基于LSTM-WGCNA聚类与负荷准线的用户需求响应潜力评估OA

Against the backdrop of the carbon neutrality strategy and the development of new power systems,effectively identifying and utilizing residential-side flexible resources has become a critical issue in demand response and grid regulation research.To address this,this paper proposes a novel user demand response potential evaluation method,LSTM-WGCNA,that integrates Long Short-Term Memory(LSTM)networks and Weighted Gene Co-expression Network Analysis(WGCNA).In this approach,the 24-hour load curve of residential users is analogized to a gene expression profile to construct a load co-variation network.By leveraging the topological overlap matrix,the method identifies load pattern modules with high temporal correlation,thereby enabling clustering of user electricity consumption behaviors.On this basis,a response potential evaluation framework is established by matching each user's profile to a reference customer directrix load,quantitatively characterizing both collective and individual response capabilities across different time periods.Comparative analysis with efficient algorithms such as K-means and SOM demonstrates that the proposed method significantly improves intra-cluster consistency and boundary separability.Case study results show that the method effectively identifies representative load patterns with a 43％improvement in intra-cluster purity,and accurately evaluates user flexibility potential,providing both data support and methodological insight for load aggregators to implement differentiated resource allocation and demand response strategies.