Degree Days and Weather Normalization
Why do I need degree days?
You need degree days to find out if your energy projects actually save energy or not. Suppose I tell you that my electric bill for the hot month of August was only $50. Doesn’t mean much does it? There is no reference. So I tell you that the bill was $75 last August. Then you come back and tell me that last August was much hotter than this August, so I didn’t use as much air conditioning this year as last year. So, it only makes sense that my electric bill was lower this year.
The question is, “did my energy saving projects really save energy or not?” It’s clear that we need a way to make a fair comparison of energy use this year to that of last year, independent of the weather. The key tools used to do this are degree days and weather normalization. These tools give us a way to change an apple-to-orange comparison into an apple-to-apple comparison.
Let’s start by defining degree days. But first, we need to define a term called “base temperature”.
Definition of base temperature
Base temperature, sometimes call balance point, is simply the outside temperature at which a building needs no heating or air conditioning to maintain a constant desired internal temperature. It is the balance point where the desired inside temperature is maintained without air conditioning or heat. Air conditioning would be needed if the weather was a degree warmer, and heat would be needed if it was a degree cooler.
For most buildings, the base temperature falls in the range of 55-65 °F. When the base temperature is not known, 65 °F is the default value typically used. Obviously, every building is different. Office buildings with lots of hot occupants, copy machines, and computers have a lot of internal heat, making the base temperature lower than for buildings with fewer heat loads. Lots of variables go into determining the base temperature including building insulation and construction, internal heat load and weather (windy, clear or cloudy day, etc.) but most buildings have a base temperature that is fairly constant when averaged over long periods of time.
Now, back to degree days.
Definition of degree days
For the purposes of saving energy, we deal with two types of degree days: heating degree days, and cooling degree days. When heat is required to keep a building at a constant desired temperature, we experience heating degree days. When air conditioning is needed, we experience cooling degree days. Degree days are based on 24-hour periods, hence the word “day” in the term. One heating degree day means the average temperature for one day is one degree below the base temperature. One cooling degree day means that the average temperature for one day is one degree above the base temperature. If the average temperature was 2 degrees above the base temperature, it would be 2 cooling degree days for the 24-hour period. Let’s work through an example:
- Our example is a house with a base temperature of 60 °F.
- The average temperature over the 24-hour period on Monday is 66 °F
- The average temperature for Tuesday is 68 °F
- Then on Wednesday, a cold front comes in and the average temperature is only 56 °F
- What are the heating and cooling degree days for the Monday through Wednesday period?
Monday is 6 cooling degree days (66 °F – 60 °F). Tuesday is 8 cooling degree days (68 °F – 60 °F). And Wednesday is 4 heating degree days (60 °F – 56°F).
So, over the 3-day period, we have accumulated 14 cooling degree days and 4 heating degree days. You can see that doing this over a month or year would be a great way to compare weather temperature conditions between two periods.
Cooling degree days are roughly equivalent to the amount of cooling required for a building. Doubling the cooling degree days roughly doubles the cooling requirements. Likewise, doubling the heating degree days would roughly double the heating requirements.
Heating and cooling degree days are always positive numbers. Every day has heating degree days, cooling degree days, or zero degree days (if the average temperature is exactly equal to the base temperature for that day). A day cannot have both heating and cooling degree days since calculations are based on 24-hour averages. If a day has heating degree days, then cooling degree days is zero for that day, and vice versa.
Now I’ll show you the easy way to get heating and cooling degree day data. The best way is to simply look up the historical weather data for your area on the internet. Go to http://www.degreedays.net and enter your zip code. Choose the weather station closest to you. Choose either heating or cooling for the type of degree days you’re interested in. Adjust the remaining parameters as desired. Normally you’ll want to leave the base temperature at 65 F. Click the “Generate Degree Days” button. Then, after a few seconds, you’ll see a download button at the top of the screen. Click on it to download the CSV (comma separated values) file. You can open it in Microsoft Excel or your favorite spreadsheet program. Play with it and perhaps compare a month last year to a month this year. Any signs of global warming? 🙂
There are lots of resources for degree day data on the internet. Another popular website is http://www.weatherdatadepot.com As with the above site, it allows you to choose the base temperature (called “Balance Point” on the site). If you have enough data on your building to know its exact base temperature, you can adjust it to achieve more accurate results.
Everything should be self-explanatory on the websites, except perhaps “growing degree days”. This is used by farmers to predict plant development rates based on temperatures. Think of it as heating and cooling degree days applied to plant growth instead of buildings. That subject is beyond the scope of this article, so I’ll leave it at that.

Linear Regression Chart for Last Year’s Data
Weather Normalization
In the example at the beginning of this article, we saw the need for an apple-to-apple comparison of energy used this August to that used in August a year ago. We can look up the total heating and cooling degree days for August this year and last year. But when we only compare degree days, we still have an apple-to-orange comparison because of differences in weather.
This is where “normalization” comes into play. We have to normalize our data by transforming it to a common reference period. This effectively removes variations in weather from the comparison.
