Optimal preventive policies for parallel systems using Markov decision process: application to an offshore power plant

•Application of the Markov decision process for maintenance backlog management.•Optimal prevention levels for varying operating scenarios (i.e., combinations of failure and repair rates).•Integrated O&M policies for offshore parallel machinery systems•Capacity utilization rate optimization as a practical alternative to cost minimization.•Optimal stationary policies can be non-intuitive even in a 4-unit parallel system.