Research on Smart Grid Users’ Power Consumption Behavior Classification Based on Improved k-Means Algorithm

Abstract

Now, the number of smart gird users is increasing. The classification of smart grid users has become the basis of user behavior analysis, load forecasting and demand response. This paper improves the traditional k-means algorithm that is the most common users’ classification algorithm. This improved k-means algorithm uses a new distance calculation method to replace the Euclidean distance. The paper firstly uses two distance calculation methods to process the same set of users load data. Then two different cluster results of smart grid users are acquired. Finally, we uses the cluster validity index Mean Index Adequacy (MIA) to evaluate two results that respectively get from traditional k-means algorithm and improved k-means algorithm. This simulation verifies that the improved kmeans algorithm is better than the traditional k-means algorithm. The improved k-means algorithm not only eliminates the normalization of all samples but also makes the clustering result better. And our improved k-means algorithm can solve the smart grid users’ classification problem better. 