How Batteries Bid in ERCOT: A Field Guide to Timeline, Dispatch, and Settlement

Every battery selling into ERCOT lives inside a specific machine: a voluntary day-ahead auction, a five-minute real-time dispatch engine, and a settlement rule that nets one against the other. This guide maps the machine as it stands after the RTC+B redesign, with an emphasis on what a storage operator decides when, what is knowable at each point, and where the catastrophic risks live.

Two markets, one machine

ERCOT runs a two-part market structure. The Day-Ahead Market (DAM) is a voluntary, financially binding forward market: participants lock in positions the day before delivery, with energy co-optimized against ancillary services. The real-time market then re-trues everything: SCED (Security-Constrained Economic Dispatch) solves roughly every five minutes during the operating day, dispatching resources and setting locational marginal prices at every node.

Since RTC+B (Real-Time Co-optimization plus Batteries, live December 5, 2025, the largest market change in fifteen years), SCED also co-optimizes ancillary services with energy in every five-minute interval, producing real-time AS awards and clearing prices alongside energy prices. Batteries are modeled as single unified Energy Storage Resources with one bid and offer structure, and their telemetered state of charge (SOC) is an explicit constraint the dispatch engine respects.

Battery specifications, since they anchor everything

A "100 MW / 200 MWh (2-hour)" battery has a power rating of 100 MW (the maximum instantaneous charge or discharge rate: the width of the pipe), an energy rating of 200 MWh (the storable amount: the size of the tank), and a duration of 2 hours (their ratio: full-power runtime from full to empty). The battery may run slower for longer, but never faster than its power rating or longer than its energy allows. Duration determines which benchmark applies (a 2-hour battery is scored against the top-bottom 2-hour spread), power determines ancillary capability, and real systems carry SOC buffers, roughly 85 to 90 percent round-trip efficiency, and degradation that erodes usable capacity over life.

The timeline: what you decide when

The morning before delivery. The observable information set: ERCOT's load forecast for the operating day, wind and solar production forecasts, the outage scheduler showing planned generator and transmission outages, the ancillary service requirements and the Ancillary Service Demand Curves that both DAM and SCED will use (so the price of scarcity in each reserve product is known before bidding), morning weather model runs, day-ahead natural gas prices (gas sets the thermal fleet's marginal cost and hence the baseline power price), and full price history through yesterday. Not observable: tomorrow's actual weather, actual renewable output, forced outages that have not happened, and competitors' bids.

10:00 AM: DAM submissions close. The deepest-uncertainty, largest-position decision of the cycle, made 14 to 38 hours before delivery. Through its Qualified Scheduling Entity (QSE), a battery may simultaneously submit energy offers to sell in chosen hours (each a monotone curve of price-quantity pairs), energy bids to buy in cheap hours, and ancillary service offers per product. The day-ahead question is a portfolio decision under uncertainty: how much of tomorrow's expected value to lock at day-ahead prices versus leave open for real time. Selling the evening peak day-ahead hedges against a real-time crash but caps the upside of a real-time spike.

Around 1:30 PM: DAM results publish. Awards are known: which hours carry committed discharge, charge, or AS responsibility, at what clearing prices. From here forward the participant maintains a Current Operating Plan (COP), an hour-by-hour declaration of expected status, limits, and AS responsibilities used in ERCOT's reliability studies.

The adjustment period. After DAM results, offers for hours not already committed may be submitted or revised up until shortly before each operating hour (exact cutoffs live in the current protocols). This is a genuine rolling decision loop, consuming the second information wave: updated weather runs, evolving load and renewable forecasts, overnight forced outages, and the real-time price signal itself revealing how tight the day actually is.

The operating day: nobody bids. During delivery there are no interval-by-interval decisions. SCED dispatches each battery against its standing offer curves, subject to transmission constraints and telemetered SOC, issuing a base point (a MW instruction) seconds after each solve, 288 times a day. The real-time decision is entirely embedded in the shape of the curves on file: offer discharge steeply and SCED calls only in scarcity; offer flat and dispatch is frequent. Optimizer craft is exactly this: encoding price beliefs, SOC trajectory preferences, and opportunity costs into piecewise-linear curves that a mechanical optimizer interacts with all day.

Summarized as a cascade: the 10 AM commitment (deepest uncertainty, biggest positions: a forecasting problem), the rolling curve revisions (uncertainty partially resolved: a control problem), and the five-minute dispatch (no decision at all: pure consequence of curves already filed). Skill concentrates in the first two layers.

When information arrives: awards versus dispatch

An asymmetry runs through the market's epistemics. In the DAM, positions are known before delivery: submissions at 10 AM, awards around 1:30 PM, delivery the next day. In real time, everything is learned only at the moment of delivery: there is no advance notice of dispatch. The base point is the dispatch is the award, with instruction and obligation collapsed into the same five-minute instant. Hour-level dispatch accumulates as twelve successive base points and is fully known only when the hour ends.

RTC+B added real-time AS awards without changing the epistemics: each SCED solve tells a battery what reserves it is carrying at what clearing price, but these behave like dispatch, not like DAM awards. Zero advance notice, revealed interval by interval, changeable at the next solve.

