, required utilities to purchase power from "qualifying facilities" (cogenerators and small renewables) at the utility's *avoided cost*. Avoided cost is, of course, an economist's concept—the marginal cost the utility escapes—and its implementation launched a thousand rate proceedings in which economists testified about how to measure it. PURPA's deeper significance was unintended: the wave of non-utility generation it induced (especially in California, New York, and Texas) demonstrated that independent power producers could finance, build, and operate generation, destroying the empirical premise that generation required vertically integrated monopoly. Economists, including Joskow, documented both the demonstration effect and the pathologies of administratively determined avoided-cost contracts—which, when set too high (as in California's Standard Offer 4 contracts), saddled ratepayers with above-market costs and taught a lasting lesson about administered prices versus auctions.
+
+### 4.2 The Energy Policy Act of 1992 and FERC Orders 888/889 (1996)
+
+The Energy Policy Act of 1992 created "exempt wholesale generators," freeing independent generation from the Public Utility Holding Company Act, and empowered FERC to order transmission access case by case. FERC generalized this in Order No. 888 (1996), which required all transmission-owning utilities to file open-access tariffs offering service to third parties on the same terms they provided themselves, and Order No. 889, which established the OASIS information system and functional unbundling. The economic logic—that the bottleneck monopoly (wires) must not be permitted to leverage its position into the potentially competitive segment (generation)—was the essential-facilities/vertical-foreclosure reasoning of industrial organization, argued to FERC by a small army of economists on all sides. Order 888 also blessed the recovery of "stranded costs," itself the subject of a substantial economic literature (with William Baumol, Alfred Kahn, and others debating whether recovery was efficient transition compensation or an unwarranted bailout).
+
+### 4.3 State restructuring: California AB 1890 and the Northeast
+
+Restructuring of *retail* markets was a state affair. California moved first and most ambitiously: AB 1890 (1996), implementing CPUC decisions shaped by years of hearings in which academic economists were central witnesses, created the Power Exchange (PX) and the California ISO (launched March 31, 1998), required the utilities to divest most fossil generation, froze retail rates at 10% below prior levels, and—fatefully—discouraged long-term contracting, pushing procurement into the day-ahead and real-time markets. The design reflected a compromise between the bilateral/decentralized school and the pool school that satisfied neither; several prominent economists (including members of what became the ISO's Market Surveillance Committee) warned before launch about the combination of tight supply, uncontracted demand, capped retail prices, and market-power exposure. Massachusetts, Rhode Island, Pennsylvania, New York, New Jersey, Maryland, Illinois, Ohio, Michigan, and Texas followed with their own restructuring statutes between 1996 and 1999, typically pairing retail choice with generation divestiture and transition-cost charges. By 2000, roughly half the states had enacted or opened restructuring proceedings.
+
+### 4.4 Order 2000 (1999) and the Standard Market Design episode (2002–05)
+
+FERC's Order No. 2000 encouraged (but did not compel) formation of Regional Transmission Organizations—independent grid operators with defined characteristics and functions. The intellectual case for the ISO/RTO institution—an independent, non-profit operator running open, bid-based markets—had been developed by Hogan, Joskow, and others as the resolution of the transmission-independence problem. In 2002, FERC proposed to go further: the Standard Market Design (SMD) NOPR would have mandated the full nodal architecture—LMP, FTRs, day-ahead/real-time settlements, market monitoring, and resource adequacy—everywhere in the country. SMD was the high-water mark of economists' direct influence on federal policy: its text reads like a summary of the design literature. It also provoked a fierce political backlash from the South and West (regions with cheap regulated power and no appetite for markets), and FERC formally withdrew the proposal in 2005. The result is today's map: organized nodal markets covering roughly two-thirds of US load (PJM, MISO, CAISO, ERCOT, NYISO, ISO-NE, SPP), with the Southeast and much of the West remaining vertically integrated—an enduring natural experiment that empiricists would later exploit (Section 10).
+
+### 4.5 The later federal orders
+
+Subsequent FERC rulemakings, each examined in later sections, continued to translate economic analysis into rules: Order 719 (2008) on demand response and ISO governance; Order 745 (2011) on demand-response compensation; Order 755 (2011) on performance-based regulation payments; Order 1000 (2011) on transmission planning and cost allocation—a subject on which Joskow and Jean Tirole had written foundational theory; Order 841 (2018) on electric storage participation; and Order 2222 (2020) on distributed energy resource aggregation. The pattern is consistent: a design question surfaces; economists write papers and file affidavits; FERC codifies a (contested) resolution; litigation and further papers follow.
+
+---
+
+## 5. Trial by Fire: The California Crisis and the Rise of Market Monitoring
+
+### 5.1 The crisis
+
+Between May 2000 and June 2001, average wholesale prices in California rose by a factor of five to ten relative to prior years; the state suffered rolling blackouts; Pacific Gas & Electric entered bankruptcy; Southern California Edison approached it; the Power Exchange ceased operations; and the state spent tens of billions of dollars buying power. The proximate causes included drought (reduced Northwest hydro), rising natural gas and NOx permit prices, demand growth, and a genuinely tight market. But the crisis became an economics event because of what happened *on top of* scarcity.
+
+### 5.2 The diagnosis: economists as forensic analysts
+
+A group of empirical industrial-organization economists—principally Severin Borenstein, James Bushnell, and Frank Wolak, working through the University of California Energy Institute and the CAISO Market Surveillance Committee (which Wolak chaired)—developed and applied a method for *measuring* market power directly: simulate the competitive counterfactual price by stacking actual marginal costs against actual demand (accounting for hydro opportunity costs and imports), and attribute the gap between actual and competitive prices to the exercise of market power. Their central publication, Borenstein, Bushnell, and Wolak, "Measuring Market Inefficiencies in California's Restructured Wholesale Electricity Market" (*American Economic Review*, 2002), found that market power accounted for a large share of the price increases during summer 2000—unilateral withholding by a handful of suppliers facing inelastic, uncontracted demand, not necessarily illegal conspiracy. Joskow and Edward Kahn (2002) reached similar conclusions with complementary methods, documenting substantial output withholding. Scott Harvey and William Hogan contested the magnitudes and the attribution in a series of papers, arguing that cost mis-measurement and operational constraints could explain much of the gap—a debate that, whatever one's verdict, hardened the methodological standards of the entire field. Subsequent enforcement actions and the Enron tapes ("Death Star," "Fat Boy," "Ricochet"—strategies that exploited precisely the seams between California's zonal market and its neighbors that the pool school had warned about) settled the public argument.
+
+The economists' account of *why* California failed became canonical: (i) retail price freezes severed demand from wholesale prices, making demand almost perfectly inelastic; (ii) the utilities were prohibited/discouraged from forward contracting, leaving them exposed to spot prices and giving suppliers maximal incentive to raise them (the Allaz–Vila logic in reverse); (iii) the separation of the PX energy market from the ISO's congestion and reliability functions, and zonal rather than nodal pricing, created arbitrage seams; and (iv) soft price caps invited strategic exports and re-imports. Each element had been flagged in advance by parts of the profession; the crisis converted those warnings into design orthodoxy: contract cover matters, demand response matters, nodal beats zonal, and market power mitigation must be built in, not bolted on.
+
+### 5.3 Institutionalizing economics: the market monitors
+
+The most durable institutional legacy of California is the **market monitor**. Every ISO/RTO now has an independent internal and/or external market monitoring unit—Potomac Economics (led by economist David Patton) serves as independent monitor for several markets (including ERCOT, NYISO, and MISO), Monitoring Analytics (led by economist Joseph Bowring) monitors PJM, and CAISO maintains a Department of Market Monitoring and a Market Surveillance Committee that has been chaired by academic economists (Wolak, Bushnell, and colleagues). These units run conduct-and-impact mitigation screens (automatic offer-capping when suppliers with local market power bid far above reference costs), publish annual State of the Market reports dense with concentration and price-cost-margin analysis, and refer cases to FERC enforcement. Market monitoring is applied industrial organization practiced as a permanent regulatory function—a professional niche that simply did not exist before economists created it.
+
+---
+
+## 6. Resource Adequacy: Missing Money, Capacity Markets, and Scarcity Pricing
+
+### 6.1 The "missing money" problem
+
+In the theoretical ideal, energy-only markets finance capacity through scarcity rents: prices spike toward the value of lost load in shortage hours, and the expectation of such rents induces efficient investment in exactly enough capacity that the marginal plant recovers its fixed costs. In practice, US markets systematically suppress those rents: offer caps (long $1,000/MWh in the East), out-of-market reliability actions by operators who dispatch reserves rather than let prices rise, and mitigation rules together truncate the right tail of the price distribution. The resulting shortfall between what generators need to cover fixed costs and what suppressed energy prices deliver was dubbed the **"missing money" problem**—a term popularized through the work of Peter Cramton and Steven Stoft and central to Joskow's influential treatments (e.g., Joskow 2008). The diagnosis reframed reliability from an engineering mandate into a pricing failure, and the profession split over the remedy: fix the energy prices (scarcity pricing) or bolt on a separate market for capacity.
+
+### 6.2 Capacity markets and the demand-curve innovation
+
+The Eastern ISOs chose capacity markets: mandatory forward auctions in which load-serving entities procure enough accredited capacity to meet a reserve-margin target. Early "vertical demand" designs (a fixed requirement, price collapsing to zero when supply barely exceeded it) proved pathologically volatile and manipulable. The fix—first implemented in NYISO in 2003 and now universal—was the **sloped administrative demand curve**: a downward-sloping schedule anchored at the estimated "cost of new entry" (CONE) that makes the capacity price a smooth function of the surplus margin. The demand curve is pure economic engineering—an administratively constructed marginal-value schedule for reliability—and its parameters (CONE, the shape and anchor points) are set through recurring proceedings that are essentially econometric litigation. Cramton and Stoft's design work (much of it commissioned by the ISOs), together with Joskow's analyses, shaped PJM's Reliability Pricing Model (implemented 2007) and ISO-NE's Forward Capacity Market (first auction 2008), including features such as three-year-forward procurement, locational capacity zones, and—later—"pay for performance" penalty structures (ISO-NE, and PJM's Capacity Performance reforms after the January 2014 polar vortex exposed the weakness of capacity that fails to show up). The subsequent generation of disputes—minimum offer price rules (MOPR) aimed at state-subsidized entrants, buyer-side market power, and the accreditation of intermittent and duration-limited resources—has kept capacity-market economics among the most litigated areas of the field.
+
+### 6.3 The energy-only alternative: ERCOT and the operating reserve demand curve
+
+Texas took the other road. ERCOT—intrastate, outside FERC jurisdiction, restructured under Senate Bill 7 (1999)—runs an energy-only market with no capacity mechanism and, historically, a high offer cap (reaching $9,000/MWh). Its signature mechanism, adopted in 2014, is the **Operating Reserve Demand Curve (ORDC)**, based directly on William Hogan's proposal ("Electricity Scarcity Pricing Through Operating Reserves," 2013): a real-time price adder equal to the loss-of-load probability (as a function of the current reserve level) multiplied by the value of lost load, so that prices rise smoothly and *automatically* toward VoLL as reserves shrink—administratively replicating the scarcity rents that an ideal market with responsive demand would generate. The ORDC is perhaps the purest example in any industry of a formula from an economics working paper being written into the settlement software of a major market. The 2021 crisis (Section 9.4) subjected the design to its most severe test.
+
+### 6.4 The unresolved debate
+
+The capacity-versus-energy-only debate remains live and has been sharpened by decarbonization. Capacity-market advocates emphasize investment-risk reduction and political robustness (regulators will never tolerate true VoLL pricing for long); energy-only advocates (Hogan, and in a different register Wolak, who is skeptical of both constructs and emphasizes contracting obligations) counter that capacity markets pay for "iron in the ground" rather than performance, invite endless administrative gaming, and mute the scarcity prices needed to reward flexibility and demand response. The rise of storage and renewables—whose capacity value (measured by effective load-carrying capability, ELCC, another economics-adjacent construct) depends on penetration—has turned resource-adequacy design into one of the most active current research areas (Section 9).
+
+---
+
+## 7. Mechanism Design in the Engine Room: Auctions, Non-Convexities, and Virtual Bidding
+
+### 7.1 The market as an optimization problem
+
+An ISO's day-ahead market is a sealed-bid, multi-unit, multi-product auction cleared by security-constrained unit commitment and economic dispatch—a mixed-integer program whose scale (tens of thousands of constraints, thousands of resources) makes it among the largest auctions run anywhere. The economic questions are classic mechanism design: What prices support the efficient allocation? What are the incentives to bid truthfully? How should linked products (energy, regulation, spinning reserve) be co-optimized? Economists—often working with operations researchers—have supplied the answers now embedded in the software.
+
+### 7.2 Non-convexities and uplift: the pricing frontier
+
+Power plants have non-convex costs—start-up costs, minimum run levels, minimum up/down times—so a Walrasian equilibrium supporting the efficient commitment may not exist: at the LMP, some committed units lose money and some uncommitted units would profit. US markets patch this with "uplift" or make-whole side payments, which are discriminatory, opaque, and blunt price signals. The theoretical response has been one of the liveliest applied-theory literatures of the past two decades: O'Neill, Sotkiewicz, Hobbs, Rothkopf, and Stewart (2005) showed how to construct market-clearing prices with non-convexities using integer-activity prices; Gribik, Hogan, and Pope (2007) proposed **convex hull pricing** ("extended LMP"), which minimizes total uplift by pricing off the convex envelope of the cost function. Variants of these ideas have been implemented: MISO adopted approximations of convex-hull pricing, and FERC's fast-start pricing orders (2016–2020) pushed ISOs to let inflexible fast-start units set prices. This is an area where economic theory, operations research, and settlement software co-evolve in real time.
+
+### 7.3 Virtual bidding and the two-settlement system
+
+Hogan's multi-settlement architecture separates a financially binding day-ahead market from real-time balancing. To keep the two aligned, markets admit **virtual (convergence) bidding**: purely financial positions that arbitrage expected day-ahead/real-time spreads. The theory says arbitrageurs should converge the two prices, improving unit commitment; the empirical literature broadly confirms it while documenting nuances—Jha and Wolak's work on California finds convergence bidding improved price convergence and productive efficiency, while episodes like the JP Morgan make-whole manipulation cases (settled with FERC in 2013) and studies of loss-related arbitrage in MISO illustrate how financial products interact with market seams. The design of FTR auctions, their chronic revenue underperformance relative to auction prices, and the question of who should bear FTR-portfolio default risk (brought to a head by the 2018 GreenHat default in PJM) have similarly been analyzed principally by economists.
+
+### 7.4 Procurement auctions and default service
+
+Beyond ISO markets, economists designed the auctions through which restructured states procure default retail service—most prominently New Jersey's Basic Generation Service **descending-clock auction**, designed with the involvement of auction economists (including Ausubel and Cramton's circle) and running annually since 2002, and Illinois' procurement events. Milgrom's and Wilson's broader auction-design consulting practices, celebrated in the 2020 Nobel citation, include electricity procurement among their applications. Contract design for renewable procurement (indexed PPAs, contracts-for-differences) is the current frontier of this tradition.
