Balancing energy: AI-supported price forecast increases revenues by up to 37 percent

Balancing energy helps support the power grid when imbalances or bottlenecks occur. This occurs on the balancing energy market, where short-term increases or reductions in demand for electricity are traded and called upon as needed – so those who can flexibly adjust their electricity needs can also earn money.

Flexible industrial electricity demand could reduce peak load by up to 15 percent

The price forecast is crucial for the revenue generated. "Control power is traded on a so-called pay-as-bid market," explains Alexander Sauer, Director of the Fraunhofer Institute for Manufacturing Engineering and Automation. "There, a bidding process applies in which each provider is paid the price they submitted their bid." Therefore, anyone who significantly underbids the actual electricity price forgoes money. Anyone who bids higher than the price misses out completely.

AI instead of static bids

Scientists at Fraunhofer IPA have now succeeded in developing a forecasting method that, with the help of artificial intelligence, enables significantly more accurate predictions, thus increasing revenue by up to 37 percent, according to a press release from the institute. By using various machine learning methods, the scientists at Fraunhofer IPA have now succeeded in forecasting this price more accurately.

Want to stay up to date on the energy transition? Subscribe to the free newsletter from ERNEUERBARE ENERGIEN – Germany's largest non-affiliated magazine for renewable energies!

For example, many industrial companies participating in the balancing energy market have so far relied on simple, static bidding strategies, it continues. They set their bid once and then stick to it. Or they base their bid on the previous day's or week's price.

Malte Schön from Tesvolt: “The storage system must be controllable by a flexibility marketer.”

Four submarkets for balancing energy examined

In a second step, they supplemented their AI-supported price forecast with a specially developed offsetting process. "This is, in a sense, the post-processing of the forecasted electricity price, so that the submitted bid is slightly lower," explains Vincent Bezold from the Data-Driven Energy System Optimization research team at Fraunhofer IPA. "This has to do with the rules of the pay-as-bid market." It is most worthwhile to deliberately undercut the actual electricity price. "That's exactly what we achieve with our offsetting process."

In a paper, the scientists examined four submarkets of German balancing energy and showed that even a forecast error of one euro per megawatt hour can – depending on the market – generate up to 3,631 euros in additional annual revenue per megawatt. To further improve this result, the researchers at Fraunhofer IPA plan to apply even more complex AI models in the future, taking external factors such as weather data and probability-based forecasts into account to further improve forecast quality.