On Bloomberg’s Masters in Business podcast, Fred Fox shares insights–gained over 21 years of studying the relationship between sales figures and weather data–into the dollar cost of weather impacts on consumer behavior.
For example, 36% of the world’s GNP is directly affected by the weather, and the domestic Do-It-Yourself industry took a $3 billion hammering during this year’s unseasonably wet and cold March. On certain consumer product categories, filtering out weather volatility can improve forecast accuracy between 20-50%.
When Fred talked about the predictive ability of weather analytics to forecast consumer demand, we heard echoes…weather normalization and temperature response are the scaffolding for accurate energy efficiency measurement and verification and utility budget forecasting.
At any given place and time, the weather only recurs about 20% of the time from one year to the next. Without weather-normalizing their data, the retailer or business forecaster gives him or herself an 80% chance of being wrong.
It’s no different for your building. And this is why building operators use Snapmeter to gauge energy efficiency returns and forecast costs.
Compared to last year, this building used less energy in response to changes in temperature.
Weather-normalized energy use
In nearly every case, weather has an impact–and sometimes quite a big impact–on how much energy the building uses. And just like comparing the sales figures for Lennox furnaces or Patagonia fleeces, it’s important to quantify the change in product demand or energy use that can be attributed to the weather, versus other factors such as price or baseload use.
For the building pictured above, imagine if September 2016 had above average temperature and humidity while September 2017 was below average. On the face of it, the two months are each an apple and an orange. But by filtering out weather effects–by correcting for observed differences in temperature–you can isolate the changes down to a degree of fidelity that makes the whole exercise worthwhile.
Weather-driven energy use
Weather-driven use is one of the isolated drivers of total energy use that falls out from a variance analysis. It relies on the combination of two quantifiable facts: weather conditions and your building’s response to changes in the weather.
Snapmeter estimates the net effect of weather conditions on your building’s energy use. Compared to September 2016, the building pictured above used 7.9% less energy in response to changes in Mother Nature.
Energy use due to weather conditions
The net effect of weather-driven use is further split between two related analytical facts: temperature response and weather conditions.
Weather conditions variance assumes that nothing has changed in the current billing period, versus the same billing period last year, about how the building responds to changes in the weather. Put another way, Snapmeter holds constant the building’s temperature response, and it does this by drawing an average temperature response curve for the building and by using that single temperature response curve in both the current billing period and the billing period from last year. By using one, average, temperature response curve, Snapmeter can calculate the change in energy use associated only to the change in temperature.
Imagine a weekday average temperature reading in September 2016 of 88 degrees and, in September 2017, a weekday average temperature reading of 84 degrees. With a single temperature response curve, it’s possible to compare the resulting energy use figure at 88 degrees (apple_A) with another estimate of the energy use figure at 84 degrees (apple_B). This math isolates the change in use due to conditions external to your building.
Energy use due to changes in temperature response
To calculate the temperature response, Snapmeter runs a variant of the weather conditions math.
Instead of using a single temperature response curve across the two billing periods, we’ll now use two different temperature response curves. That is, Snapmeter calculates the temperature response curve for the billing period last year, and a second temperature response curve for the current billing period, and returns the difference between those two response curves as the change in temperature response.
If you upgrade the efficiency of that Lennox furnace, or other HVAC hardware in the building, you will see an improvement in the temperature response curve. The weather-effect will still be there, but the amount of energy your building uses in response to the weather will be lower. Refinements to BMS scripts can have similar results. This math isolates the change in use due to conditions internal to the building.
The difference between weather and temperature
Weather is a combination of temperature (dry-bulb) and humidity (wet-bulb). The main factor here is almost always temperature. For now, this doesn’t include cloud cover or whether or not it’s raining.
The difference between Minneapolis and Miami
Another of Fred’s insights is that weather drives needs, not wants. This dynamic can be seen in the difference between the point on the thermometer when thermal underwear starts to sell in Minnesota (25°F) and when it starts to sell in Florida (65°F). Apparently people, just like buildings, have different temperature responses!
Let us know if you need help filtering out your building’s weather effects.
