Joint Forecasting of Energy Consumption and Generation in P2P Networks Using LSTM–CNN and Transformers

Electric energy is an essential resource in modern society; however, most current distribution systems are centralized and dependent on fossil fuels, posing risks of shortages and a potential energy crisis. The transition to renewable sources represents a sustainable alternative, though it introduces challenges associated with intermittency and generation variability. In this context, peer-to-peer (P2P) networks and artificial intelligence (AI) emerge as strategies to promote decentralization, self-management, and efficiency in energy operation. This research proposes an AI-based knowledge discovery model to predict electricity generation and consumption in a P2P network. The study was developed in four phases: exploration of AI techniques for energy prediction; analysis of the most widely used techniques in the Knowledge Discovery in Databases (KDD) process; construction of the predictive model; and validation using real energy generation and consumption data from renewable energy sources. The LSTM–CNN and Transformer models achieved an R2 greater than 80% and mean absolute errors (MAE) of less than 0.02 kWh, demonstrating high prediction accuracy. The results confirm that integrating the KDD approach with deep LSTM–CNN and Transformer architectures significantly improves energy management in P2P networks, providing a solid foundation for the development of innovative and sustainable electrical systems.