The parametric uncertainty is a basic type of epistemic uncertainty in probabilistic power system risk assessment（PRA）. The second order Taylor expansion for PRA index at the mean values of component reliability parameter is derived， through which the expected risk indices can be analytically calculated after multicomponent reliability parameters have been changed. Furthermore， using the kernel density estimation theory and sampling technique， the probability density distributions of expected risk indices under the uncertainty of reliability parameters can be obtained. It provides visual specification of quantitative uncertainty analysis. Finally， the RBTS and IEEE-RTS 79 power systems are used to validate the proposed approach.