Anchor Graph-Based Subspace Clustering for Large-Scale Residential Load Profiling

Residential load profiling clustering is crucial for optimizing energy use and enhancing efficiency of power distribution system. However, current methods face three limitations like reliance on manual similarity measures, high computational costs, and inflexibility to new load profiles. To address these challenges, this paper proposes a large-scale anchor graph-based subspace clustering method for residential load profiling. Specifically, the distance matrix is directly learned from the load profile through subspace clustering to handle high-dimensional data and uncover hidden load patterns. Then, an anchor graph is constructed to capture the intrinsic structure of the learned subspace, which can effectively reduce the computational complexity. Finally, k-Nearest Neighbor (KNN) is adopted to assign labels to new load profiles based on selected anchor points, eliminating the need for re-clustering the entire dataset. Tests on real-world London household data confirm the method’s effectiveness and efficiency. The code is available at https://github.com/U-T-G/AGSC.