Energy Optimization Management Of Integrated Energy System Considering Flexible Resources

With the increasing complexity of the current energy system operation,integrated energy,as an advanced energy technology with more flexible resources and greater regulatory potential,is an important means to deal with the uncertainty and volatility of the system operation process and ensure the safe and stable operation of the system.However,the complex characteristics of integrated energy system operation and the introduction of flexible resources put forward higher requirements for system energy management,and the traditional solution algorithms have gradually been unable to meet the current trend of more complex integrated energy system model and more operational subjects.Therefore,based on reinforcement learning,this paper carries out a series of research work on energy optimization management of integrated energy system considering flexible resource and related algorithms.The main work is as follows:(1)An operation optimization method of park integrated energy system considering flexible resources is studied.Firstly,an integrated energy system model with flexible resources is established from both sides of the system source load.Secondly,taking operation economy as the objective function,an optimization model of park integrated energy system considering flexible resources is established.Aiming at the above model,this paper proposes a solution algorithm based on twin delayed deep deterministic policy gradient(TD3),and proposes a noise dynamic balance improvement mechanism based on the problem characteristics,which improves the performance of the algorithm.The introduction of flexible resources improves the operating economy of the system and promotes the system’s consumption of renewable energy.Compared with the traditional algorithm,the proposed algorithm has a faster solution rate and better adaptability to uncertain environments on the premise of ensuring the solution accuracy.(2)An energy management method of area ntegrated energy system cluster considering energy complementary is studied.Firstly,based on the characteristics of system cluster operation,a distributed scheduling framework of coordination between area energy coordination center and subsystems in the area is proposed.Secondly,according to the operation characteristics of area energy cluster,the energy interaction mechanism and energy transaction pricing strategy are proposed,and an energy management model aiming at economic optimization is established on this basis.Aiming at the above model,this paper proposes a solution algorithm based on multi-agent deep deterministic policy gradient(MADDPG),and proposes an initial state space compression mechanism to improve the stability of the algorithm.The proposed area energy complementarity takes full account of the operation differences of subsystems in the cluster,improves the area energy self-sufficiency level while improving the system operation economy,which is beneficial to the stable operation of the area energy cluster.The introduction of multi-agent reinforcement learning algorithm overcomes the shortcomings of traditional distributed optimization algorithm,such as slow solution time and poor convergence performance,and realizes the rapid solution of multi-agent distributed problem.(3)A hierarchical regulation method of integrated energy cluster considering demand response is studied.Firstly,in the context of demand response market,this paper analyzes the corresponding market responsibilities of integrated energy clusters,and proposes a bidding mechanism for clusters to participate in the real-time demand response market;Secondly,starting from the different interest demands of each subject,and taking the minimization of the cluster’s own operating cost and the maximization of social welfare as the objective function,an "area cluster-main network" hierarchical regulation model considering demand response is proposed.This paper proposes a solution method based on reinforcement learning algorithm,and introduces multi-process mechanism and prioritized experience replay mechanism to improve the performance of the algorithm,so as to ensure the adaptability of the algorithm to high-latitude,large-scale problems.The introduction of real-time demand response not only alleviates system congestion and improves system security and stability,but also takes into account the interests of all parties in the hierarchical model.It not only reduces the operating cost of integrated energy cluster,but also reduces the unit cost of main network generator set and improves system energy efficiency.The two-layer problem solving method based on reinforcement learning not only makes up for the shortcomings of the traditional algorithm and improves the problem solving speed,but also can effectively carry out system scheduling and guide the cluster to participate in the real-time demand response market reasonably in the case of the continuous expansion of the system scale.