The main obstacle to the wide application of the measurement-based load modeling approach（MBLMA） lies in the serious insufficiency of the effective data provided by disturbance normally occurring in a power system. This is because the currently used MBLMA can only identify the load parameters from three-phase symmetric disturbance. But most of the faults occurring in a power system are asymmetric， and it is hard to get permission to do field test. An approach to load dynamic modeling during asymmetric disturbance（LMAD） is proposed. The problem of insufficient data is solved by applying Park's equations in solving the induction motor's dynamic process. And the problem of divergence in MBLMA is solved with the modified immune algorithm in the optimization of the model parameters instead of the least square estimation（LSE） method. Finally， the identified load model's influence on the critical clearing time（CCT） in Shanghai Power Grid is discussed.