Participants are not blind before the hour, though. Since dispatch is mechanical, one's own curve plus a forecast of the price distribution is a probabilistic dispatch forecast, and sophisticated operators run this expectation continuously because it feeds the thing that actually needs steering in real time: the SOC trajectory. If early intervals dispatch harder than expected, the energy remaining for later hours shrinks, and later hours' curves may need immediate revision.

Financial confirmation lags far behind the physics: settlement statements arrive days later with resettlements over weeks and months, but operators compute near-real-time P&L from base points, telemetered output, and published prices. Statements are reconciliation, not news.

Two-settlement: how money actually flows

The netting rule that makes every strategy question tractable: DAM awards settle at day-ahead prices; real time settles only the deviation between physical delivery and the day-ahead position, at the real-time price of each interval.

The offer curve is never what gets settled. Curves gate whether dispatch happens (they are reservation prices); the market price is what gets paid. A resource offering at $150 when the price clears at $200 runs and receives $200.

Three canonical cases for a battery facing hour 19:

• No DAM position. Capacity left open, dispatched 100 MW at a $200 real-time price: settlement is 100 MWh at $200, all real time.
• DAM position, delivery matches. Sold 100 MW day-ahead at $120, discharged exactly 100 MW: the DAM leg pays $12,000, the deviation is zero, and the realized price is $120/MWh regardless of whether real time cleared at $50 or $2,000. This is what "hedged" means.
• DAM position, delivery diverges. Sold 100 MW day-ahead at $120, dispatched for only 60 MW while real time cleared at $200: $12,000 from the DAM leg minus 40 MWh bought back at $200, netting $4,000, worse than delivering. Had real time crashed to $20, the buyback would cost $800 and the undischarged energy would still be in the tank: deviation as a profitable trade.

The strategic content: the day-ahead decision is a forward sale against one's own real-time forecast. Sell day-ahead when day-ahead looks rich relative to expected real time; stay open when real time should out-pay.

Awards are honored financially, never physically

A DAM award is unconditionally honored as a trade: settlement occurs regardless of what happens physically, and nothing in real time can un-award it. It is not a dispatch schedule, and leaving curves untouched does not guarantee delivery will match the award. Physical dispatch is always determined by SCED against whatever curve is standing. If real-time prices come in below the standing offer, SCED simply does not run the resource, producing a deviation bought back at the (low) real-time price, typically benign or profitable. The system never forces delivery; it prices non-delivery, asking every five minutes: still want to deliver, or buy it back at this price?

Three obligations do attach. DAM commitments must be reflected in the COP, and a resource that takes awards then shows itself unavailable draws scrutiny. Deviation exposure into scarcity is the real financial risk (next section). And ancillary awards are stricter: an AS award creates a supply responsibility, with SOC to back it (verified via telemetry post-RTC+B), and failure to provide is a compliance matter with ERCOT, not merely a financial buyback. AS awards are the closest thing in the design to a physically binding commitment; energy awards are finance wearing a physical costume.

The dangerous quadrant: short into a spike

The scenario that ends careers: holding a day-ahead sale, failing to deliver, while real-time prices exceed the sale price. The loss on the shortfall is the price gap times the undelivered volume, unbounded up to the offer cap. 100 MWh short into a $3,000 hour against a $120 sale is a $288,000 hit from one mistake in one hour.

Because SCED is a price-merit machine with no discretion, "prices were high and I was not dispatched" can only occur through four failure modes:

• Priced out (self-inflicted). The curve said $600; the market cleared at $500. SCED honored the stated reservation price. The high offer curve that brilliantly captures spikes on open capacity becomes dangerous against a committed sale: the award becomes a forced short into a rising market. The DAM hedge and the real-time curve strategy cannot be optimized separately.
• Physically unable (the empty tank). An outage, or the storage-specific killer: depleted SOC. Energy burned in hours 17 and 18 cannot serve a committed sale in hour 19; SCED sees the telemetered SOC and cannot dispatch energy that does not exist, leaving the full buyback at spike prices. Winter Storm Uri's sustained $9,000/MWh hours are the canonical lesson in what the sign of a deviation means during scarcity.
• Congestion split the node from the market (geography). Dispatch and settlement both occur at the resource node, and congestion can pin a local price far below the hub everyone watches. Hub at $500, node at $40: SCED correctly does not run the resource, and, the redeeming symmetry, the day-ahead buyback also settles at the node's cheap price if the sale was made at the node. Selling day-ahead at a different settlement point than the resource node introduces basis risk, a real and underappreciated loss channel.
• Plumbing (telemetry and limits). Understated telemetered limits, QSE system errors, or ramp-rate caps that prevent reaching full power fast enough to catch a short spike. Prosaic, and why boring telemetry hygiene sits inside the competitive moat.

The risk doctrine

Four rules fall out, and catastrophic battery P&L stories are, almost without exception, violations of one of them:

• Cover the SOC backing day-ahead sales first; that energy is spoken for and must not be spent chasing earlier opportunities.
• Price committed hours' curves low: the revenue is already locked, and delivery merely closes the position.
• Reserve aggressive high-priced curves for uncommitted capacity.
• Monitor SOC-at-risk against day-ahead obligations as a hard limit: the battery world's margin check.

One caveat on everything above: segment counts, price caps and floors, and submission cutoffs are protocol parameters that change. Verify against the current ERCOT Nodal Protocols before treating any specific number as current.