An Intelligent Profiling and Classification Method for Load Adjustment Potential of Multi-Type Demand-Side Resources Considering Adjustment Willingness

The rapid development of new energy has caused a sharp increase in the stochasticity on the source side of the new power system (NPS), and extreme weather along with climate variability have also led to increased stochasticity in power demand on the load side; thus, how to achieve source-load matching and enable the load to track the source under the new situation is the key to the efficient operation of the power system. Aiming at the problem that existing load regulation potential evaluation mainly focuses on physical capacity, making it difficult to reflect users’ subjective willingness to participate as well as the dynamic changes in regulation capability under different operating scenarios, this paper proposes a two-stage dynamic profiling classification method for multi-type power user loads considering regulation willingness. First, an evaluation index system is constructed from three dimensions, physical reliability, execution reliability, and behavioral willingness, to achieve the unified characterization of the regulation capabilities of heterogeneous resources such as industrial loads and electric vehicle (EV) aggregators. Second, the DBSCAN algorithm is adopted to identify typical annual operating scenarios. Finally, the Dynamic Time Warping (DTW) distance is introduced to improve the K-Means++ algorithm, achieving the profiling classification of user regulation potential. This paper takes a certain NPS demonstration park as an example for verification, and the results show that the annual operating scenarios can be divided into 4 types of typical days; the proposed DTW-K-Means++ method has better classification performance compared with traditional Euclidean distance clustering, can effectively identify the differences and dynamic migration characteristics of user regulation potential under different operating scenarios, and stably classifies users into three types of profiles: deep regulation type, agile response type, and rigid constraint type. The research results aim to provide reliable data support for the refined dispatch of the power grid by effectively quantifying the dynamic migration patterns of heterogeneous resources under variable scenarios.