The identified values of parameters in power system models often differ owing to different optimization algorithms， convergence intervals and iterative numbers and so on， making it difficult to determine and select a rational parameter set. A new method of dynamic consistency test is proposed based on the stochastic response surface method（SRSM） as well as the relative consistency test criteria. Since SRSM can solve the limitations in the conventional methods of quantitative uncertainty analysis， such as immense computation time-consumption and large model runs， the uncertainty of responses arising from uncertainty of models and parameters can be quickly calculated. Therefore the expectation value and confidence interval can be obtained， providing useful information for checking parameters validity and selecting a rational parameter set. The method proposed is verified by load measuring data in a certain power grid.