Artificial Neural Networks applied in the predictive maintenance of Electricity Generating Units through the Partial Discharges Classification

Electrical equipment in transmission lines and power distribution oftenexperiences faults due to insulation failures. Studies have shown that insulation defectscause localized electrical discharges, known as Partial Discharges (PD). PD analysisis an effective method for monitoring electrical equipment. Repeated PDs weaken theinsulation and may eventually cause system failure. Regulations identify seven typesof PDs, each linked to specific insulation failures and criticality levels. This articlepresents promising results using Artificial Intelligence algorithms, specifically ArtificialNeural Networks, to classify PD types with over 99% accuracy. The algorithm operatesas a microservice, automatically handling search, classification, and event generation.