Commercial and Industrial Electricity Consumption Pattern Recognition Using Smart Meter Data: Unsupervised Clustering Method, Application, and Case Study in China

Abstract

Smart meter deployment in commercial and industrial sector provides an amount of power data that allows analyze the behaviors of commercial and industrial consumers. However, few studies pay attention to commercial and industrial end-users of energy consumption pattern. Besides, conventional methods for load profiling faces the problem of extracting features of smart meter data for making the clustering practical as well as confirming the appropriate clustering parameters. This paper resolves these problems via the unsupervised clustering algorithm that combines significant statistics and behavioral characteristics, and particle swarm optimization and simulated annealing algorithms are implemented to improve the global searching ability of clustering algorithm. A case study of the electricity consumption behaviors of 560 commercial and industrial consumers for a southwest Chinese city is investigated from 2018-2022. Relying on the visual test and validity indices scoring, the algorithm obtains an effective classification of three groups. The results show that electricity consumption can be robustly modeled using improved fuzzy c-means algorithm.