Génération de politiques transactionnelles pour les agrégateurs de réponse à la demande

The increase in energy needs of the growing global population and the concern with its associated Greenhouse Gas emissions cause significant challenges for conventional power systems.In this regard, the smart grid concept is proposed as a key enabler for clean energy generation and efficient energy consumption.Under the smart grid paradigm, the emergence of transactive energy systems brings about a remarkable opportunity for realizing a modernized power grid through enhanced Energy Management Systems.Particularly, these new systems offer innovative Demand Response (DR) programs in order to improve energy efficiency and flexibility and facilitate renewable resources and energy storage integration.This is achieved by leveraging advanced metering infrastructure, two-way communication networks, and distributed control systems.The smart grid also frames a group of mechanisms for DR characterized by generating incentives or pricing policies in an adaptive and more real-time manner.Consumers can react by modifying their load profiles in order to minimize energy costs while maintaining comfort desires.On the grid side, the operator can manage system congestion and minimize operational costs by reducing the peak demand and deferring the construction of new power plants and power delivery systems.However, these DR programs face significant challenges in terms of modeling and management of decision-support information in dynamic and non-homogeneous environments.Indeed, the incomplete information on the dynamics of the behind-the-meter resources, the inherent issues of user data confidentiality, the potential failures in the communication channels, and the emergence of intelligent loads (including storage) create a complex and uncertain environment for the decision-making process.As a result, a new entity is emerging, the demand response aggregator.This aggregator acts as a mediator between consumers and the electricity market to explore the flexibility iv opportunities offered by the residential sector.This new entity will then seek to offer benefits to both parties (distributors and users) by exploiting the policies of demand response programs.The mentioned role translates into an interaction between players seeking to maximize their gains and thus ends up being framed by game theory.However, the various sources of uncertainty mentioned above considerably complicate the process of generating optimal policies.With this in mind, reinforcement learning methods emerge, offering the possibility of managing uncertainty through a trial-and-error process.In other words, this approach takes advantage of the interactions between the various players in the system, in order to achieve an optimized generation of transactive policies.This thesis proposes to develop an automated agent (meeting the needs of network managers) for generating optimized transactive policies through interactions in a residential environment.The proposed approach considers a transactive environment composed of rational residential agents and a demand response aggregator agent interacting in a game theoretic framework.The aggregator's adaptability and ability to handle uncertainty are considered through reinforcement learning techniques.The results demonstrate the effectiveness of the proposed method in managing residential consumption.The aggregator agent is able to offer economic incentives to users through the development of pricing policies while respecting users' privacy, in order to exploit the potential of residential flexibility.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.