Personalized Electricity Retail Plan Recommendation via User Clustering and Partial Ordinal Priority Approach

Growing competition in retail electricity markets creates a need for recommendation methods that account for heterogeneous consumption behavior, incomplete preference information, and interpretability. This study proposes a personalized framework that integrates Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) and the partial ordinal priority approach with Pareto optimality identification (OPA-P). Seven descriptors derived from monthly load data define user profiles, and HDBSCAN identifies stable clusters across heterogeneous density levels while isolating atypical profiles. For each cluster, a multi-attribute evaluation system represents incomplete preferences among users, criteria, and candidate plans as partial-order relations. OPA-P jointly derives criterion and plan weights; integrated weights and an adversarial Hasse diagram then characterize both the total ranking and the hierarchical partial-order structure of the plans. New users are assigned to the nearest stable cluster and receive a primary recommendation, alternatives, and a structured explanation. In an illustrative case study, the framework identified four stable clusters while isolating the atypical profiles, with a mean silhouette coefficient of 0.673. These results demonstrate the methodological feasibility of the framework under controlled conditions and its potential to support interpretable, personalized recommendations; they do not establish predictive validity for an actual electricity retail market.