The research and application of WNN in the fault diagnosis technology of electric locomotive Main Transformer

The paper proposes an optimized wavelet neural network (WNN) to diagnose faults of electric locomotive main transformer. The optimization algorithm introduces the concepts of quantum, affinity, concentration and chaos sequence. In the process of network training, the chromatographic data and electrical test data worked as the inputs of orthogonal wavelet neural network, the network´s hidden layer used orthogonal db4 function as basis function, the hybrid particle swarm algorithm can be used to obtain the initial values of orthogonal wavelet neural network and optimize the network parameters. The test results show that the proposed HPSO-WNN do effectively improve the traction transformer fault diagnosis speed and accuracy.