Recommending electricity plans: A data-driven method

Yuan Zhang, Ke Meng, Dong Xu, Mingyong Lai, Yu Zheng

Abstract: In this paper, a customized electricity retailing plan recommendation strategy is developed. Cost-effective retailing plans are recommended according to customer's electricity usage pattern. Usage pattern is recognized through classifying daily load profile (DLP) into a certain class, which is obtained in DLP clustering process. Two DLP clustering skills, plan rank-oriented and load feature-oriented clustering are investigated in this paper. Each DLP class is attributed to a typical plan rank, plans in which are more likely to supply good deal for DLPs in this class. Correlative features of DLP are evaluated in DLP classification phase, which enables recommendation accessible to customers with limited features information other than full load profile record. Promising results on plan recommendation for 60 residential customers demonstrate that the proposed strategy could supply effective recommendation to electricity users.