+
+---
+
+## 8. The Demand Side: Dynamic Pricing and the Demand-Response Wars
+
+### 8.1 The economists' oldest complaint
+
+From Boiteux onward, economists have insisted that the demand side is half the market: if retail consumers face time-invariant prices, the price elasticity that should discipline wholesale markets is absent, scarcity cannot clear the market efficiently, and (as California proved) the system is fragile. Borenstein's work in the 2000s (e.g., "The Long-Run Efficiency of Real-Time Electricity Pricing," 2005; Borenstein and Holland 2005 on the distortions when only some customers face real-time prices) made the modern welfare case for **dynamic retail pricing**; Wolak designed and evaluated field experiments (including in Anaheim and internationally) measuring household response to critical-peak pricing; and a large subsequent experimental literature (including Jessoe and Rapson's work on information feedback) quantified how technology (smart meters, automation) raises effective elasticity. Actual adoption remains limited—default flat tariffs persist almost everywhere—which economists attribute to political economy and behavioral frictions rather than to the economics; the gap between the theory's clarity and retail practice is a standing embarrassment the profession continues to probe (including through work on the distributional effects of dynamic pricing).
+
+### 8.2 Order 745 and *EPSA v. FERC*: economists on both sides
+
+Wholesale **demand response**—paying consumers to curtail—produced the sector's most famous economics dispute. FERC Order 745 (2011) required ISOs to pay demand response the full LMP. Alfred Kahn, in one of his final interventions, filed in support of full-LMP compensation; William Hogan argued forcefully that the correct payment is LMP minus the retail rate the customer avoids ("LMP−G"), since paying full LMP to someone who also saves the retail price double-compensates curtailment and subsidizes inefficient demand reduction. The dispute—at bottom about the definition of the counterfactual and the treatment of the retail-rate distortion—went to the Supreme Court as *FERC v. Electric Power Supply Association* (2016), where the Court upheld FERC's jurisdiction and Order 745. Economists' briefs and papers populated both sides; the episode is now a teaching case in how second-best reasoning cuts in opposite directions depending on which distortion one takes as fixed.
+
+---
+
+## 9. The Modern Frontier: Renewables, Storage, Distributed Resources, and Texas 2021
+
+### 9.1 Zero-marginal-cost entry and price formation
+
+Subsidized and mandated wind and solar—driven by production tax credits, investment tax credits, and state renewable portfolio standards—entered wholesale markets in volumes that transformed price formation: negative prices in windy regions (a predictable consequence of the PTC paying generators to produce, analyzed early by economists), the California "duck curve," depressed mid-day energy prices, and merit-order displacement of thermal units. The economics literature responded on several fronts: measuring the market value decline of variable renewables as penetration rises; quantifying the emissions and price effects of renewable entry; analyzing how out-of-market subsidies interact with capacity markets (the MOPR wars); and pressing the first-best alternative—carbon pricing—in ISO markets, including formal proposals for carbon-adder designs in NYISO and PJM analyzed by economists at Resources for the Future and elsewhere. The broad professional consensus that technology-neutral carbon pricing dominates technology-specific mandates has had, it must be said, limited legislative success; economists have therefore increasingly turned to analyzing second-best instrument design (clean-energy standards, ELCC-based accreditation, hybrid procurement).
+
+### 9.2 Storage and distributed resources: Orders 841 and 2222
+
+FERC Order 841 (2018) required ISOs to create participation models letting storage set prices as both buyer and seller; Order 2222 (2020) extended participation to aggregations of distributed energy resources. Both orders resolve questions economists had framed: storage arbitrage and its welfare effects, the double-charging of storage for transmission, and the treatment of behind-the-meter resources. A growing empirical literature (on ERCOT and CAISO battery fleets) now measures how quickly storage competes away its own arbitrage margins—a clean test of entry economics playing out in real time.
+
+### 9.3 Transmission, again
+
+Transmission planning and cost allocation—the subject of Joskow and Tirole's theoretical work on merchant transmission and of Hogan's beneficiary-pays principle—returned to the center with Order 1000 (2011) and Order 1920 (2024), which require forward-looking regional planning with benefit-based cost allocation. The "beneficiary pays" standard that courts (notably the Seventh Circuit in the *Illinois Commerce Commission* cases, in opinions by Judge Posner—an economist-judge) have enforced is straight cost-benefit economics.
+
+### 9.4 Winter Storm Uri (February 2021): stress-testing the energy-only design
+
+The Texas blackout of February 2021—days of rotating outages, prices administratively held at the $9,000/MWh cap for over four days, roughly 200+ deaths, and tens of billions of dollars of financial redistribution—triggered the most intense economic post-mortem since California. The economics discussion separated several issues: the failure was primarily one of *supply availability* (unweatherized gas supply and generation) rather than market design per se; scarcity pricing performed as designed but the duration of cap-level prices exposed the absence of retail hedging (the Griddy customers on wholesale pass-through rates) and raised the question whether VoLL-level prices sustained for days are politically or contractually tolerable; and the PUCT's decision to hold prices at the cap after load shed ended became a celebrated dispute about out-of-market intervention. Peter Cramton, then an ERCOT board member (who, with other out-of-state directors, resigned after the storm), Hogan, Wolak, Bushnell, and many others produced dueling analyses; the aftermath brought a lower price cap, a modified ORDC, weatherization mandates, and a new Performance Credit Mechanism proposal—each debated in explicitly economic terms. Uri also revived academic interest in the interaction between gas and electricity market design, and in mandatory retail/load-serving hedging obligations of the kind Wolak has long advocated.
+
+### 9.5 Resource adequacy for a decarbonized grid
+
+The current research frontier asks whether the LMP-plus-scarcity architecture, designed for a fuel-burning fleet with meaningful marginal costs, remains adequate for a system dominated by zero-marginal-cost, weather-driven, and duration-limited resources. Questions under active study by economists include: price formation when the marginal resource is storage or curtailed renewables (opportunity-cost pricing); capacity accreditation via marginal ELCC; long-duration adequacy and tail risk (climate-correlated outages); the financing of capital-intensive clean firm resources under volatile revenue streams (motivating contract-for-differences and reliability options, the latter a Cramton–Stoft design adopted in Colombia and ISO-NE's FCM in modified form); and whether organized markets should evolve toward centralized long-term procurement—"hybrid markets"—a debate joined by Joskow (whose recent papers argue the pure energy-market model is under increasing strain), Schmalensee, Wolak, Hogan, and a younger generation. The wheel, in other words, is turning again, and economists are once more on both sides of it.
+
+---
+
+## 10. Did It Work? The Empirical Assessment Literature
+
+Restructuring created a natural experiment—organized-market regions versus traditionally regulated regions; divested plants versus utility-retained plants—that empirical economists have mined for two decades. The headline findings:
+
+- **Operating efficiency.** Fabrizio, Rose, and Wolfram (*AER*, 2007) found that investor-owned plants in restructuring states reduced labor and nonfuel operating expenses several percent relative to those in non-restructuring states—evidence that competition (or its anticipation) sharpened cost discipline. Davis and Wolfram (*AEJ: Applied*, 2012) found that divestiture of nuclear plants to independent merchant operators raised capacity factors by roughly 10 percentage points, a large efficiency gain concentrated in reduced outage durations. Cicala (*AER*, 2015) showed deregulated plants procured coal more cheaply once cost pass-through incentives were removed.
+- **Dispatch efficiency.** Cicala (*AER*, 2022) used the staggered expansion of ISO markets to estimate that market-based dispatch reduced production costs on the order of billions of dollars per year by reallocating output toward lower-cost plants and increasing gains from trade across utility boundaries—the cleanest evidence that the LMP machinery does what its designers promised.
+- **Investment and technology mix.** Merchant investment responded to price signals with technology choices (fast, modular gas; later renewables and batteries) very different from the regulated era's, and the literature documents both the responsiveness and the boom-bust cycles that pure merchant exposure produces.
+- **Retail competition.** The retail literature is more mixed: evidence from Texas and elsewhere documents meaningful search frictions, price dispersion, and confusion pricing in residential retail choice (with Hortaçsu, Madanizadeh, and Puller's 2017 study of Texas a benchmark), tempering the more expansive claims for retail liberalization even where wholesale gains are accepted.
+- **Prices.** Simple regulated-versus-restructured retail price comparisons are confounded (restructured states were high-cost to begin with; gas price cycles dominate), and the profession has largely converged on the view that the welfare gains of restructuring show up in costs, dispatch, and plant performance rather than in unambiguous retail price declines—an honest, if politically unsatisfying, verdict.
+
+Alongside evaluation, electricity became empirical IO's favorite laboratory: because engineering marginal costs are measurable, researchers can compute markups directly rather than infer them, and the industry has consequently generated foundational studies of oligopoly bidding (Wolfram on the England–Wales pool; Hortaçsu and Puller on ERCOT bidders' deviations from optimal bidding), forward contracting, auction behavior, and environmental policy interactions. The methodological traffic runs both ways: electricity data disciplined IO theory, and IO tools staffed the market monitors.
+
+---
+
+## 11. Critiques, Countercurrents, and the Limits of Design
+
+An honest survey must record that economists' stewardship of electricity restructuring has critics, including within the profession.
+
+First, **the promised consumer savings were oversold** in the 1990s political campaigns, and the profession's more careful voices (Joskow prominent among them) spent years distinguishing what the evidence supports from what advocates claimed. Second, **complexity itself is a cost**: the layered edifice of energy, ancillary, capacity, FTR, and virtual products—each patching the previous layer's incentive problems—has been criticized (from the left as designed-for-traders opacity, from parts of the engineering community as fragile, and by public-power advocates as an expensive detour). Third, **the governance critique**: ISO stakeholder processes and FERC litigation allocate rents through processes in which well-funded incumbents are structurally advantaged—a Stiglerian observation that sits uncomfortably with the design tradition's technocratic self-image. Fourth, **the decarbonization critique**: state clean-energy policies now drive most investment, and the resulting collision between subsidized entry and market price formation (MOPR, hybrid-market proposals) suggests to some scholars that the 1990s architecture is being quietly superseded by a return to planned procurement with markets relegated to short-run balancing—"markets for dispatch, planning for investment." Whether that constitutes the failure of the design program or its adaptive success is perhaps the central interpretive question of the field's next decade, and—fittingly—it is being argued out by the same cast of economists, their students, and their students' students.
+
+---
+
+## 12. Conclusion
+
+The arc surveyed here runs from Boiteux's peak-load tariffs at EDF, through the Chicago and MIT critiques of regulation, Schweppe's spot prices, and Hogan's contract networks, to the five-minute nodal prices that today clear most American wholesale electricity. Along the way economists built institutions (ISOs, market monitors, capacity auctions), fought and refereed crises (California, Uri), created a permanent professional apparatus of market design and surveillance, and produced one of empirical economics' richest bodies of evidence on what competition and regulation each do well. Three lessons generalize beyond electricity. First, where physics or technology preclude decentralized exchange, markets are engineered artifacts, and the quality of the engineering—the mechanism design—determines whether competition delivers its textbook benefits. Second, transitions are governed by second-best problems (stranded costs, retail-price distortions, subsidized entry) at least as much as by first-best blueprints, and the profession's most valuable interventions have often been diagnostic rather than architectural. Third, market design is never finished: each solved problem (transmission access, congestion, missing money) exposes the next (non-convexities, accreditation, deep decarbonization). Forty years in, the US electricity markets remain what they were at the start—the most ambitious ongoing experiment in applied economics anywhere in the world.
+
+---
+
+## References (selected)
+
+- Allaz, B., and J.-L. Vila (1993). "Cournot Competition, Forward Markets and Efficiency." *Journal of Economic Theory* 59(1): 1–16.
+- Averch, H., and L. Johnson (1962). "Behavior of the Firm under Regulatory Constraint." *American Economic Review* 52(5): 1052–1069.
+- Boiteux, M. (1949). "La tarification des demandes en pointe." *Revue Générale de l'Électricité*; English translation, "Peak-Load Pricing," *Journal of Business* 33 (1960): 157–179.
+- Boiteux, M. (1956). "Sur la gestion des monopoles publics astreints à l'équilibre budgétaire." *Econometrica* 24(1): 22–40.
+- Borenstein, S. (2005). "The Long-Run Efficiency of Real-Time Electricity Pricing." *The Energy Journal* 26(3): 93–116.
+- Borenstein, S., J. Bushnell, and F. Wolak (2002). "Measuring Market Inefficiencies in California's Restructured Wholesale Electricity Market." *American Economic Review* 92(5): 1376–1405.
+- Borenstein, S., and S. Holland (2005). "On the Efficiency of Competitive Electricity Markets with Time-Invariant Retail Prices." *RAND Journal of Economics* 36(3): 469–493.
+- Christensen, L., and W. Greene (1976). "Economies of Scale in U.S. Electric Power Generation." *Journal of Political Economy* 84(4): 655–676.
+- Cicala, S. (2015). "When Does Regulation Distort Costs? Lessons from Fuel Procurement in US Electricity Generation." *American Economic Review* 105(1): 411–444.
+- Cicala, S. (2022). "Imperfect Markets versus Imperfect Regulation in US Electricity Generation." *American Economic Review* 112(2): 409–441.
+- Cramton, P., and S. Stoft (2005). "A Capacity Market that Makes Sense." *The Electricity Journal* 18(7): 43–54.
+- Davis, L., and C. Wolfram (2012). "Deregulation, Consolidation, and Efficiency: Evidence from US Nuclear Power." *American Economic Journal: Applied Economics* 4(4): 194–225.
+- Demsetz, H. (1968). "Why Regulate Utilities?" *Journal of Law and Economics* 11(1): 55–65.
+- Fabrizio, K., N. Rose, and C. Wolfram (2007). "Do Markets Reduce Costs? Assessing the Impact of Regulatory Restructuring on US Electric Generation Efficiency." *American Economic Review* 97(4): 1250–1277.
+- Green, R., and D. Newbery (1992). "Competition in the British Electricity Spot Market." *Journal of Political Economy* 100(5): 929–953.
+- Gribik, P., W. Hogan, and S. Pope (2007). "Market-Clearing Electricity Prices and Energy Uplift." Harvard Electricity Policy Group working paper.
+- Hogan, W. (1992). "Contract Networks for Electric Power Transmission." *Journal of Regulatory Economics* 4(3): 211–242.
+- Hogan, W. (2013). "Electricity Scarcity Pricing Through Operating Reserves." *Economics of Energy & Environmental Policy* 2(2): 65–86.
+- Hortaçsu, A., and S. Puller (2008). "Understanding Strategic Bidding in Multi-Unit Auctions: A Case Study of the Texas Electricity Spot Market." *RAND Journal of Economics* 39(1): 86–114.
+- Hortaçsu, A., S. Madanizadeh, and S. Puller (2017). "Power to Choose? An Analysis of Consumer Inertia in the Residential Electricity Market." *American Economic Journal: Economic Policy* 9(4): 192–226.
+- Joskow, P. (2008). "Capacity Payments in Imperfect Electricity Markets: Need and Design." *Utilities Policy* 16(3): 159–170.
+- Joskow, P., and E. Kahn (2002). "A Quantitative Analysis of Pricing Behavior in California's Wholesale Electricity Market During Summer 2000." *The Energy Journal* 23(4): 1–35.
+- Joskow, P., and R. Schmalensee (1983). *Markets for Power: An Analysis of Electric Utility Deregulation.* MIT Press.
+- Joskow, P., and J. Tirole (2005). "Merchant Transmission Investment." *Journal of Industrial Economics* 53(2): 233–264.
+- Kahn, A. (1970–71). *The Economics of Regulation: Principles and Institutions.* Wiley (2 vols.).
+- Klemperer, P., and M. Meyer (1989). "Supply Function Equilibria in Oligopoly under Uncertainty." *Econometrica* 57(6): 1243–1277.
+- Laffont, J.-J., and J. Tirole (1993). *A Theory of Incentives in Procurement and Regulation.* MIT Press.
+- O'Neill, R., P. Sotkiewicz, B. Hobbs, M. Rothkopf, and W. Stewart (2005). "Efficient Market-Clearing Prices in Markets with Nonconvexities." *European Journal of Operational Research* 164(1): 269–285.
