Decentralized Electric Microgrid Optimization by Reinforcement learning

The goal of this project is to improve the management of electric community microgrids. In contrast to traditional power grids with large centralized power plants that provide energy in a top down fashion to consumers, the introduction of renewable energy such as solar has given rise to bottom-up electric microgrids of prosumers (i.e., consumers that also produce energy) where energy flows in a bottom up fashion from the edge of the grid. Furthermore, the intermittent nature of solar generation and its lack of synchronization with energy consumption creates important challenges for energy management. We will develop a decentralized agentic framework for grid management. More precisely, distributed reinforcement learning agents will dynamically manage energy production, storage and purchase/sale based on load measurements, weather information, demand patterns and spot prices. This decentralized agentic framework will help reduce peak loads, reduce costs, improve energy self-sufficiency, increase resilience to grid outages and scale to increasingly large communities of prosumers.

Faculty Supervisor:

Pascal Poupart;Yuntian Deng

Student:

Partner:

Vector Institute

Discipline:

Computer science

Sector:

Professional, scientific and technical services

University:

University of Waterloo

Program: