Fault diagnosis method for intelligent substation secondary system communication network based on improved TCN

Abstract

Aiming at the low accuracy of the Fault Diagnosis (FuDg) methods used in the Intelligent Substation Secondary System Communication Network (ISSSCN), a FuDg method of the ISSSCN based on the improved Time Convolution Network (TCN) is proposed. Firstly, the overall structure of the ISSSCN and the mathematical description method of the fault signal are analyzed. Secondly, a basic model for f FuDg of the ISSSCN is built based on the TCN including the Extended Causal Convolution (ECC) and the ResNet. Finally, the TCN model is optimized by the improved Hybrid Attention Mechanism (H-AM) based on channel and spatial attention, which greatly improves the accuracy of ISSSCN FuDg. The experiment shows that the accuracy of the proposed method for four different types of FuDg reaches 95.72%, 96.83%, 95.32% and 95.28% respectively, which is higher than the other two comparison methods. Therefore, the proposed method has good FuDg ability