In recent years, the electric utility industry worldwide has been facing pressure to be deregulated. Along with it, risk of blackout in large area will also increase. Actually, it is still vivid in our memory that the northeastern US and southern Canada suffered the worst blackout in history. Consequently, a method to find the optimal solution rapidly is needed all the more. In this paper, we propose a new multi-agent method for a bulk power system restoration. In order to demonstrate the capability of the proposed multi-agent system, it has been applied to a model bulk power system, which consists of three local areas including twelve generating units and twelve loads, and three remote areas with twelve loads. A large number of simulations are carried out on this model network with changing conditions. The simulation results show that the proposed multi-agent approach is effective and promising.