+- Peltzman, S. (1976). "Toward a More General Theory of Regulation." *Journal of Law and Economics* 19(2): 211–240.
+- Schweppe, F., M. Caramanis, R. Tabors, and R. Bohn (1988). *Spot Pricing of Electricity.* Kluwer Academic Publishers.
+- Steiner, P. (1957). "Peak Loads and Efficient Pricing." *Quarterly Journal of Economics* 71(4): 585–610.
+- Stigler, G. (1971). "The Theory of Economic Regulation." *Bell Journal of Economics and Management Science* 2(1): 3–21.
+- Turvey, R. (1968). *Optimal Pricing and Investment in Electricity Supply.* MIT Press.
+- Vickrey, W. (1961). "Counterspeculation, Auctions, and Competitive Sealed Tenders." *Journal of Finance* 16(1): 8–37.
+- Williamson, O. (1966). "Peak-Load Pricing and Optimal Capacity under Indivisibility Constraints." *American Economic Review* 56(4): 810–827.
+- Wilson, R. (2002). "Architecture of Power Markets." *Econometrica* 70(4): 1299–1340.
+- Wolak, F. (2003). "Measuring Unilateral Market Power in Wholesale Electricity Markets: The California Market, 1998–2000." *American Economic Review* 93(2): 425–430.
+- Wolfram, C. (1999). "Measuring Duopoly Power in the British Electricity Spot Market." *American Economic Review* 89(4): 805–826.

us_market/economists_survey_zh.md

@@ -0,0 +1,272 @@
+# 经济学家如何塑造美国电力市场：理论、实践与关键里程碑
+
+**一篇综述论文**
+
+---
+
+## 摘要
+
+很少有哪个行业像美国电力部门那样，被学院派经济学如此彻底地重新设计。在大约四十年的时间里，经济学家提出了瓦解发电环节"成本加成"监管体制的思想批判，发展出支撑美国所有集中式电力批发市场的定价理论，诊断了第一代市场设计的失败，并持续塑造着有关容量、需求响应、储能以及可再生能源并网的市场规则。本综述从四个维度追溯这一影响：（一）电力市场化改革在监管经济学中的思想渊源；（二）现代市场设计的理论基础，尤其是峰荷定价、现货定价与节点边际电价（LMP）；（三）从《公用事业监管政策法》（PURPA，1978）到联邦能源监管委员会（FERC）第888、2000、745、841和2222号法令等一系列将经济学思想转化为监管规则的制度里程碑；（四）经济学家借以诊断危机（2000–01年加州危机、2021年得州危机）、建立市场监测机构、并评估市场化改革是否兑现其效率承诺的实证与应用文献。我们认为，美国电力市场化改革最恰当的理解不是"放松管制"（deregulation），而是一场持续的应用机制设计实践——经济学家在其中从监管的批评者转变为制度的设计师。本文最后讨论了经济理论再次被推向前台的若干悬而未决的设计问题：深度脱碳背景下的资源充裕性、非凸定价问题以及"混合市场"之争。
+
+---
+
+## 1. 引言
+
+1996年4月，美国联邦能源监管委员会（FERC）颁布第888号法令，要求对州际输电网实行开放、非歧视性接入，由此启动了自20世纪30年代以来美国电力工业最深刻的一次重组。不到十年，美国三分之二的电力需求已通过独立系统运营商（ISO）和区域输电组织（RTO）运营的集中式批发市场来满足。在这些市场中，电能价格每五分钟在数千个不同的网络节点上计算一次，其求解的是一个大规模优化问题，而该问题的目标函数——在电力潮流物理规律约束下最大化社会剩余——正是直接来自福利经济学。
+
+这最后一点正是本综述的核心主题。在大多数经历放松管制的行业——航空、货运、电信、天然气——经济学家提供的是支持竞争的*论证*，市场随后通过分散的企业家活动自行演化。电力则不同。因为电力潮流服从的是基尔霍夫定律而非合同条款，因为电能（直到最近）在经济上无法储存，因为供需必须每秒每刻保持平衡以避免连锁停电，一个可运转的电力"市场"不可能自发涌现，它必须被*设计*出来。其结果是，经济学家不仅仅是电力竞争的倡导者——他们撰写了电力市场的宪法。美国所有集中式市场核心的定价算法（节点边际电价）、对冲阻塞风险的金融工具（金融输电权）、采购未来容量的拍卖机制、在得克萨斯州替代容量市场的稀缺定价机制，以及监督上述一切的市场力缓解机制，在异乎寻常的程度上都是学院派经济学研究的直接产物。
+
+本综述为一般经济学读者梳理这段历史。第2节追溯市场化改革在战后监管经济学与自然垄断理论中的思想渊源。第3节展开理论核心：峰荷定价、现货定价以及Hogan的节点边际电价框架。第4节回顾这些思想成为法律的立法与监管里程碑。第5节考察2000–01年加州危机——正是在这一事件中，实证经济学家证明了自己作为市场"诊断医师"的价值，并使经济学市场监测永久性地制度化。第6节讨论资源充裕性问题——"缺失的货币"（missing money）、容量市场与纯电能量市场的替代方案。第7节综述机制设计的贡献：拍卖、非凸定价与虚拟报价。第8节转向需求侧：动态零售定价与围绕需求响应的论战。第9节讨论当代前沿——可再生能源、储能、分布式资源与2021年得州危机。第10节回顾评估市场化改革效果的实证文献。第11节讨论各种批评意见以及正在兴起的"混合市场"之争，第12节为结论。
+
+三点说明。第一，本文聚焦于美国；1990年英国电力私有化尽管对美国的思考影响巨大（其本身也深受Stephen Littlechild等经济学家的塑造），在此仅作为背景出现。第二，"经济学家"作宽泛理解，包括麻省理工学院（MIT）等机构的工程经济学传统——Fred Schweppe是电气工程师，但《电力现货定价》（*Spot Pricing of Electricity*）是一部经济学著作。第三，如此篇幅的综述不可能面面俱到；我们的目标是准确勾勒出主要的思想脉络与制度转折点。
+
+---
+
+## 2. 思想渊源：监管经济学及其不满
+
+### 2.1 旧体制及其理论依据
+
+从20世纪20年代到80年代，美国电力工业的组织形式是：纵向一体化的投资者所有制公用事业公司持有排他性的零售特许经营权，价格由州公用事业委员会按"服务成本"（cost-of-service）原则核定，批发与州际事务则由1935年《联邦电力法》管辖。其理论依据是自然垄断理论：由于固定成本巨大，且（当时人们相信）发电、输电、配电各环节都存在普遍的规模经济，竞争被认为是浪费的或不可持续的，受监管的垄断则是"最不坏"的制度安排。
+
+经济学家参与了这一体制的创建——"公允价值基础上的公允回报率"这套方法源自制度学派经济学以及20世纪初的费率基数诉讼——但经济学界最重要的贡献出现在其后，其形式是批判。
+
+### 2.2 战后的批判
+
+战后经济学的四条思想脉络汇聚在一起，动摇了人们对受监管垄断模式的信心。
+
+**激励扭曲。** Averch与Johnson（1962）形式化地证明，回报率监管会诱使公用事业公司过度投资于资本——即所谓"镀金"效应或A–J效应。无论该效应的实证规模如何（至今仍有争议），这篇论文将监管重新定义为一个激励问题而非会计问题，并为后来的激励性监管文献（Laffont与Tirole 1993）播下了种子，这一文献后来又影响了绩效导向的费率制定。
+
+**监管俘获。** Stigler（1971）与Peltzman（1976）指出，监管是提供给政治上有影响力的利益集团的，而非服务于公共利益；Demsetz（1968）则追问：既然"服务权"本身可以通过拍卖出售——即以特许经营权竞标实现"为市场而竞争"——自然垄断为何还需要监管？这些芝加哥学派的贡献侵蚀了"监管可靠地矫正市场失灵"这一预设。
+
+**关于规模经济的证据。** 实证研究——最有影响的是Christensen与Greene（1976）——发现到20世纪70年代，美国大多数大型发电企业在电厂与企业层面的规模经济已经耗尽。如果发电不再是自然垄断，那么对其实施监管的理由便不复存在，即便电网环节仍然是自然垄断。这一"拆分"（unbundling）洞见——在可行之处引入竞争（发电、零售），在必要之处保留监管（输电、配电）——成为整个市场化改革的总蓝图。
+
+**放松管制运动的实践。** Alfred Kahn的两卷本《监管经济学》（1970–71）提供了理论综合，而Kahn本人作为卡特总统任内的民用航空委员会主席，证明了一位学院派经济学家可以从体制内部拆解一个监管体制。航空（1978）、货运（1980）、铁路（1980）与天然气（1978–92）的放松管制，为攻克电力这一公认最难啃的硬骨头，同时提供了政治模板与职业自信。
+
+### 2.3 直接诱因与那本关键著作
+
+20世纪70年代摧毁了公用事业行业的战后均衡：石油危机、核电建设经济性的崩溃（伴随著名的成本超支事件与华盛顿公共电力供应系统WPPSS违约）、需求增长停滞，以及两位数的电价上涨，使监管者与消费者一同背弃了原有模式。1978年的《公用事业监管政策法》（PURPA，详见第4节）无意间进行了一场自然实验，证明非公用事业发电商同样能够建设并运营电厂。
+
+正是在这一背景下，Paul Joskow与Richard Schmalensee的《电力市场：电力公用事业放松管制分析》（*Markets for Power*，1983）问世，成为有史以来对电力市场化改革最具影响力的经济学分析。Joskow与Schmalensee做了两件事。第一，他们系统性地驳斥了"现状即有效率"的论断，翔实记录了服务成本监管的激励问题。第二——这一点常被遗忘——他们对批发竞争持深度*审慎*态度，逐一列举了天真的放松管制将遭遇的技术障碍（环流、可靠性外部性、输电受限地区的市场力）。这本书的持久贡献在于界定了研究议程：他们指出的每一个障碍，在随后二十年里都变成了经济学家着手求解的设计问题。Joskow此后四十年的研究产出——涉及输电投资、合同结构、容量市场，直至最近的脱碳问题——构成了整个领域的连接组织；而他作为顾问、董事会成员和无数监管程序中的专家证人的角色，使他堪称电力行业历史上最具影响力的一位经济学家。
+
+---
+
+## 3. 理论基础：从峰荷定价到节点边际电价
+
+### 3.1 边际成本定价与峰荷问题
+
+深层理论始于战后法国。法国电力公司（EDF）的经济学家兼工程师Marcel Boiteux（Maurice Allais的学生）在20世纪40年代末至50年代研究出：当需求随时间波动且容量成本高昂时，如何按边际成本为电力定价——非高峰时段用户只需支付电能成本，高峰用户则需支付电能成本加上容量的边际成本（Boiteux 1949、1956；其英译本主要经由Nelson 1964介绍给英语世界）。Steiner（1957）在美国文献中独立发展了峰荷定价模型，Williamson（1966）完善了其福利分析。Boiteux关于预算约束下次优定价的姊妹篇研究（拉姆齐–布瓦特定价）成为以最小扭曲回收固定成本的标准框架。EDF于1956年实施的"绿色电价"（tarif vert）证明这些绝非黑板上的空谈。
+
+另外两项要素补齐了古典工具箱。Turvey（1968）等人将峰荷定价与投资规划联系起来。而不确定性与配给条件下的定价理论——当容量短缺时，有效率的价格是什么？——则指向了后来主导资源充裕性论争的核心概念：**失负荷价值**（Value of Lost Load, VoLL），即消费者为避免停电而愿意支付的金额。当系统逼近非自愿限电时，有效率的稀缺价格应向VoLL攀升；从ERCOT的价格上限到容量市场的需求曲线，一切设计都建立在这一个思想之上。
+
+### 3.2 现货定价：Schweppe与MIT学派
+
+从电价理论迈向市场设计的决定性一步，是在20世纪70年代末至80年代由MIT的Fred Schweppe——一位电气工程师——与经济学家及工程师Michael Caramanis、Richard Tabors、Roger Bohn合作完成的。他们的研究计划以《电力现货定价》（Schweppe, Caramanis, Tabors, and Bohn 1988）一书达到顶峰，提出的问题是：如果承认电力的边际成本不仅随时段变化，而且是连续的、随机的、并且——关键地——随电网位置而变化，那么电力的*真实*边际成本究竟是什么？
+
+他们的答案是：每一时刻、电网每一母线上的有效率现货价格，等于系统lambda（参考节点的电能边际成本）经边际输电损耗和有约束力的输电约束的影子价格调整后的值。一旦电网发生阻塞，各地价格便会不同，因为基尔霍夫定律使廉价电力无法送达高价地区。Schweppe等人设想以这些价格为基础，实现电力系统的去中心化"自稳态"（homeostatic）控制：发电商和（原则上）消费者对实时价格作出反应，就能复现中央调度工程师历来计算的最小成本调度——但这一次是通过市场实现。这本书写成之时，尚不存在任何能够实施它的机构；它后来成为此后所有市场设计的技术圣经。
+
+### 3.3 Hogan：节点边际电价与金融输电权
+
+将现货定价理论转化为可实施市场架构的人，是哈佛大学肯尼迪学院的William Hogan。从《电力输送的合同网络》（Hogan 1992）开始，Hogan解决了长期困扰去中心化方案的难题：在交流电网中，电力潮流无法按合同路径流动，如何定义可交易的输电权？
+
+Hogan的洞见分为两部分。第一，拥抱而非对抗集中调度：让独立系统运营商执行*基于报价的安全约束经济调度*——这与公用事业公司历来求解的优化问题相同，只是报价曲线改由相互竞争的发电商提交——并让该优化问题的对偶变量（影子价格）定义每个节点的市场出清价格，即**节点边际电价（LMP）**。节点的LMP恰好按Schweppe的公式分解为：电能成本＋网损＋阻塞。第二，以*金融*而非物理方式定义输电权：一份从节点A到节点B的**金融输电权（FTR）**，赋予其持有者按指定电量获得阻塞租金差额LMP(B) − LMP(A)的权利。只要基础注入量在物理上可行，FTR就同时可行（"收入充足性"定理）；对于匹配的物理计划，FTR能够完美对冲阻塞风险；并且FTR可以拍卖或分配给历史电网用户——从而在解决对冲这一经济问题的同时，也解决了转型的政治问题。
+
