Fault Diagnosis and Prediction of Large Power Model Driven by Multi-Modal Data

The multi-modal data in power system comes from a wide range of sources, and the data from different sources have different structures and representations. It is very difficult to integrate these heterogeneous data into a unified framework for analysis. Moreover, the power system contains many different types of equipment, so it is difficult to extract power fault factors. Therefore, a fault diagnosis and prediction method of large power model driven by multi-modal data is proposed. Firstly, the power multi-modal data is collected and fused, the power fault fault factors are extracted based on the power model, and the power fault is diagnosed based on DS evidence theory. Based on this, feature variable screening is introduced to predict power failures. The experimental results show that the proposed method can accurately extract the operating signal features of power equipment, and the sensitivity of power fault diagnosis and prediction is higher.