abstract |
The invention discloses a medium- and short-term forecasting method for multi-factor power demand based on multi-model fusion. By integrating multiple models, the power demand under different time-length conditions is forecasted, and a large amount of forecasting time is saved on the premise of effectively improving the forecasting accuracy. The three models of GBDT, XGBoost and LightGBM belong to the Boosting model, which is a decision-making model based on a decision tree. The answer to the target problem is solved by finding the optimal solution in the decision tree. The stacking framework is used to predict the power data in two layers. After the prediction is completed by the three Boosting models of GBDT, XGBoost and LightGBM, the prediction result is corrected and output by the LR model, which enhances the accuracy and credibility of the prediction. The three models are combined to jointly predict power demand, so that the three models can complement each other and bring about higher-precision prediction results. This method belongs to the decision search type optimization solution algorithm, so the time required for decision-making is much lower than that of artificial intelligence algorithms. |