Abstract: For the market demands of electricity retailers in providing differentiated electricity selling services and improving user stickiness, an electricity plan recommendation method based on the implicit score of electricity plan and user portrait is proposed.First, by extracting specified labels to represent the characteristics of electricity plans, the historical purchase behaviors of users are introduced as the implicit score of the corresponding plans, and a label-based user portrait model considering the attenuation of user preference is constructed. Then, the Pearson correlation coefficient and Euclidean distance are used to evaluate the similarity of time-sharing load and total load demand between users, respectively, and a label weighting method based on the two-scale similarity clustering of users′ load profiles and the silhouette coefficient is proposed. On this basis, a user portrait similarity evaluation model is constructed based on the weighted Euclidean distance, and a collaborative filtering based electricity plan recommendation method is proposed to select and recommend the most economical electricity plans for target users. The electricity plan recommendation simulation is carried out for users with different load demands and electricity consumption habits. The results show that the proposed electricity plan recommendation method can explore consumption preferences of users according to their historical purchase information, so as to improve the accuracy of electricity plan recommendation of electricity retailers.