Abstract: Maintenance and spare parts inventory management are two fundamental sequential procedures in offshore wind farm operation and maintenance（OM）. To improve the OM efficiency of offshore wind turbines（OWTs）in the design life cycle,a joint optimization strategy of condition-based maintenance and spare parts inventory control is proposed. Firstly,the OWT is constructed as a series system of subsystems including blades,gearbox,electrical,yaw,wheel hub,braking,transmission chain,generator,the deterioration process of each subsystem is modeled as a multi-state Markov process,and the coupling model of maintenance and spare parts inventory is established. Secondly,the representation and update method of subsystems deterioration,spare parts inventory are formulated as the framework of deep reinforcement learning. Through the iterative training of the deep networks,the optimal maintenance and spare parts ordering decision making are obtained. Finally,taking an OWT of a practical offshore wind farm as an example,the effectiveness of the proposed joint optimization method is verified. The impacts of the exploration rate and the accessibility of OWTs on the OM cost are further discussed.