An Algorithm for Abnormality Detection and Failure Diagnosis of Electric Power System Data Collection Based on Deep Learning

Aiming at the complexity and spatio-temporal characteristics of electric power system (EPS) data, this article puts forward an abnormality detection model based on Autoencoder, and a fusion failure diagnosis model based on CNN+LSTM (Convolutional neural network + Long-term and short-term memory network). By introducing DL (Deep learning) technology, the characteristic information in power data is fully mined, and the anomalies and faults are accurately identified and classified. In terms of methods, this article first constructs an abnormality detection model based on Autoencoder, learns the normal distribution of power data by using the symmetrical structure of encoder and decoder, and realizes abnormality detection through reconstruction error. Then, a fusion model based on CNN+LSTM is designed, which combines the spatial feature detection ability of CNN and the time series processing ability of LSTM to accurately classify power faults. The model adopts a multi-task learning strategy to realize the joint optimization of abnormality detection and failure diagnosis tasks. The results show that the proposed model has achieved remarkable results in abnormality detection and failure diagnosis tasks. The proposed model provides strong technical support for the safe and stable operation of EPS.