Electricity Consumption Behavior Analysis Based on DE-DBSCAN
Shu Yifei, Bo Fan, Kang Jieying, Zeng Lai
Abstract
Analyzing the behavior of power users is of great significance to the co-ordination of demand and supply of power system. There are usually lots of noised power consumption data (PCD) of power users. The traditional DBSCAN algorithm has strong anti-noise ability, but it needs to manually preset the appropriate parameters. To handle this issue, this paper proposes a parameter adaptive DBSCAN algorithm to cluster PCD. Firstly, based on the original PCD, the feature indexes of PCD are extracted; Then, the proper parameters are adaptively determined by using differential evolution algorithm, and used in DBSCAN algorithm to implement the clustering; Finally, according to the cluster results, we divide power users into 5 categories, each category has different electricity habits, electric power utility can adjust the demand-side response strategies on the basis of this. We also compare our method with traditional DBSCAN, KANN-DBSCAN and other algorithms, the effectiveness of our method is proved by the results.