A learning method to evaluate a generation company''s bidding strategy in the electricity market

Abstract

In the electricity market, generation companies (GenCos) are usually faced with the problem of choosing a better bidding strategy. They often have to evaluate each possible strategy according to its potential reward. In a competitive market environment, the electricity price is stochastic and volatile, and the GenCo's mixed strategies also make the problem more complicated. In this paper, we model the market price with a Markov regime-switching model and propose the temporal difference learning method in the Markov decision process to approximate the expected reward over an infinite horizon. The simulations based on this method have achieved the evaluation of 2 mixed strategies. The results show the difference of the expected rewards between the strategies, which could be important evidence for choosing a better strategy.