Transformer fault diagnosis based on homotopy BP algorithm

Power transformer fault diagnosis is the key technology of electric power system. To solve the problem that BP neural network easily trapped in local minima points, a non-linear homotopy based BP neural network is introduced in power transformer fault diagnosis. The neural network parameters were chosen after several experiments. LM optimization algorithm trained the non-linear homotopy BP neural network. DAG data was processed by cumulative frequency method and sent to BP neural network. The neural network proposed in this paper had a better performance on convergent speed and avoid trapped in local minima points. The power transformer fault diagnosis experiments and gases regression curve analysis both demonstrate that fault diagnosis precision of non-linear BP neural network was higher than standard BP network.