There are problems in existing PV power generation system maintenance strategy research, since it usually only considers the deterioration state and leads to unreasonable maintenance decisions. Thus this paper proposes a PV power generation system visual maintenance model and optimal overhaul strategy that takes into account the deterioration state and random faults. First, based on the Markov process in the discrete state, the influence of random faults on the maintenance strategy is introduced into the PV power generation system visual maintenance modeling. Then a visual maintenance model of the system taking into account the deterioration state and random fault state is established, the state of the system is detected periodically. Incomplete maintenance is carried out when the state of the system is more than the threshold value of the preventive maintenance; the minimum maintenance is carried out when there is a random fault; and replacement is carried out when it reaches the fault state. Then, an improved viscous bacteria optimization algorithm is used to find the optimal detection period and preventive maintenance threshold of the PV power generation system with the objective of the minimum total cost rate of the system for long-term stable operation. Finally, taking a PV power generation system as an example, the effectiveness and feasibility of the proposed model and strategy are verified through sensitivity analysis of the model parameters and a comparative study of the traditional maintenance model without considering the random faults.