The invention relates to a prediction method for reliability of a power generation system and a power transmission system. The method is characterized in that the method comprises: step one, first-stage sampling is carried out by using a Monte Carlo simulation method; step two, whether an overall sampling flow system index convergence condition is met is determined; step three, if so, a reliability prediction result is calculated according to obtained system indexes LOLP and EPNS; if not, a step four is carried out; and step four, whether a kth-stage sampling flow system index convergence condition is met is determined; if so, sampling at a (k+1)th stage is carried out and then whether an overall sampling flow system index convergence condition is met is determined until sampling carried out at a plurality of stages meets the overall sampling flow system index convergence condition. According to the method, consumption of calculation resources for multi-stage important sampling is low; the total number of times for sampling is reduced; the calculation time is short; and the efficiency and accuracy are high.