Research On Optimization Operation Strategy Of Integrated Energy System Based On Deep Reinforcement Learning Algorithm

In recent years,China continues to promote the energy revolution and transformation,with the characteristics of renewable,clean and pollution-free renewable energy has been developed rapidly and widely used.Although renewable energy represented by wind power and photovoltaic can effectively promote the process of "carbon peaking and carbon neutralization",the randomness and volatility of renewable energy power generation reduce the security and stability margin of power system,restrict the flexible dispatching of power grid,and further affect the consumption capacity of renewable energy.The large-scale development and utilization of renewable energy promotes the interaction and coupling among various energy systems,which makes the related technologies of integrated energy system get attention and rapid development,and provides an opportunity to ensure the security,stability and flexible dispatching of power grid.Therefore,this paper introduces a comprehensive energy system including heating system and gas supply system.In order to ensure the smooth access of large-scale wind power to the power grid,the heating network and gas supply network are used as the energy storage of the power grid,improve the system’s ability to absorb renewable energy,and ensure the stability and economy of power grid operation.With the development of information technology and artificial intelligence technology,the algorithm based on deep reinforcement learning has gradually become a new optimization algorithm.It has advantages in solving the uncertainty problem of renewable energy,and can realize the control mode of "real-time calculation and real-time control".Aiming at the intelligent scheduling problem of integrated energy system based on deep reinforcement learning,the main research contents of this project include: 1(1)Research on renewable energy intelligent scheduling strategy of integrated electricity and heating energy system.Considering the economy in the large-scale wind power integrated energy system,the scheduling strategy takes 100% wind power consumption as the premise,takes the lowest operation cost as the optimization objective,takes the real-time conversion rate of wind power as the optimization object,and considers multiple uncertainties such as wind turbine output,user load and upper level electricity price,The proximal policy optimization algorithm is used to solve a real-time renewable energy scheduling strategy to minimize the operation cost of the electric thermal integrated energy system.(2)Research on Intelligent Scheduling of integrated electricity and natural-gas energy system for load shifting model.In the power gas integrated energy system with large-scale wind power,the inverse peak regulation characteristics of wind power are considered,and the time-space translation of wind power output and effective smoothing of net load curve are taken as multi optimization objectives.The real-time output of multi energy coupling elements is taken as optimization object,and the uncertainty of wind power output,user load and natural gas energy price is considered.An intelligent energy management strategy for optimizing the output of coupling components in real time is solved by using the deep deterministic policy gradient algorithm.(3)Research on multi-objective energy scheduling optimization of integrated electricity,heating and natural-gas energy system.The scheduling strategy aims to improve the economy of system operation and power supply reliability.The real-time output of multi energy coupling components is taken as the research object.Considering the uncertainty of fan output,user load and battery real-time energy storage,the soft actor-critic algorithm with maximum entropy mechanism is adopted to realize the real-time scheduling of coupling components.In order to reduce the operating cost of the system and improve the reliability of power supply(4)For each research,after theoretical analysis,a detailed simulation test is carried out.Taking the actual output data of a regional wind power station as an example,the feasibility and universality of the model and method are verified.Compared with the traditional optimization algorithm,the energy scheduling strategy based on deep reinforcement learning is proved to be superior.