Electricity Price Forecasting: Bridging Linear Models, Neural Networks and Online Learning

TL;DR

This paper introduces a hybrid neural network model with online learning and forecast aggregation for electricity price prediction, achieving higher accuracy with lower computational costs in European markets.

Contribution

It presents a novel partial online learning method and a multivariate hybrid neural architecture with Bernstein Online Aggregation, improving accuracy and efficiency over existing models.

Findings

Reduces RMSE by 11-12% compared to benchmarks.

Decreases MAE by 14-17% relative to current state-of-the-art.

Significantly lowers computational time in six-year European market study.

Abstract

Precise day-ahead forecasts for electricity prices are crucial to ensure efficient portfolio management, support strategic decision-making for power plant operations, enable efficient battery storage optimization, and facilitate demand response planning. However, developing an accurate prediction model is highly challenging in an uncertain and volatile market environment. For instance, although linear models generally exhibit competitive performance in predicting electricity prices with minimal computational requirements, they fail to capture relevant nonlinear relationships. Nonlinear models, on the other hand, can improve forecasting accuracy with a surge in computational costs. We introduce a novel partial online learning approach, the key contribution of this work, which substantially reduces computational time. In addition, we propose a multivariate hybrid neural architecture that combines linear and nonlinear feed-forward neural structures. Unlike previous hybrid models, our approach integrates forecast combination using Bernstein Online Aggregation (BOA) to further improve forecasting accuracy. Compared to the current state-of-the-art benchmark models, the proposed forecasting method significantly reduces computational cost while delivering superior forecasting accuracy (11-12% RMSE and 14-17% MAE reductions). Our results are derived from a six-year forecasting study conducted on major European electricity markets.