+这一"电力库＋FTR"或称"一体化"架构——集中调度、节点电价、金融权利、双结算（日前与实时）、电能与辅助服务联合优化——在20世纪90年代与另一种"去中心化/双边"愿景（与部分加州设计者及英国电力库的批评者相关联）展开了激烈较量，后者主张自计划、分区电价与物理输电权。经验给出的裁决毫不含糊：PJM在分区电价遭到投机套利后于1998年放弃分区改用节点电价；纽约于1999年以LMP（及其FTR变体——输电阻塞合同TCC）启动市场；新英格兰于2003年转轨；中西部ISO于2005年以节点制启动；加州在危机之后于2009年围绕完全节点电价重组市场；得克萨斯于2010年从分区转向节点；西南电力库（SPP）于2014年跟进。如今美国所有集中式市场都运行在Hogan的架构之上。很难在其他任何行业找到这样的例子：一位经济学家的设计方案，每五分钟在一个大陆经济体的大部分区域内被执行一次。
+
+Hogan后续的工作将该框架延伸到几乎每一个有争议的设计维度：多重结算体系、FTR期权与义务、经由运行备用需求曲线实现的稀缺定价（见第6.4节）、附加费用（uplift）与非凸定价（与Gribik和Pope合作）、输电投资激励，以及——延续至今的——脱碳电网的定价问题。Harvey与Hogan同伯克利实证学派围绕加州问题的持续论战（见第5节），也为市场力测度确立了方法论标准。
+
+### 3.4 理论的配角群像
+
+另有三支理论文献汇入了设计典籍：
+
+- **拍卖与机制设计。** Vickrey（1961）的奠基性工作；Wilson–Milgrom关于共同价值与多单位需求拍卖的研究计划；以及Robert Wilson的《电力市场的架构》（*Architecture of Power Markets*，Econometrica，2002）——该文将电力设计表述为一个"市场架构"问题，即把一个经济问题分解为一系列相互衔接的拍卖子市场。Wilson本人曾为国际及加州的电力市场设计提供咨询，Paul Milgrom的咨询实践也将拍卖设计延伸至电力采购领域。2020年授予Milgrom与Wilson的诺贝尔经济学奖，其表彰的工作明确将电力列为应用领域之一。
+- **供给函数均衡。** Klemperer与Meyer（1989）建模了不确定性下以供给曲线展开的竞争——这正是发电商提交给ISO的东西；Green与Newbery（1992）将供给函数均衡应用于英国电力库，为电力市场提供了标准的寡头模型，并较早发出了有影响力的警告：双寡头电力库的定价将远高于边际成本。这一跨大西洋的结论深刻影响了美国对市场结构与资产剥离要求的思考。
+- **合同与纵向安排。** Joskow早年关于长期煤炭合同与资产专用性的实证研究（属威廉姆森传统），以及Allaz–Vila（1993）关于远期合约可以缓解现货市场力的结论，共同影响了过渡性合同的设计，以及一项经久不衰的政策偏好——保持高合同覆盖率。这是经济学家从加州（无合同覆盖，酿成灾难）与其他市场的对比中提炼出的最清晰教训之一。
+
+---
+
+## 4. 制度里程碑：思想成为法律
+
+### 4.1 PURPA（1978）：一场意外的实验
+
+1978年《公用事业监管政策法》（PURPA）作为能源危机立法获得通过，它要求公用事业公司按其*可避免成本*（avoided cost）购买"合格设施"（热电联产及小型可再生能源）的电力。可避免成本当然是一个经济学概念——即公用事业公司因此免于发生的边际成本——其落地引发了上千场费率听证，经济学家在其中就如何测算该成本出庭作证。PURPA更深层的意义是无心插柳：它引致的非公用事业发电浪潮（尤其在加州、纽约与得克萨斯）证明独立发电商有能力融资、建设并运营电厂，从而摧毁了"发电必须依托纵向一体化垄断"的经验前提。包括Joskow在内的经济学家们，既记录了这一示范效应，也记录了行政核定可避免成本合同的种种弊端——当定价过高时（如加州的"标准报价4号"合同），电力用户被迫背负高于市场水平的成本，这留下了一个关于"行政定价对拍卖定价"的长久教训。
+
+### 4.2 1992年《能源政策法》与FERC第888/889号法令（1996）
+
+1992年《能源政策法》创设了"豁免批发发电商"，将独立发电从《公用事业控股公司法》中解放出来，并授权FERC逐案下令开放输电。FERC在1996年第888号法令中将其普遍化：所有拥有输电资产的公用事业公司必须提交开放接入电价表，以其向自身提供服务的同等条件向第三方提供输电服务；第889号法令则建立了OASIS信息系统与职能拆分要求。其经济逻辑——瓶颈垄断环节（电网）不得利用其地位向潜在竞争性环节（发电）延伸市场势力——正是产业组织学中"关键设施/纵向封锁"的推理，由各方阵营中的一大批经济学家向FERC陈述。第888号法令还认可了"搁浅成本"的回收，这本身就是一个可观的经济学文献主题（William Baumol、Alfred Kahn等人曾就回收究竟是有效率的转型补偿还是不当纾困展开论战）。
+
+### 4.3 州层面的重组：加州AB 1890法案与东北部各州
+
+*零售*市场的重组属于各州事务。加州行动最早、最为激进：AB 1890法案（1996）落实了加州公用事业委员会经多年听证形成的决定（学院派经济学家是听证的核心证人），设立了电力交易所（PX）与加州独立系统运营商（CAISO，于1998年3月31日启动），要求公用事业公司剥离大部分化石能源发电资产，将零售电价冻结在此前水平的九折，并且——祸根就此埋下——不鼓励长期购电合同，把采购推向日前与实时市场。这一设计是双边/去中心化学派与电力库学派之间一个双方都不满意的折中；数位知名经济学家（包括后来组成CAISO市场监察委员会的成员）在市场启动前就已警告：供给紧张、需求无合同覆盖、零售价格封顶与市场力暴露的组合十分危险。马萨诸塞、罗德岛、宾夕法尼亚、纽约、新泽西、马里兰、伊利诺伊、俄亥俄、密歇根与得克萨斯在1996至1999年间相继通过各自的重组法案，通常将零售选择权与发电资产剥离及转型成本附加费配套推行。到2000年，约半数州已颁布重组法案或启动相关程序。
+
+### 4.4 第2000号法令（1999）与"标准市场设计"插曲（2002–05）
+
+FERC第2000号法令鼓励（但未强制）组建区域输电组织（RTO）——具备特定特征与职能的独立电网运营机构。ISO/RTO这一制度安排的理论依据——由独立、非营利的运营商运行开放的、基于报价的市场——是Hogan、Joskow等人为解决输电独立性问题而发展出来的。2002年，FERC试图更进一步：《标准市场设计》（SMD）规则提案拟在全国范围内强制推行完整的节点架构——LMP、FTR、日前/实时结算、市场监测与资源充裕性机制。SMD是经济学家对联邦政策直接影响力的最高峰：其文本读起来就像市场设计文献的摘要。它也激起了南部与西部各州（电价低廉、由监管体制供电、对市场毫无兴趣的地区）的激烈政治反弹，FERC于2005年正式撤回提案。其结果便是今日的版图：集中式节点市场覆盖美国约三分之二的用电负荷（PJM、MISO、CAISO、ERCOT、NYISO、ISO-NE、SPP），东南部与西部大部分地区仍维持纵向一体化——这一格局构成了一场持久的自然实验，实证研究者日后将充分加以利用（见第10节）。
+
+### 4.5 后续联邦法令
+
+此后的FERC规则制定（各项将在后文相应章节讨论）继续把经济分析转化为规则：第719号法令（2008）涉及需求响应与ISO治理；第745号法令（2011）涉及需求响应补偿；第755号法令（2011）涉及按绩效付费的调频补偿；第1000号法令（2011）涉及输电规划与成本分摊——Joskow与Jean Tirole曾就该主题撰写奠基性理论；第841号法令（2018）涉及储能参与市场；第2222号法令（2020）涉及分布式能源资源聚合。模式始终如一：一个设计问题浮出水面；经济学家撰写论文、提交专家意见；FERC以规则形式确立一个（有争议的）解决方案；随之而来的是诉讼与更多论文。
+
+---
+
+## 5. 火的考验：加州危机与市场监测的兴起
+
+### 5.1 危机
+
+2000年5月至2001年6月间，加州批发电价均值升至此前数年水平的五到十倍；全州遭遇轮流停电；太平洋燃气电力公司（PG&E）进入破产程序；南加州爱迪生公司濒临破产；电力交易所停止运营；州政府为购电耗资数百亿美元。危机的直接原因包括干旱（西北水电出力下降）、天然气与氮氧化物排放许可价格上涨、需求增长，以及市场供需的确紧张。但这场危机之所以成为一个经济学事件，在于稀缺之*上*又发生了什么。
+
+### 5.2 诊断：经济学家充当法医
+
+一批实证产业组织经济学家——以Severin Borenstein、James Bushnell与Frank Wolak为代表，依托加州大学能源研究所以及由Wolak担任主席的CAISO市场监察委员会——发展并应用了一种直接*测度*市场力的方法：以实际边际成本叠加实际需求（并考虑水电机会成本与外购电力），模拟出完全竞争情形下的反事实价格，再将实际价格与竞争性价格之间的差距归因于市场力的行使。他们的核心论文——Borenstein、Bushnell与Wolak，《测度加州重组后批发电力市场的市场无效率》（*American Economic Review*，2002）——发现市场力解释了2000年夏季价格上涨的很大一部分：在需求缺乏弹性且无合同覆盖的情况下，少数供应商单边持留产能所致，未必是非法共谋。Joskow与Edward Kahn（2002）用互补的方法得出类似结论，记录了大量的产出持留行为。Scott Harvey与William Hogan则在一系列论文中对上述测算的量级与归因提出质疑，认为成本测度误差与运行约束足以解释差距的大部分——无论人们对这场论战的裁决如何，它都锻造并抬高了整个领域的方法论标准。随后的执法行动与安然（Enron）录音带（"死星""胖小子""回旋镖"——这些策略所利用的恰恰是电力库学派早已警告过的加州分区市场与邻近市场之间的制度缝隙）为这场公共论争画上了句号。
+
+经济学家对加州*为何*失败的解释后来成为经典叙事：（一）零售价格冻结切断了需求与批发价格的联系，使需求几乎完全无弹性；（二）公用事业公司被禁止或不被鼓励签订远期合同，使其完全暴露于现货价格，从而给予供应商最大的抬价激励（Allaz–Vila逻辑的反面）；（三）PX电能市场与ISO阻塞管理及可靠性职能的分离，加之分区而非节点定价，制造了套利缝隙；（四）软性价格上限诱发了策略性外送再回购。上述每一要素事前都已有部分经济学家指出；危机将这些警告转化为设计正统：合同覆盖至关重要，需求响应至关重要，节点优于分区，市场力缓解必须内建于设计而非事后补丁。
+
+### 5.3 经济学的制度化：市场监测机构
+
+加州危机最持久的制度遗产是**市场监测机构**（market monitor）。如今每个ISO/RTO都设有独立的内部和/或外部市场监测单位——Potomac Economics公司（由经济学家David Patton领导）担任多个市场（包括ERCOT、NYISO与MISO）的独立监测方；Monitoring Analytics公司（由经济学家Joseph Bowring领导）监测PJM；CAISO则设有市场监测部及市场监察委员会，后者历来由学院派经济学家（Wolak、Bushnell及其同事）担任主席。这些单位执行"行为与影响"缓解筛查（当拥有局部市场力的供应商报价远超参考成本时自动实施报价封顶），发布充满集中度与价格–成本毛利分析的年度《市场状况报告》，并向FERC执法部门移交案件。市场监测是作为一项常设监管职能来实践的应用产业组织学——这一职业生态位在经济学家将其创造出来之前根本不存在。
+
+---
+
+## 6. 资源充裕性："缺失的货币"、容量市场与稀缺定价
+
+### 6.1 "缺失的货币"问题
+
+在理论理想中，纯电能量市场（energy-only market）通过稀缺租金为容量融资：短缺时段价格飙升至失负荷价值（VoLL）水平，对此类租金的预期恰好诱导出有效率的容量投资，使边际机组能够回收其固定成本。而在实践中，美国各市场系统性地压制了这些租金：报价上限（东部市场长期为1,000美元/兆瓦时）、系统运营商宁可调用备用也不让价格上涨的场外可靠性操作，以及市场力缓解规则，三者共同截断了价格分布的右尾。发电商回收固定成本所需的收入与被压制的电能价格实际提供的收入之间的缺口，被称为**"缺失的货币"问题**——这一术语经由Peter Cramton与Steven Stoft的工作得以流行，并成为Joskow若干重要论述（如Joskow 2008）的核心。这一诊断把可靠性从一项工程指令重新表述为一种定价失灵，而经济学界在药方上产生了分裂：是修复电能价格（稀缺定价），还是外挂一个独立的容量市场。
+
+### 6.2 容量市场与需求曲线创新
+
+东部各ISO选择了容量市场：强制性的远期拍卖，售电主体必须采购足以满足备用裕度目标的经认证容量。早期的"垂直需求"设计（固定需求量，一旦供给略微超过需求价格便崩至零）被证明具有病态的波动性且易被操纵。补救方案——2003年首先在NYISO实施、如今已被普遍采用——是**倾斜的行政需求曲线**：一条以估算的"新建机组成本"（CONE）为锚点的向下倾斜曲线，使容量价格成为容量盈余的平滑函数。这条需求曲线是纯粹的经济工程——一条为可靠性行政构造的边际价值曲线——其参数（CONE数值、曲线形状与锚点）通过周而复始的监管程序确定，而这些程序本质上是计量经济学的诉讼战。Cramton与Stoft的设计工作（其中许多受各ISO委托完成），连同Joskow的分析，塑造了PJM的"可靠性定价模型"（RPM，2007年实施）与ISO-NE的"远期容量市场"（FCM，2008年首次拍卖），包括提前三年采购、分区容量定价，以及后来的"按绩效付费"惩罚结构（ISO-NE首创；2014年1月极地涡旋暴露了"有容量却顶不上"的弱点后，PJM推出容量绩效改革）。此后一代的争端——针对受州补贴新进入者的最低报价规则（MOPR）、买方市场力、以及间歇性与时长受限资源的容量认证——使容量市场经济学始终位居该领域诉讼最密集的战场之列。
+
+### 6.3 纯电能量市场的替代方案：ERCOT与运行备用需求曲线
+
+得克萨斯走了另一条路。ERCOT——州内电网，不受FERC管辖，依据1999年参议院第7号法案完成重组——运行一个没有容量机制的纯电能量市场，其报价上限历史上很高（一度达到9,000美元/兆瓦时）。其标志性机制是2014年采纳的**运行备用需求曲线（ORDC）**，直接基于William Hogan的方案（《通过运行备用实现电力稀缺定价》，2013）：一个实时价格加价项，等于失负荷概率（作为当前备用水平的函数）乘以失负荷价值，使价格随备用收缩而平滑地、*自动地*向VoLL攀升——以行政方式复现一个需求侧可响应的理想市场本应产生的稀缺租金。ORDC或许是所有行业中最纯粹的例证：一篇经济学工作论文中的公式被直接写进了一个大型市场的结算软件。2021年的危机（见第9.4节）使这一设计经受了最严酷的检验。
+
+### 6.4 悬而未决的论战
+
+容量市场对纯电能量市场之争至今仍在继续，并因脱碳而愈发尖锐。容量市场的支持者强调投资风险的降低与政治上的稳健性（监管者绝不会长期容忍真正的VoLL定价）；纯电能量市场的支持者（Hogan，以及立场不同但同样持批评态度的Wolak——后者对两种机制都心存疑虑，而强调合同义务）则反驳说：容量市场付费买的是"立在地上的铁疙瘩"而非实际绩效，招致无休止的行政博弈，并削弱了奖励灵活性与需求响应所必需的稀缺价格信号。储能与可再生能源的兴起——其容量价值（以"有效带载能力"ELCC测度，又一个与经济学相邻的构造）随渗透率变化——使资源充裕性设计成为当下最活跃的研究领域之一（见第9节）。
+
+---
+
+## 7. 机器舱里的机制设计：拍卖、非凸性与虚拟报价
+
+### 7.1 作为优化问题的市场
+
+ISO的日前市场是一个密封报价、多单位、多产品的拍卖，通过安全约束机组组合与经济调度出清——这是一个混合整数规划，其规模（数万条约束、数千个资源）使其跻身世界上运行规模最大的拍卖之列。其中的经济学问题是经典的机制设计问题：什么价格能够支撑有效率的配置？如实报价的激励何在？相互关联的产品（电能、调频、旋转备用）应如何联合优化？经济学家——常与运筹学家合作——给出了如今已嵌入软件的答案。
