The paper presents a Monte-carlo method for reliability evaluation of large scale composite generation transmission systems. The method, which possesses coputational rapidity and sufficient accuracy,is based on combining the dispersed sampling techique with the analysis approach and the optimization model of power not supplied. Because the amount of calculation increases only linearly with, the number of elements, it is specially sui- table for large scale systems. The Monte-carlo sampling draws more near the real situation than determinate state enumeration in simulating occurrence of power system outages. Also, the method can be easily generalized to involve multi-states of elements, multi-levels and stochastic distribution of loads. The results of IEEE-RTS, as example, are given and compared with ones obtained by using the other two methods.