This paper describes a prototype expert system for aiding dispatchers to deal with faults occurring in a bulk power system. Compared with previous expert systems for power systems restoration, this one has more powerful functions for the recognition of post-fault status. The breadth-first and depth-first search techniques of artificial intelligence (AI) have been used to detect power system separation, determine network topology, and look for synchronization paths for power system reintegration. While previous restoration expert systems only relieve overloads in radial power systems, this expert system can relieve overloads in loop power systems. An optimal power flow (OPF) program for overload relief has been embedded in the expert system. To solve the OPF efficiently both in terms of computer memory space and speed, a new decoupled approach for overload relief OPF has been presented. This hybrid expert system has been tested off-line, and will be used in the North East China Electric Network.