Research on LSTM-XGBoost Integrated Model of Photovoltaic Power Forecasting System

Abstract

In view of the insufficient feature extraction that affects the accuracy of photovoltaic forecasting, a photovoltaic power generation power forecasting model is presented, which integrates the Long Short-Time Memory (LSTM) algorithm and the Extreme Gradient Boosting (XGBoost) algorithm. In this paper, the advantages and disadvantages of LSTM algorithm and XGBoost algorithm are analyzed, and the advantages of the two forecasting models are integrated to obtain a more accurate forecasting model, XGBoost-LSTM; and compare the model with the popular Gated Recurrent Unit (GRU) and Deep Belief network, DBN). The experimental results show that the PV power forecasting model based on XGBoost-LSTM integration has higher forecasting accuracy, which has guiding value for photovoltaic grid-connected and off-grid.