The way we do that is to first graph, day-by-day, the energy used last year against last year’s temperature for each day. For electricity, this is typically kWh/day on the y-axis versus average temperature/day on the x-axis of the graph. Then, we can pick a temperature and use the curve to show how many kWh our home used based on that temperature. The year whose data makes up this graph, last year in our example, is called the “base year” (no relation to the base temperature). The base year is the reference period all future energy usage is converted back to for an apple-to-apple comparison. And we write an equation, usually a simple equation for the straight linear regression line, that describes the energy to temperature relationship shown on the graph.
We now have an equation that represents the energy versus temperature relationship for our building last year. If nothing changes in the building, that same relationship should hold true for this year. Of course, temperatures are different this year, so the kWh will be different each day.
So this is what we have at this point:
- For last year, an equation that gives us the energy used for any given temperature.
- For this year, we can go to the internet and look up historical weather to determine the cooling and heating degree days for the period in question.
We have all we need! Remember that degree days are roughly proportional to cooling or heating requirements. This gives heating or cooling requirements over a period of time. So we can combine the two bullet points by inserting degree days into the equation. This lets us determine how much energy we would have used this year based on last year’s graph and this year’s temperatures. Got that important point? Degree days is handy because it is a single number we can use in place of calculating energy usage each day and summing it over the period.
That’s the key concept explaining how we get an apple-to-apple comparison. Since we know how much energy we would have used this year if no energy improvements had been made, we can compare that to what we actually used this year with our energy improvements included. This holds all variables constant except temperature, which is by far the most influential variable in determining energy use.
When doing weather normalization we compare how much energy we would have used this year to how much energy we actually did use this year.
Our example uses last year as the “base year”. It provides that important reference that we were initially missing. A base year can be 1 year or 5 years ago, it doesn’t matter. Base years are chosen to match each individual situation.
We make the assumption that everything in the building stays the same as it was in the base year, so if the temperature this year was identical to the base year, the energy used would also be identical. If anything changes (like building additions, new or added equipment, more people, etc.) then we would need to adjust for those changes.
How Professionals Do the Calculations
I hope you’re getting a feel for the concept of weather normalization. Grinding through the actual calculations is tedious, and there are some statistical pieces that I didn’t bring up. For example, when making the graph for energy versus temperature you’ll find that every building has a base load that never goes away. Therefore, the degree days relationship portion of the curve doesn’t start at zero, it starts at the base load energy usage and goes up from there. The base load is not temperature sensitive. It is entered into the equation as a constant. Intersection of the base load and the temperature sensitive degree day curve is then used to determine the base temperature for a building.
Scattered data points create another complication in making graphs. A straight line rarely connects the points perfectly. A “best-fit” straight line must be used and linear regression analysis is used to create the best fit. Data points tend to be scattered as depicted in the Energy vs Temperature chart shown above. Linear regression usually provides an equation that fairly represents this type of data.
So you can see why I say the calculations are tedious. Even with a spreadsheet, it is quite tedious. Most energy professionals use specialized software to do these calculations. Using the software still requires a solid understanding of the concepts discussed in this article.
How accurate is weather normalization?
The variables that affect energy usage are mind boggling. Weather normalization typically only considers one variable – temperature. What about solar gain? A clear, sunny day heats a building much more than a cloudy day at the same temperature. Wind has an effect on convective cooling and heating. The type of construction, windows, shading, insulation, lighting and HVAC systems all affect heating and cooling. Building utilization is also a big factor. It determines the number of occupants and quantity and types of equipment found in the building. Heat-producing equipment will greatly offset the base temperature in an office building. In contrast, a residence typically has a low internal heat load.
All the variables, not just temperature, would be taken into consideration in an ideal world. But that complicates the calculations dramatically and increases the cost of finding reliable data. Fortunately, back-testing data over the last 20 years shows no significant improvement in accuracy when additional variables are taken into consideration. So, at least for now, we settle for “very good” rather than “near perfect” normalization results.
Similarly, there is room to improve degree day calculations. The daily temperature is typically calculated as the average of the high and low temperature for the day. But that average might not match with what nature presents to us. In a 24-hour period, the temperature might stay high for much longer than it stays low, offsetting the average by producing more cooling degree days.
A much more accurate way to calculate degree days is to do it on an hourly basis and then sum the 24 hourly calculations. Many of the online services actually do this using hourly weather data from airports or other reliable weather stations. So if hourly is good, how about 15-minute data? Sure, it would be even more accurate, but at some point, the juice is no longer worth the squeeze. Considering how fast temperature changes, hourly data is “excellent”, and using the average of high and low temperatures is “quite satisfactory”.
Summary
In this article, we looked at the issue of comparing energy used by a building for two different periods. We saw the need to find an apple-to-apple method to compare the two periods. So we defined base temperature followed by a discussion of degree days. This provided the building blocks we used to explain weather normalization. Specialized software performs weather normalization calculations in nearly all applications today. The results are good, but not perfect because calculations typically account for only one variable, temperature, out of many variables. Indeed, temperature is the most significant variable, and over the long term, we get very good results from using it and ignoring all the other variables. Looking to the future, incremental improvements to normalization calculations will occur as software improves and additional data becomes more commonly available.
Let me know your thoughts on degree days and weather normalization. And let me know about related topics you would like me to cover.
Leave A Comment