Research On Operation Strategy Optimization Of Power Generators In Power Market Based On Multi-Agent

Under the dual background of the high penetration rate of new energy and the further liberalization of the power market in the future,the optimization model of the operation strategy of power generation entities in the power market is constructed to explore how power generation entities in the market such as wind power,photovoltaic,thermal power and new energy storage participate in the coordinated clearing of the spot power energy market and the auxiliary service market.It is of great significance to find the bidding strategy when the system cost is minimized and the profit of each market power generation entity is maximized,which is to promote the construction of power market trading mechanism adapting to the development of new energy,and further play the role of market mechanism in promoting the absorption of new energyBased on the multi-agent modeling framework,this paper establishes the operation strategy optimization model of power generators in spot market.The environment module is a joint clearing stochastic optimization model of spot power and deep peaking market,including a unit combination model considering safety constraints,an economic scheduling model considering safety constraints and a node price model.The agent module is a bidding decision-making model for power generation entities in each market.The behavior of the agent seeking the best offer is described as a semi-Markov decision-making process.A multiagent reinforcement learning algorithm composed of two-layer reinforcement learning algorithm and hierarchical structure of strategies is constructed to solve the optimal strategy behavior of the unit,including the upper macro-strategy algorithm and the lower micro-action execution algorithm.At the same time,the original concept of environment in the algorithm is rewritten,and the random optimization model of power market clearing is taken as the environment in the algorithm.In the strategy optimization model,the calculation result of the power market clearing optimization model is taken as the data input of the agent learning,and the bidding decision output of the agent module is taken as the bidding data input of the market clearing model.The environment module and the agent module continue to interact and iterate,and the unit will learn the optimal bidding strategy.Firstly,the HRP-38 node system is taken as an example to simulate the market,and the trading strategy and income of thermal power units are studied under different new energy penetration,different prediction errors,and with or without energy storage.The results verify the validity of the model and clarify the market trading strategy of thermal power enterprises under the high proportion of new energy penetration.The results show that with the increase of new energy penetration,some thermal power units are gradually withdrawn from the market,but some thermal power units with unique locations and cost advantages are still competitive.With the increase of prediction error,the bidding strategy of large capacity thermal power units tends to be conservative,while the bidding strategy of small capacity units is opposite.Thermal power units have implicit collusion tendency in all kinds of situations.Secondly,the above optimization model is applied to carry out short-term simulation and strategy optimization for Western Inner Mongolia power spot market,and carry out research on new energy participation in spot market quotation.Finally,the paper puts forward some suggestions on the power spot market model and the operation strategy of each market power generation entity.