Jose is the next awesome IEL student to successfully defend his PhD entitled Multi-Agent Reinforcement Learning for Demand Response and Load Shaping of Grid-Interactive Connected Buildings. We congratulate him for this great piece of work and wish him good luck for his future career!
Abstract
Buildings account for over 70% of the electricity use in the US. Renewable energy resources can help reduce our need for fossil fuels, and distributed generation has the potential to make buildings less dependent on the electrical grid. However, integration of renewable energy resources has certain challenges regarding grid stability and security of supply. Demand response can help buildings play an active role in the generation and storage of electricity by increasing demand flexibility. Demand response programs must also allow the coordination of multiple buildings such that the peaks of net electrical demand are not only shifted but shaved. However, buildings are dynamic energy systems, in constant change due to diverse factors, e.g., refurbishment measures, installation of PV panels, changes in the energy supply, variations in consumption patterns, or the future integration and charging of electric vehicles. Adaptive model-free control approaches have the potential to overcome these challenges, as they can constantly learn from, and adapt to, their changing environment. In this research, we explore the use of multi-agent reinforcement learning (RL), an adaptive and potentially model-free control algorithm, for multi-agent coordination of several buildings in simulated demand response scenarios using our OpenAI Gym environment CityLearn.
Over 70 people from all over the world connected to his defense.