Abstract
Because of the accidents, the outages of electrical power service occur inevitably. How to restore the out-of-service zones after a fault occurs is called the power restoration problem. In this thesis, we treat the electrical power grid as a smart grid so that devices in the smart grid can be treated as agents. We build a stochastic negotiation protocol based on the decision theory. This stochastic negotiation protocol can avoid trapping into local optimal in sequential decision making. Agents in the smart grid can negotiate with each other using the stochastic negotiation protocol to solve the power restoration problem. We show that our system is workable and capable to solve subtle cases. We also show that the performance of our system is better than a committee-based negotiation multi-agent system. Under some conditions, the power restoration problem is emergent and has a time limit. For this reason, we finally analyze the relationship between the restoration rate and the execution time so that we may find a more proper time limit.