Weather Correction and Normalization for Number of Days

Does Metrix do weather correction?

Yes.  This is one of the main reasons people choose Metrix. 

What is weather correction?  And why would anyone want it?

First, let me explain what could happen if you didn’t have weather correction.  Suppose you installed an energy efficient lighting retrofit in your store.  You would expect to see energy and cost savings from this equipment.  For June you might expect your utility bills to look like this:

But what if, instead you see this?

Your boss, when seeing the numbers, could conclude that the energy efficient lighting was actually using more electricity than the older less efficient lighting.  Your recommendations might carry much less weight, and you could find difficulty funding your next “energy savings” project.  No one wants that.

There are several reasons this could have occurred.  One likely reason could be that June 2002 was much hotter than June 2001, and since it was so much hotter, the air conditioning load would have been much greater, and this could have resulted in increased utility bills.  If weather was that much a factor, then how could you ever measure savings from energy efficiency projects?  You would be at the mercy of weather.  In fact, if you were a gambling man, you could just do nothing and hope the summers got colder each year, then you would always be able show savings to management.

It would be much easier to prove energy savings if you could somehow remove weather from the entire equation.  Then it wouldn’t matter if it was hot one summer or the next.  If you saved energy you would see energy savings in your comparisons. 

Many companies do just this thing.  They remove weather from the savings equation because they want to be able to prove that they are actually saving energy (if indeed they are).  Imagine if you were a performance contractor and you were going to share cost savings with your customer.  During hot years, there would be no savings to share.  You would be at the mercy of the weather.  No company can take those risks.  As a result, many of the top performance contracting companies weather normalize their savings.

This process has many names, and is called weather correction, weather normalization, tuning to weather, or weather regression.

What about number of billing days?  Does Metrix correct for that as well?

In the graphs above, one of the likely reasons that 2001 consumption was higher than 2000 consumption could have been that the June 2001 bill had 40 days in it, while the June 2000 bill had only 30 days in it.  It isn’t fair to compare bills when each bill represents a different number of billing days.  Metrix always corrects for number of billing days.

How does the weather normalization work?

I want to warn you that this answer, although I am trying to present it as simply as I can, may be difficult to understand.  Proceed with caution.

The entire concept of weather normalization centers on the concept of Baseline, which we will have to define first.  “Baseline” consumption is the amount of energy your building would have used given today’s conditions, where conditions mean billing periods and weather data and how the building is operated now.  (Let’s keep it simple for now, and leave billing days and building operations out of it for now.)

So how do you calculate baseline?  To calculate the baseline, you select one or two years of bills from before the retrofit took place.  Once you have your bills, you can graph them against weather as shown below.  In the graph, you may notice that kWh and Cooling degree days are divided by number of days.  This is done so that all bills are represented fairly.  If one bill had 45 days, and another had 30 days, because the bills are being divided by number of days, we are looking at the average kWh/day or the average CDD per day.  (Cooling degree days is a way to measure weather data.)  In the graph below, the blue dots represent bills.  The red line is the “best-fit line” through the blue dots.  This “best-fit line”, then, represents the energy usage of the meter before the retrofit takes place.

For example, given the graph above, if there was a billing period, say June 2001, with 30 days and 750 CDDs (which would be 750 CDD / 30 days = 25 CDD/Day), then our fit line predicts that the meter would have used about 7500 kWh/Day or 7500 kWh/Day * 30 Days = about 225,000 kWh.

This can also be calculated mathematically, rather than graphically.  But before we do this, first we have to introduce the fit line equation.  You may remember in those high school algebra classes that every line has an equation.  The equation for a line is

y = mx + b

where m represents the slope, and b represents the y-intercept (where the line crosses the y-axis). 

In our graph above, the y-intercept is 5619.3 kWh/Day.  You can see that the line does cross the y-axis at about 5620.  The slope of the line is 74.16 kWh/CDD.  (You will have to take my word on that.)  If we insert these numbers into the fit line equation above, we get

Where N_CDD = number of CDDs in a billing period, and N_days is the number of days in the billing period. 

In our example that we did earlier graphically, assume there was a 30 day billing period and 750 CDDs in the 30 day billing period.  Then we could use the fit line equation and calculate: 

Which is just about the same number we estimated when we did this graphically. 

Since we are calculating numbers, I might as well restate what we have just done.  We took some bills from some prior period in time, and we graphed them against weather.  We then found a “best-fit line” that represents those bills, and we found an equation that represents that line.  So, if the line represents the bills, and the equation represents the line, then the equation represents the bills, doesn’t it?  So that fit line equation, then, represents the bills from some prior period of time, or in other words, they represent how the building used to consume energy during that base year.

So, suppose someone installed a lighting retrofit on this building.  And we expect that the building should now be using less energy for some post-retrofit month, say June 2001.  In fact, suppose we received a bill for 197,295 kWh.  Remember from above, we cannot just compare old bills to new bills, because weather could foul the whole calculation up.  So instead, we compare “how much energy would have been used given June 2001’s weather” against “how much energy actually was used in June 2001.”  Or in other words, Baseline – Actual. 

To figure out how much energy “would have been used” given June 2001’s weather, well, we just figured that out.  We take the number of days in the June bill and the number of CDDs.  We plug those into the fit line equation, and then we calculate the kWh that would have been used during the base year given June 2001’s conditions.  We call this Baseline usage.