Large-Scale Generator Maintenance Scheduling using Simulated Evolution Method

Abstract

This paper addresses a new effective method for the generator maintenance scheduling. As the scheduling is a 0-1 integer problem, no excellent approach to find optimal solution has been presented for a large scale problem because of the combination explosion. The proposed method combines a simulated evolu-tion and a branch and bound method. The idea of the simulated evolution is based on an analogy of the natural competition and selection process. The proposed method is not guaranteed to find an optimal solution. It has a capability to find sub-optimal ones with less CPU time by means of an effective search, however. It features (1) the high-speed search of a feasible solution by using depth priority search, (2) the effective improvement of the solution by the re-construction of the branching tree, and (3) requirement less memory size and CPU time. The proposed method was applied to a small scale 10 generator system and large scale 38 generator system. With comparing to branch and bound method, the ability of finding the optimum solution and of the effective search were verified.