Sistema Preditivo para Diagnóstico de Transformadores a Seco Baseado em Escala de Degradação de Isolamento

Abstract

The aging of power transformers is a natural operational effect, mainly due to insulation degradation caused by heating and mechanical stresses. Additionally, power transformers are subject to short circuits and overvoltages, which increase the dielectric stress on the insulation materials. Partial Discharges (PDs) are directly related to the degradation of winding insulation, which can evolve into Inter-Turn Short-Circuits (ITSC). This study proposes a monitoring and diagnostic system based on periodic PDs and ITSC measurements to infer a dry-type transformer’s operational state. The methodology uses Machine Learning (ML) algorithms to correlate these events and predict possible failures. A hypothetical scale of insulation system degradation levels is proposed to allocate the operational state of the transformer, supporting the decision-making system. The system prototype is validated in a relevant industrial environment, indicating that the integration of online detection methods of PDs and ITSC with ML techniques shows promise for continuously monitoring transformer conditions.