Parameter identification for load modeling is analyzed and the conventional single-objective optimization model for load modeling is modified with a multi-objective optimization model. Then the improved strength Pareto evolutionary algorithm（SPEA2） combined with the parallel genetic algorithm（PGA） is used to identify the parameters. Compared with the conventional one， the proposed algorithm has such features as fast convergence and powerful optimization capability. It can obtain multiple sets of parameters for load modeling， which makes it convenient for customers to choose an appropriate solution according to their own specific demands. This efficient and high-precision algorithm is also a preparation for load modeling on the grid platform. Finally， the proposed method is validated by successful tests in the Shanghai power system.