Research On Optimal Dispatching Strategy Of Integrated Energy System Considering Multi-agent Benefits

The dual pressures of the energy crisis and environmental deterioration have prompted mankind to accelerate the process of energy transformation.Traditional power systems,thermal systems,and natural gas systems are individually planned,designed,and operated independently,which ignores the coupling relationship between different energy types and greatly limits the flexibility of energy systems.Through the coupling of different energy types,it realizes the cascade utilization and efficient integration of energy,with the advantages of clean,low-carbon,safe,reliable,flexible and efficient,which conforms to the future energy development trend.In the context of a new round of scientific and technological revolution,the integrated energy system is expected to break through the traditional energy supply system and industry barriers,become a specific way to meet the multiple supply system,and has become the focus of the energy transformation and development in the world.This thesis mainly takes the integrated energy system as the research subject,explores the multi-equilibrium optimal scheduling and its supporting technology,furthermore,taking the multi-load scenario generation technology,the integrated demand response strategy,the optimal scheduling method as major object.The specific work content is as follows:1)Due to the limitation of data privacy,data security and acquisition cost,large-scale acquisition of load data is still a big challenge.A multi-load scenario generation method based on generative adversarial network(GAN)is proposed.GAN consists of two independent networks: a generator and a discriminator.They are trained with each other until they reach a balance,so that the generator learns the mapping relationship between the noise distribution and the scene set.The actual load data is used to test the proposed method,and the results show that the proposed method can generate realistic load data in different modes without losing diversity.2)With the deepening of marketization,the status of users as serviced subjects will significantly improve.Aiming at this trend,this paper proposes a bilevel optimization model that considers the interests of both parties.First of all,the demand response characteristics and the comprehensive demand response strategies of energy operators and energy consumers are analyzed.On this basis,a bilevel optimal scheduling model of integrated energy system considering the integrated demand response is proposed: in the up-level optimization,the goal is to maximize the profit of integrated energy operators;in the low-level optimization,the end user adjusts the energy consumption strategy according to the response compensation signal given by the operator to balance energy purchase expenditure and energy experience.Combining KKT conditions and Big-M method,the bilevel optimization problem is transformed into a single-layer mixed integer lineal problem for solving.As a result,the proposed method can effectively coordinate and balance the interests of both parties.3)With climate change,it is expected that the frequency,intensity and duration of natural disasters and extreme weather events will continue to increase in the future,which will pose a great threat to the safe operation of integrated energy systems.Therefore,it is of great significance to improve the survivability of integrated energy system in islanded operation mode by optimizing local micro-source energy supply and demand response after high impact low probability events.In this paper,the stochastic dynamic optimization problem of integrated energy system under islanded operation mode is described as a Markov decision process,and Q-learning algorithm is introduced to solve this complex problem.In order to overcome the disadvantages of Q-learning algorithm,two improvements have been made to the typical Qlearning: the Q table initialization method is improved and the upper bound convergence algorithm is adopted for the action selection.The simulation results show that the Q learning algorithm ensures better convergence while solving the problem,and the improved initialization method and the upper bound convergence algorithm can significantly improve the computational efficiency and make the results converge to a better solution.Moreover,compared with the conventional mixed integer linear programming model,Q-learning algorithm achieves better optimization results.