Diagnosis of Power Transformer Internal Fault Based on Denoising Autoencoder and Optimized Convolutional Neural Network

For power transformer protection, differential protection has long been engaged as the main protection scheme. However, it still suffers from the risk caused by inrush currents, which may result in relay mal-operation. For the purpose of improving the reliability of differential protection and decreasing the risk of false trips, a data driven fault diagnosis method based on differential current is presented in this paper, using denoising autoencoder (DAE) and convolutional neural network (CNN) optimized by the improved particle swarm optimization (IPSO), to make a discrimination between internal fault and inrush currents. The feasibility of the proposed method is verified on the data generated by a simulation model established in PSCAD/EMTDC and data collected from laboratory experiment. The simulation results indicate that even under certain extreme conditions such as current transformer (CT) saturation, series capacitor (SC) compensation, superconducting fault limiter (SFCL), the proposed method still exhibits an outstanding performance. Meanwhile, an extensive comparison with several state-of-the art methods demonstrates the superiority of the proposed method.