The Artificial Intelligence for Energy (AI4E) team conducts research on energy applications, e.g., smart grids, mostly using deep learning based models, including reinforcement learning based demand response. A central theme of our research is the development of creative new algorithms for effective control of flexible assets in the grid, both in residential and industrial segments of the smart grid. We closely link/align our research with industry partners through collaborative projects.
The evolution of power grids to smart grids is characterized by the integration of renewable energy sources (RES) and a distributed network of prosumers. This transition requires managing the ever more challenging demand-supply balance, given the high penetration of RES in power generation, as well as increased electrification on the demand side (e.g., electric vehicle charging, heat pumps). Artificial intelligence (AI) and machine learning facilitate this transition through predictive models and control systems (e.g., models trained on data collected from EV charging stations, solar farms). AI techniques we recently have been working on include reinforcement learning (e.g., DQN, DDQN, DDPG), physics informed neural networks, deep learning, graph neural networks, etc. We use such methods to address demand response for EV charging, home energy management (with a strong focus on HVAC), flexibility identification, load profile clustering, etc.
For samples of our recent work, follow us on LinkedIn!
