Analysis of Distribution System Considering Load Uncertainties: a K-Means Clustering Approach

Abstract

This paper presents an analysis of distribution system performance considering load uncertainties. A multi-scenario approach is proposed in order to preserve simultaneity and chronological combination of different loads. K-means clustering algorithm is used to reduce an extensive amount of load profiles data, where a sensitivity analysis of the number of clusters is performed, based on power flow studies. The results were compared considering an annual database and a load duration curve approach, as well as K-means clustering methods, taking into account the accuracy and computational burden points of view. Simulation results shows that the introduced method significantly reduced the computational burden while keeping the accuracy, with a lower number of multi-scenarios when compared to other approaches.