Developing an adaptive thermostat control algorithm for occupant-centric demand response in residential buildings

In cold-climate regions such as Quebec, Canada, electric heating imposes substantial strain on the power grid during winter peak hours. To manage this load and ensure grid stability, utilities have adopted thermostat-based direct load control (DLC) programs as a demand-side management solution. Conventional static DLC strategies often underperform in practice because they fail to account for diverse occupant comfort preferences and varying building thermal characteristics. This mismatch results in high override rates, reduces the potential of peak demand reduction, and leads to user dissatisfaction. This study contributes to addressing this gap by proposing an adaptive, occupant-centric thermostat control algorithm that dynamically adjusts preheating and setback strategies during DLC events based on real-time observations of occupant overrides and zone-level heating system runtime. Notably, the algorithm operates without requiring prior knowledge of occupant type or building properties. Using EnergyPlus simulations and its Python API across various scenarios combining two occupant profiles (prefer warmer/cooler) and two envelope types (good/poor envelope), the algorithm iteratively refines control parameters by learning from zone-level thermal and behavioural response patterns. Compared to static DLC, the adaptive strategy reduces peak demand for space heating by up to 72%, depending on house type and occupant behaviour, minimizes overrides, and lowers electricity costs, with savings ranging from $77 to $126 per winter under Québec’s dynamic pricing. By tailoring control parameters to both user tolerance and building thermal inertia, the algorithm supports context-aware demand shifting, whether through deeper or shallower setbacks or reduced or enhanced preheating, without compromising occupant comfort.