Probabilistic electricity price forecasting for profit maximization during energy crisis periods

Abstract

ABSTRACT In Europe, electricity prices and volatility have risen considerably following the COVID-19 pandemic and energy crisis, increasing the importance of probabilistic forecasting and risk management. This paper proposes a profit-maximization strategy using predictions from LightGBM quantile regression (LightGBM-QR), whose forecasting performance is evaluated using the Winkler score. The analysis covers the crisis and COVID-19 periods and includes updated data through 2024. We simulate an electricity supplier that diversifies trades across hours in the day-ahead market and integrate the forecasts into portfolio optimization strategies for risk minimization, profit maximization, and Sharpe ratio maximization. During the two-year test period, the best-performing probabilistic strategies improved observed profits by approximately 6% in Germany, 5% in France, and 4% in Iberia relative to the point-forecast benchmark. In the last quarter of 2021, the peak of the energy crisis, profits increased by 15%, 9%, and 10% in the German, French, and Iberian markets, respectively. These findings suggest that integrating probabilistic forecasts with profit-oriented and risk-aware optimization can help electricity suppliers, policymakers, and market operators improve trading decisions and risk management under changing market regimes.