Optimization algorithm for fault diagnosis of electrical power system is to describe fault diagnosis as optimization problem and then solve that by optimization method. A new approach to fault diagnosis based upon developed Tabu Search algorithm is proposed in the paper. Based on set covering theory, alarm processing problem is first expressed as 0-1 integer programming problem, a new criterion on describing alarm processing problem is proposed to ensure global optimal solution and extremely satisfy the need of network celerity. Eventually, a typical 152 node systemic case is served to demonstrate probability of reliability and celerity quality of computed results, the test results suggest that evolutionary TS based method is more efficient than existing optimization algorithm.