A clustering method for charging behaviors of electric vehicle users based on an improved k-means algorithm

Abstract The uncertainty of Electric vehicle users’ (EVUs) charging behaviors affect the power grid safety. Therefore, mining the charging behavior characteristics of EVUs can help the power grid to make optimal operation plans previously to smooth the load curve and improve stability. In this regard, a clustering strategy for EVUs is proposed, using an improved K-means algorithm based on the minimum variance theory(MVT). Additionally, the Monte Carlo algorithm is utilized to obtain the simulation results of EVUs. Finally, a numerical simulation based on actual data in China is established to validate the effectiveness and reliability of the proposed cluster algorithm. The results indicate that the proposed cluster algorithm can accurately extract the characteristics of EVUs’ off-on grid behaviors and obtain the accurate daily charging demand representing the main charging rules among EVUs. The proposed algorithm results in a reduction of 69.2% of solution time compared to the traditional K-means algorithm.