Maintenance is regarded as one of the effective methods to enhance the reliability of nuclear power plant (NPP) system. The existing research works on NPP maintenance optimization problem mainly focus on the preventive maintenance (PM) scheduling, with the common assumption that maintenance activities are performed on the periodic scheduled time. Though the straightforward asset management method is brought by PM scheduling, its inflexibility could result to either unnecessary maintenance activities or undiscovered failures. In this research, a nonperiodic maintenance scheduling for NPP subsystem is considered. Effective age is used to determine the performed maintenance activities. To deal with this problem, a Markov decision process is utilized with the value function of minimizing the total maintenance cost for each state. The proposed model is applied to an NPP feed water subsystem. Via the dynamic programming, the optimal action and minimized total cost of each possible state over the short scheduling horizon is obtained.