Short-term Load Forecasting Based on GBDT Combinatorial Optimization

When using a single model for short-term electric load forecasting, the prediction results are easily affected by unexpected events and fall into local optimization. In this paper, a short-term load forecasting method based on combinatorial optimization of Gradient Boosting Decision Tree (GBDT) is proposed. Based on the single model, the optimal combination is established, and a GBDT load forecasting model is established. The above method is used to predict a total of 96 point loads in a certain area in southern China with 15 minutes as time interval. Compared with the Autoregressive Integrated Moving Average Model (ARIMA), Support Vector Machine (SVM) and Back Propagation Neural Network (BPNN) methods, the effectiveness of the method is proved.