A campus wastes energy and spoils its own air on the same schedule.
GEST sees the load two hours before it arrives and takes about 20 kW off the peak. SABZPOOST is a surface that oxidises airborne organics using the light a room already has. Both are prototypes, and every figure on this site states how it was arrived at.
Loading campus model…
One simulated campus day. Demand and room air on the same clock — simulation, not metering.
Peak demand removed
5 kW lighting + 8 kW HVAC setback + 7 kW stored cooling, against a 257 kW modelled peak
ModelledVOC removed in four hours
500 → 300 ppb · 300 m³ room · 100 m² coated · 400 lux
BenchForecast error, 2 h ahead
MAE · 4.0% MAPE against a 176 kW mean
ModelledAnnual consumption avoided
9% of a 1,546 MWh modelled baseline
ModelledSensing nodes at 1 Hz
MQTT over the campus VLAN
Design targetConsumables in the air path
No filter media, no adsorbent to regenerate
BenchBoth systems close the same loop
Measure continuously, work out what happens next, change something while there is still time, then check the change against the following measurement. GEST runs that loop on electrical demand and SABZPOOST runs it on air chemistry. The loop is the shared part — the sensors and the coating are only where it touches the building.
Measure
What is the building doing right now?
312 nodes at 1 Hz: sub-metering per feeder, CO₂ and occupancy per zone, flow and temperature difference across the chilled-water loop.
Photoionisation detector for total VOC and an NDIR head for CO₂, one node per 25 m², sampled at 1 Hz.
Predict
What will it do in the next two hours?
An LSTM trained per building on 15-minute aggregates, run forward two hours. Backtest MAE 7 kW.
A first-order decay fitted per room from the last hour of samples, giving the time to cross 400 ppb.
Act
What can be changed before it happens?
Lighting setback, a 1.5 °C cooling setpoint shift, and discharge of the phase-change store — armed, not forced.
Nothing to switch. The surface oxidises continuously whenever there is light on it; the control is where you paint.
Check
Did it work, and by how much?
The next metered interval is compared against the forecast it was supposed to beat. The residual is logged either way.
Concentration is re-measured against the pre-coating baseline for the same room, season, and occupancy.
The two systems
One manages what a building draws. The other manages what it breathes. Neither is finished, and the pages below say exactly how far along each one is.
GEST
Campus energy intelligenceGEST reads 312 metering and environmental points once a second, runs a demand forecast two hours ahead of the campus, and spends that lead time taking about 20 kW off a 257 kW peak — by easing lighting and cooling, and by discharging cooling banked overnight in a phase-change store.
SABZPOOST
Photocatalytic air chemistrySABZPOOST is a 20-nanometre doped anatase film applied by sol-gel. Under ordinary interior lighting it generates hydroxyl radicals at the surface and oxidises volatile organics where they touch it, taking a 300 m³ room from 500 to 300 ppb over four hours with no fan, no filter, and nothing to replace.
Ask for the working behind any number here
Every figure on this site is either bench work, a model output, or a design target, and it is labelled as such. If you want the model, the raw run, or the assumptions behind one of them, email us and we will send them.