A Data-Driven Hybrid Optimization Model for Short-Term Residential Load Forecasting

Abstract

Analyzing the characteristic of residential load and forecasting are beneficial for electric power companies to provide higher quality, more reliable electricity services to the customers. This paper is focusing on the research of short-term load forecasting based on the residential load data of Shanghai, China. Firstly, the time series mining technology was studied and also its disadvantages such as the meteorological conditions were not taken into consideration were pointed out. Secondly, we analyzed the most competitive method of the existing forecasting methods that taking meteorological conditions into account, namely the similar day method. In order to consider comprehensively the effect of meteorological conditions and fully utilize the advantages of the time series mining technology, this paper proposed a hybrid optimization model that combines the time series method and the similar day method. The extensive experimental results show that the hybrid optimization method can significantly improve the accuracy and robustness of short-term residential load forecasting.