Weather normalization, CUSUM and RETScreen M&V: proving savings from utility data

When a building's gas consumption drops 15 percent the winter after a boiler retrofit, the retrofit gets the credit. When it drops 15 percent because the winter was mild, the retrofit still gets the credit, and the following cold winter is blamed on something else. Weather normalization is the discipline that separates the two, and CUSUM is the chart that shows the result over time. Together they are the core of whole-building measurement and verification.

This article explains the method as it is used in RETScreen and in the VE-MAP monthly review, and why utility bills are enough data to run it.

Option C: the whole building as the meter

The IPMVP, maintained by the Efficiency Valuation Organization, describes four ways to verify savings. Options A and B isolate the retrofit with its own measurements. Option D uses a calibrated simulation. Option C uses the meters measuring the supply of energy to the whole facility, and applies when several measures interact or when the goal is to manage overall facility performance using utility-level data. For a portfolio of buildings with monthly bills and, increasingly, interval data through Green Button, Option C is the practical choice.

Step 1: the baseline model

A baseline is not last year's consumption. It is a model of what consumption should be given the drivers that change independently of the building's efficiency. For heated and cooled buildings the main driver is weather, expressed as degree days. Environment and Climate Change Canada defines heating degree-days for a day as the number of degrees Celsius that the mean temperature is below 18 degrees, and cooling degree-days as the number above 18. Sum them per billing period and you have the driver for that period.

The baseline model is a regression of consumption against degree days over a baseline period, typically twelve months before the change being measured. Its intercept is the weather-independent load (domestic hot water, base electrical), its slope is the building's sensitivity to weather. Other drivers (occupancy, production, operating hours) can be added where they matter.

Step 2: expected versus actual

For each period after the baseline, plug the period's actual degree days into the model to get expected consumption, then subtract actual. A positive difference is a saving; a negative one is excess. This period-by-period comparison is what the M&V chart in the monthly review shows: actual and baseline plotted from zero, with the baseline period shaded.

Step 3: CUSUM

CUSUM is the cumulative sum of those differences. The Bureau of Energy Efficiency's monitoring and targeting chapter describes it as the difference between the baseline (expected) consumption and actual consumption accumulated over time; the line oscillates around zero until something changes the pattern, such as an energy saving measure or a worsening in efficiency from poor control, housekeeping or maintenance. A BC Hydro and Prism Engineering paper on monitoring, targeting and reporting calls the same line the cumulative savings total and notes that trends in it indicate consumption patterns.

Where RETScreen fits

RETScreen Expert, developed by the Government of Canada and maintained by Natural Resources Canada, is the reference tool many Canadian energy managers use for this analysis. Its performance analysis builds the regression baseline, produces the actual versus expected series and the CUSUM, and supports a portfolio of facilities. Viewer mode is free of charge and premium features are by annual subscription; NRCan notes it is used in 36 languages.

VE-MAP does not replace RETScreen; it carries its results into the monthly workflow. A RETScreen export (a workbook or a project file) is uploaded per facility, and the platform renders the same graphs the analyst sees in RETScreen: monthly consumption and cost per commodity, annual totals, the M&V actual versus baseline chart, and the CUSUM cumulative savings chart, with each period labelled by the same reporting month rule as the bills. Those charts and an M&V table with bill types and a total row then appear in every monthly review report.

What can go wrong, and the fixes

• Bad baseline data. A duplicated or mis-parsed bill in the baseline year shifts the whole model. Validate the baseline before fitting it; see the validation checks.
• Misaligned periods. A bill labelled by its end date lands in the wrong month and pairs with the wrong degree days. Label periods consistently, the way the platform and RETScreen do.
• Estimated reads. An estimated bill followed by a true-up creates a false saving then a false loss. The M&V table marks bill type so the reviewer can discount the pair.
• Non-weather changes. A new tenant, a vacancy or a change in operating hours changes the baseline. Add the driver or reset the baseline, and record it in the BAS notes.
• Weather stations. Use a station near the building. Degree days from the wrong region make a good building look erratic.

Why monthly bills are enough

Sub-metering has its place, but for the question that boards and lenders ask (did the building save what the project promised) whole-building utility data is both sufficient and authoritative, because it is the data the money is paid on. Twelve months of validated bills give a baseline that covers both seasons. Interval data, where Green Button provides it, adds the daily and hourly shape that explains why a slope changed; see interval data heat maps. The two together are what the measurement and verification post calls a repeatable pipeline instead of a one-off scramble.

FAQ

How long a baseline do we need?

Twelve consecutive months of validated data is the practical minimum, so both heating and cooling seasons are represented. Longer baselines are better if operations were stable.

Can CUSUM be run on water?

Yes, against occupancy or simply against time, since water is weakly weather dependent in most residential buildings. Water is often where CUSUM finds leaks first.

Do we need a RETScreen licence?

Viewer mode is free per NRCan. The VE-MAP service produces the M&V and CUSUM views from your bill and meter data, and imports RETScreen files where your analyst already uses it. See the about page for how the pieces fit.