Crunching the Numbers
When I set out to determine whether my 145-year-old home was a good candidate for electrification, I knew I needed to go beyond assumptions. So, like any good data nerd, I fired up R Studio and got to work. What I found challenged a lot of conventional wisdom and gave me a much clearer picture of what my house actually needed to stay warm in the winter.
Step One: Finding the Correlation Between Temperature and Therm Usage
The first thing I wanted to know was how closely my furnace’s natural gas consumption (in therms) correlated with the outside temperature. I took my daily therm usage and plotted it against the average temperature for that day. As expected, there was a strong negative correlation—the colder it got, the more gas my furnace used. But I wasn’t satisfied just knowing that.
I then ran the same analysis using the two-day average temperature instead of just the single-day average. The results were even stronger. It turns out that my furnace’s gas usage was more closely tied to the previous two days' average temperatures rather than just the day itself. This made sense—the thermal mass of my home doesn’t instantly adjust to temperature swings; it responds gradually.
To confirm, I also checked the three-day rolling average temperature. But that correlation wasn’t as strong as the two-day. That was my first real insight: my home’s heating needs are best predicted by the temperature over the past 48 hours, not just the present moment.
Step Two: Checking for Exponential Energy Use in Colder Temps
Next, I wanted to see if my furnace’s energy use ramped up exponentially as temperatures dropped—basically, did my house need disproportionately more heat on frigid days? My assumption was that it would, but the data told a different story.
I ran a regression model comparing therm usage to the two-day average temperature and looked for signs of an exponential curve. Instead, I found a tight linear relationship. That was unexpected. My furnace wasn’t struggling significantly more on the coldest days than on just moderately cold ones—it was just plugging along, burning fuel in a steady, predictable way.
I also added in weather conditions—sunny vs. cloudy days—just to see if solar gain was making any impact on therm usage. But the results didn’t show any strong indication that cloud cover affected heating needs in a way I could measure. If the sun was helping at all, it was subtle enough that it didn’t show up against the larger factors at play.
Step Three: Spotting the Overkill
One of the most eye-opening discoveries came when I examined how long my furnace actually ran on the coldest days of the year. I expected it to be working overtime, barely keeping up with the demand. Instead, I found that even on the most brutal winter days, my furnace only ran for about 13 hours total. That meant it was producing nearly twice as much heat as my home actually needed.
Would the furnace have been overstressed if it had to run for 24 or 36 hours straight in extreme cold? Possibly. But the key takeaway was this: I didn’t need as many BTUs as my furnace was designed to deliver.
The Takeaway: R Helped Me See Through the Noise
Most homeowners (including me, before this analysis) rely on rough estimates or contractor recommendations when choosing heating equipment. But after crunching the numbers in R, I had hard data telling me that my house wasn’t as needy as I thought.
The two-day rolling temperature average was the best predictor of gas usage. There was no exponential spike in heating needs at lower temps. And my furnace had way more power than I actually needed.
But could air-source heat pumps do it?