Abstract: To solve the increasing random disturbance problems faced by power grid under the integrated energy system, a deep reinforcement learning algorithm with a multiple incentive actor-critic（MIAC） was proposed for the control strategy of automatic generation control. Considering the optimization target in the control process, the AC strategy with the incentive-inspired update mechanism was utilized to improve the quality of strategy mining and efficiency of experience exploration. Simultaneously, an action value update method with relatively minimizing the Q-function was utilized to reduce the deviation of optimization, and guide the strategic goal achieve the equilibrium of exploration and utilization, finally the optimal coordinated control of automatic generation control was obtained. Through the simulation of the improved IEEE standard two-area power system model and the integrated energy system model, the results show that MIAC has great dynamic control performance and migration generalization ability, can realize rapid adaptation and stable optimization to the complex grid environment, and then it can effectively solve the random disturbance problems witnessed by the integrated energy system.