The role of probabilistic load and renewable prediction in enhancing day-ahead electricity price forecasts

Abstract

The increasing penetration of renewable energy sources (RES) has amplified volatility in electricity markets, while demand variability continues to challenge system stability. Traditional day-ahead electricity price forecasting (EPF) models rely on point forecasts of load and RES generation, which fail to capture the uncertainty inherent in weather-driven supply and dynamic demand. This study introduces a novel framework that integrates probabilistic forecasts of load, wind, and solar generation into EPF models. Using data from the German EPEX market between 2015 and 2023, we generate quantile forecasts of fundamental variables via historical simulation, conformal prediction and quantile regression, and incorporate them into both parsimonious expert and high-dimensional LASSO-based models. Empirical results show that probabilistic inputs substantially enhance forecast accuracy, reducing root mean square errors by up to 13% compared with point-forecast benchmarks. Notably, extreme quantiles of load and RES forecasts emerge as the most influential predictors, underscoring the importance of rare but system-critical scenarios. These findings demonstrate that accounting for uncertainty in both load and renewables is crucial for reliable electricity price forecasting and offers practical value for system operators, traders, and policymakers navigating renewable integration.