Electrical Load Forecasting Based on Multi-model Combination by Stacking Ensemble Learning Algorithm

Abstract

Load forecasting is helpful to achieve the goals of emission reduction and the balance of power generation and consumption. In this paper, a load forecasting method based on multi-model combination by Stacking ensemble method was proposed. The most appropriate basic models were chosen as the basic learners in order to achieve the optimal performance of Stacking model. The second layer choose the model based on a simple algorithm to prevent over fitting. Some representative load data are selected to verify the feasibility of the algorithm. The results show that the Stacking learning framework improves the overall prediction accuracy by optimizing the output results of multiple models, has a good application effect in power load prediction.