A Novel LSTM Optimized by Neurodynamic Network for Heat Prediction of Power Equipment

With the rapid development of smart grid and substation equipment, the heating problem of power equipment has become an important factor affecting the stability and safety of equipment. The traditional equipment monitoring method has lag, and it is difficult to predict the heating trend of the equipment in time. Therefore, a long short-term memory based on neurodynamic network(NN-LSTM) is proposed to improve the accuracy of predicting the heating trend of power equipment. In this paper, the predictive ability of the LSTM network is improved by combining the neurodynamic optimization strategy. And it is applied to power load forecasting and transformer oil temperature forecasting. Experimental results show that NN-LSTM is superior to the traditional LSTM model on the two datasets, the mean square error is effectively reduced, and the prediction accuracy is higher.