A Projection-based Clustering of Flexibility Resources for Peak Load Shaving

Demand response is an effective means for peak load shaving in power systems. Clustering flexible resources is necessary to implement demand response due to the inability to directly manage resources with small capacity and large quantity. However, existing clustering models cannot accurately portray the adjustable region of flexible resources, resulting in poor performance in peak load shaving. Hence, this paper proposes a projection-based cluster approach of flexible resources. First, the high-dimensional adjustable region of each flexible resource is projected to the low-dimensional region corresponding to peaking periods. Second, a large number of resources are clustered into several aggregators using the extreme points of low-dimensional regions. Last, these aggregators are scheduled by the dispatch center to shave peak load. The numerical results show that the proposed approach achieves a better performance in peak load shaving compared to traditional approaches.