This paper describes the optimization of multiple power plant maintenance based on the quantitative risk evaluation. In the plant maintenance decision, an optimal maintenance strategy is decided so that the total maintenance cost and possible default cost through the lifetime of the plants are minimized. In this calculation, the maintenance strategy should satisfy the plant operation constraints such as acceptable maintenance budget and maximum probability of plant default in every year. This problem becomes a nonlinear optimization problem defined on a high dimensional space, to which metaheuristic search method can be applied effectively. In this paper, a memetic algorithm that combines Genetic algorithm and Nelder-Mead simplex method is proposed. Through the comparison with an ad hoc method that has been used conventionally, we show that the proposed metaheuristic algorithm has an effective solution power for real-world plant maintenance problems.