Focusing on the design aspect of the commutation strategy of intelligent commutation switch, this paper puts forward a more practical and robust commutation strategy, i.e. using the differential evolutionary algorithm to make it optimized. At the same time, it is improved based on the traditional differential evolution algorithm and the problem of the actual phase change. The initial solution (initial population) of the algorithm is not simply generated randomly, but is produced by using greedy algorithm according to the diversity of genes, and then the mutation, crossover and selection operation are completed, which not only the problem dimension is reduced, but also the convergence speed of the algorithm is accelerated. Finally, the reliability and effectiveness of the algorithm in the treatment of three-phase imbalance are verified by MATLAB simulation.