Electrical peak load clustering analysis using K-means algorithm and silhouette coefficient

Abstract

Nowadays, data analysis widely used in many fields especially in engineering. Clustering is one of data analysis methods to organize the amount of data into groups with similarity characteristics. One powerful analysis method to learn information by grouping data is clustering algorithms. The clustering advantages for electrical power utilities is to learn load behavior and provide information for power plant operation and also generation cost. In this paper, a simulation concept is proposed for analysis of peak load data by K-means clustering algorithm based on historical dataset. The results show electrical peak loads clustering by K-means algorithm are optimum classified into three clusters. This cluster evaluated by silhouette scores which high, intermediate, and low load level interpretation. One cluster has centroid during January, June, and July are relatively lower than another cluster caused by Indonesia national holiday. This concept also evaluates the load level affected by Covid-19 pandemic condition.