The new power system consists of four major elements such as source, network, load and storage. The distributed“energy stor?age”characteristics of electric vehicles can complement new energy generation, alleviate the impact of intermittent and fluctuation of newenergy output on the power grid, and effectively reduce the capacity of storage allocation in the power grid. However, the disorderly charg?ing behavior of EVs can aggravate the difference between new energy output and load, so it is necessary to control the charging of EVs inan aggregated manner. Firstly, the disordered charging behavior of different types of electric vehicles is simulated using the Monte Carlomethod, and the impact of disordered charging on the balance between source and load is analyzed. Secondly, the improved k-means++clustering algorithm is used to divide the result of disorderly charging and the difference of new energy into four clustering periods, name?ly,“sharp, peak, flat and valley”. Finally, an optimization strategy for charging aggregation control during overlapping clustering periods isproposed, which is verified in combination with the“Lingfeng Zero-Carbon Evolution Park”project in Ningbo, Zhejiang Province, and theresults show that, under the premise of impairing the charging satisfaction of EV owners as less as possible, this strategy can effectively re?duce the imbalance between new energy output and load, and combined with fixed-energy storage configurations, it can achieve a new typeof electric power and load imbalance. Combined with the fixed energy storage configuration, it enables source-load balancing in new powersystems.