Research on Model Integration Strategy in Shortterm Power Load Forecasting

Multi-model ensemble model is an effective method to improve forecast accuracy in short-term load forecasting which is becoming popular in recent years. However, the research on model integration method is relatively limited. This paper investigates the performance of combining different load forecasting method based on multiple integration method. Nine integration strategies are considered and the effectiveness of these strategies are validated based on ISO New England public dataset. Case studies demonstrate that the accuracy of ensemble models significantly outperform that of the benchmark methods in terms of mean absolute percentage error.