Computational Game Theory Applied in Electricity Market - Games Played with Heterogeneous Beliefs

Name: Zhifeng Qiu

Prof. dr. ir. Geert Deconinck 

In the context of liberalized markets, market outcomes generally resultfrom the strategic interactions of all market players. Generationcompany (Genco), as the distributed players, build their subjectivedemand evaluations (SDFs) about market for optimal bidding purpose. Thepicture of a real electricity market game in Genco#s eye is #playing isbelieving#. Therefore, a question naturally comes to the table: howthose SDFs with the heterogeneous manner impact individual player#sdecision and game results. To answer this question, this thesis relaxesa conventional assumption, commonly used in the classical oligopolisticequilibrium model, that a correct and uniform demand knowledge is sharedby all Gencos. The investigations are carried out from threeperspectives:Firstly, the impact of SDFs is analyzed in a dynamic Cournot game and atransmission constrained supply function bidding market gamerespectively. The economic value of perfect information in the two gamesettings is discussed.Secondly, to alleviate the system oscillations caused by SDFs in theconventional conjectural variation (CV)-based learning bidding method, adata filter is designed to make bidding system stable.Finally, the stochastic learning technique, i.e. agent-basedreinforcement learning, is employed to answer the research questionsfrom another point of view. The comparisons are made between theequilibrium-oriented (EO) approach mentioned above and the multi-agentlearning (MAL) approach.The findings from this thesis provide insights on several interestingissues aroused by assuming players# heterogeneous beliefs incomputational game theory under the background of electricity market.The results can provide important guidelines in multiple perspectives,e.g. the effective design of a dynamic bidding learning processconsidering players# heterogeneous demand knowledge; the policyimplications of certain bidding behaviors caused by the imperfect demandevaluations, etc.

https://lirias.kuleuven.be/handle/123456789/410996

•   Prof. dr. ir. Geert Deconinck (promotor)
•   Prof. dr. ir. Hendrik Van Brussel (voorzitter/chairman)
•   Prof. dr. ir. Joseph Vandewalle (secretaris/secretary)
•   Prof. dr. ir. Johan Driesen
•   De heer Leonardo Meeus
•   Prof. dr. Ettore Bompard , University Politecnico di Torino