In this paper, a methodology for measurement-based load modeling in power systems with distributed generation (DG) as wind power sources is presented. A ZIP load model in parallel with an asynchronous machine is used as load model and particle swarm optimization algorithm is applied in parameter estimation. The proposed methodology is applied to obtain several composite load models in a Western Systems Coordinating Council (WSCC) 9-bus test system considering different levels of power injection from a DG source. The results show that the response of the developed load models can fit accurately to the system measurements for several operating conditions.