Data layer
Three measured streams — CT consumption, anonymized occupancy and HVAC operating status — flow into a single score that tells you, every minute, how efficiently each zone and system is running.
Current transformers on every device and circuit meter real power and energy at 1-minute resolution — pumps, heat pumps, ventilation, DHW, lighting and plug loads. No estimates, no nameplate values.
Presence and people-count signals per zone, fully anonymized before storage. Occupancy tells the model whether consumption is serving real demand or running into empty rooms.
Operating mode, setpoints, fan speeds, valve positions and fault states from all 8 parallel HVAC systems across 14 brands — normalized into one semantic point model.
All three streams are time-aligned on a common 1-minute grid and mapped onto the same zone and asset topology.
A learned baseline predicts expected consumption for the current mode, weather and occupancy pattern.
Deviations between CT-measured consumption and baseline are attributed to operating-state changes — not noise.
Each zone and system receives a 0–100 efficiency score, updated every minute, with the drivers of every point lost.
Move the three inputs and watch the zone efficiency score respond — over-consumption pulls it down, correct operating mode and real occupancy push it up.
Simplified illustration of the scoring model. In production the score is computed per zone and per system every minute from live telemetry, using learned baselines per operating mode, weather and occupancy pattern.
The reference hotel streams this data layer continuously — 640 m² of instrumented envelope, 8 parallel HVAC systems, 14 brands compared.
Request access