Transmission Line Fault Location Based on the Stacked Sparse Auto-Encoder Deep Neural Network

Given the current situation that a large amount of data is available for transmission line faults, and supervised learning is mainly used to extract fault features, a fault location method based on Stacked Sparse Auto-Encoder (SSAE) is proposed. Firstly, the maximal information coefficient is used to reduce the dimension of the data. The data interval with good information density is selected, the deep network is used to extract features, and the network parameters are initialized by greedy layer-wise unsupervised pretraining. Finally, the linear regression layer is connected and fine-tuned with the fault distance data. The simulation data show that, compared with the traditional BP neural network method, the proposed fault location method has higher accuracy and obvious feature extraction ability.