A Data-Driven Based Smooth Charging Method for Supercharger of Electrical Vehicle

Superchargers can not only significantly reduce the charging time of electric vehicles (EVs), enhancing the user experience, but also effectively alleviate the range anxiety problem of EVs, thus further promoting the expansion of the EV market. However, the high-power charging of superchargers leads to a decline in charging efficiency or even cause charging interruptions. In this work, a data-driven based optimal control method is proposed for superchargers to realize a smooth charging process. Specifically, the neural network is designed based on the dynamic model of the supercharger circuit to make the black-box of machine learning explainable. Finally, compared to the benchmark system, the proposed method can reduce more than $\mathbf{7 5 \%}$ of the voltage disturbance.