An electricity retail plan recommendation method based on NMIDP clustering and collective optimal preference

Existing Electricity Retail Plan (ERP) recommendation methods often overlook the diversity of customer evaluation information and the influence of customer confidence, leading to significant deviations in recommendation results. To address this issue, this paper proposes an ERP recommendation method based on Neighborhood Mutual Information Density Peak (NMIDP) clustering and Progressive Collaborative Model (PCM). First, an improved NMIDP clustering: algorithm is proposed to cluster customers based on typical daily load curves, laying a theoretical foundation for similar customers identification. Second, a multi-dimensional evaluation matrix of customers’ perceptions of plan attributes that considers confidence levels is constructed to comprehensively capture customers’ true preferences. Next, the maximizing deviation method is used to determine the attribute weights of ERPs, and a collective optimal power consumption preference vector is constructed based on PCM to ensure group consensus in recommendation results. Finally, a case study is conducted for verification using ERP consumption data of 2164 customers in a development zone in eastern China. The results show that the proposed method improves the accuracy of ERP recommendations, provides new ideas for electricity retail companies, and holds important theoretical and practical significance.