Research on Power Load Forecasting Based on Deep Learning

Abstract

In order to fully explore the time-series correlation of power load data and improve the prediction accuracy of power load, this paper proposes a neural network-based deep learning approach for power load prediction. Firstly, the relevant electric power data are obtained and divided into appropriate sample sizes, and the samples are normalized; then, a prediction model based on LSTM is built to explore the correlation between different features, and the corresponding model of this neural network is further trained and validated on the data test set; finally, a comparison between LSTM and other algorithms such as SVM, ANN, GAOS and GM are performed. The results show that the LSTM prediction algorithm can better track the trend of power load change, with higher prediction accuracy and efficiency.