Abstract:With the promotion of the "double carbon" policy, the construction of an integrated energy system with power grid as the core and complementary and coordinated supply of electricity, heat and gas is an important means to realize the "double carbon". Integrated energy system where electricity, heat, and gas operate coordinately is rapidly developing. The economic and stability issues that exist in the operation of the system need to be addressed. This paper is committed to using machine learning algorithms to solve the economic issues in operation of integrated energy system while taking stability into account. Firstly, this paper proposes operation model of integrated energy system that includes energy storage devices and P2G. In the proposed model, considering the general framework of optimal operation, optimization objective and constraints are included. Power balance, output limitations, ramp rate limitations, and capacity limitations of each device are considered in the constraints. Then, this paper designs a solution strategy for the optimal operation of integrated energy system based on DRL . The DRL algorithm combines advantages of reinforcement learning on strategy selection and advantages of deep learning on environment simulation. In the DRL algorithm design, five modules are considered in detail: action space, return function, state space, DRL algorithm, and DRL network. Finally, four cases are designed in this paper under the typical daily operation conditions of integrated energy system including electricity, heat, and gas devices. The simulation verifies that the combination of the electric, heat and gas devices in an energy network can effectively achieve multi-energy complementarity and reduce energy costs. The DRL method designed in this paper can effectively solve the optimization operation problem of integrated energy system including electricity, heat, and gas devices.