In order to meet the requirements of prior power system simulation, it is necessary to summarize the load characteristics at different time. Since the measured data are usually scarce, this paper proposes a new strategy to build the comprehensive load model. Firstly, single measured data are used to build the load model at every sampling time. In order to reduce the dispersibility of the parameters, only the parameters of high sensitivity are identified. And then the probability density functions of load model parameters are estimated based on the information spread principle. The correctness of the probability density functions is verified using K-S testing method. Finally, the parameters of the comprehensive load model are evaluated according to the probability density functions. This method can expand the limited measured data information. The general characteristics of the model parameters are estimated by deep analysis of the sampling information. A case study is presented to verify that the comprehensive load model established by the proposed mathods has good generalization ability.