Industry-adaptable explainable AI based methodology for forecasting electricity prices

The COVID-19 pandemic and the volatility of the energy market triggered by Russia’s invasion of Ukraine have highlighted the critical importance of the reliability and transparency of electricity price forecasting. The use of artificial intelligence models based on explainable AI has become essential for market participants to develop more efficient and informed strategies by making predictions more meaningful. Within the framework of this study, we developed artificial intelligence models based on open-source data and models for forecasting the day-ahead electricity prices in 19 European countries and analyzed the importance of different features in the models for the forecasting using SHAP values. Our results showed that for both 2015–2020 and 2020–2024 periods, the tree-based machine learning models performed best in price forecasting. By analyzing our models using SHAP, we show how much feature importance has changed from 2020 to 2024, demonstrating the increased complexity of electricity price forecasting due to the energy crisis and structural changes in the electricity system. Using Germany and France as case studies, we present detailed results for the two distinct electricity markets across both modelled periods.