A hybrid optimal closed-loop grid scheduling algorithm with carbon emission restriction was proposed to solve the problems of low economic scheduling efficiency of power grid. While considering the combined dispatching of thermal power generating units and multi-distributed power generation aggregates, with the actual demand response of the grid and carbon emission limits as constraints, a grid economic dispatching model was established. By transforming the economic dispatching problem into the optimal control of the hybrid system, the equivalent consistency between the economic dispatching of the power grid and the optimal control sequence of the hybrid system was established. The optimal cost function satisfying the Bellman equation was obtained by deriving the hybrid system, and then, the optimal value was approximated by neural network, which the optimal closed-loop scheduling sequence was obtained. Simulation experiments verify the effectiveness of the proposed closed-loop scheduling algorithm, which can still obtain the optimal scheduling sequence under different initial conditions and external disturbances. Simultaneously, the simulation results confirm that reducing carbon emission limits or increasing carbon emissions trading prices can help reduce the total carbon emissions of power grid system.