Speaker
Description
Distributed energy resources (DERs) and energy communities (ECs) are becoming increasingly popular as a way to decarbonize the housing sector. With DERs, traditional end-consumers become prosumers, who both consume and produce electricity. ECs aim at balancing their own energy demand and generation, by engaging in peer-to-peer (P2P) trading. After having their daily demand met, prosumers can sell their energy surplus to other community members who have been unable to fulfill their energy needs.
The P2P trading system challenges the traditional centralized electricity system structure and market. Suddenly, prosumers start playing a more active role in the energy market, while having no knowledge on how to optimally trade. Reinforcement learning and, in particular, multi-agent reinforcement learning (MARL), can be used to support the decision-making process of the market participants, optimizing the trading process and guaranteeing the maximum possible rewards (i.e. profits/savings). The choice of market clearing mechanism, as well as physical characteristics of the grid and the physics of the battery storage unit affect the financial aspects of P2P trading.
This project’s main aim is to understand how profit in P2P trading is affected by the degradation in the battery energy storage units. A MARL algorithm will be used to find the optimal policy under different market clearing mechanisms, with and without battery degradation.