+
+### 7.2 非凸性与附加费用：定价的前沿
+
+发电机组的成本是非凸的——启动成本、最小出力水平、最小开停机时间——因此支撑有效率机组组合的瓦尔拉斯均衡可能不存在：在LMP价格下，一些已开机机组亏损，而一些未开机机组本可盈利。美国各市场用"附加费用"（uplift）即"补齐"侧支付来打补丁，但这种支付具有歧视性、不透明，并且钝化了价格信号。理论界的回应构成了过去二十年最活跃的应用理论文献之一：O'Neill、Sotkiewicz、Hobbs、Rothkopf与Stewart（2005）展示了如何利用整数活动定价在存在非凸性时构造市场出清价格；Gribik、Hogan与Pope（2007）提出了**凸包定价**（convex hull pricing，又称"扩展LMP"），即基于成本函数凸包络定价以最小化总附加费用。这些思想的各种变体已付诸实施：MISO采纳了凸包定价的近似算法，FERC的快速启动机组定价系列法令（2016–2020）则推动各ISO允许缺乏灵活性的快启机组参与定价。在这一领域，经济理论、运筹学与结算软件正在实时协同演化。
+
+### 7.3 虚拟报价与双结算体系
+
+Hogan的多重结算架构把具有财务约束力的日前市场与实时平衡市场分离开来。为使二者保持一致，市场允许**虚拟（收敛）报价**：纯金融头寸，套利日前与实时价格的预期价差。理论预言套利者将促使两个价格收敛，改善机组组合；实证文献大体证实了这一点，同时也记录了诸多微妙之处——Jha与Wolak关于加州的研究发现收敛报价改善了价格收敛与生产效率，而摩根大通"补齐支付"操纵案（2013年与FERC和解）以及关于MISO网损套利的研究，则展示了金融产品如何与市场缝隙相互作用。FTR拍卖的设计、FTR收益长期低于拍卖价格的痼疾，以及FTR组合违约风险应由谁承担的问题（2018年PJM市场GreenHat违约事件将其推向风口浪尖），同样主要由经济学家进行分析。
+
+### 7.4 采购拍卖与保底供电服务
+
+在ISO市场之外，经济学家还设计了重组各州采购保底零售供电服务的拍卖机制——最著名的是新泽西州"基本发电服务"（BGS）**降价时钟拍卖**，其设计有拍卖经济学家（包括Ausubel与Cramton圈子）的参与，自2002年起每年举行；伊利诺伊州的采购活动亦然。Milgrom与Wilson更广泛的拍卖设计咨询实践（2020年诺贝尔奖颁奖词对此予以表彰）将电力采购列为其应用领域之一。可再生能源采购的合同设计（指数化购电协议、差价合约）是这一传统当前的前沿。
+
+---
+
+## 8. 需求侧：动态定价与需求响应之战
+
+### 8.1 经济学家最古老的抱怨
+
+从Boiteux开始，经济学家就坚持认为需求侧是市场的另一半：如果零售用户面对的是不随时间变化的价格，那么本应约束批发市场的需求价格弹性便不复存在，稀缺无法有效率地出清市场，而且（如加州所证明的）整个系统将变得脆弱。Borenstein在21世纪头十年的研究（如《实时电价的长期效率》，2005；Borenstein与Holland 2005关于仅部分用户面对实时电价时产生扭曲的分析）为**动态零售定价**建立了现代福利经济学论证；Wolak设计并评估了测度家庭对尖峰电价响应的现场实验（包括在阿纳海姆及海外进行的实验）；其后大量的实验文献（包括Jessoe与Rapson关于信息反馈的研究）量化了技术手段（智能电表、自动化）如何提高有效弹性。实际采用仍然有限——默认的固定费率几乎无处不在——经济学家将其归因于政治经济学与行为摩擦，而非经济学本身有误；理论之清晰与零售实践之落差，是经济学界一个持续存在的尴尬，学界至今仍在探究（包括关于动态定价分配效应的研究）。
+
+### 8.2 第745号法令与"EPSA诉FERC案"：对垒双方都是经济学家
+
+批发侧的**需求响应**——付费让用户削减用电——引发了电力行业最著名的一场经济学争论。FERC第745号法令（2011）要求各ISO按完整的LMP向需求响应付费。Alfred Kahn在其晚年最后的公共介入之一中提交意见支持全额LMP补偿；William Hogan则强力主张，正确的支付额应为LMP减去用户所规避的零售电价（"LMP−G"），因为向一个同时省下零售电费的人支付全额LMP，是对削减用电的双重补偿，会补贴无效率的需求削减。这场争论——其实质在于反事实基准的定义以及对零售电价扭曲的处理方式——最终以"FERC诉电力供应协会案"（*FERC v. EPSA*，2016）的形式打到最高法院，法院维持了FERC的管辖权与第745号法令。双方阵营中都布满了经济学家的法庭之友意见书与学术论文；这一事件如今已成为教学案例，示范"次优"推理如何因将哪一种扭曲视为给定而得出截然相反的结论。
+
+---
+
+## 9. 当代前沿：可再生能源、储能、分布式资源与2021年得州危机
+
+### 9.1 零边际成本的进入与价格形成
+
+受补贴与政策强制驱动的风电与光伏——依托生产税抵免（PTC）、投资税抵免（ITC）及各州可再生能源配额制——以改变价格形成规律的体量进入批发市场：多风地区出现负电价（PTC付钱让机组发电的可预见后果，经济学家很早就作了分析）、加州的"鸭子曲线"、被压低的午间电价，以及火电机组在报价排序中被挤出。经济学文献从多条战线作出回应：测度可变可再生能源的市场价值随渗透率上升而衰减；量化可再生能源进入对排放与价格的影响；分析场外补贴与容量市场的相互作用（MOPR之战）；并力推首优方案——在ISO市场中实行碳定价，包括NYISO与PJM的碳加价正式设计方案，"未来资源研究所"（RFF）等机构的经济学家对此进行了分析。必须承认，"技术中性的碳定价优于特定技术强制令"这一广泛的职业共识，在立法层面收效有限；经济学家因此日益转向次优工具设计的分析（清洁能源标准、基于ELCC的容量认证、混合采购机制）。
+
+### 9.2 储能与分布式资源：第841号与第2222号法令
+
+FERC第841号法令（2018）要求各ISO建立参与模型，允许储能同时作为买方和卖方参与定价；第2222号法令（2020）将参与资格扩展到分布式能源资源的聚合体。两项法令回应的都是经济学家提出的问题框架：储能套利及其福利效应、储能被双重收取输电费的问题，以及表后资源的处理。日益增长的实证文献（关于ERCOT与CAISO电池机群的研究）如今正在测度储能以多快的速度竞争掉自身的套利利润——这是进入经济学在现实中上演的一场干净利落的检验。
+
+### 9.3 输电，再一次
+
+输电规划与成本分摊——Joskow与Tirole关于商业化输电的理论研究以及Hogan"受益者付费"原则的主题——随着第1000号法令（2011）与第1920号法令（2024）重回舞台中央，后者要求开展前瞻性的区域规划并按受益分摊成本。法院（特别是第七巡回上诉法院在"伊利诺伊商务委员会"系列案件中，判决书出自经济学家出身的Posner法官之手）所执行的"受益者付费"标准，就是不折不扣的成本收益经济学。
+
+### 9.4 冬季风暴"乌里"（2021年2月）：对纯电能量设计的压力测试
+
+2021年2月的得克萨斯大停电——连续数日的轮流停电、电价被行政性地按在9,000美元/兆瓦时的上限超过四天、约二百余人死亡、数百亿美元规模的财务再分配——引发了自加州危机以来最激烈的经济学复盘。经济学讨论区分了几个层面的问题：这场失败首先是*供给可用性*的失败（未做防寒改造的天然气供应与发电设施），而非市场设计本身的失败；稀缺定价按设计运转了，但上限价格持续的时长暴露了零售侧对冲的缺失（使用批发价格直通费率的Griddy公司用户），并提出了一个问题——持续数日的VoLL水平价格在政治上或合同上是否可以承受；而得州公用事业委员会（PUCT）在限电结束后仍将价格维持在上限的决定，则成为一场关于场外干预的著名争议。时任ERCOT董事会成员的Peter Cramton（与其他州外董事一道在风暴后辞职）、Hogan、Wolak、Bushnell等众多学者发表了针锋相对的分析；危机之后，得州降低了价格上限、修改了ORDC、强制实施防寒改造，并提出了新的"绩效信用机制"方案——每一项都以明确的经济学语言进行辩论。"乌里"还重新激发了学界对天然气与电力市场设计相互作用的兴趣，以及对Wolak长期倡导的零售/售电主体强制对冲义务的兴趣。
+
+### 9.5 脱碳电网的资源充裕性
+
+当前的研究前沿在追问：为拥有可观边际成本的燃料发电机群设计的"LMP＋稀缺定价"架构，对于一个由零边际成本、气象驱动、时长受限资源主导的系统而言是否仍然适用。经济学家正在积极研究的问题包括：当边际资源是储能或被弃的可再生能源时的价格形成（机会成本定价）；基于边际ELCC的容量认证；长时充裕性与尾部风险（与气候相关的相关性停机）；资本密集型清洁可控资源在收入高度波动下的融资问题（由此催生差价合约与"可靠性期权"——后者是Cramton–Stoft的设计，已在哥伦比亚采用，ISO-NE的FCM也以修改形式采纳）；以及集中式市场是否应演化为集中化的长期采购——即"混合市场"。Joskow（其近年论文认为纯电能市场模式正承受日益加大的压力）、Schmalensee、Wolak、Hogan以及更年轻的一代学者已加入这场论战。换言之，车轮再次转动，而经济学家又一次站在了车轮的两侧。
+
+---
+
+## 10. 改革成功了吗？实证评估文献
+
+市场化改革创造了一场自然实验——集中式市场地区对传统监管地区；被剥离的电厂对公用事业保留的电厂——实证经济学家对此挖掘了二十年。主要发现如下：
+
+- **运营效率。** Fabrizio、Rose与Wolfram（*AER*，2007）发现，重组各州的投资者所有制电厂相对于未重组各州的电厂，其人工与非燃料运营支出降低了数个百分点——竞争（或对竞争的预期）强化了成本纪律的证据。Davis与Wolfram（*AEJ: Applied*，2012）发现，将核电厂剥离给独立的商业化运营商使容量因子提高了约10个百分点，这一巨大的效率增益集中体现在停机时长的缩短上。Cicala（*AER*，2015）表明，一旦成本转嫁激励被移除，放松管制的电厂采购煤炭的价格更为低廉。
+- **调度效率。** Cicala（*AER*，2022）利用ISO市场分批扩张的准实验设计估计，基于市场的调度通过将出力重新配置给低成本电厂并扩大跨公用事业边界的贸易利得，每年节约生产成本达数十亿美元量级——这是LMP机制兑现其设计者承诺的最干净的证据。
+- **投资与技术结构。** 商业化投资以与监管时代迥异的技术选择（快速、模块化的燃气机组，其后是可再生能源与电池）响应价格信号；文献既记录了这种响应性，也记录了纯商业化风险暴露所产生的繁荣–萧条周期。
+- **零售竞争。** 零售侧的文献结论更为复杂：来自得克萨斯等地的证据记录了居民零售选择中显著的搜寻摩擦、价格离散与"迷惑定价"（Hortaçsu、Madanizadeh与Puller 2017年关于得州的研究是这方面的标杆），这为零售自由化的宏大主张降了温——即便批发侧的收益已被接受。
+- **价格。** 简单地比较监管州与重组州的零售价格存在严重的混淆因素（重组各州本来就是高成本州；天然气价格周期主导一切），经济学界已大体形成共识：市场化改革的福利收益体现在成本、调度与电厂绩效上，而非明确无误的零售降价上——这是一个诚实的、尽管在政治上不能令人满意的裁决。
+
+在评估之外，电力还成了实证产业组织学最钟爱的实验室：由于工程口径的边际成本可以直接测量，研究者能够直接计算加成率而无需推断。这个行业因此产出了关于寡头报价行为（Wolfram关于英格兰–威尔士电力库的研究；Hortaçsu与Puller关于ERCOT报价者偏离最优报价的研究）、远期合约、拍卖行为以及环境政策交互作用的一系列奠基性研究。方法论的交流是双向的：电力数据锤炼了产业组织理论，而产业组织的工具则为市场监测机构输送了人才。
+
+---
+
+## 11. 批评、逆流与设计的限度
+
+一篇诚实的综述必须记录：经济学家对电力市场化改革的主导地位不乏批评者，其中也包括经济学界内部的声音。
+
+第一，20世纪90年代的政治动员中**对消费者节省的承诺被夸大了**，学界更为审慎的声音（Joskow是其中的代表）花了多年时间区分证据支持什么与倡导者宣称什么。第二，**复杂性本身就是一种成本**：电能、辅助服务、容量、FTR与虚拟产品层层叠加的大厦——每一层都在给上一层的激励问题打补丁——招致了多方批评（左翼视之为为交易商设计的不透明体系，部分工程界人士视之为脆弱，公共电力倡导者则视之为昂贵的弯路）。第三，**治理批判**：ISO的利益相关方程序与FERC诉讼是通过资金雄厚的在位者拥有结构性优势的程序来分配租金的——这一斯蒂格勒式的观察，与设计传统的技术官僚自我形象格格不入。第四，**脱碳批判**：各州清洁能源政策如今驱动着绝大部分投资，受补贴的进入与市场价格形成之间由此产生的碰撞（MOPR、混合市场方案）使一些学者认为，20世纪90年代的架构正在被悄然取代，回归"计划采购、市场只管短期平衡"——即"调度靠市场、投资靠规划"。这究竟意味着设计纲领的失败，还是其适应性的成功，或许是这个领域未来十年的核心解释学问题；而颇为应景的是，参与这场论争的仍是同一批经济学家、他们的学生，以及学生的学生。
+
+---
+
+## 12. 结论
+
+本文所综述的历史弧线，从Boiteux在法国电力公司的峰荷电价，经芝加哥学派与MIT对监管的批判、Schweppe的现货价格、Hogan的合同网络，一直延伸到今天为美国大部分批发电力出清的五分钟节点电价。沿途，经济学家建立了制度（ISO、市场监测机构、容量拍卖），参与并裁判了危机（加州、"乌里"），创造了一套市场设计与市场监督的常设职业体系，并产出了实证经济学中关于竞争与监管各自擅长什么的最丰富证据之一。有三条教训可以推广到电力之外。第一，当物理规律或技术条件排除了去中心化交易时，市场是被设计出来的人工制品，而设计的质量——机制设计——决定了竞争能否兑现其教科书上的收益。第二，转型受制于次优问题（搁浅成本、零售价格扭曲、受补贴的进入），其程度至少不亚于受制于首优蓝图；经济学界最有价值的介入往往是诊断性的，而非建筑师式的。第三，市场设计永无完工之日：每解决一个问题（输电开放、阻塞、缺失的货币），下一个问题（非凸性、容量认证、深度脱碳）便随之显现。四十年过去，美国电力市场依然是它开始时的样子——全世界规模最宏大的一场正在进行中的应用经济学实验。
+
+---
+
+## 参考文献（精选，保留英文原文）
+
+- Allaz, B., and J.-L. Vila (1993). "Cournot Competition, Forward Markets and Efficiency." *Journal of Economic Theory* 59(1): 1–16.
+- Averch, H., and L. Johnson (1962). "Behavior of the Firm under Regulatory Constraint." *American Economic Review* 52(5): 1052–1069.
+- Boiteux, M. (1949). "La tarification des demandes en pointe." *Revue Générale de l'Électricité*; English translation, "Peak-Load Pricing," *Journal of Business* 33 (1960): 157–179.
+- Boiteux, M. (1956). "Sur la gestion des monopoles publics astreints à l'équilibre budgétaire." *Econometrica* 24(1): 22–40.
