Analysis on residential electricity consumption behavior using improved k-means based on simulated annealing algorithm

It is indispensable for electric power companies to classify the electricity consumption behavior of residential customers, in order to understand the user's personalized demands and provide them with targeted services. K-means Clustering algorithm is one of the most popular methods for grouping the consumption patterns in previous studies. However, in traditional K-means algorithm, the initial clustering centroids are selected randomly, which makes it susceptible to local optima and has difficulty in converging to the global minimum. In view of this drawbacks, an improved K-means based on the simulated annealing algorithm is proposed in this paper. By employing the simulated annealing algorithm, the optimal cluster centers can be obtained. By experiments using a large-scale dataset including about 216 houses' consumption records in American, it is shown that the proposed method has a better performance than traditional K-means algorithm and be able to extract typical consumption patterns.