One of the main reasons behind power market design flaws is the lack of comprehensive quantitative models that take into account important market features. This deficiency coupled with an absence of computationally viable methods for solving stochastic games (e.g., obtaining Nash Equilibrium) that can be used to model market behavior has resulted in lower economic reliability for power markets. The objectives of this research are: 1) to develop a stochastic game theoretic modeling framework consisting of features like energy bidding and price settlement policies, FTR auction, demand elasticity and other evolving market rules, 2) to formulate a reinforcement learning (RL) based solution methodology and develop its convergence analysis, 3) to conduct testing and benchmarking with power market industry data, and 4) to develop plans to train students and also offer continuing education for power market professionals. The proposed model and its solution methodology would allow us to identify the equilibrium state of a market, which is characterized by steady state bidding strategies of the market participants and the resulting energy prices. Thus the economic reliability of alternative market designs can be assessed before implementation. The RL based solution methodology can also be extended to problems in other sectors of economy. The broader impacts would include 1) enhanced economic reliability of power markets through dissemination of knowledge gained from the research, 2) increased competition due to better market design resulting in process innovation and price reduction, and 3) development of a diverse group of professionals capable of designing and analyze deregulated power markets.
