Field Notes · Step 04 · Findings

What the building told us before the model did

Before any model was calibrated, the install and the sensor data turned up missing equipment, an unevenly loaded service, an after-hours load pattern and a hot equipment room.

The energy model isn’t calibrated yet. These findings come from the install and the sensor data alone. Finding 1 was observed during install; findings 2 to 6 come from measured data.

1. A rooftop unit believed to have an economizer doesn’t have one. It runs on fixed outside air. This came to light on install day, when the planned duct sensors couldn’t be placed. It is now a retrofit candidate for the model to evaluate. Equipment that isn’t what the records say is a finding in its own right.

2. The main service is unevenly loaded. One phase runs about 30% lighter than the other two under load, which makes it a candidate for a circuit rebalancing review. Monthly bills can’t show this.

3. The night base load is on the plug and process panels. Overnight, the HVAC panel is quiet while the plug and process panels draw 10–25 amps continuously. After-hours savings here depend on what stays plugged in overnight.

4. A recurring weekend load. A midday increase of about 65 amps appears on the downstream panels on weekends. Attribution is in progress. A repeating pattern like this can become an automated rule.

5. The equipment room overheats. A sensor placed in the equipment room recorded 34 °C (93 °F) on a hot day, which gives a ventilation discussion real data.

6. The building’s schedule is now measured. People counts and zone CO₂ show how the building is used. They replace assumed schedules in the model, so calibration adjusts physical inputs rather than fitting curves.

Lessons learned

  • Interval data with verified circuit attribution produces findings before any model exists.
  • Instrument your own installation. The equipment-room sensor was the cheapest insurance in the project.
  • Log as-found conditions during install. Missing equipment is a deliverable.

Next in this series: calibrating the energy model against utility and measured data.