Schedule batteries at three California trading hubs on real-time price forecasts

To schedule a battery for the next day, an operator must forecast the price at the battery location. CAISO clears a separate price at each trading hub, so each location needs its own forecast.

At ZP26, prices average USD 8/MWh from 09:00 to 15:00 and USD 45/MWh during the evening peak. Prices fall below zero in 14% of hours at SP15 and ZP26, and in 5% of hours at NP15.

To compare forecasts on equal terms, the backtest schedules one 100 MW / 400 MWh battery per trading hub on each forecast. The battery has 85% round-trip efficiency, one cycle per day, and a degradation cost of USD 10/MWh.

The schedule is fixed at 10:00 Pacific on the day before delivery and settles at the local price. Real-time prices are 15-minute prices averaged to hours, and the naive schedule repeats the latest complete day.

eomer adds Site Agent, which forecasts the output of the 100 largest solar parks and the 100 largest wind farms in CAISO.

Day-ahead prices follow a regular daily shape, so the naive schedule captures 86% to 94% of perfect-foresight value and eomer captures 91% to 96%. In real time, eomer captures 68% to 83%, compared with 46% to 74% for the naive schedule.

At NP15, eomer earns 49% more battery revenue than the naive schedule. Site Agent lowers the real-time mean absolute error by 7% across the three hubs. This experiment does not measure the additional battery revenue from Site Agent.