The increasing amount of Distributed Energy Resources (DERs), which have recently been integrated in power systems, and the more proactive role of consumers have transformed the classical centralized power system operation by introducing more uncertainty and decentralization in the decisions. Following this trend, electricity markets are starting to restructure, from a centralized market design in which all the operations were managed by a central (global) market operator, modeled as a classical constrained optimization problem, to more decentralized designs involving local energy communities which can trade energy by the intermediate of the global market operator or in a peer-to-peer setting. This latter design requires the introduction of game-theoretical approaches to model the complex interactions between the agents and provide guidelines regarding decentralized system evolution.
The information (historics of the agents’ past actions, private information captured through types, etc.) that each agent can access is capital, as it impacts the agent’s strategy definition and the dynamic evolution of the underlying game. Various learning mechanisms can be implemented by the agents, in a distributed fashion (Bayesian inference, regret minimization, etc.). Questions on convergence of these distributed learning algorithms, as well as the capability for these algorithms to efficiently approximate equilibrium (Generalized Nash in case of shared resource, correlated equilibrium in case Nature is introduced) resulting from such complex underlying games, raise tricky questions.
In this PhD work we propose to:
· Develop and analyse a game-theoretic model for peer-to-peer energy trading where peers may have various level of information, the performance of which will be compared to centralized and hierarchical market designs. These latter will be used as benchmarks, and formulated using control-theoretic and Stackelberg game approaches in a context of sequential decision making. Endogenous risk impact on the market behavior will also be studied. To perform the comparison of the different market designs, performance indicators will be considered, reflecting various points of view on the system behavior (social welfare, efficiency loss relying on the Price of Anarchy, etc.)
· Provide methodological approaches for the design of a smart contract on top of a decentralized demand and supply matching platform, guaranteeing the automatic detection of a consensus reaching. Careful analysis of distributed trust-based communication mechanisms in a peer-to-peer market design setting should be provided.
· Analyse the robustness of the smart contract in case of malicious behavior of some agents, attacks or faults. The detection of these latter in a decentralized energy trading system will also be considered as a methodological approach.
Background: the candidate should be pursuing or hold a Master 2 and have a solid background in applied mathematics, computer sciences, economics and interest for energy questions.
Practicalities and organisation: The PhD work will last 3 years, with a possible start in October 2019. The candidate is expected to present his/her workplan in front of a jury in Belgium, before the official attribution of the grant.
The PhD candidate will be registered at a doctoral school from ENS Ulm, Paris, France. The PhD is a collaboration between INRIA Paris, France and VITO, Belgium. The PhD grant will be provided by VITO/EnergyVille, Belgium. The candidate should be prepared to allocate his/her time between Paris, France and Genk, Belgium, with some possible stays abroad (conferences, research stays for possible collaboration with DTU, Copenhagen , Denmark, etc.).
Contact: Please provide CV and motivation letter to Dr Ana Busic (ana.busic@inria .fr) and Dr Hélène Le Cadre (email@example.com)