Akshi Energy · Live for Texas (ERCOT)
Tomorrow's load, prices and battery revenue, with the record to check them.
Akshi Energy forecasts ERCOT load by zone from the weather, builds net load and headroom from wind, solar, outages and batteries, and prices the day-ahead and real-time markets from them. Every forecast is issued before the day-ahead deadline and scored against what cleared. Launched for ERCOT; other markets are next.
Load
Four zones, each on its own weather station; hourly shapes from analog days; one system total.
Prices
A supply-stack model: price follows headroom above a gas-driven level. Ranges, spike odds, negative-price odds.
Record
Every forecast kept as issued and scored in public: load against ERCOT, prices against the clear.
What is in it
Akshi ERCOT: from weather to price, in one place.
Load forecast
Next-day system and zonal load from the weather forecast, beside ERCOT’s own forecast and, after the day, the actual.
Net load and headroom
Load minus wind and solar; thermal capacity minus outages, plus expected battery discharge. The quantity that moves price.
Next-day prices
Day-ahead and real-time by hub with P10–P90 ranges, the odds of a spike and of a negative hour, and the scarcity curve the model learned.
Scenarios
Replay a past heat dome or winter storm onto future dates with your load growth, wind, solar and battery build-out, outage and gas assumptions.
Battery
Forecast-driven backtests for 1, 2 and 4-hour batteries: what a plan made before the deadline earned against perfect foresight and the TB4 spread.
Basis
Hub and load-zone spreads by shape and season, worst days, and the all-pairs matrix across the 13 settlement points.
The forecast record
Scored every day, in public.
At 09:30 Central each morning Akshi Energy issues tomorrow’s forecast: hourly load by zone, prices with ranges for each hub, and a battery dispatch plan. The issue is timestamped against the 10:00 day-ahead deadline. Once the day has cleared, it is scored: load against ERCOT’s own forecast and the actual, prices against what cleared and against a “same as yesterday” baseline, the battery plan against perfect foresight.
Nothing is recomputed after the fact. A forecast made after the deadline is marked late. Walk-forward backtests use only what was known on the day, including the weather forecast that existed then. The scorecard and every day’s forecast are on the site.
Model limits are stated where they apply. Hours tighter than any in the calibration data, and hub-specific congestion, are the current ones.
Who it is for
Three desks, one model.
Power traders
The morning sheet for tomorrow: prices, spike odds and headroom by hour; basis by shape; a record of what the desk was told and what happened.
Load schedulers
Hourly load by zone, the day-ahead versus real-time decision by hour, and the cost of forecast error.
Battery operators
A dispatch plan from the forecast, revenue against the ceiling, and the forecast-driven backtest for the duration you own.
Markets
Live: Akshi ERCOT (Texas). Other North American power markets will follow as their own modules; the models, the record and the workflow carry over. Tell us which market you trade.
Coming next: a strategy tester (sell day-ahead blocks against real-time, with the forecast as the signal), nodal basis on a watchlist, a seven-day load view and ancillary-service revenue in the battery model.
See tomorrow's forecast now.
The app is open. Other markets, a private deployment or your own data in the models: talk to us.