A saving is only as good as the evidence behind it. Measurement and verification (M&V) is how savings are shown in a way an owner's engineer, lender or auditor will accept, and it is why Gigabiome's core deliverable is a monthly verified report, not a dashboard.
The problem M&V solves
Savings can't be metered directly. They are the difference between what the plant used and what it would have used without the change. Weather, occupancy and load change month to month, so "this July versus last July" proves little. M&V builds a baseline model from data before the change and uses it to predict the counterfactual.
IPMVP options, and why Option B fits a chiller plant
The International Performance Measurement and Verification Protocol (IPMVP), published by the Efficiency Valuation Organization, defines four options:
- Option A, retrofit isolation with key parameter measurement: measure some parameters and stipulate others.
- Option B, retrofit isolation with all parameter measurement: measure the affected system's energy directly.
- Option C, whole facility: use the utility meter.
- Option D, calibrated simulation.
Plant optimization changes how a well-defined system runs, so Option B fits best. Draw the measurement boundary around the chillers, cooling-tower fans and plant pumps. Meter their electricity and cooling output. Model the baseline against the variables that drive them: cooling load and outdoor wet-bulb (and dry-bulb for air-side effects). Whole-building meters (Option C) bury plant savings in everything else the building does.
What makes a baseline credible
- Enough data, covering the conditions. A baseline built only on mild weeks can't predict a heat wave. It should span the load and weather range the reporting period will see.
- The right variables. For a plant, load (ton-hours) and wet-bulb explain most of the variation. Calendar and occupancy effects matter where schedules change.
- Stated goodness of fit. ASHRAE Guideline 14 sets statistical acceptance criteria for baseline models, chiefly the coefficient of variation of the root-mean-square error, CV(RMSE), and the normalized mean bias error, NMBE. Savings should also be reported with their fractional savings uncertainty, not as a bare number.
- Clean data. Mislabelled or swapped points, frozen sensors and unit mix-ups will ruin a baseline quietly. Check the physics before modeling: supply colder than return, wet-bulb at or below dry-bulb, flows and powers that add up.
What belongs in a monthly verified report
For a chiller plant with both energy and water in scope:
- Energy. Measured kWh against the adjusted baseline, with the uncertainty band. Plant kW/ton by load and wet-bulb band, so a change in weather isn't mistaken for a change in performance.
- Water. Makeup and blowdown gallons, cycles of concentration, and savings against a baseline of the same evaporation.
- Guardrails. Scaling index and chemistry within limits, biocide residuals, Legionella results where they are required (NYC rules). Energy savings that come at the cost of fouled tubes or a compliance breach aren't savings.
- Actions and acceptance. What was recommended, what operators accepted, what changed. Advisory systems only save what people act on.
- Dollars at the site's tariffs. Last, and only after everything above holds up.
LBNL's Smart Energy Analytics Campaign found median savings of about 9% among organizations using fault detection and analytics. Getting from a finding to a verified saving takes follow-through, and a monthly report built this way makes that follow-through visible.
To see how your plant runs today, the first step of any baseline, try Plant Check. To talk about a measured pilot, see the pilot program.