In a year of storage market flux, many Texas battery owners were either too risky or too conservative in how they operated their systems. And they missed out on potential revenue, and potential up time, because of it.
- The top line: Using artificial intelligence-based forecasting could have led to a 32% increase in revenue across ERCOT’s storage fleet in 2023, according to a report released this week from power market platform Gridmatic. As both storage markets and public interest in AI have evolved, optimization platforms — like those from Gridmatic itself, Fluence, and Stem — have increasingly been cited as an example of how automation can benefit the grid.
- The market grounding: 2023 was a year full of uncertainty for ERCOT storage owners, involving shifting state of charge rules, peak demand growth of 7%, and the hottest summer on record in Texas. All of those elements, according to Gridmatic, led operators to manage their batteries more conservatively — though some had the opposite problem, operating too aggressively and being dinged for “failure to provide.”
- The current take: The concept of risk management in a battery fleet isn’t something that most deployments on the market today account for, Gridmatic’s vice president of business development David Miller told Latitude Media. “What we’re seeing is these batteries don’t look like they’re being quantitatively managed — they’re being managed more manually,” he said. “And that’s something that we expect to change over time as it becomes clear what the penalties for taking risks are, and the need for a more quantitative risk management framework is more evident for battery operators.”
This is Gridmatic’s third annual report, based on public data from the ERCOT Day-Ahead Market and SCED Disclosure reports. Gridmatic, for its part, offers AI-backed tools for predicting energy supply, demand and pricing, and targets battery storage resources owned by third parties.
Traditionally, price forecasting happens via power flow modeling, which relies on physics-based approximations. But Miller said the company’s approach is to pour lots of data, from weather data to grid data, into its machine learning models, then use those to forecast prices. That system “now works better than the traditional ways of doing it,” Miller said, a conclusion they validated by trading in the markets directly.
The data Gridmatic collected from ERCOT showed that, on the whole, the total revenue per storage system increased by more than 50% over the course of the last year.
However, some operational strategies resulted in significantly higher monthly revenue: batteries with the highest revenue throughout the year were bidding large quantities of ERCOT Contingency Reserve Services — a new ancillary service product released in June of last year. Operators that hit those top tiers of revenue did so either by forecasting state of charge demands, or by bidding aggressively on that new service.
A lower-revenue battery managed with a conservative operational strategy that bid only into the real-time market and essentially eliminated the risk of “failure to provide” incidents could have doubled monthly revenue, Gridmatic found, by incorporating a diversified mix of ancillary services.
But, Miller said, the current market framework seems not to be sufficiently incentivizing that ancillary revenue stream.
“The concept of risk in managing a battery is really not well accounted for among the batteries in ERCOT,” Miller said. “There is this risk-return trade-off, where you can get in trouble for not managing your battery…and you can make quite a lot of money by taking very little risk. And so it’s a framework that’s not really being used effectively.”
