Accurate short-term electricity price forecasting is essential for cost optimization, strategic planning, and operational decision-making in the energy sector. This paper presents a case study focused on the Iberian day-ahead electricity market, aiming to address two main research challenges. The first consists of constructing a unified dataset by integrating heterogeneous data sources, namely OMIE (marginal electricity prices), REN Datahub (load and renewable generation), and Copernicus Climate Data Store (meteorological variables). The second involves developing a machine learning framework capable of delivering accurate, real-time, and scalable electricity price predictions for seven days ahead (168 hours). Extensive time-series feature engineering was applied, including lag features, rolling statistics, and calendar encodings, to enhance model learning. Several machine learning models were tested, with LightGBM selected as the final predictor due to its superior accuracy and generalization performance. A complete pipeline was implemented in Python, and FastAPI endpoints were created to enable future deployment and real-time forecasting capabilities. The results demonstrate how machine learning can effectively integrate environmental, market, and load data to improve electricity price forecasting, providing a replicable methodology for similar challenges in energy markets.
Khalid et al. (Thu,) studied this question.