Degree Days

Weather Data for Energy Professionals

There are a few questions that we get asked fairly frequently - we've tried our best to answer them clearly below:

The # symbols just mean that you need to make the column a little wider in Excel. Excel shows # symbols when a column isn't wide enough to display the dates or numbers contained within it.

Position your mouse between the labels for column A and column B, hold the left button down, and drag the column wider. The animated screenshot on the right should help to make this process clearer. Alternatively you can double-click the mouse between the two columns to make Excel auto-fit column A.

We say "base temperature(s)" because, for buildings with heating and cooling, you'll usually want one base temperature for heating degree days and a different base temperature (usually higher) for cooling degree days.

The optimal base temperature(s) depend on the building. You can never determine them perfectly (degree-day-based calculations can never be *that* accurate), but using degree days with base temperature(s) that are approximately right for your building can significantly improve the accuracy of your calculations.

If you're not sure what "base temperature" actually means, we suggest you start with our introduction to degree days.

Next, we recommend this article about choosing base temperatures - it should help you make a rough estimate of the base temperature(s) of your building.

To get a more accurate estimate we recommend regression analysis. This is easiest done with our regression tool - just go to the Degree Days.net web tool, select "Regression" as the data type, and follow the instructions from there. This will give you a shortlist of regression models (each with different base temperatures) that give the best statistical fit with your energy-consumption data. Use your estimates from the above-mentioned article on estimating base temperatures to help you choose the regression model with base temperatures that make most logical sense for your building.

We've written an article explaining how to calculate/prove energy savings. This is a very common use of degree days for people who have made efforts to reduce energy consumption in their building(s).

*Ideally* you'd have your heating and cooling energy consumption metered separately. But, if that's not an option (e.g. because you have a all-electric building with one meter only), you'll need to make estimates...

"Baseload" consumption is a term that's used to describe the energy consumption that *doesn't* depend on the weather, and that's typically pretty constant throughout the year. If you can estimate your baseload energy consumption, you can deduct it from your recorded consumption figures to get an estimate of the consumption that's heating or cooling only.

You can often estimate the baseload by investigating the consumption over a period when the heating/cooling was switched off. Though the approach that's usually recommended is regression analysis.

Baseload energy consumption is a very approximate concept for most buildings. It's often important to estimate it and remove it from consumption figures, but it can be difficult to do this well. Our Degree Days - Handle With Care! article explains more about the relevant issues.

The best option would be to separate the heating and cooling energy consumption by metering them separately. But, if that's not possible, we recommend multiple regression of HDD and CDD together. Our regression tool does this automatically, and will also help you find appropriate HDD and CDD base temperatures by testing thousands of multiple regressions with different HDD/CDD base-temperature combinations to find the ones that give the best statistical fit.

Just go to the Degree Days.net web tool, select "Regression" as the data type, and follow the instructions from there.

We don't actually *have* a database of degree days... Each set of degree-day data that you download from Degree Days.net has been freshly generated from raw temperature data, just for you. It's this *on-demand* approach that enables us to offer all the options that we do (thousands of weather stations, lots of base temperatures, a choice of timescales etc.).

This might not be the best analogy, but here goes anyway... Most sources of degree days are like times tables - the results are calculated in advance for a fixed set of numbers (or locations and base temperatures). Degree Days.net is more like a calculator: you tell it exactly what data you want, and it calculates it for you there and then.

Anyway, analogies aside, if you want to buy a database of degree days, we suspect that you'll be interested in the following two questions and answers:

Yes! The Degree Days.net API makes it easy to automate the process of getting data out of our system.

Click here for more information about our API.

Yes! With Degree Days.net Desktop you can specify a big list of locations and download data for all of them into a single spreadsheet.

If you've seen this page before you might remember us talking about the possibility of a Degree Days.net Excel add-in. That was our original thought, as we wanted to take advantage of our experience developing this Excel-based energy-management solution. But we decided in the end that a standalone desktop app would be better for downloading data - more user friendly and more powerful. And we think it's turned out very well!

Click here to find out more about our desktop app.

Outside temperatures vary throughout the day, so keeping a building at a constant internal temperature can often require both heating *and* cooling on the same day (e.g. heating in the morning, cooling in the afternoon). Degree Days.net calculates data in a way that takes into account the temperature variations *within* each day. This makes the calculation process more complicated, but it results in degree days that correlate better with real-world energy consumption.

If you want to understand this better, please read our page on calculating degree days. Though there's no real need to understand the calculation process if you're just using the data for your own calculations.

On a related note: using heating *and* cooling to maintain a constant internal temperature is bad for energy efficiency. Typically you want the internal temperature above which the air conditioning comes on to be a few degrees higher than the temperature below which the heating comes on. Assuming you're doing this in your building(s) you will typically also want HDD and CDD with different base temperatures.

Firstly, temperatures recorded by different weather stations at nearby locations can vary by quite a surprising amount. This is why it's good to use data from a weather station that's as close as possible to the building that's energy consumption you're analyzing.

Secondly, it's likely that the two sets of data you're comparing were calculated using slightly different methods. We calculate degree days using a more sophisticated method than many other sources.

Finally, it's possible that you're actually comparing the data from our site with population-weighted degree-day data. Population-weighted data can be useful for utilities that are doing simple aggregate analysis of energy consumption across a large region (e.g. the state of California), but it's not so good for analysis of the energy consumption of individual buildings.

We've made a whole separate page about degree-day-data calculation.

We are big fans of fine-grained energy data such as hourly or half-hourly data - its analysis is really useful for identifying and quantifying patterns of energy consumption that are impossible to find in courser data like daily, weekly, or monthly figures. We even make a software package called Energy Lens that is designed specifically for the analysis of fine-grained energy data.

However, heating/cooling energy consumption is a lot more complicated than the energy consumption of simple on/off electrical equipment like lighting. At a fine-grained level the relationship between the temperature outside a building and the energy consumption inside a building is neither simple nor direct. Buildings retain heat/cool in a manner that varies with construction and usage and is difficult to model accurately. There are complicated time lags (typically of the order of hours) between changes in outside temperature and the increases or reductions in heating/cooling energy consumption that result inside.

Heating/cooling systems often cycle too, operating in bursts that are unlikely to fall regularly within fine-grained metering intervals. And intermittent heating/cooling (e.g. a building that is only heated/cooled during office hours) complicates the fine-grained temperature/energy-usage relationship more as buildings that cool down (or warm up) when unoccupied need an intense period of heating/cooling (and energy consumption) to bring them back to temperature again. The energy consumption of that intense period of heating/cooling is only loosely related to the outside temperature during that period - it's much more dependent on the temperature changes that came beforehand.

These factors can be modelled, but doing so is the realm of sophisticated building-simulation software that requires detailed information about a building's construction, equipment, and usage patterns. If you try a simple regression of detailed energy data against detailed temperature data, the noise and complexity in the temperature-consumption relationship means you are unlikely to get useful results.

However, if you combine your detailed energy data into, say, weekly totals (a good option since most buildings operate on a weekly cycle), the noise is smoothed away and you will typically see a simple, direct relationship with degree days.

Also note that properly-calculated degree days encapsulate all the relevant variations within the detailed temperature data that they came from. So you're not losing detail in that regard.