Energy waste in commercial buildings: where it hides and how to cut it.
Milvian AI teamFacilities and infrastructure intelligence
Energy waste in commercial buildings is energy used when it does no useful work: equipment running outside occupied hours, heating and cooling fighting each other, faulty economizers, drifting setpoints and avoidable demand peaks. You find it by comparing metered use with a weather-normalized baseline, fixing the cause, and confirming the saving at the meter.
What counts as energy waste?
Energy waste is not the same as high energy use. A hospital, a data hall or a cold store uses a lot of energy because it has to. Waste is the part of the bill that buys nothing: air handlers running for an empty floor, chillers cooling while boilers heat the same space, lights on in a closed car park, or a demand peak that pushes the site into a higher tariff band for the whole month.
In most commercial buildings the waste is not one big failure. It is many small drifts that add up. A schedule overridden for a weekend event and never reset. A sensor that reads two degrees off. A damper stuck at minimum. A setpoint nudged by a tenant complaint. Each one is easy to fix once found. The hard part is finding them among thousands of points.
Waste also persists because nobody owns the whole picture. The BMS knows the setpoints but not the bill. The utility knows the bill but not the equipment. The maintenance team knows the work orders but not whether the fix saved anything. Each system holds part of the answer, and none of them closes the loop on its own.
Where does energy waste hide?
The usual suspects, and the signal each one leaves.
| Source | Typical cause | Signal in the data |
|---|---|---|
| After-hours operation. | Schedules overridden or never set. | A night and weekend floor that never drops. |
| Simultaneous heating and cooling. | Competing setpoints, valve faults. | Heating and cooling both active in one zone. |
| Economizer faults. | Stuck dampers, bad sensors. | Mechanical cooling running on mild days. |
| Setpoint drift. | Manual overrides, comfort complaints. | Setpoints outside the agreed range. |
| Demand peaks. | Loads starting together. | Short spikes that set the monthly demand charge. |
| Poor power factor. | Motor loads, uncorrected circuits. | Apparent power well above active power. |
| Plug and process loads. | Equipment left on. | Feeders that never ramp down. |
Why can the bill not tell you?
A utility bill reports one total for the month. It cannot say which floor, feeder or unit drove it. It also mixes your actions with the weather. A building that cut its waste in a hot summer can show a higher bill than last year. A building that wasted more in a mild winter can look like it improved.
That is why a raw year-over-year comparison is a weak test. The fair comparison is against a baseline: what the building would have used in this weather and at this activity level if nothing had changed. The standard method is regression on past usage against cooling degree days and heating degree days. The gap between actual and expected use is the number that matters. Below the baseline is a saving. Above it is waste.
Rate configuration turns that gap into money. Utility bills carry the tariff, so the same model that gives you the energy delta can price it, including demand charges and time-of-use bands. A finding that says a site is a certain amount above baseline this month, and what that is worth, gets acted on faster than a chart.
Which waste is not about kilowatt-hours?
Two kinds of waste do not show up as extra energy at all. The first is peak demand. Many commercial tariffs charge for the highest demand in a billing period, so a short spike when refrigeration, chillers and lighting all start together can cost as much as hours of steady use. The fix is sequencing: know when the site peaks, stagger the loads, and pre-cool or shed where the operation allows.
The second is power factor. When current and voltage drift out of phase, apparent power rises above the power doing useful work. The site uses more of its electrical infrastructure than it should and may pay a reactive penalty. Per-feeder metering shows which circuits are responsible, so correction goes where it pays back.
Where should an existing building start?
Start with the main meter at short intervals. A daily load profile alone shows a lot: how high the base load sits overnight and at weekends, when the building ramps up and down, and when the peak arrives. A base load that barely changes between occupied and unoccupied hours is usually the first and cheapest waste to fix.
Next, sub-meter the largest loads. HVAC feeders, central plant, lighting and major process equipment are the usual candidates. Sub-metering turns one total into a set of numbers that can each be compared with a baseline, so the investigation goes straight to the feeder or unit that changed.
Then connect the BMS. Setpoints, schedules, damper positions and valve commands explain why a load behaves the way it does. Meter data says how much is being used. BMS data says why. Together they turn a high reading into a specific fix that a technician can act on, and that the meter can later confirm.
