Data-Driven Fraud Detection in Active Participation Programs of End-Use Energy Consumers Based on a Hybrid Fuzzy Approach

Innovations in smart electricity grids are essential for decarbonization, integrating renewable sources and improving the balance between energy supply and demand, as well as establishing a more effective relationship between prosumers and the electricity system. However, the expansion of the energy infrastructure brings new challenges in terms of the security and integrity of the data collected; without robust validation methods, the system becomes vulnerable to manipulation. This article therefore presents a hybrid fuzzy approach, which combines a data-based evolving model, called Evolving Takagi-Sugeno Plus, with a classic knowledge-based Mamdani system. The evolving model, applied to modeling and predicting the patterns of involvement of individuals in active participation programs, evolves dynamically, adjusting parameters and reconfiguring its structure based on the samples received. In the training and validation stages, the model showed competitive performance, especially in terms of execution time. In turn, the Mamdani system uses the residual calculated between the output of the evolving model and the real flexibility data, together with generation and consumption information, to generate a degree of alert when anomalous behavior is identified. The results obtained indicate that the system can detect fraud in both isolated events and prolonged patterns, demonstrating its potential for application in practical contexts.