444 5th
Avenue SW
A 23-storey office tower in downtown Calgary, built in 1972 and still running most of the control philosophy it was commissioned with. Five studies, one building, and a set of findings that cost nothing to discover and a great deal to keep ignoring.
Five studies,
one building.
Each study was scoped to answer a question the others could not. Together they cover the envelope that admits the load, the plant that serves it, the controls that decide when, the instruments that watch, and the framework that proves whether any of it worked.
Building envelope
Drone and handheld thermography across the roof assembly and all four facades, mapping insulation gaps, air leakage paths and moisture risk.
Chiller plant
Coefficient of performance measured against metered flow, temperature and power. Staging, pump control and heat rejection all assessed.
Building optimization
Six measures against BAS controls, sensor calibration, equipment staging and VAV integration.
Monitoring & tracking
A portable toolkit — temperature loggers, clamp-on power meters, ultrasonic flow — deployed to verify the BAS independently.
Measurement & verification
An IPMVP Option C plan with fitted baseline models, stated uncertainty and a twelve-month reporting period.
Phase III
Extending the proven method across the rest of the Aspen portfolio under BOMA Enspire Retrofit Ready.
A 53-year-old
curtain wall,
seen in infrared.
Roughly 40% of this building's HVAC load is driven by its envelope. You cannot see any of that from the street. Flown with a thermal camera on a cold morning, the failures resolve immediately — and they are not subtle.
What the survey found
Five deficiencies, in priority order. None of them require replacing the curtain wall.
Measures do not simply add up.
Seal the windows and you cut infiltration. Film the glass and you cut solar gain. Do both and you do not get the sum of the two, because both act on the same HVAC load. Switch them on and off below.
The finding that
reframed the
engagement.
Two Trane screw chillers on R-134a, a primary chilled water header, condenser pumps and a forced-draft cooling tower. On paper, an ordinary plant. Measured over six days, it returned a coefficient of performance of about 1.0 — against a benchmark of 5.0 to 6.0 for machines of this type.
A result that low, that consistently, is not a bad week. It is how the plant runs.
Four valid days, all in the same place
Two weekend days were excluded. The remaining four returned 0.88, 1.08, 1.03 and 0.99 — a spread of 0.2 across a week.
| Date | Energy input | Cooling out | COP | Status |
|---|---|---|---|---|
| 22 Aug | 1,304 kWh | 1,148 kWh | 0.88 | Below benchmark |
| 23 Aug | — | — | — | Weekend, excluded |
| 24 Aug | — | — | — | Weekend, excluded |
| 25 Aug | 2,506 kWh | 2,701 kWh | 1.08 | Below benchmark |
| 26 Aug | 2,074 kWh | 2,129 kWh | 1.03 | Below benchmark |
| 27 Aug | 1,824 kWh | 1,811 kWh | 0.99 | Below benchmark |
Work the arithmetic yourself.
Cooling output is flow times temperature difference times a constant. Divide by input power and you have COP. The model below is loaded with the measured values — a steady 350 GPM, a ΔT of 2.4 °C, and 225 kW going in.
Four contributors, ranked
- Low cooling load, narrow ΔT. Output is proportional to ΔT. Move water without moving heat and even normal power draw yields a poor COP.
- High input power. Dirty heat exchangers, compressor or VFD faults, non-condensables, or staging both machines at part load.
- Poor flow conditions. Pumps at 100% with no modulation observed anywhere in the campaign.
- Condenser water too warm. Entering and leaving tower temperatures running close together — the signature of poor heat rejection.
Maintenance before capital
Clean the evaporator and condenser tubes. Inspect the compressor and verify VFD function. Tune staging to actual load. Recalibrate the outdoor air sensors. Reset the chilled water setpoint. Rebalance the loop.
Only then evaluate permanent metering, pump VFDs, tower control sequences and — if the numbers still justify it — replacement machines above COP 6.0. The interim target after maintenance and controls work alone was a COP of 4.0 to 5.0.
One stuck sensor,
six failures
downstream.
The outdoor air sensor was reading a fixed 24 °C and 88% relative humidity. Every sequence that depends on knowing the weather — economizer, supply air reset, boiler lockout, staging — was therefore operating on a number that had stopped being true some time ago.
Six measures, in the order we would do them.
