MONITORING AND WARNING OF SMART GRID FAULT INFORMATION USING DEEP LEARNING

Abstract

To ensure the stable operation of a smart grid, it is necessary to monitor fault information and release timely warnings. In this paper, a monitoring algorithm based on a convolutional neural network (CNN) was designed using the deep learning method to determine the fault category of transformers and monitor them. An improved particle swarm optimization-long short-term memory (IPSO-LSTM) model was established to realize early warning of faults based on the prediction of dissolved gas concentration in oil. Experiments were conducted on the data of a power company. It was found that the CNN model had higher fault monitoring accuracy than the support vector machine (SVM) model and other methods; compared with the LSTM model, the IPSO-LSTM model had a smaller error in gas concentration prediction. Taking H2 as an example, the root-mean-square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) of the IPSO-LSTM model was 0.0412, 0.0432, and 6.2145%, superior to the LSTM model, so the IPSO-LSTM model could achieve the early warning of faults. The experimental results prove the effectiveness of the proposed method. The method can be further applied to the actual smart grid.