Abstract

This study investigates price formation and forecasting in the intraday and mFRR electricity markets, with a focus on the growing challenges of integrating intermittent renewable energy sources. The first part examines the development of the intraday market, particularly the introduction of intraday auctions (IDAs). While IDAs were intended to support system flexibility, trading volumes remain low, and their influence on the intraday continuous (IDC) market appears limited but increasing throughout the trading day.

The second part analyzes price dynamics in the IDC market and develops forecasting models for intraday prices. The day-ahead (DA) price was identified as a strong baseline model. Forecasting was conducted using both autoregressive models and artificial neural networks (ANNs), with the ANN models generally outperforming traditional approaches.

The third part focuses on the mFRR balancing market, analyzing the direction of activated balancing energy and mFRR prices. Logistic regression and ANN models were developed to predict activation direction, with ANNs achieving the highest accuracy. However, class imbalances and weak signals limited practical applicability.

Finally, mFRR up- and down-regulation prices were modeled separately. While the DA price proved to be a useful baseline, ANN and ARX models offered improved performance, particularly for the down price. Overall, the results underscore the potential of machine learning approaches in electricity market forecasting, while highlighting the importance of data selection and market specific dynamics.