Because of the complex composition of distributed generators and various operational modes in the microgrid, the voltage restoration control of an islanded microgrid will be affected by uncertain disturbances. In this paper, a distributed secondary microgrid voltage optimization control method based on cooperative reinforcement learning algorithm applied to the secondary voltage restore control problem of the microgrid under uncertain disturbances is proposed. This can realize the voltage regulation control of an islanded microgrid. First, a distributed consensus cooperative voltage control algorithm of the islanded microgrid is constructed, and a stability judgement method based on the Lyapunov function is established. Second, from the relationship between controller performance and gain parameters, the upper bound of the voltage controller gain of the islanded microgrid is derived. This can be used to constrain the action space of the reinforcement learning agent. Third, the reinforcement learning algorithm is applied to optimize the secondary voltage controller gain parameters, and the corresponding reinforcement learning state space and cooperative global reward function are presented. Finally, the effectiveness and adaptability of the proposed control are verified by simulation experiments on Matlab/Simulink. This work is supported by the Natural Science Foundation of Anhui Province (No. 2108085QE237).