PHASE 1 — PROOF OF CONCEPT

A supervised deployment designed to prove — or disprove — that AI guidance cuts cooling peak load.

Phase 1 is a 16-week programme on real Dubai buildings. It is deliberately small, fully supervised by the operator, and structured so the result stands up to independent review. CoolGrid has not yet run a Phase 1 deployment; this page sets out exactly what one would involve so partners and funders can judge it before committing.

HOW IT STAYS HONEST

Four rules the deployment is built around.

Advisory first, never autonomous

CoolGrid issues recommendations. The building operator or district cooling control room accepts, edits or ignores every one. No setpoint is written to plant equipment without a human action, and comfort limits agreed at kick-off override the model at all times.

Measured against a real baseline

Weeks 1–4 run with no guidance at all, only data collection. The baseline model relates measured cooling load to outdoor conditions and occupancy, so later savings are compared to what the building would have drawn on the same day, not to last year.

Alternating test days remove luck

During the intervention period, guided and unguided days are alternated on a fixed schedule set before the pilot starts. Two comparable populations of days under similar weather is the cleanest evidence a short pilot can produce.

Supervised by the partners, not by us

A named operator lead, a named academic lead and a CoolGrid engineer review every fortnight. The academic lead signs off the measurement method before any result is published or submitted to a funder.

16 WEEKS, FOUR STAGES

What actually happens on site.

Weeks 1–2
Stage 01
Instrumentation and data audit
Confirm which meters exist and what they actually report: chilled water flow, supply and return temperature, plant electrical demand, and building-level kWh. Fill gaps with temporary loggers where a signal is missing. Agree data ownership, retention and who may see what.
Weeks 3–6
Stage 02
Baseline period — no intervention
Nothing changes in how the building is run. CoolGrid ingests readings alongside Dubai weather and builds the baseline load model. Model quality is reported openly: if it cannot predict the building within an agreed error band, the pilot is re-scoped rather than pushed forward.
Weeks 7–14
Stage 03
Supervised intervention
On guided days the operator receives a day-ahead peak plan and intraday alerts — pre-cooling windows, staging order, delta-T corrections. Each recommendation is logged with whether it was accepted, so the analysis can separate 'the model was wrong' from 'the advice was not followed'.
Weeks 15–16
Stage 04
Measurement, verification and report
Peak demand, energy use and comfort exceedances are compared across guided and unguided days. The result is written up with the method, the raw uncertainty range and the failure cases — including a clear statement if the reduction is not statistically meaningful.

LIVE PROGRAMME STATUS

Measured from the host building's own meter readings.

Loading the current programme status…

WHAT GETS MEASURED

Six metrics, agreed before the first recommendation is sent.

Peak reduction is the headline, but a pilot that cuts peak while breaking comfort has failed. All six are reported together, with the measurement window fixed in writing at kick-off so nothing can be chosen after the fact.

Metric	How it is defined	Unit
Peak cooling demand	Highest 30-minute average plant demand per day, guided days vs matched unguided days.	kW / TR
Peak-hour energy	Electricity consumed during the local network peak window agreed at kick-off.	kWh
Total cooling energy	Plant electrical energy per unit of cooling delivered, weather-normalised.	kW/TR
Delta-T performance	Time spent below the design temperature difference, a direct indicator of hidden waste.	% of hours
Comfort compliance	Zone temperature and humidity excursions outside agreed limits. Any increase counts as a failure.	excursion count
Recommendation uptake	Share of CoolGrid recommendations the operator accepted, and how quickly.	%

WHAT A SITE NEEDS

Entry conditions.

One or more real Dubai buildings or a district cooling plant serving them, with an operator willing to run a 16-week programme.
Interval data at 15-minute resolution or better for chilled water and electrical demand — historic data helps but is not mandatory.
A named Dubai-affiliated academic lead to own the measurement method and independent review.
A written data sharing agreement covering ownership, confidentiality and publication rights.
A safety and comfort envelope signed off by the operator before any guidance is issued.

WHAT PHASE 1 UNLOCKS

Exit criteria.

Phase 1 succeeds if the guided days show a measurable reduction in peak demand with no increase in comfort excursions, and the method survives independent review. It also succeeds, in a different way, if it shows the effect is small — that answer is worth knowing before anyone scales it.

Under the Dubai Resilience Challenge Fund, Phase One awards are up to AED 500,000 and progression to a larger Phase Two is explicitly not automatic — it depends on delivered results. CoolGrid has no affiliation with the fund and no application outcome to report; figures are taken from the fund's published guidelines.

Pilot agreement terms Phase 2 savings ledger Grant pathway

Have a building, a plant or a research group that fits?

Applying starts a scoping conversation. Nothing is binding until a written pilot scope and agreement are signed by both sides.

Start a Phase 1 application Partner with us