Turning HVAC into a revenue stream

Every conversation about HVAC starts from the same assumption: it is a cost. You budget for it, you defer maintenance on it, and once a decade you replace something large and unhappy in the penthouse. The equipment appears in the operating statement in exactly one column.

That framing is now incomplete. In Ontario, capacity is scarce at specific, predictable moments — a handful of hot weekday afternoons when air conditioning across the province peaks at once. The system operator would rather pay buildings to use less at those moments than build generation that runs for forty hours a year. That payment is real, it is recurring, and the asset that earns it is one you already own.

The obstacle is almost never the equipment. It is that most buildings cannot prove what they did.

The load you already have

When we instrumented a 23-storey office tower in downtown Calgary last year, we clamped current transformers onto the motor control centre feeding the chiller plant and logged three phases continuously for thirty days. The peak came in at 280 kW — both screw chillers and every associated pump drawing simultaneously.

That number was originally a problem. It was evidence of a staging failure: two machines and all their auxiliaries running flat out regardless of load, which is a large part of why the plant measured a coefficient of performance of about 1.0 against a benchmark of 5.0 to 6.0.

But read it the other way and 280 kW is the size of the lever. A building with a 280 kW controllable thermal load has something the grid genuinely wants for three hours on an August afternoon. The same concentration of load that makes a plant inefficient makes it dispatchable.

The properties that make a chiller plant a liability in August are precisely the properties that make it an asset at 5 p.m. on the worst day of the year.

Why buildings have thermal flexibility and most other loads don't

A server room cannot be turned down. Lighting can, but occupants notice within seconds. Elevators are non-negotiable. HVAC is different, and the difference is thermal mass.

A large commercial building is a heat battery. Concrete floor slabs, structure, furniture and the air itself store an enormous amount of energy relative to the rate at which the space loses comfort. Push the building a degree cooler than setpoint over the two hours before an event and you can back the plant substantially off for the two hours during it, while indoor temperature drifts up by a fraction of a degree per hour. Nobody in the building experiences an event. They experience a Tuesday.

The strategies are unglamorous and well understood:

• Pre-cooling. Charge the mass ahead of the event window, so the plant can coast through it.
• Setpoint relaxation. A one to two degree drift within the comfort band, applied gradually rather than as a step change.
• Staged equipment shedding. Drop the lag chiller, then tower fans, then trim pump speed — in an order defined in advance, not improvised on the day.
• Deferring non-critical loads. Domestic hot water recirculation, garage exhaust, and anything else with a tolerant duty cycle.

None of this is new. What is new is that there is now a counterparty willing to pay for it.

Try it on real numbers

The model below runs a curtailment against a summer demand profile scaled to that measured 280 kW plant peak. Change the depth, the duration and the pre-cooling lead, and watch two things at once: the peak reduction, and the net energy cost of achieving it.

The second readout is the one worth sitting with. Push the curtailment deep with no pre-cooling and you get a large peak reduction and a rebound spike the moment the event ends — which, depending on timing, can land you right back at a system peak. Add pre-cooling and the rebound flattens, but you have spent energy in advance to do it.

Demand response pays for the kilowatt at the peak, not the kilowatt-hour across the day. Confusing the two is the most common reason a programme underdelivers.

The part that actually stops people

Here is where most buildings fall out of the process, and it has nothing to do with chillers.

To be paid for a reduction, you have to demonstrate it against a counterfactual — what the building would have drawn had nothing been dispatched. That requires a defensible baseline methodology, interval data of adequate quality, and the ability to separate your curtailment from the weather, the occupancy and everything else that moved that day.

This is the same discipline as measurement and verification on a retrofit, compressed from twelve months into four hours. And it fails for the same reasons: a building automation system nobody trusts, sensors that have quietly stopped reporting, no interval metering at the equipment that actually moved, and no record of what the operator did or when.

At the Calgary tower, the outdoor air sensor had been reading a fixed 24 °C and 88% relative humidity for long enough that every weather-dependent sequence in the building was running on fiction. A building in that condition cannot participate in demand response in any meaningful way — not because it lacks flexibility, but because it cannot prove it used any.

Sequence matters

The order of operations we recommend is consistent, and it is not the order most vendors propose:

1. Instrument first. Interval metering at the plant, verified against something independent of the building automation system. Portable loggers are enough to start — no capital request required.
2. Fix the controls. If sequences are wrong, curtailment is unpredictable, and unpredictable curtailment is worse than none. It also tends to be where the largest permanent savings are hiding anyway.
3. Write the curtailment sequence down. What sheds, in what order, at what indoor temperature limit, with what override. Before the season, not during an event.
4. Then enrol. With a baseline you can defend and a plant that does what you tell it.

Steps one and two pay for themselves regardless of whether you ever dispatch a single event. That is the argument we make to owners who are sceptical of demand response as a category: the prerequisites are things you should do anyway, and the revenue is what you get for having done them properly.

What we do with it

We aggregate commercial buildings into the IESO Save on Energy Peak Performance Program through our own platform, Q-Peak. It forecasts which afternoons are likely to matter, dispatches curtailment across the enrolled portfolio, measures the delivered reduction against baseline, and handles settlement.

Aggregation matters more than it sounds. A single building's curtailment is lumpy and its baseline is noisy. A portfolio's is neither — one building's tenant event is another building's quiet Tuesday, and the aggregate behaves far more predictably than any of its parts. That predictability is what makes the reduction saleable.

Programme structures and payment rates change between years and between programmes; the IESO publishes the current terms, and we will walk you through what your specific load profile is likely to be worth under them. What does not change is the prerequisite. You need a plant that responds to instruction, and data good enough to prove it did.