Abstract
As cities move toward climate resilience, the way buildings use and share energy is becoming increasingly important. This doctoral research presents a Smart Community Energy Management System (SCEMS) that uses artificial intelligence to make buildings more energy-efficient, flexible, and resilient to climate change.
The system combines two advanced control methods: Reinforcement Learning (RL), which teaches buildings to cooperate and make long-term energy decisions, and Model Predictive Control (MPC), which manages heating and cooling demand locally to save energy while keeping indoor comfort. A Long Short-Term Memory (LSTM) model predicts indoor temperature, enabling the controller to plan ahead without relying on complex physical models.
The SCEMS was tested on real data from buildings in Norway and Cyprus, representing very different climates. Results show up to 34% higher solar self-consumption, 45% lower daily peak demand, and 5% better thermal resilience under extreme conditions compared to the conventional control.
This work bridges the gap between theory and practice, offering a flexible and open-source framework that can be scaled from single buildings to entire urban districts. The research provides a pathway for the next generation of smart, autonomous, and climate-friendly energy communities.