Multi-Objective Optimization for Flexible Building Space Usage

Globally, more than 30% of the world's energy consumption arises in buildings. Optimization of buildings is a key opportunity for reducing energy consumption and carbon emissions, improving operational efficiency and occupant well-being and comfort. While building generative design and control systems have received considerable research attention, optimizing space utilization, particularly for flexible spaces is an underdeveloped research area that is relevant to existing buildings. Flexible spaces, for example, rooms with movable walls, are increasingly common in modern building designs where space requirements are dynamic. In this paper, a novel space usage optimization framework is proposed, including a practical task formulation that enables room reallocation, combination and removal, a machine learning model for energy cost estimation (XGBoost) based on real sensor data and a multi-objective optimization component to minimize energy consumption and maximize room thermal comfort simultaneously (NSGA-II). Its effectiveness is tested and discussed through two representative problem scenarios. Our case studies show that we can reduce energy cost substantially by around 40% in comparison with the original space usage setting, while additionally improving thermal comfort for the occupants. This work shows great potential of using AI techniques for optimizing building space usage.