Power Load Forecasting Model Based Long Short-Term Memory Neural Network

Abstract

In order to fully exploit the correlations of time-series features in power load data and improve the accuracy of power grid load forecasting, this paper proposes a neural network load forecasting method based on long short-term memory (LSTM) network. A feature set is constructed with climate factor, date factor, and characteristic daily load factor as input. First, the data are normalized to eliminate the effect of dimensionality; then the LSTM network is used to extract the connection between the electric load and the feature values in high-dimensional space and construct the high-dimensional feature vector of the time series; finally, the parameters of the LSTM network model are trained and the load prediction values are output. Using this method to predict the electric load data of a plant area in Jiangsu Province, the results show that the proposed prediction method has significant advantages in prediction accuracy compared with BP network model, ANN network model and RF network model.