Bayesian Neural Networks with Monte Carlo Dropout for Probabilistic Electricity Price Forecasting
Abhinav Das, Stephan Schlüter
TL;DR
The paper tackles the volatility of electricity prices in deregulated markets by developing a probabilistic forecasting framework based on Bayesian neural networks with MC dropout. It trains hour-specific models using a rich 248-dimensional feature set to capture diurnal and exogenous effects, producing full predictive distributions rather than point estimates. The proposed approach outperforms traditional benchmarks (GARCHX and LEAR) in both point forecasts and interval quality (CRPS, MPIW) on German market data, albeit with some undercoverage in prediction intervals, signaling calibration opportunities. The work demonstrates the practical value of probabilistic neural models for risk-aware bidding and resource planning in energy markets and outlines directions to improve interpretability and cross-hour dependencies.
Abstract
Accurate electricity price forecasting is critical for strategic decision-making in deregulated electricity markets, where volatility stems from complex supply-demand dynamics and external factors. Traditional point forecasts often fail to capture inherent uncertainties, limiting their utility for risk management. This work presents a framework for probabilistic electricity price forecasting using Bayesian neural networks (BNNs) with Monte Carlo (MC) dropout, training separate models for each hour of the day to capture diurnal patterns. A critical assessment and comparison with the benchmark model, namely: generalized autoregressive conditional heteroskedasticity with exogenous variable (GARCHX) model and the LASSO estimated auto-regressive model (LEAR), highlights that the proposed model outperforms the benchmark models in terms of point prediction and intervals. This work serves as a reference for leveraging probabilistic neural models in energy market predictions.