Five of the six are controls work. Together the package was assessed at 15–30% of HVAC energy.
Measure 04 is the one worth understanding. The penthouse air handler was cooling 100% outdoor air down to 5 °C in the middle of a Calgary summer, and the floor-level fan coil units were reheating it back up to something habitable. The building was paying for the cooling, then paying again for the heat to undo it. Simultaneous heating and cooling is the most expensive failure mode in commercial HVAC, and it is almost always a controls fault rather than a broken machine.
Never trust a
single source
of truth.
The building already had a Trane BAS. It was also the system reporting 24 °C outdoors in every season. A portable toolkit was deployed alongside it — not to replace it, but to check it, with no permanent hardware and no capital request.
What the instruments said that the BAS did not
Supply air cycling between 9 and 14 °C
Independent confirmation of excessive cooling, measured at the air handler discharge rather than read from a controller.
Peak of 280 kW at MCC-1
Both chillers and all pumps drawing simultaneously — the staging failure, visible in the electrical data.
Steady ~350 GPM on weekdays
Flat, unmodulated flow with a clean overnight shutoff. The input to every COP calculation in Study 02.
Consolidate the BAS, loggers, power meters and flow data into a single analytics layer with automated fault detection, so that a sensor stuck at 24 °C raises an alarm rather than waiting for an engineer with a drone.
Proving it,
afterwards.
Recommendations are easy. The hard part is demonstrating, twelve months later, that the savings were real and not a mild winter. The verification plan for this building follows IPMVP Option C against ASHRAE Guideline 14, with both baseline models fitted, validated and published with their uncertainty.
11,963 GJ/yr
Nov 2022 – Oct 2023. Heating-driven, and strongly correlated with degree-days.
10,139 GJ/yr
Jul 2024 – Jun 2025. Flat, baseload-driven, and uncorrelated with weather in either direction.
Two fuels, two drivers
Elec (GJ/mo) = 29.31 × days
- Gas R²
- 0.96 (limit > 0.70)
- Gas CV(RMSE)
- 10.24% (limit < 25%)
- Elec CV(RMSE)
- 13.6% (limit < 25%)
- Elec NMBE
- << 0.005%
- HDD reference
- 17 °C
The number beside the number
Expected accuracy of ±13.6% on electricity and ±9.7% on gas, both at a 68% confidence level. Publishing this is not a caveat — it is what makes the savings figure meaningful. A 5% saving claimed against a ±13.6% model is not a saving, it is noise.
Boundary: whole facility, utility-level meters. Retail tenants are sub-metered for electricity. Retail gas metering was excluded as insufficiently accurate. Reporting period: twelve consecutive post-retrofit months.
Adjust the baseline, not the results.
This is the mechanism that makes Option C credible. The baseline is re-computed at the reporting period's own weather before anything is compared. Change the winter below and watch the baseline move with it.
Eight measures under verification.
Together they project 5,522 GJ per year, or 25% of total baseline consumption — comfortably clear of the 10% floor that makes whole-building analysis defensible in the first place. Switch measures off and watch that margin disappear.
One building
proves the
method.
Phase III applies it across the rest of the portfolio under BOMA Enspire Retrofit Ready, funded by NRCan's Deep Retrofit Accelerator Initiative. The emphasis is deliberately on low-cost and no-cost operational measures — schedule alignment, setpoint optimization, controls tuning — because that is where the return is, and because the program funds recommissioning at 80%.
Key dates
Program figures published by BOMA Enspire (bomaenspire.ca) under the NRCan Deep Retrofit Accelerator Initiative. Payments are issued directly to service providers.
Fix the plant, and you still own the largest controllable load in the building.
Everything that makes a chiller plant a liability in August makes it an asset on a peak afternoon. Q-Peak is our aggregator platform for the IESO Save on Energy Peak Performance Program: it forecasts the peak, dispatches curtailment, proves the reduction against baseline and settles the payment.
Published with the client's agreement. Figures on this page are drawn from the BOMA Phase II deliverables for 444 5th Avenue SW (project references ENS-BPE-A00099, A00100 and A00296). Commercial terms, capital costs and payback figures have been omitted. Energy, intensity, uncertainty and carbon figures are reproduced as assessed. Interactive models on this page run the same formulas and fitted coefficients as the underlying reports; inputs you change are your own, not measurements.