Photovoltaic Generation Forecasting for Intelligent Management in Energy Communities

Energy Communities (ECs) promote decentralized energy production and consumption with active citizen participation. Currently, Universidad de La Laguna is designing and implementing an energy community on its campus. A key element of this design is defining an Energy Management System (EMS) for efficient energy control. Accurate demand and generation forecasting will be essential for this system to make informed decisions. The main objective of this paper is to present a photovoltaic generation forecasting model tailored to the installation at Universidad de La Laguna. Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) architectures were analyzed, and different model configurations for each structure were compared. A feature selection based on model performance was conducted. Results indicated that the LSTM model, trained with data from the previous 72 h, effectively predicted the next 24-hour power generation. This work represents a crucial step toward designing an EMS for the EC at Universidad de La Laguna.