LSTM-AdaBoost Electricity Sales Forecasting Model Based on One-Dimensional Time Series Input

With the rapid development of power system, the research of artificial intelligence method for short-term electricity sales prediction has made remarkable achievements. As long short-term memory (LSTM) neural network can fit the time series and nonlinear characteristics of electricity sales data, the short-term electricity sales prediction based on LSTM has a remarkable effect. However, in the case of one-dimensional input, LSTM prediction method has some problems such as insufficient prediction accuracy. Therefore, adaptive boosting (AdaBoost) ensemble learning method is used to train several LSTM strong prediction models, and a short-term electricity sales prediction method based on LSTM-AdaBoost is proposed, and a case is analyzed. By comparing with a single LSTM and other single and integrated models, the results show that LSTM-AdaBoost has lower prediction error and higher prediction accuracy. This method has certain application value for the intelligent development of power system.