Grid Fault Diagnosis Based on Deep Pyramid Convolutional Neural Network

Tian Lan^*Yuezhou WuChen Wang

• Power Dispatching and Control Center of Guangdong Power Grid, Guangzhou, China

The traditional power grid fault d iagnosis method relies on manual experience w hen dealing with massive alarm information, ha s a complex modeling process and insufficient generalization ability, and lacks direct diagnosti c research on alarm information text. Therefore, we propose an intelligent fault diagnosis meth od based on the Deep Pyramid Convolutional Neural Network (DPCNN). This method builds an end-to-end fault classification model and key information extraction model to directly mine the implicit fault features from the alarm infor mation text, achieving accurate classification of fault types and rapid location of faulty equip ment. Comparative experiments show that the proposed method performs well in complex po wer grid scenarios and noisy data environments, with the highest fault classification accuracy o f up to 100%, and can effectively identify mult iple fault types such as simple faults, switch fai lure to operate, and protection failure to opera te. In addition, we also integrate this method with the time-sequence-priority fault equipment identification strategy to further improve the a ccuracy of fault location. The case study verific ation shows that our method has a fault recog nition rate of up to 99.5%, and can achieve 98. 7% accurate positioning after eliminating one b y one through the identification strategy, signifi cantly reducing manual intervention and having a high application value in the actual power g rid.