Fault diagnosis of high-frequency synchronous full-power data based on multi-source data acquisition and deep neural network

Abstract

To meet the need for high-frequency synchronous full-power data fault diagnosis in new power systems, this study proposes an innovative method combining multi-source data acquisition technology and deep neural networks for accurate power system fault identification and efficient fault location. Firstly, it integrates multi-source heterogeneous data from WAMS, SCADA, and meteorological sensors to form a holistic sensing network covering electrical parameters, environmental conditions, and equipment operating conditions, creating a multi-dimensional feature space. Secondly, deep neural networks extract features and recognize patterns in the collected full-power data to identify fault types, locate faults, and analyze fault causes. Finally, the research shows that this method has made significant breakthroughs in data synchronization accuracy, diagnosis accuracy, and adaptability to complex scenarios.