Measurement of Electrical Power Usage Performance using Density Based Clustering Approach

Abstract

Density-based clustering is related to the value space surrounding non-data points with data points. This algorithm uses a multi-resolution grid data structure and uses grid density to form clusters. The density-based clustering algorithm starts by determining the size or threshold of cluster density. In this study, density-based clustering is used to group the electrical power usage dataset into three density clusters (low, medium, and high density). The electrical power usage dataset has two attributes: the actual use and ideal use. The generation of the ideal use data for both UOL (Usage Off-peak Load) and UPL (Usage Peak Load) is using two scenarios: worst and best scenario. The application of these two scenarios is expected to provide a significant difference in performance. The cluster density threshold is determined based on the selection of the extreme distance range between data points (min and max). The purpose of the use of this clustering technique is to obtain the pattern of electrical power usage per month represented by the density level of each cluster. All the members of the high-density cluster are then used to measure its performance. The results of the study showed that the average performance of −17.48% (over kWh). The total performance of the usage load between the worst and best scenario was not so significantly different (25.15% of the best scenario) compared to the generation results of the ideal use data for both scenarios (682% of the best scenario). This result can be an indication of other factors contributing to these conditions which need to be analyzed in more depth, perhaps one of which is the aspect of the feasibility of existing electrical installations.