Ultra-Short Term Power Load Prediction Based on Gated Cycle Neural Network and XGBoost Models

Abstract

Load forecasting is an important guarantee for power system design, planning and operation. In order to further improve the accuracy of power load forecasting, a combined forecasting model based on gating recurrent neural network (GRU) and limit gradient lifting (XGBoost) is proposed. Firstly, according to the load data and the input structure of GRU and XGBoost models, the data are preprocessed, and the preprocessed data are input into the corresponding model respectively. Then, the model is weighted by entropy weight method to obtain the predicted value of the final combination model. Finally, the first mock exam is used to compare the combination forecasting model with the common forecasting models. The results show that the proposed method can effectively combine the advantages of the two models, and take into account the continuous time series and discontinuous feature variables, which is more accurate than the single model and the common forecasting models.