A Time-of-Use Electricity Consumption Decomposition Model for Agent Power Purchase Users Based on HDBSCAN-UMAP Multidimensional Clustering

To address the challenges of incomplete time-of-use (TOU) electricity consumption collection for agent power purchase users and the high-dimensional, non-linear data encountered in electricity consumption pattern analysis, this paper proposes a TOU electricity consumption restoration model that integrates HDBSCAN-UMAP multidimensional clustering and constraint optimization. The model aims to enhance data integrity, deepen the analysis of electricity consumption behaviors, and ensure the efficiency of electricity market transactions. Taking the data of an agent power purchase user in the textile industry from March to August 2023 as an example, the verification results show that the model significantly outperforms traditional fitting methods in both cloudy and sunny scenarios. It effectively solves the problem of missing TOU electricity consumption data for agent power purchase users, provides accurate data support for spot market transactions, and is particularly suitable for analyzing the electricity consumption behaviors of industrial users with distributed photovoltaics.