Image courtesy of Flickr user Hakan Dahlstrom

Removing weather and calendar effects shines a spotlight on your building's true nature

A statistical model of energy use is crucial for understanding how your building is performing over time. All Snapmeter reports show you how your use and demand compares to a baseline that takes into account weather and other sources of noise, so you can understand how you’re performing against your goals. As you might imagine, computing this baseline is complex, especially in buildings whose energy use fluctuates due to changes in occupancy or because of shifting IT, lab or process loads.

A good energy model compensates for three major elements:

  • Weather: all buildings (even data centers) respond to weather. We use temperature, humidity, and other factors to help explain how.
  • Calendar effects: most buildings have subtle patterns that reflect worker and equipment schedules.
  • Special days: holidays and demand response events are visible in the load curve.

A great energy model also compensates for shifts in building demand due to occupancy or process loads.

Let’s see how this plays out in the real world, with two different buildings. A basic temperature response plot helps to illustrate the power of trend normalization.

This chart shows daily use plotted against average daily temperature, with individual days color coded by type (open, closed, demand response event, and holiday). In the building below, we see a clear relationship between temperature and energy use when the building is open. The relationship between temperature and energy use is less clear when the building is closed. Likewise, the balance point, the temperature at which cooling load starts to affect energy use, is not readily apparent.

Daily energy analytics on use for closed, holiday, DR, and open days

The energy pattern on open days is clear, but what about closed days? Do warmer closed days use more energy?

Now, let’s use Gridium’s statistical model to normalize for non-weather trends to see if we can better highlight the relationship between energy use and temperature. Suddenly the pattern is a little more clear. Adjusted for trends, there is a clear temperature effect on weekends, which might indicate a savings opportunity. The balance point also shines through — in this case the balance point is below 50 degrees, a low figure that perhaps indicates an economizer malfunction.

Normalized daily energy use compared to mean temperature

Trend normalization brings greater clarity.

The next example is even more powerful. Look carefully and try to draw simple relationship between temperature and energy consumption. Despite the clear upward trend, the huge spread suggests a weak relationship between temperature and energy use. The hottest closed days have energy consumption profiles as high as open days.

Data analytics on the relationship between weather and energy
Building B, no trend normalization

What is the temperature response? Your guess is as good as mine.

Now let’s look at the same building with the energy use normalized. The temperature response is as clear as day, balance points are visible, and we have a good understanding of how the building behaves.

Normalized daily energy use compared to mean temperature closed, open, and demand response days.
Building B, with trend normalization

Why do you care?

Gridium’s trend normalization allows for much more accurate models, many times more accurate than those produced by the major ESCOs, energy consultants, and utilities. And all the stats happens under the hood. What matters to you are the implications for how you use Snapmeter.

Because the models are accurate, you can be sure that you are getting solid feedback on your energy management activities. Rather than tuning into to the “noise” of your building, you hear the true signal of how your building is working. This allows you to experiment with different operating procedures and immediately see the effect on the building, rather than collecting months of data and trying to compare year-on-year bills.

Further, you can take the savings to the bank. The results we report are statistically significant, and accurate models are highly sensitive to even small changes in behavior.

So yes, we’re nerds about the stuff. But that means the models are rock solid and your savings are certain. Not bad for only a few bucks a month.

About Tom Arnold

Tom Arnold is co-founder and CEO of Gridium. Prior to Gridium, Tom Arnold was the Vice President of Energy Efficiency at EnerNOC, and cofounder at TerraPass. Tom has an MBA from the Wharton School of Business at the University of Pennsylvania and a BA in Economics from Dartmouth College. When he isn't thinking about the future of buildings, he enjoys riding his bike and chasing after his two daughters.

0 replies on “How Snapmeter removes noise from your load curve”

You may also be interested in...