Occupant-Responsive Control

Occupant-responsive control begins with knowing something about the occupants, and there is a ladder of richness. The simplest is presence: a passive-infrared (PIR) occupancy sensor answers is anyone here? - cheap, private, and enough to switch lights and setback HVAC in a room that has emptied. Richer is count: how many people, which drives ventilation and cooling load. Count can be inferred from CO2 concentration (people exhale it, so it is a good proxy for how crowded a room is), from people-counting sensors at doorways, from Wi-Fi or Bluetooth device counts, from desk and camera-based systems, or from badge and room-booking data. Richest of all is identity and preference - who specifically is here and what they like - which powers personal comfort but raises the sharpest privacy questions.

Each source trades accuracy, cost and intrusiveness. PIR is blind to someone sitting still; CO2 lags real occupancy by minutes as gas accumulates and disperses; camera and Wi-Fi counting are accurate but far more privacy-laden. A robust system usually fuses several - PIR plus CO2 plus bookings - so no single blind spot governs the building. The design instinct that matters is proportionality: use the least intrusive signal that answers the question you actually need. If all you need is to turn off lights in an empty room, a PIR sensor is both sufficient and respectful; reach for cameras only when the value genuinely demands it and consent and governance are in place.

With occupancy known, control gets to adapt. Lighting is the easiest and most established: occupancy sensors switch or dim lights when a space empties, and daylight harvesting dims electric light as daylight rises, holding a target level at the desk while cutting energy - the two combined routinely cut lighting energy substantially. HVAC is higher-value and slower: an unoccupied zone can be set back (a wider temperature deadband, reduced or paused conditioning) and then recovered before people return - which is where the predictive control of Lesson 7.2 pairs beautifully, using occupancy forecasts so a room is comfortable on arrival rather than lurching to catch up after.

The subtlety in HVAC is that thermal systems are slow and people are impatient, so naive occupancy control backfires: cut cooling the instant a room empties and it will be uncomfortable when someone returns, breeding complaints and overrides that destroy the savings. Good design uses setback rather than hard-off, sensible timers, and predictive pre-conditioning. There is also a real limit: in a densely, continuously occupied space, occupancy control saves little because the room genuinely is always full - the savings live in the intermittently used spaces (meeting rooms, perimeter offices, back-of-house, out-of-hours) that dominate most real buildings. Targeting those, not chasing occupancy control everywhere, is where the return actually is.

The signature application is demand-controlled ventilation (DCV). Ventilation exists to dilute the CO2, odours and pollutants people produce, so codes size fresh air for design occupancy - the room assumed full. But a meeting room rated for twelve, sitting with two, is being massively over-ventilated, and every cubic metre of outdoor air must be heated or cooled at real energy cost. DCV modulates fresh air to actual demand, typically using CO2 as the proxy: as concentration rises toward a setpoint (commonly kept in the region of 800-1000 ppm), the fresh-air damper opens; as the room empties, it closes. The result is good indoor air quality and lower conditioning energy, because you ventilate for the people present, not the people assumed. Design and codes such as ASHRAE's ventilation standards explicitly permit DCV, though minimum-air floors must always be respected.

At the finest grain sits personal comfort. People differ - thermal comfort is individual, and a single zone setpoint satisfies no one perfectly - so systems increasingly let occupants shape their immediate environment: app-based "warmer/cooler" requests that nudge a zone, personal desk fans or radiant panels, task lighting, and models that learn preferences over time. Beyond energy, this serves wellbeing and productivity, and it is the territory that frameworks like the WELL Building Standard foreground. The point is not to hand every person a private thermostat war, but to give people agency within sensible bounds - which, as the next section argues, is also what makes efficiency stick.

Occupant-responsive control gets far more powerful when it stops merely reacting to who is present and starts anticipating them - which is where this lesson meets the predictive control of Lesson 7.2. Because thermal systems are slow, the highest-value move is predictive pre-conditioning: using an occupancy forecast - from booking systems, recurring weekly patterns, badge histories or a learned model in the digital twin - so a meeting room is already comfortable and properly ventilated the moment people arrive, rather than lurching to recover after they walk in. Occupancy becomes one more forecast feeding the building's plan, alongside weather and price.

This reframes occupancy data as an input to optimisation, not just a switch. A twin that knows the building's typical rhythms can stage plant, set back empty zones with confidence they will not be needed soon, and recover them just in time - capturing the savings of setback without the discomfort that makes occupants revolt. It also lets DCV and cooling be coordinated rather than fighting: ventilation, temperature and lighting all responding to one shared, forecast-aware picture of occupancy instead of three sensors making three uncoordinated decisions. The caveat carries straight over from the predictive lessons - a forecast can be wrong, so pre-conditioning must degrade gracefully to sensible reactive behaviour, and the occupancy patterns being learned are, again, data about people, to be governed with the same restraint.

Occupant-responsive control lives on a genuine tension: the building wants to minimise energy, and people want to be comfortable and in control - and the two do not always agree. The failure mode is a building optimised so hard for efficiency that it feels like it is fighting its occupants: lights that snap off while you sit still, air that feels stuffy because the CO2 setpoint was set high to save fans, a temperature you cannot influence. When people feel out of control, they rebel - propping doors, taping over sensors, bringing space heaters, filing complaints - and those overrides typically cost far more than the control ever saved. Perceived agency is not a nicety; it is what makes the efficiency durable. The best systems give occupants a real, visible say inside bounds the building sets, and lean toward comfort when the two conflict, because a system people trust is one they leave switched on.

And then privacy, which is not optional. Occupancy data is data about people - when they are at their desk, who is in which room, patterns of presence that can reveal a great deal - and it can drift into surveillance fast. The obligations are concrete: collect the minimum needed (favour presence and count over identity and cameras), aggregate and anonymise wherever possible, be transparent with occupants about what is sensed and why, secure the data as the operational-technology asset it is, and comply with data-protection law - which is a matter for qualified legal and privacy professionals, not a default you can assume. The line to hold: use occupancy sensing to serve the people in the building, never to quietly watch them. Get that balance right and occupant-responsive control is one of the most humane and highest-return things a smart building does; get it wrong and it is both creepy and self-defeating.