A Study of Fault Diagnosis Methods for Substation Primary Equipment Based on Improved Deep Convolutional Neural Networks
Xiangsheng Lei, Xinghua Wang, Jingyi Wang, Shibo Dong, Xiaodong Wang · 2024
Due to the diversity of fault states of primary equipment in substation, it is difficult to guarantee the effect of its diagnosis. Therefore, the research on fault diagnosis method of primary equipment in substation based on improved deep convolution neural network is proposed. Retinex algorithm is introduced to enhance the original status data of primary equipment in substation; Combining the cross entropy function, the deep convolution denoising self encoder is constructed to reduce the dimension of the status data of the primary substation equipment. After the mapping relationship between the fault characteristics and the fault types is determined by using the deep convolution neural network, the data of the primary substation equipment to be diagnosed is input into the trained model, and the diagnosis results are output. The test results show that, the designed diagnosis method has always been more than 0.95 ACC for different fault diagnosis, which has obvious advantages compared with the control group, and has higher learning efficiency.