Combined probability density model for medium term load forecasting based on quantile regression and kernel density estimation. (February 2019)

• Record Type: Journal Article
• Authors: Wang, Shaomin; Wang, Shouxiang; Wang, Dan
• Is Part Of: Energy procedia. Volume 158(2019)
• Page Start: 6446
• Page End: 6451
• Publication Date: 2019-02
• Subjects: Load forecasting -- probability density forecasting -- quantile regression -- kernel density estimation; Power resources -- Congresses; Power resources -- Periodicals; Power resources; Conference proceedings; Periodicals; 333.7905
• DOI: 10.1016/j.egypro.2019.01.169
• Languages: English
• ISSNs: 1876-6102

Abstract

Abstract: The medium term load forecasting is the basis of power grid planning and electricity transaction in power market. Current medium term load forecasting researches mainly focus on point forecasting, whereas with the development of smart grid and energy interconnection, numerous stochastic factors are emerging which affect the preciseness of deterministic point method. This paper proposes a combined probability density model for medium term load forecasting based on Quantile Regression (QR). The combined model combines three individual models of Random Forest Regression(RFR), Gradient Boosting Decision Tree(GBDT) and Support Vector Regression (SVR). Then a Kernel Density Estimation (KDE) method is used to achieve the load probability density distribution. The model is testified by an actual monthly data set from United States, and it proves that the proposed combined model can not only achieve more accurate point forecast result than individual models, but also effectively obtain the probabilistic result of load forecasting.