+- Borenstein, S. (2005). "The Long-Run Efficiency of Real-Time Electricity Pricing." *The Energy Journal* 26(3): 93–116.
+- Borenstein, S., J. Bushnell, and F. Wolak (2002). "Measuring Market Inefficiencies in California's Restructured Wholesale Electricity Market." *American Economic Review* 92(5): 1376–1405.
+- Borenstein, S., and S. Holland (2005). "On the Efficiency of Competitive Electricity Markets with Time-Invariant Retail Prices." *RAND Journal of Economics* 36(3): 469–493.
+- Christensen, L., and W. Greene (1976). "Economies of Scale in U.S. Electric Power Generation." *Journal of Political Economy* 84(4): 655–676.
+- Cicala, S. (2015). "When Does Regulation Distort Costs? Lessons from Fuel Procurement in US Electricity Generation." *American Economic Review* 105(1): 411–444.
+- Cicala, S. (2022). "Imperfect Markets versus Imperfect Regulation in US Electricity Generation." *American Economic Review* 112(2): 409–441.
+- Cramton, P., and S. Stoft (2005). "A Capacity Market that Makes Sense." *The Electricity Journal* 18(7): 43–54.
+- Davis, L., and C. Wolfram (2012). "Deregulation, Consolidation, and Efficiency: Evidence from US Nuclear Power." *American Economic Journal: Applied Economics* 4(4): 194–225.
+- Demsetz, H. (1968). "Why Regulate Utilities?" *Journal of Law and Economics* 11(1): 55–65.
+- Fabrizio, K., N. Rose, and C. Wolfram (2007). "Do Markets Reduce Costs? Assessing the Impact of Regulatory Restructuring on US Electric Generation Efficiency." *American Economic Review* 97(4): 1250–1277.
+- Green, R., and D. Newbery (1992). "Competition in the British Electricity Spot Market." *Journal of Political Economy* 100(5): 929–953.
+- Gribik, P., W. Hogan, and S. Pope (2007). "Market-Clearing Electricity Prices and Energy Uplift." Harvard Electricity Policy Group working paper.
+- Hogan, W. (1992). "Contract Networks for Electric Power Transmission." *Journal of Regulatory Economics* 4(3): 211–242.
+- Hogan, W. (2013). "Electricity Scarcity Pricing Through Operating Reserves." *Economics of Energy & Environmental Policy* 2(2): 65–86.
+- Hortaçsu, A., and S. Puller (2008). "Understanding Strategic Bidding in Multi-Unit Auctions: A Case Study of the Texas Electricity Spot Market." *RAND Journal of Economics* 39(1): 86–114.
+- Hortaçsu, A., S. Madanizadeh, and S. Puller (2017). "Power to Choose? An Analysis of Consumer Inertia in the Residential Electricity Market." *American Economic Journal: Economic Policy* 9(4): 192–226.
+- Joskow, P. (2008). "Capacity Payments in Imperfect Electricity Markets: Need and Design." *Utilities Policy* 16(3): 159–170.
+- Joskow, P., and E. Kahn (2002). "A Quantitative Analysis of Pricing Behavior in California's Wholesale Electricity Market During Summer 2000." *The Energy Journal* 23(4): 1–35.
+- Joskow, P., and R. Schmalensee (1983). *Markets for Power: An Analysis of Electric Utility Deregulation.* MIT Press.
+- Joskow, P., and J. Tirole (2005). "Merchant Transmission Investment." *Journal of Industrial Economics* 53(2): 233–264.
+- Kahn, A. (1970–71). *The Economics of Regulation: Principles and Institutions.* Wiley (2 vols.).
+- Klemperer, P., and M. Meyer (1989). "Supply Function Equilibria in Oligopoly under Uncertainty." *Econometrica* 57(6): 1243–1277.
+- Laffont, J.-J., and J. Tirole (1993). *A Theory of Incentives in Procurement and Regulation.* MIT Press.
+- O'Neill, R., P. Sotkiewicz, B. Hobbs, M. Rothkopf, and W. Stewart (2005). "Efficient Market-Clearing Prices in Markets with Nonconvexities." *European Journal of Operational Research* 164(1): 269–285.
+- Peltzman, S. (1976). "Toward a More General Theory of Regulation." *Journal of Law and Economics* 19(2): 211–240.
+- Schweppe, F., M. Caramanis, R. Tabors, and R. Bohn (1988). *Spot Pricing of Electricity.* Kluwer Academic Publishers.
+- Steiner, P. (1957). "Peak Loads and Efficient Pricing." *Quarterly Journal of Economics* 71(4): 585–610.
+- Stigler, G. (1971). "The Theory of Economic Regulation." *Bell Journal of Economics and Management Science* 2(1): 3–21.
+- Turvey, R. (1968). *Optimal Pricing and Investment in Electricity Supply.* MIT Press.
+- Vickrey, W. (1961). "Counterspeculation, Auctions, and Competitive Sealed Tenders." *Journal of Finance* 16(1): 8–37.
+- Williamson, O. (1966). "Peak-Load Pricing and Optimal Capacity under Indivisibility Constraints." *American Economic Review* 56(4): 810–827.
+- Wilson, R. (2002). "Architecture of Power Markets." *Econometrica* 70(4): 1299–1340.
+- Wolak, F. (2003). "Measuring Unilateral Market Power in Wholesale Electricity Markets: The California Market, 1998–2000." *American Economic Review* 93(2): 425–430.
+- Wolfram, C. (1999). "Measuring Duopoly Power in the British Electricity Spot Market." *American Economic Review* 89(4): 805–826.

us_market/field_guide_en.md

@@ -0,0 +1,370 @@
+# US Power Markets: A Field Guide
+
+*Bidding, trading, and the impact of AI — consolidated from a series of working discussions, July 2026.*
+
+---
+
+## Contents
+
+1. [How the market clears](#1-how-the-market-clears)
+2. [How participants compete](#2-how-participants-compete)
+3. [Where edge actually lives](#3-where-edge-actually-lives)
+4. [How AI changes the picture](#4-how-ai-changes-the-picture)
+5. [Building a bidding system](#5-building-a-bidding-system)
+6. [China: market structure and deployment](#6-china-market-structure-and-deployment)
+7. [Open questions](#7-open-questions)
+8. [Sources and further reading](#8-sources-and-further-reading)
+
+---
+
+## 1. How the market clears
+
+The landscape splits into two worlds: **market operators** (the seven US ISOs/RTOs — PJM, CAISO, ERCOT, MISO, NYISO, SPP, ISO-NE) who *clear* the market, and **market participants** (generators, batteries, traders, hedge funds) who *bid into* it. Everything in this guide sits on one side or the other of that line.
+
+### 1.1 The clearing algorithms (ISO/RTO side)
+
+The core is large-scale **mixed-integer programming**, not ML:
+
+- **SCUC (Security-Constrained Unit Commitment)** clears the day-ahead market — a massive MILP solved with commercial solvers (Gurobi / CPLEX / Xpress), deciding unit commitments subject to network and reliability constraints.
+- **SCED (Security-Constrained Economic Dispatch)** runs every 5 minutes in real time (LP/QP), producing **Locational Marginal Prices (LMPs)** decomposed into energy + congestion + losses.
+- Energy and ancillary services are co-optimized in the same clearing.
+
+**Active frontiers on the clearing side:** convex hull pricing / extended LMP for non-convexities; stochastic and robust unit commitment for renewable uncertainty; and ML-accelerated optimization — learned warm starts, active-constraint prediction, transmission constraint screening (pushed by ARPA-E grid optimization competitions).
+
+### 1.2 The two-settlement structure: day-ahead vs. real-time
+
+All seven ISOs run a two-settlement (day-ahead + real-time) design. **Short answer: day-ahead is where the volume and money is; real-time is where the risk is.**
+
+**Day-Ahead Market (DAM)**
+- Primary market in most designs; ~90–95%+ of physical energy settles financially at the DA price in US ISOs
+- Clears once daily via auction with full network modeling
+- Reference price for the forward curve; most hedging keys off it
+- The classic academic electricity-price-forecasting (EPF) problem
+
+**Real-Time (Balancing) Market**
+- Settles only deviations from day-ahead positions — small volume, huge volatility
+- Prices can move from $30 to $3,000+/MWh (or deeply negative) in minutes
+- Critical for: risk management, batteries, fast-ramping peakers, demand response, DA–RT spread traders
+- Battery operators often earn most energy-arbitrage revenue from RT volatility
+
+Expected RT prices anchor DA prices (virtual bidding arbitrages them on average), but realized RT prices are far noisier. The practical forecasting hierarchy for most participants: (1) day-ahead hourly prices, (2) the DA–RT spread, (3) intraday/real-time — *reversed* if you operate storage or fast-response assets.
+
+Distinctively US features on the participant side: **virtual bidding** (INCs/DECs), **FTR/CRR auctions**, and the fast-growing **battery-storage bidding** specialty.
+
+---
+
+## 2. How participants compete
+
+### 2.1 Price forecasting: the foundation
+
+Electricity is not storable at scale, so prices reflect instantaneous supply-demand balance and can spike or go negative. Forecasting approaches fall into three families, usually blended:
+
+**Fundamental (structural) models** simulate the market itself: load forecast vs. the supply stack (merit order). Key inputs: fuel prices (especially gas, which often sets the margin), renewable output forecasts, planned/unplanned outages, transmission constraints, imports/exports. Tools: PLEXOS, Aurora, in-house production-cost / dispatch simulations. Used heavily by utilities, large generators, and ISOs.
+
+**Statistical / econometric models** — ARIMA/ARIMAX, GARCH for volatility, regressions on temperature, gas, load. They exploit strong seasonality (hour-of-day, day-of-week, season) and are best for "normal" conditions and short horizons.
+
+**Machine learning models** — gradient-boosted trees, neural nets (LSTMs, transformer-based), trained on weather, load, renewables, prices, fuel, and outage data. Sophisticated shops forecast the **full distribution**, not just the mean — the tails are where money is made or lost.
+
+**In practice: hybrid.** Fundamental model for structural shape + ML/statistical corrections + human trader judgment, especially for events models handle poorly (cold snaps, plant trips, unusual grid conditions).
+
+**Universal key inputs:** weather forecasts (the single biggest driver), natural gas prices, renewable generation forecasts, outage schedules, transmission/congestion conditions.
+
+**State of the art.** The benchmark progression runs **LEAR (regularized linear/LASSO) → deep neural nets → temporal architectures (LSTM, Transformer)**. Transformer models now forecast DA–RT price spreads in volatile markets (e.g., ERCOT), using load / solar / wind forecasts and temporal features, with walk-forward retraining. Well-tuned linear models remain surprisingly competitive; hybrid linear+nonlinear architectures with online learning are a current frontier. Practical SOTA is increasingly **probabilistic** — quantile regression, distributional deep nets, conformal prediction — because decisions need the full distribution, not a point forecast.
+
+### 2.2 Strategic bidding optimization
+
+Two state-of-the-art frameworks for price-makers: **bi-level optimization (MPEC-style)** and **reinforcement learning**. Deep RL (DDPG + prioritized experience replay, PPO, SAC) handles continuous state/action spaces and non-convex operating characteristics. But **stochastic programming and robust optimization remain the workhorses** for co-optimizing energy + ancillary service offers under uncertainty, with CVaR the standard risk overlay.
+
+### 2.3 Battery storage arbitrage (the hottest area)
+
+- Classical: stochastic dynamic programming, MPC over price scenarios, degradation costs in the objective.
+- Modern: deep RL for charge/discharge policies; **decision-focused learning** (forecasts trained on profit, not accuracy — see §4.4).
+- Research frontier: hierarchical / multi-agent RL coordinating DA + RT bidding under the two-settlement structure, with risk-adjusted rewards (fixed policies struggle under regime shifts).
+- Industry: optimizer firms (Habitat Energy, Gridmatic, Tyba, Fluence Mosaic) run ML forecasting + optimization stacks commercially.
+
+### 2.4 Virtual bidding and financial trading (INCs, DECs, FTRs)
+
+Essentially quant trading: gradient-boosted trees (XGBoost/LightGBM) and deep nets predicting DA–RT spreads at nodal granularity, with features drawn from weather, load/renewable forecast errors, outages, and congestion patterns. Portfolio construction runs under risk limits; FTR auctions add congestion-rent scenario optimization.
+
+### 2.5 Emerging directions
+
+- **Online / no-regret learning** — bidding policies updated directly from market outcomes, with sub-linear regret guarantees.
+- **Multi-agent RL** — Nash equilibrium approximation for market simulation and market-power analysis.
+- **LLM-enhanced trading frameworks** — very early, appearing in the literature.
+- **Open-source benchmarks** — Grid2Op, RL2Grid, new two-settlement bidding environments.
+
+### 2.6 Industry reality check
+
+Most real money is still made with **strong probabilistic forecasting + classical optimization (stochastic MIP/MPC) + human trader judgment**. RL is in production mainly for storage dispatch at a handful of sophisticated shops. Deep learning has clearly won the *forecast* layer; the *decision* layer remains dominated by optimization with ML inputs.
+
+> **One-sentence summary of the frontier:** DL/transformers have won the forecasting layer; RL is contesting the optimization layer but hasn't displaced stochastic optimization yet.
+
+Two nuances: the two moves are at **different maturity stages** — DL in forecasting is a completed takeover (and increasingly probabilistic), while RL in decisions is still a challenger vs. deployed stochastic optimization / MPC. And the **boundary between the layers is dissolving**: decision-focused learning trains forecasts directly on trading profit, collapsing forecast-then-optimize into one learned pipeline — arguably the most interesting current research direction (§4.4).
+
+### 2.7 Quick reference table
+
+| Layer | Deployed standard | Research frontier |
+|---|---|---|
+| Market clearing (ISO) | MILP SCUC + LP SCED, LMP | Convex hull pricing, stochastic UC, ML-accelerated optimization |
+| Price forecasting | Probabilistic deep nets, GBTs, tuned linear models | Transformers, hybrid+online learning, conformal prediction |
+| Bidding/dispatch decisions | Stochastic programming, MPC, bi-level MPEC | Deep RL (DDPG/PPO/SAC), hierarchical & multi-agent RL, no-regret online learning |
+| Forecast↔decision interface | Two-stage predict-then-optimize | Decision-focused / end-to-end learning |
+| Storage arbitrage | MPC + probabilistic forecasts | DFL, risk-aware hierarchical RL |
+| Virtual bidding / FTRs | GBT/DNN spread models + portfolio risk limits | Transformer spread forecasting, LLM-enhanced frameworks |
+
+---
+
+## 3. Where edge actually lives
+
+### 3.1 The hierarchy of durable edge
+
+Any bidding system decomposes into three functional layers: a **data and forecasting layer** that converts weather, grid, and market data into calibrated probability distributions; an **optimization layer** that converts distributions into offer curves and positions under risk constraints; and an **execution and risk layer** that submits bids, monitors positions, settles, and enforces discipline.
+
+The competitive value of these layers is not equal:
+
+1. **Forecast quality and calibration** dominate. Two participants running identical optimizers on different forecasts diverge widely in P&L, while two participants running different (competent) optimizers on identical forecasts land close together.
+2. **Execution and risk discipline** ranks second. In crowded strategies, the shop that re-optimizes every interval, never misses a submission window, and cuts losers per its own rules outperforms the one that doesn't, even with no analytical edge.
+3. **Optimizer formulation** ranks third. The underlying mathematics (quantile offering, two-stage stochastic programming, CVaR constraints, storage MILPs) is published and commoditized; what differs between shops is craft, not theory.
+
+Build implication: spend the best engineering on forecasting and backtesting, use boring proven math in the middle, and treat operational reliability as a feature.
