The Application Of Reinforcement Learning In Electricity Market

With the establishing of electricity market, the power company's competition has increasingly attracted public attention. In the electricity market environment, the power companies'task change from the traditionally completing the task of generating to bidding the power load to obtain maximum profits, and its essential purpose and behavior has changed. The power companies is prior to declare the price of electricity and other information services. With power company's generating capacity , and the spirit of "fair, just and open ", the Business Information Cente arranges for the companies'generating capacity. How to use the information reasonablly registering to obtain maximum profits, is the company's first quotations to solve .In analysis of the power companies how to obtain the maximum profit, an effective tool is game theory. Classical game theory's basic assumptions is perfect rationality, but, in fact, when returning the electricity prices, the manufacturers could not fully grasp of market information, and it is impossible to put all the reality various factors into account. Therefore, the model based on the bounded rationality is more realistic, it allows policy-makers can be streamlined, misunderstanding, lack of capacity, miscalculations, forget and estimating depending on the surface, unrelated to the issue of what constitutes the details of things. The game mode based on bounded rationality is more attention to explain how people learn , adapt or evolution to balance.This paper applicates participants'bounded rationality into the power companies bidding strategy study. Before the power company chooses a decision, they hope to be granted a profit and use this profit comparing with the actual profits to constantly modify the strategy. Due to incomplete information, power company's aspiration profit level is empirical and subjective, when many individuals use the aspiration profit to study, the interaction of power company's subjective desires will bring what impact to the macroscopic market .Which is the theme of this paper. For these, the paper puts reinforcement learning approach based on aspiration profit , which can better reflect their own bounded aspiration profit approximation to real level. The paper simulats the reinforcement learning algorithm which based on the aspiration. Along with the simulation, generator keeps correcting their strategies, and at last the strategy will convergence to a balanced position, we may experience a variety of market equilibrium.