In the evolving landscape of energy markets, accurate electricity price forecasting is crucial for informed decision-making and efficient operational management. Researchers Souhir Ben Amor and Florian Ziel from the University of Stuttgart have developed a novel approach to improve short-term electricity price predictions, potentially benefiting energy traders, grid operators, and renewable energy integrators.
The researchers have introduced a new recurrent neural network architecture specifically designed for day-ahead electricity price forecasting. This model combines the strengths of both linear and non-linear structures by embedding traditional linear models, such as expert models and Kalman filters, into recurrent networks. This hybrid approach aims to enhance computational efficiency and model interpretability while capturing relevant price characteristics in power markets.
The model considers various factors influencing electricity prices, including calendar effects, autoregressive patterns, and external influences from load demand, renewable energy generation, and related fuel and carbon markets. To validate their approach, Ben Amor and Ziel used hourly data from the largest European electricity market, spanning from 2018 to 2025. Their model was compared against state-of-the-art linear and neural network models, demonstrating approximately 12% higher accuracy.
The researchers also evaluated the contributions of the interpretable model components, highlighting the benefits of combining linear and non-linear structures. This innovative approach could lead to more accurate price forecasts, enabling better decision-making in energy trading, risk management, and grid operation. The research was published in the journal Applied Energy, offering a promising advancement in the field of energy forecasting.
For the energy industry, this research presents a practical application in improving the accuracy of short-term electricity price predictions. Enhanced forecasting capabilities can support energy traders in making informed decisions, optimize grid operations, and facilitate the integration of renewable energy sources. By providing more reliable price forecasts, this model can contribute to the overall stability and efficiency of energy markets.
This article is based on research available at arXiv.