Deep learning-based prediction models for spot electricity market prices in the Spanish market

This paper explores deep learning-based prediction models for spot electricity market prices in the Spanish market. Electricity prices in deregulated markets, such as the Spanish spot market, exhibit significant volatility. This study highlights the importance of accurately predicting electricity prices to optimize decision-making for trading companies. The focus is on artificial intelligence models, particularly neural networks, due to their ability to capture nonlinear behaviors. The research utilizes extensive data from the Spanish electricity market, including demand forecasts, wind power production, solar generation expectations, gas prices, and more. A correlation analysis reveals that the impact of these variables on electricity prices varies across different years. Several deep learning models, including feedforward, convolutional, and long short-term memory (LSTM) neural networks, are developed with hyperparameter tuning. The best-performing model is a convolutional neural network model, achieving a relative Mean Absolute Error (rMAE) of 13.29%, demonstrating its effectiveness in short-term price prediction. The study also evaluates the impact of individual variables on model performance, underscoring the importance of renewable energy sources and gas prices. The proposed model shows strong potential for accurately predicting spot market prices with a 1-day horizon, providing valuable insights for market participants.