New research from AAU Energy highlights advancements in wind power forecasting through artificial intelligence. Conducted in collaboration with Hong Kong Polytechnic University, China Agricultural University, and the École Polytechnique Fédérale de Lausanne, the study demonstrates how combining different AI tools can improve both short-term and long-term forecasting accuracy. This enables wind farm operators and grid authorities to make more precise predictions, ultimately leading to enhanced energy management and efficiency.

According to Birgitte Bak-Jensen, professor in intelligent control of electrical distribution systems at AAU, it is important to use both long- and short-term prediction methods.  “We need to have as good a forecast as possible, so we have checked different forecast methods to find one which is predicting the wind power production as well as possible. This is rather difficult, since the wind is changing all the time, both from day to day and from minute to minute,” explains Bak-Jensen.

The study evaluates four explainable AI techniques, each with unique benefits for wind power forecasting. One such technique, Shapley Additive Explanations (SHAP), determines the importance of individual factors in data analysis, while Partial Dependence Plots illustrate how variables influence predictions.

According to Bak-Jensen, improved forecasting will not only help optimize energy consumption but also contribute to better sizing of energy storage solutions, ultimately leading to reduced overestimation of wind power production. This will benefit both energy providers and consumers: “In the future, consumers will be more active, trying to use energy at the lowest cost, and with good forecasts, they will be better able to predict when prices are low,” explains Bak-Jensen.