+
+### 3.2 The insider-information myth (PJM / SCUC-SCED)
+
+A natural worry: wouldn't a vendor with inside knowledge of PJM's clearing algorithm — or of the bids flowing into it — crush all competitors? The concern dissolves under inspection.
+
+**What is NOT really secret.** The *algorithm*: SCUC/SCED are standard MIP/LP formulations, exhaustively documented (PJM Manual 11/12, the OATT, business practice manuals), solved with commercial CPLEX/Gurobi-class solvers. Competent quant teams build "shadow SCED" models from public documentation.
+
+**What IS secret — and legally fenced:**
+
+| Item | What it is | Why it's decisive |
+|---|---|---|
+| **Other participants' bids/offers** | Released only masked, ~4-month lag | DA prices are an emergent function of them |
+| **Network state** | Live state-estimator view: flows, voltages, topology, constraint headroom | Knowing which constraint binds next = seeing nodal price divergence before it happens |
+| **Derates** | Real-time reductions in unit capability (e.g., a 900 MW unit down to 600 MW) | Supply tighter than the market believes → position ahead of the spike; classic MNPI in FERC cases |
+| **Dispatch instructions** | 5-min SCED setpoints + out-of-market operator actions (reliability commitments, manual dispatch) | Reveals decisions invisible to outside models; reliability commitments distort local prices |
+
+Using these is a **federal crime**, not a vendor edge: FERC anti-manipulation rules (post-EPAct 2005) cover trading on material non-public information; PJM staff and systems vendors are bound by confidentiality and trading prohibitions; the Independent Market Monitor (Monitoring Analytics — external to PJM) and FERC's Office of Enforcement screen for anomalous profitability. There is precedent — enforcement cases against individuals trading on non-public grid information, and the Powhatan/UTC saga showed FERC pursues even aggressive rule exploitation.
+
+**Why perfect algorithm knowledge wouldn't get you far anyway.** SCED is deterministic *given its inputs*, but the inputs are unknowable in advance — even to PJM. DA prices depend on bids not yet submitted; RT prices depend on forced outages, weather errors, interchange, and binding constraints. Price is an emergent output of thousands of private decisions plus physical randomness.
+
+**Where real differentiation lives instead:** better weather ensembles and load/renewable models; **predicting which transmission constraints will bind** (the genuinely hard, high-value problem in nodal markets); modeling bidding behavior from lagged public data; reconstructing network topology (CEII model access, FTR results, historical shadow prices).
+
+**The honest residual:** EMS/market-software vendors and ex-ISO employees carry legal "soft" insider knowledge — operator behavior, solver quirks, out-of-market actions. It's valuable, common, and legal. The real moat: *people who know how the control room actually behaves at 6 PM on a July scarcity day.*
+
+### 3.3 Is accurate price prediction even the most valuable thing?
+
+**No — for most participants, point forecasts are an input, rarely the differentiator.**
+
+**Generators: optionality and risk, not prediction.** The right tail of the distribution matters far more than the mean — a peaker earns its year in a handful of scarcity hours; the question is 3 vs. 30 scarcity hours, not ±$2/MWh on the average. **Volumetric-price correlation risk** dominates hedging: being forced out exactly when prices spike (Winter Storm Uri: forward-sold generators with frozen plants buying back at $9,000/MWh). And the economics run on **spreads, not prices**: spark spread (gas), dark spread (coal), top-bottom spread (batteries).
+
+**LSEs / large buyers: load forecasting beats price forecasting.** Exposure is squared — a hot day means high load AND high price. Most valuable: your own load forecast, hedge-ratio policy, and shape-risk management. The ERCOT retailer failures (Griddy) were unhedged structural exposure, not bad price forecasts.
+
+**Financial traders: relative value, tails, speed.** The money is in **spreads and congestion** (DA–RT virtuals, UTCs, FTRs, hub-to-node basis) — the skill is predicting which constraints bind. **Calibration beats accuracy**: well-calibrated tails plus good sizing out-earn a sharper point forecast, and survive the blowups. **Speed** matters: reacting to unit trips, weather model runs, and interchange changes faster than the market reprices.
+
+**What arguably matters more than any forecast:**
+
+1. **Risk management and capital discipline** — the graveyard is full of risk-control failures, not forecast failures (the GreenHat FTR default, $180M+ socialized; the Uri casualties). *Approximately right and definitely solvent* wins.
+2. **Structural optionality** — batteries need a good response *policy*, not a forecast; market makers earn the spread; structured desks earn risk-transfer margin.
+3. **Understanding the plumbing** — settlement rules, uplift/make-whole, credit mechanics, FTR auction quirks, emergency operator discretion. Rule knowledge is durable; forecast edges decay.
+4. **Weather as the upstream input** — many firms' single most valuable proprietary asset is their weather-ensemble processing; price forecasts are largely derivative of weather + fuel + outages.
+
+> **Synthesis: price forecasting is necessary but low-moat; risk structuring and constraint/network insight are the actual differentiators.** The scarce capabilities are (a) calibrated tail distributions, (b) node-level congestion insight, (c) position-sizing discipline to survive being wrong, and (d) rule/plumbing expertise — hence firms pay up for ex-control-room operators, network engineers, and risk managers as eagerly as for forecasters. **The forecast is the ante; the other things are the game.**
+
+---
+
+## 4. How AI changes the picture
+
+The three-layer decomposition of §3.1 is the right lens for evaluating the current AI wave, because recent advances land on the layers very unevenly — and, conveniently, they land hardest on the layer that matters most.
+
+| Layer | Edge rank | AI impact | Maturity | Adoption posture |
+|---|---|---|---|---|
+| Forecasting | 1st | Step change (AI weather ensembles) | Operational now | Adopt aggressively; edge is decaying |
+| Execution / risk ops | 2nd | Large but unglamorous (LLM ops) | Production-ready with human review | Adopt quietly; compounding advantage |
+| Optimizer | 3rd | Incremental (decision-focused learning) | Research → early practice | Experiment; keep auditable structure |
+
+### 4.1 Layer 1 — Forecasting: the genuine step change
+
+**The AI weather model transition.** The most consequential AI development for power trading is not reinforcement learning or language models; it is the replacement (or augmentation) of physics-based numerical weather prediction (NWP) with learned atmospheric models. The milestone sequence is short but steep: Huawei's **Pangu-Weather** (Nature, July 2023) demonstrated forecasts roughly 10,000× faster than conventional ensemble systems in peer-reviewed tests; Google DeepMind's **GraphCast** (Science, December 2023) outperformed ECMWF's flagship HRES deterministic model on roughly 90% of 1,380 verification targets; ECMWF moved its AI-based **AIFS** model to operational status in 2024 — the first major meteorological agency to run a learned model operationally; Microsoft's **Aurora** foundation model (Nature, May 2025), pretrained on over a million hours of geophysical data, extended the approach across weather, air quality, and wave prediction at a fraction of traditional compute cost; and DeepMind's **GenCast** pushed the architecture family into probabilistic ensemble forecasting.
+
+**How learned atmospheric models work.** Neural networks are trained — mostly on the ERA5 reanalysis, ~45 years of hourly global atmospheric state — to learn the mapping "atmosphere at t → atmosphere at t+6h" directly, replacing numerical integration of physics equations with a single forward pass. Architectures span 3-D vision transformers (Pangu), icosahedral-mesh GNNs (GraphCast), graph/transformer hybrids (AIFS), diffusion models sampling plausible futures (GenCast — natively probabilistic), and pretrained geophysical foundation models (Aurora). Forecasts roll out autoregressively, which compounds errors and — under MSE training — produces progressive smoothing that underestimates extremes. Critical dependency: they still require physics-pipeline initial conditions; data assimilation is not replaced. Global resolution (~25–31 km) means site-level downscaling/calibration matters more, not less.
+
+**Why the economics matter more than the leaderboard.** Traditional NWP solves differential equations across 3-D grid cells on supercomputers, refreshing two to four times daily at thousands of dollars per cycle. Learned models run on GPUs in seconds to minutes, at accuracy parity in the 12–48h day-ahead window. Three consequences follow:
+
+- **Speed → probability.** When a forecast run costs seconds of GPU time, generating dozens or hundreds of ensemble members becomes trivial. The entire downstream bidding pipeline consumes probability distributions — production quantiles for the day-ahead offer, spread distributions for imbalance risk — so ensemble economics translate directly into better-calibrated inputs and therefore better offers. This is the cleanest causal path from "recent AI" to bidding P&L.
+- **Speed → freshness.** Commercial AI-weather providers now advertise up to 24 runs per day versus the 2–4 of traditional NWP. In a two-settlement market, fresher forecasts matter most between the day-ahead close and real-time delivery: real-time re-offering, intraday position adjustment, and imbalance management for wind and solar portfolios.
+- **Speed → iteration (and entry).** Cheap reforecasting powers backtests, and open model weights plus public initial-condition data put a near-state-of-the-art global forecast within reach of any shop with GPUs — infrastructure that previously required a national weather service or an expensive vendor contract.
+
+**Caveats and failure modes.** Three caveats keep this honest:
+
+1. **Extremes are the weak spot.** Models trained with mean-squared-error objectives produce smoothed fields that systematically underestimate sharp gradients — peak wind speeds in severe storms, for example. Power markets make and lose money precisely in the tails (scarcity events, ramps, icing), so a model that wins on average RMSE can still be the wrong tool for the hours that dominate annual P&L. Ensemble and generative approaches (GenCast-style) mitigate but do not eliminate this. Any adoption plan should include tail-specific verification against your own asset history, not just headline skill scores.
+2. **Vendor claims require independent verification.** The commercial AI-weather space is young and marketing-heavy. The practical test is a paid pilot scored against your incumbent provider on *your* nodes, *your* variables (hub-height wind, plane-of-array irradiance, temperature-driven load), and *your* loss function — ideally the downstream trading metric, not meteorological RMSE.
+3. **The edge decays.** These models are open or cheaply accessible, so the advantage from merely using them erodes as adoption spreads. Public AI raises the floor for everyone. What it cannot commoditize is what you combine it with: proprietary asset telemetry for local calibration, downscaling to your specific sites, and — above all — the translation from weather to *nodal price*, which runs through congestion.
+
+**Downstream of weather: prices, load, congestion.** Price forecasting has moved from classical time-series methods to gradient-boosted trees and increasingly to transformers and time-series foundation models; the practical gains are largest in probabilistic (quantile) forecasting of the DA/RT spread, the input that prices a renewable's imbalance risk. Load forecasting benefits from the same architectures plus improved temperature inputs. **Congestion and nodal-basis prediction remains the hardest and least commoditized problem** — learning the mapping from system conditions to binding transmission constraints — and, precisely because almost nobody does it well, it is where forecast-layer investment retains the longest-lived edge. Graph neural networks that encode grid topology are the active research direction; no vendor sells a turnkey solution worth having.
+
+*Layer 1 synthesis: adopt AI weather ensembles early and aggressively, verify tails independently, and reinvest the freed budget into the two things public models cannot provide — local calibration against your own telemetry, and nodal congestion modeling.*
+
+### 4.2 Layer 2 — Execution, operations, and risk: the quiet LLM win
+
+This is the layer vendor marketing ignores, which is exactly why it is underrated. It is unglamorous back-office work — but it attacks the second-ranked source of edge, carries low model risk because a human reviews the output, and its benefits compound.
+
+**Market-rule intelligence.** US ISO participation is governed by thousands of pages of tariffs, business practice manuals, and a continuous stream of market notices, protocol revisions, and FERC filings. Engineers building bidding systems consistently identify this — not the mathematics — as the dominant cost. It is a nearly ideal LLM workload: a retrieval corpus over the tariff and manual set for each ISO you trade; automated triage of daily market notices, flagging anything that touches your bid parameters, settlement formulas, or ancillary product definitions; and change-diffing of protocol revisions against the assumptions encoded in your optimizer. The failure mode to engineer against is hallucinated rule citations — mitigations are standard (retrieval-grounded answers, mandatory citation to source paragraphs, human sign-off on anything that changes system behavior).
+
+**Compliance documentation and the hedge/spec boundary.** For any shop running both physical assets and financial positions, FERC manipulation risk makes documentation a first-order concern: every physical bid should be independently defensible as profit-maximizing for the asset, on the record. LLMs are well suited to generating that record — drafting daily bid-rationale documentation from the optimizer's own inputs and outputs (forecast quantiles used, constraints binding, deviation from the neutral baseline and why) in consistent, auditable language. This converts a compliance burden that trading shops chronically under-resource into a largely automated byproduct of the bidding run, and materially strengthens the architectural separation between hedge and speculative books (§5.2).
+
+**Settlement, monitoring, and incident response.** Shadow settlement — independently recomputing what the ISO owes you and disputing discrepancies — is high-value, detail-heavy work mixing structured data with unstructured rule text: again a natural LLM-plus-tools workload (parsing settlement statements, reconciling against internal calculations, drafting dispute filings with rule citations). Overnight operations monitoring (telemetry anomalies, missed dispatch instructions, forecast-feed failures ahead of submission deadlines) similarly benefits from an agentic layer that triages, summarizes, and escalates rather than paging a human for everything.
+
+**Engineering velocity.** A second-order but real effect: agentic coding tools compress the build timeline of the entire system — ISO API integrations, backtesting harnesses, data pipelines — which disproportionately benefits small teams competing against incumbents with large engineering staffs. The state-of-the-art bidding shop of 2026 is not necessarily the one with the most exotic model; it is often the one whose five engineers ship like twenty.
+
+*Layer 2 synthesis: deploy LLMs as retrieval-grounded analysts and drafters across market rules, compliance, settlement, and monitoring, always with human review at the point of action. The advantage is quiet, defensible, and — because competitors under-invest in exactly these functions — durable.*
+
+### 4.3 Layer 3 — The optimizer: incremental by design, and that's fine
+
+The optimization mathematics was never the bottleneck. Newsvendor-style quantile offering for renewables, two-stage stochastic programs with CVaR, and MILP/dynamic-programming formulations for storage are published, taught, and tractable with commercial or open solvers. They are also **auditable**: every offer can be traced to a forecast input and a constraint, which matters enormously for both internal risk governance and regulatory defense. Any AI proposal for this layer must beat well-tuned classical methods *and* preserve explainability *and* bound its failure modes. Two candidate technologies are worth tracking, at very different maturity levels.
+
+### 4.4 Decision-focused learning: the credible upgrade
+
+The conventional pipeline trains forecasts to minimize statistical error (MSE, pinball loss) and then optimizes against them independently — without considering how forecast errors propagate into decision quality. **Decision-focused learning** (DFL, also called value-oriented forecasting or "smart predict-then-optimize") integrates the downstream optimization into the training loop, so the forecaster is trained against a regret-style loss measuring the sub-optimality of the *decisions* its forecasts induce. Founding results: Donti, Amos, and Kolter's task-based end-to-end learning (NeurIPS 2017) and Elmachtoub and Grigas's "Smart Predict-then-Optimize" (Management Science 2022); the energy literature has applied the framework to day-ahead scheduling against energy and reserve markets, storage arbitrage and predict-then-bid frameworks, and robust microgrid operation.
+
+The core insight is directly relevant to bidding: **the most accurate forecast is not necessarily the most valuable one.** A price forecast can improve its average error while getting worse in exactly the high-priced hours where offer decisions have consequences; the literature has repeatedly shown that a specific quantile choice, not the most accurate point forecast, maximizes trading value. DFL formalizes and automates that intuition. Critically — unlike end-to-end RL — it preserves the auditable optimizer: you change what the forecast is trained *for*, not who makes the decision.
