Comfort-aware Optimal Space Planning in Shared Workspaces

Energy consumption in office buildings, especially in shared office spaces, can be substantially reduced through joint optimization of space use and heating and cooling demands. This paper addresses this underexplored research problem in a coworking space that offers long-term and daily plans. We train an input convex neural network to estimate the energy consumed by the HVAC system in a single day to condition a given zone of the building. Due to the convexity of this model in its inputs, we formulate a convex mixed-integer program to optimize HVAC energy consumption by deciding how to assign desks to occupants and adjust zone temperature setpoints. Considering a medium-sized office building as the coworking space, we show that this optimization problem can be solved to near-optimality relatively quickly, hence it can be used to make decisions regarding long-term bookings. For daily bookings, we design heuristic algorithms that take the solution of the optimization problem and assign the remaining space, while ensuring the satisfaction of thermal comfort constraints. By incorporating these algorithms in the workspace reservation system, energy consumption can be reduced by up to 11.7% while maintaining individual thermal comfort.