Research On Generator’s Balanced Bidding Strategy In Carbon And Electricity Coupling Market

Carbon emission reduction actions have become a social consensus with the goals of "carbon peak" and "carbon neutrality" in China.Various industries have taken various emission reduction measures to respond the goals of "carbon peak" and "carbon neutrality".In order to optimize resource allocation and reasonably guide carbon emission enterprises to participate in carbon emission reduction actions in an orderly manner,China officially launched the carbon trading market in July 2021.Reducing carbon emissions in the power industry,a large carbon emitter,is the only way to achieve the goals of "carbon peak" and "carbon neutrality".At the same time,the electricity market will guide all entities to engage in healthy competition and enhance their market vitality with the promotion of Chinese electric power reform.Because dual attributes of electricity and carbon is contained in power industry,its associated carbon trading market and electricity market have also formed a more closely connected electricity and carbon coupling market.In this market,generators get the winning bids through strategic quotations in order to obtain their maximum revenue;on the other hand,generators decide their own carbon trading in the market based on the on grid electricity,and their bidding strategies in the market will be affected by both the electricity and carbon markets.In this thesis,we propose to study the equilibrium bidding strategy of generators in the electricity and carbon coupling market.Considering the limitations of traditional multi-intelligent reinforcement learning for solving high-dimensional and continuous models,this study combines the Prioritized Experience Replay(PER)mechanism and the Multi-Agent Deep Deterministic Policy Gradient(MADDPG)algorithm to form the PER-MADDPG algorithm.which is used to solve the model of generator’s balanced bidding strategy in carbon and electricity coupling market.The research contents of this thesis include:First,it is difficult to effectively control carbon emissions and increase the incentive of generators to reduce emissions in view of the same incentives and penalties for purchasing carbon quota in a typical stepwise carbon trading reward and punishment model.This research project proposes an improved stepwise carbon trading reward and punishment model based on the domestic carbon market trading mechanism,differentiating the length of the positive and negative intervals and the carbon trading price to improve the binding force of the market for carbon emissions,and establishes an model of generator’s balanced bidding strategy in carbon and electricity coupling market based on the coupling relationship between the electric power market and the carbon market.Second,the lack of environmental information in the electricity and carbon coupling market makes it difficult for generators to formulate bidding strategies through environmental information.Because of the exploration capability of reinforcement learning for unknown environments and the communication and coordination capability of multi-agent systems,this research project using multi-agent reinforcement learning algorithm that combines reinforcement learning and multi-agent systems.Considering that the decision variables of the proposed model are high-dimensional and continuous,the existing MADDPG algorithm has the problems of low exploration efficiency and slow convergence in solving this model,the problems of MADDPG for solving this model can be solved by introducing prioritized experience replay mechanism in the experience extraction process of the algorithm to improve the efficiency of experience utilization.Finally,the IEEE30 node system is used to simulate the model of generator’s balanced bidding strategy in carbon and electricity coupling market.Three scenarios are set for comparison and analysis: no carbon trading mechanism is considered,a typical stepwise carbon trading mechanism is considered,and an improved stepwise carbon trading mechanism is considered.The simulation results show that improved stepwise carbon trading model can effectively reduce carbon emissions and increase the on-grid electricity of low carbon generators.To verify the advantages and disadvantages between the proposed PER-MADDPG and MADDPG algorithms,the two algorithms are used to solve the proposed model and make a comparative analysis,and the results show that the proposed PER-MADDPG algorithm can significantly improve the exploration efficiency and accelerate the convergence speed.