Modelling the impact of occupant behaviour on direct load control of HVAC systems

Direct load control (DLC) algorithms for HVAC systems are automated temporary interventions to the sequences of operation to reduce on-peak electricity demand. While DLC of HVAC systems has the potential to dramatically reduce the economic, societal, and environmental burden of electrification, occupant behaviour, specifically thermostat use, accounts for major uncertainty on this potential [1]. This study first develops a thermostat use behaviour model upon longitudinal field data of office occupants. The model represents both the stochasticity of an individual’s thermostat use patterns and the inter-occupant diversity. The model is then incorporated to EnergyPlus through its Python API. Seven DLC algorithms are examined at varying setback/setup intensities. Of them, three were without preconditioning and four were with preconditioning. Simulations were conducted with the EnergyPlus model of a small commercial building in Toronto, Canada. The results indicate that occupant behaviour can reduce the median on-peak demand savings by up to 20%, particularly with DLC algorithms with more than 2°C setback/setup and without preconditioning. Preconditioning could significantly reduce the risk of occupant overrides and improve the robustness of DLC to occupant behaviour.