This is the first part of a two-part blog.
If your building has multiple energy meters dedicated to floors or systems and your team can access daily or hourly energy data, you’re in really great shape to judge energy performance of your building. You can monitor and detect relatively small changes in energy use that can guide your detective work and document improvements.
What can you do if your building only has whole-building energy meters and monthly utility data?
Option 1 Use ENERGY STAR Portfolio Manager information
Portfolio Manager tracks whole building energy use summed over all energy sources. You can see weather-adjusted energy intensity—12 months of energy use per square foot, month by month. For buildings that have an ENERGY STAR rating, you can see the ENERGY STAR score, which accounts for a set of use factors, including hours of operation. One drawback of the ENERGY STAR system is that you can’t disentangle electricity from natural gas or other energy sources and you can’t get individual month weather-adjusted energy use.
Option 2 Do It Yourself
A health system client who uses Energy Stewards to organize their energy management of hospitals and clinics asked me a reasonable question. How could they estimate recent impact from a series of retro-commissioning changes, disentangling electricity and natural gas?
I’ll show you how to detect small but meaningful changes using monthly utility data—changes in the range of 5 to 10% and over a time frame less than one year. While I did the analysis in the statistical package R, you can do the work in any statistical package or even in Excel®.
Method to Check Whole Building Changes in Energy Use
- Get the right data.
- Set a baseline period.
- Plot the data to understand patterns and unusual values.
- Model the energy use as a function of temperature or degree days.
- Predict the energy use beyond the baseline period.
- Compare the actual energy use and predicted use. Use a control chart to judge if there are savings.
- If step 6 gives you a signal of savings, estimate avoided energy use and costs
This post will cover steps 1 through 3.
Get the Right Data
A building’s energy use is driven by many factors, some under your control and others not. The main factor not under your control that has a big impact on energy use is weather—and for most parts of the temperate world, weather reduces to outside air temperature.
Here’s a view of the monthly electricity and natural gas use of our client’s hospital building, located in the northern half of the United States. We clipped the view from our Energy Stewards application. The use of electricity peaks in the summer, driven by air-conditioning; the use of natural gas peaks in winter, driven by space heating.
It looks like the summer of 2013 has a lower peak in summer electricity use and the winter peaks in 2011-2012 and 2012-2013 natural gas look lower than the winter of 2010-2011.
Is the reduction in summer electric energy use really from energy efficiency? Or was the summer just cooler? How can we start to answer these questions?
At the very least, we need to take outside temperature into account to judge whether the energy team has made any improvements.
The client also shared additional data on monthly surgical cases and patient days. We checked whether changes in these numbers matched changes in energy use. We couldn’t see any relationship, so we set these numbers aside.
Energy data: we downloaded electricity and natural gas data from Energy Stewards—not the original utility data with bill periods misaligned with calendar months but adjusted, complete month data. For electricity use, we have complete month data from January 2010 through September 2013; for gas use, , we have complete month data from January through August 2013.
Energy Stewards breaks up each bill into energy use per day and then reassembles the energy per day according to the calendar. While Energy Stewards knows about leap years in building the energy records for the month of February, it doesn’t do anything about the length of the month. At 28 days, February is almost 10% shorter than a 31 day month like January. That’s a problem when working with natural gas use, high in cold in months like February. There’s a work-around, keep reading.
Temperature data: We got temperature data from the Midwest Regional Climate Center. The MRCC provides free downloads of National Weather Service data by weather station. We used the weather station closest to the building. In addition to the average of the daily mean temperatures for each month, we got the “heating degree days” and “cooling degree days” for each month.
Essentially, a heating degree day assesses how cold each day is and then adds up the coldness to get a total for each month. A 28 day month will have 28 days of coldness, added up. A 31 day month will have 31 days of coldness, added up. The practical implication: if we use heating degree days as our measure of outside air temperature, we get around the short-month February issue. If you want to get more details on degree days, read a blog post I wrote last year at our Energy Stewards site.
When we got the temperature data from MCIS, we used the default degree day base value of 65° F. This base often works well for heating degree days but not so well for cooling degree days, mostly because some air-conditioning runs all year round in big buildings like hospitals in temperate climates. You'll see the problem if you keep reading.
Set a Baseline.
We used January 2010-December 2012, 36 months of records. The client started making changes to reduce energy use in late 2012; they want to see if 2013 shows reduced energy use, after accounting for outside temperature.
Plot the Data to Understand Patterns and Unusual Values--"You can see a lot just by observing." (Yogi Berra)
Here are plots of the baseline period monthly records: electric energy use in kilowatt-hours (kwh); gas energy use in therms; cooling degree days—base 65° F (CDD); heating degree days base-65 ° F (HDD) and average daily mean temperature.
It looks like there’s an upward trend in electricity (kwh plot, upper left panel) in addition to the seasonal swings. You can see that the cooling degree days plot has many zeros (second row, left panel), which means we may be better off relating electricity use to average monthly mean temperature.
We also made several scatter plots to begin to understand the relationship between energy use and temperature for this particular building.
The plot of kwh vs Average Monthly Mean Temperature suggests a curved relationship. The plot of kwh versus cooling degree days has a funny stack at zero cooling degree days, which means the base temperature of 65 is too high. For gas, the plot of therms versus heating degree days suggests that a straight line will represent the relationship pretty well.
Part 1 Conclusion
We’ve got the energy and temperature data organized for our baseline period. We also have some ideas on how to model the relationships between energy and temperature--we may need a curved relationship between electricity and outside air temperature, for instance.
In part 2 of this blog post, we’ll go through the remaining steps of our method and answer the question about energy use this year.