In order to solve the problem caused by paralleling operation of electric vehicles and distributed power grids, which brings strong randomness, intermittent and correlation to power system. Based on the probabilistic model of distributed power supply and electric vehicle, an optimal configuration model taking distributed power supply total cost, power supply reliability and active network loss as the objective function was established, embeding the probabilistic power flow calculation into the adaptive parameter difference based on the success history. The evolutionary algorithm was used to solve the objective function. The unscented transformation was used to describe the statistical properties of the system state variables by using the mean and covariance of the input random variables to directly deal with the random variables with uncertainty. Then RBF neural network was used to solve the power equation, which avoided calculating the Jacobian matrix and partial derivative, reducing the running time of the algorithm. Finally, the multi-objective function was calculated in parallel using the adaptive parameter differential evolution method based on the success history. The simulation of the IEEE33 node power distribution system verified the effectiveness and efficiency of the method and the method can save planning costs.