Electrical Load Profile Clustering Using Particle Swarm Optimization

Abstract

Nowadays, most of the electricity is produced from petroleum, and coal. These fuels are
depletable and produce toxic to the environment. In addition, the electricity demand is increasing.
Many researches have been focusing in energy management. Load profile clustering is one of
the energy management tools. In this paper, one of the most efficient clustering method, Particle
Swarm Optimization Clustering (PSOC), was firstly explored in load profile clustering. The results
were compared to the K-mean clustering which was a widely used method. It is a standard data
clustering method. There were total of 482 load profiles. Each load profile was energy consumption
recorded every 15 minutes for a 24-hour period. Therefore, there were total of 96 points in one
load profile. The load profiles were cleansed and normalized. Then, they were clustered by particle
swarm optimization and K-means clustering methods. The best number of groups was identified
using the Knee method. The results could be concluded that the particle swarm optimization
clustering method provided a better result than the K-means clustering method, and the particle
swarm optimization clustering method converged faster than K-means clustering method.