Research on Park User Identification Based on Improved K-MEANS and Multidimensional Potential Index of Distributed Energy Storage

With the growing share of new energy installations, traditional power systems are under increasing pressure. Distributed energy storage systems not only mitigate the volatility caused by renewable resources but also enable load balancing and adjustments in user energy consumption patterns. To provide users with a better understanding of their energy storage habits, a park user identification method is proposed based on an enhanced KMEANS algorithm and multidimensional indicators for distributed energy storage. The enhanced K-MEANS, integrated with the Borderline SMOTE algorithm, addresses data class imbalances. Selection of the optimal number of clusters K is achieved using evaluation metrics such as SSE, silhouette coefficient, and CH. Preliminary assessment of user energy potential is based on daily electricity load patterns. To address low energy storage utilization among distributed energy storage users, a comprehensive multidimensional potential index system is introduced. Weighting of these indicators is determined using the Delphi method, and users are ranked based on their energy usage potential using the TOPSIS method. The effectiveness of this proposed method is demonstrated through experimental validation.