A novel electricity sales forecasting method based on clustering, regression and time-series analysis

Abstract

The forecasting of monthly electricity sales is a fundamental job for the marketing department of State Grid Corporation of China. In this paper, we propose a novel electricity sales forecasting method. Firstly, with the visual clustering algorithm, 27 provincial electric power companies of State Grid Corporation of China are clustered into different groups according to their historical monthly electricity sales curve; secondly, as to different groups, the monthly electricity sales are predicted by specific regression or time-series algorithms. Experiments show that the proposed approach has very high prediction accuracy rate in terms of forecasting the the electricity sales of the next 12 months for the power companies of State Grid Corporation of China.