A multi-resolution system approach to power transformer insulation diagnosis

Based on the synthetic analysis of the cause and serious degree of transformer faults, a decision tree method is presented, its identification order is from up to down, and each leaf of the decision tree represents a kind of fault model. The classification machine of different branches of decision tree is composed of different artificial neural network (ANN), thus a system model of combinatorial ANN is constructed, and the multi-resolution identification for transformer faults is achieved. The shortage of the complex configuration and the slow convergence of single ANN is overcome with this method, and the diagnosis accuracy is improved simultaneously. The application results show that this method is valuable for transformer faults diagnosis.