An improved importance sampling method with splitting optimal multiplier is presented. Through a heuristic identification of the components' importance degree to system reliability, the presented technique can significantly accelerate the convergence speed of the Monte-Carlo simulation. The importance identification is realized using the product of a component's capacity and its forced outage rate. Before the Monte-Carlo simulation, the importance identification has been carried out and the optimal multipliers for important components have been achieved through golden section searching approach. Then the importance sampling is carried out among these important components while the traditional Monte-Carlo simulation is done among the remaining components. The convergence speed of the presented method is verified using the IEEE-RTS 79 test system.This work is supported by National Natural Science Foundation of China(No. 50607021).