Optimizing occupant-centric building controls given stochastic occupant behaviour

Abstract

Occupant-centric control (OCC) strategies represent a novel approach for indoor climate control in which occupancy patterns and occupant preferences are embedded within control sequences. They aim to improve both occupant comfort and energy efficiency by learning and predicting occupant behaviour, then optimizing building operations accordingly. Previous studies estimate that OCC can increase energy savings by up to 60% while improving occupant comfort. However, their performance is subjected to several factors, including uncertainty due to occupant behaviour, OCC configurational settings, as well as building design parameters. To this end, testing OCCs and adjusting their configurational settings are critical to ensure optimal performance. Furthermore, identifying building design alternatives that can optimize such performance given different occupant preferences is an important step that cannot be investigated during field implementations of OCC due to logistical constraints. This paper presents a framework to optimize OCC performance in a simulation environment, which entails coupling synthetic occupant behaviour models with OCCs that learn their preferences. The genetic algorithm for optimization is then used to identify the configurational settings and design parameters that minimize energy consumption under three different occupant scenarios. To demonstrate the proposed framework, three OCCs were implemented in the building simulation program, EnergyPlus, and executed through a Python package, EPPY to optimize OCC configurational settings and design parameters. Results revealed significant improvement of OCC performance under the identified optimal configurational settings and design parameters for each of the investigated occupant scenarios. This approach would improve OCC performance in actual buildings and avoid discomfort issues that arise during the initial implementation phases.