Despite the advantage of stochastic response surface method（SRSM） in overcoming the drawback of repeated simulation required by the conventional method by quickly estimating response uncertainty with only small amounts of simulation， it is worth further investigation for application in power system simulation because of the systems complexity and characteristics of multi-dimensionality and strong nonlinearity. In order to enable the SRSM to find even wider application， by referring to the Monte Carlo simulation， an in-depth analysis is made of such problems as the selection of collocation points， parameter distribution in obeisance to empirical probability or the uncertainty when random inputs are correlated. In addition， by consulting the trajectory sensitivity method， a parameter uncertainty analyzing strategy capable of dramatic reduction in the number of uncertain inputs is proposed. The efficiency and validity of the SRSM are shown by case study results.