+
+**Production-readiness: not broadly — late-stage research / early adoption.** What's mature: established frameworks (smart predict-then-optimize, learning-by-experience, black-box differentiable optimizers), and implicit differentiation through convex optimization layers demonstrated for economic dispatch and storage arbitrage, consistently beating accuracy-trained forecasts on realized profit for simple convex problems. Why it hasn't crossed into broad production:
+
+1. **Fragile differentiation through the optimizer** — no closed form for the backward pass; unrolling has accuracy/efficiency issues; analytical methods impose rigid problem-form requirements; mixed-integer/nonconvex bidding needs surrogates, subgradients, or perturbation estimators (brittle in practice).
+2. **Partial uncertainty coverage** — most DFL handles uncertainty in the objective only, assuming constraint parameters are known; real trading has uncertainty everywhere.
+3. **Task-specificity** — the model is welded to one decision problem; any change (asset size, market, bid format) forces retraining. Two-stage pipelines are modular: one forecast feeds many decisions.
+4. **Interpretability / risk control** — profit-trained forecasts are deliberately biased and hard to explain to risk committees; two-stage lets you audit forecast and decision separately.
+5. **Data scarcity and regime-shift fragility** — policies trained on historical profit can degrade badly exactly when markets shift.
+
+Recent evaluation work adds honest counter-evidence: DFL's benefits are application-dependent and do not always translate into higher economic value, while typically demanding substantially more computation. Adoption barriers are structural too — third-party forecast vendors cannot train against every client's private loss function, and cost-oriented forecast targets lack the intuitive interpretation (mean, median, quantile) that human reviewers rely on.
+
+**Rule of thumb:**
+
+- **Small convex inner problem + simple asset (one battery, one market):** DFL is production-viable today for a strong technical team.
+- **Integer commitments, multi-market co-optimization, ISO bid curves:** two-stage probabilistic forecasting + optimization remains the deployed standard, likely for several more years.
+- **Pragmatic middle ground** (what many shops actually do): "decision-aware" training — keep the two-stage architecture but weight forecast loss by economic consequence. Much of the benefit, little of the fragility.
+
+DFL is worth piloting *because you own your full pipeline* (an advantage of building over buying), evaluated against a strong quantile-forecast baseline on realized trading P&L, and expected to pay off most where the decision problem is asymmetric — storage bidding and imbalance-exposed renewable offers — rather than uniformly.
+
+### 4.5 Reinforcement learning: shadow mode, not production
+
+End-to-end RL — a learned policy directly emitting offer curves — remains overwhelmingly academic despite a large literature (comprehensive reviews cover 150+ papers). The structural obstacles have not moved: live trial-and-error is impossible at real-market cost, so policies train in simulators whose fidelity gap to the real market has never been closed; day-ahead auctions offer roughly 365 independent samples per year, a brutal sample-efficiency regime; and an unexplainable policy that misbehaves during a scarcity event is both a financial and a regulatory catastrophe.
+
+Where RL-adjacent methods *are* in production is instructive: **approximate/stochastic dynamic programming** — value-function methods over sequential decisions, mathematically RL's sibling — is standard inside commercial storage optimizers, and learned policies are credibly deployed in **European continuous intraday markets**, where thousands of order-placement decisions per day give RL the interaction density it needs.
+
+The sensible roadmap for a US system mirrors that pattern: classical stochastic optimization for the day-ahead auction; RL experiments confined to real-time re-offering and intraday adjustment, where decisions are frequent; mandatory shadow-mode evaluation against the production optimizer before any capital exposure; and, if RL earns its way in, deployment as a bounded residual correction on top of the optimizer's output rather than a replacement for it.
+
+### 4.6 Vendor "AI bidding" claims: substance vs. hype
+
+**Substance:** automated 24/7 bidding demonstrably works and is now *independently measurable* (Modo Energy leaderboards for storage optimizers); ML forecasting genuinely outperforms older methods; operational consistency alone beats human desks.
+
+**Hype patterns:** "AI" frequently means a standard GBM-forecast + LP-optimizer dressed up; uplift claims lack stated counterfactuals; and the edge from mere automation compresses as adoption spreads — ERCOT ancillary-services saturation being the clear example.
+
+**Due-diligence questions for any vendor:** Where do you rank on an independent benchmark? What exactly is the model architecture, and what counterfactual sits behind your uplift numbers? How did the system behave during Winter Storm Uri / the last scarcity event? What is the fee structure (revenue-share vs. fixed)?
+
+### 4.7 Cross-cutting dynamics
+
+**The floor rises; the ceiling is private.** Nearly every AI advance discussed here is public or purchasable: open weather-model weights, foundation time-series models, commodity LLMs, published DFL methods. Their diffusion compresses the edge available from any single technology — visible already in ERCOT battery markets, where algorithmic saturation of ancillary services shifted revenue toward energy arbitrage and shrank the gap between top and median operators. Durable differentiation migrates to what cannot be bought: proprietary telemetry and its use in local calibration, nodal congestion modeling, execution reliability, and the institutional discipline of a well-run risk process.
+
+**Auditability is a feature, not a constraint.** The regulatory environment (FERC manipulation doctrine, ISO market-monitor scrutiny) and internal risk governance both reward architectures whose decisions can be explained after the fact. This is a genuine, often-overlooked argument for forecast-then-optimize over end-to-end learned alternatives — and it is why the highest-value applications of the newest AI (weather ensembles, LLM operations, DFL) are precisely the ones that *strengthen* the classical architecture rather than replacing it.
+
+**Independent benchmarking is changing vendor dynamics.** Third-party leaderboards for storage optimizers have replaced self-defined vendor benchmarks with a shared reference point, and the same discipline should be applied internally: every AI adoption in the stack deserves a counterfactual (what would the incumbent method have earned?) and a tail-event stress test (how would this component have behaved during Winter Storm Uri or the most recent scarcity event?).
+
+### 4.8 Adoption roadmap
+
+Sequenced by expected risk-adjusted return on effort:
+
+1. **Now — AI weather ensembles (Layer 1).** Pilot one or more AI-weather feeds against the incumbent, scored on your assets and your trading loss function with explicit tail verification. Low integration risk, direct P&L path, and a decaying edge that rewards early movers.
+2. **Now — LLM operations layer (Layer 2).** Retrieval-grounded market-rule assistant, automated bid-rationale documentation, shadow-settlement triage, and agentic engineering tooling. Human review at every point of action.
+3. **Next two quarters — local calibration and congestion modeling (Layer 1).** Reinvest weather-layer savings into asset-specific downscaling and nodal congestion prediction; this is where forecast edge survives commoditization.
+4. **Next two quarters — DFL pilot (Layer 3).** Retrain the production-forecast loss against realized bidding regret for one asset class (storage or a single wind portfolio), evaluated against the quantile baseline on trading P&L, not forecast error.
+5. **Opportunistic — RL in shadow mode (Layer 3).** Real-time re-offering only; promotion to production contingent on sustained shadow-mode outperformance and bounded-action deployment.
+
+The through-line: recent AI does not overturn the architecture of a well-built bidding system. It makes the boring architecture better — sharper distributions in, cheaper discipline around, and a modestly smarter objective inside — which, given where the secrets actually live, is exactly the outcome a builder should want.
+
+---
+
+## 5. Building a bidding system
+
+### 5.1 Architecture for a renewables + wholesaler bidding system
+
+Five-stage pipeline: **data ingestion → probabilistic forecasting → bid optimization → submission → settlement/monitoring**, with a feedback loop retraining models on settlement outcomes.
+
+Key design decisions:
+
+- **Forecasting:** quantile forecasts (P10–P90) for production and load, built on NWP ensemble inputs blended with asset telemetry; DA LMP, RT LMP, and crucially the DA–RT *spread* distribution.
+- **Optimization core:** for a price-taking renewable, the optimal DA offer is a newsvendor quantile of the production distribution set by the expected DA/RT price ratio, wrapped in a two-stage stochastic program with CVaR. Tractable LP per asset per day (Gurobi/CPLEX/HiGHS). The wholesaler side is the mirror image: DA-vs-RT load procurement plus a hedging overlay (forwards, FTRs).
+- **Operations:** automated ISO submission with rule validation, shadow settlement, and a serious backtesting harness — the most commonly under-built component.
+- **Gotchas that dominate real P&L:** negative prices and PTC/REC-driven offer floors; nodal basis and congestion (often the highest-ROI forecasting work); PPA/hedge structures reshaping incentives; per-ISO rule differences.
+
+Note for hybrid assets: storage co-location changes the optimizer from an LP to a sequential problem (see §2.3).
+
+### 5.2 Separating hedging from speculation
+
+Separation is necessary for three reasons:
+
+1. **Performance attribution** — the two books have opposite definitions of success.
+2. **Risk governance** — bounded physical exposure vs. leveraged financial loss distributions (GreenHat as the cautionary tale).
+3. **FERC manipulation risk** — physical bids must be independently defensible; joint optimization of physical bids and financial positions is a manipulation-allegation generator.
+
+But complete separation is impossible: every DA offer quantile *is* a spread view. Resolution: define a **neutral baseline bid** (e.g., P50 offer / 100% DA load), log every deviation as "embedded alpha," and decompose P&L into baseline + embedded view + pure spec book. Architecturally: share data, forecasting, and backtesting; split optimizers, mandates, limits, and P&L; keep coupling one-directional — the physical optimizer never reads the financial book's positions.
+
+The two books consume the same forecast stack differently: the hedge book needs calibrated mid-distribution quantiles; the spec book needs tails and spread skew.
+
+Strategic note: virtuals/FTRs are crowded quant markets; embedded-view alpha inside a well-run asset book is often the better initial return on modeling effort.
+
+---
+
+## 6. China: market structure and deployment
+
+### 6.1 Market structure: different design, weaker safeguards
+
+**The market operator is NOT independent — the fundamental difference from the US.** Provincial spot markets are run by dispatch centers and trading centers *inside the grid companies* (State Grid, China Southern Grid). Trading centers are nominally independent via shareholding reform, but dispatch remains within grid organizations. Grid companies historically had commercial interests in outcomes — the "insider" is structural, not hypothetical. Dispatch/trading separation has been a reform demand since Document No. 9 (2015).
+
+**Transparency: improving but well below PJM.** A "1+6" national rule framework has been built since 2023: Basic Rules for Market Operation + spot market (2023), information disclosure (2024), medium/long-term trading, registration, ancillary services (2025), metering/settlement (2025). Provinces (Shanxi, Guangdong, Shandong) publish clearing methodology; several use centralized SCUC/SCED-style clearing, some nodal/zonal. Gaps vs. PJM: no bid-level data (even masked/lagged), little constraint-level detail, clearing-model internals undisclosed — shadow-SCED reconstruction is far harder. Rules are revised frequently by administrative notice and differ substantially by province; price caps/floors are tight, and administrative interventions are more common and less documented.
+
+**Laws and enforcement: embryonic.** No independent market monitor analogous to Monitoring Analytics — surveillance is done by trading centers/dispatch (the grid companies) plus thinly-staffed NEA bureaus. No developed body of power-market manipulation case law; general laws are untested on spot-power conduct. SOE-dominated participation and administrative dispute resolution mean little deterrence-by-precedent.
+
+**Net assessment.** All three PJM safeguards (§3.2) are weaker in China: partial algorithm disclosure, structural data asymmetry favoring grid-affiliated entities, nascent enforcement. The realistic information edge comes from **proximity to grid/dispatch institutions**, not rogue vendors. Mitigants: heavy medium/long-term contract coverage (~47% of consumption in 2024) plus price caps mean less money at stake in spot per unit of information advantage, and the reform direction clearly points toward disclosure and oversight. **Watch:** whether dispatch is ever truly separated from the grid companies. National unified market targeted ~2029; seven provinces in full spot operation as of 2025 (Shanxi, Guangdong, Shandong, Gansu, W. Inner Mongolia, Hubei, Zhejiang).
+
+### 6.2 Weather data sourcing for a China deployment
+
+Baseline reviewed stack — ECMWF HRES+EPS backbone, GFS/GEFS secondary, CMA regional models + observations, Himawari/Fengyun satellite irradiance, customer met masts, buy-first on WRF downscaling — is conventionally sound. Recommended changes:
+
+1. **Add AI weather models — the glaring omission.** AIFS open data plus open-weight models (GraphCast/Pangu) densify the ensemble at near-zero marginal cost; China's domestic AI-weather ecosystem (Pangu, Fuxi, FengWu, CMA-integrated) is both cheap and a compliance asset; GPU-cheap models unlock the intraday refresh cadence the licensed stack can't provide.
+2. **Add ERA5/reanalysis on day one** — the training substrate for calibration models and backtesting.
+3. **Make the blending/calibration layer explicit** with its own budget line — multi-model ensembles calibrated per-site against ground truth is where the money is; source diversity is calibration fuel.
+4. **Specify a fallback hierarchy** (EPS → AIFS open → GFS → persistence) as an engineering requirement.
+5. **Two-way compliance review early:** ECMWF redistribution terms × Chinese regulations on foreign-data commercial use and domestic-observation export — this constrains where the blending layer can physically run.
+6. **Add inverter/SCADA data as ground truth** (free, dense, bakes in real plant behavior), plus a QC pipeline for customer sensors.
+7. **Reframe "self-run WRF" as "self-run downscaling," method open** — learned downscaling will likely beat WRF on cost by decision time.
+8. **Design the satellite-nowcast → NWP blending handover** in the 0–4h window explicitly, to avoid a discontinuity in the intraday trading horizon.
+
+---
+
+## 7. Open questions
+
+Carried forward from the discussions:
+
+- **Which market(s)?** US ISOs vs. Chinese provincial spot pilots — gate-closure times and settlement rules change the forecast-refresh requirements.
+- **Asset mix?** Solar/wind/hybrid-with-storage — storage co-location changes the optimizer from an LP to a sequential problem.
+- **Risk mandate?** Appetite for a standalone financial (virtuals/FTR) book vs. embedded-alpha-only inside the asset book.
+
+---
+
+## 8. Sources and further reading
+
+- GraphCast: *Learning skillful medium-range global weather forecasting*, Science (Dec 2023). https://www.science.org/doi/10.1126/science.adi2336
+- Pangu-Weather: *Accurate medium-range global weather forecasting with 3D neural networks*, Nature (Jul 2023). https://www.nature.com/articles/s41586-023-06185-3
+- Aurora: *A foundation model for the Earth system*, Nature (May 2025). https://www.nature.com/articles/s41586-025-09005-y
+- ECMWF AIFS operational status (2024). https://www.ecmwf.int/en/about/media-centre/news/2024/ecmwfs-ai-forecasts-become-operational
+- Donti, Amos, Kolter: *Task-based end-to-end model learning in stochastic optimization*, NeurIPS 2017. https://arxiv.org/abs/1703.04529
+- Elmachtoub & Grigas: *Smart "Predict, then Optimize"*, Management Science (2022). https://arxiv.org/abs/1710.08005
+- Decision-focused learning for energy/reserve market participation. https://www.researchgate.net/publication/383107857
+- Decision-focused predict-then-bid for strategic storage. https://arxiv.org/abs/2505.01551
+- Statistical vs. economic evaluation of electricity-market forecasts (DFL counter-evidence). https://arxiv.org/abs/2511.13616
+- RL in deregulated energy markets: comprehensive review (150+ papers). https://www.sciencedirect.com/science/article/abs/pii/S0306261922014696
+- On-line RL for real-life energy trading (simulator-gap discussion). https://arxiv.org/abs/2303.16266
+- Modo Energy: independent benchmarking of storage optimizers. https://modoenergy.com
+
+*Prepared July 2026. Vendor performance claims cited in the underlying research should be independently verified against your own assets before procurement decisions.*