A meta-heuristic optimization based residential load pattern clustering approach using improved Gravitational Search Algorithm

More in-depth understanding about the load pattern clustering (LPC) can enhance the knowledge on end-users' electricity consumption behavior characteristics to improve the design of demand-side response schemes and service level for utility companies. However, traditional clustering methods such as K-means only take the compactness of the formed clusters into account without considering the separation, which makes the clustering results unreasonable. Moreover, K-means is sensitive to the initial centroids and easy to trap into local optimum. To address these issues, a meta-heuristic optimization based residential LPC approach is proposed in this paper. First, the density based spatial clustering of applications with noises (DBSCAN) is adopted to remove the outlier load profiles and obtain typical load patterns. Second, LPC is formulated as an optimization problem which considers both compactness and separation of the formed clusters. Then an improved Gravitational Search Algorithm (IGSA) is proposed to solve it. To improve the global search capabilities of the standard GSA, memory management strategies from PSO and multi-mutation/crossover/selection mechanisms from difference evolution are modified and adopted in this research. Finally, a case study is carried out to verify the feasibility and effectiveness of the proposed method, in which IGSA is compared with three other well-known clustering methods including K-means, PSO and GSA. The simulation results indicate that IGSA shows better performance in terms of the clustering quality.