Energy Use Intensity, How to Self-Check an Energy Model
As a building practitioner, I have been building energy models and training some of the best in the business for years. I believe it is important to be able to self-check an energy model and gauge if the results are pragmatic or far off. Often, energy modeling is done by younger engineers who are building their knowledge bank and may not have a great resource to quickly ask if a number seems high or low. Equally challenging, most energy models are built to be comparative against an energy code or compared with LEED, yet more often the results shared with a design team and owner are often assumed to be good predictions of the buildings future performance as well.
While I have explained code modeling vs performance modeling dozens of times, I find it more fruitful to just improve my own skills at making models more robust. An energy model might be intended for LEED, yet there is no reason it cannot be built to be an indication of performance energy use. The key is using robust software to capture the actual systems and building up skills for quickly self-checking analysis before a more thorough review. This can allow even a new energy modeler to rationally think about their results and hopefully land a bit closer to reality. Having worked on over 2 dozen net zero energy buildings and energy models I can say with confidence a model will never be spot on. The human factor alone in building energy use makes any number within 20% a great estimate. Although even this range can be harder to hit than it might seem.
Here are four quick self-check methods:
- Educate Yourself on Real Energy Use Data
- Check Energy Estimates with Simple Math
- Stress Test Key Efficiency Assumptions
- Ask if the Answer is Rational
Educate Yourself on Real Energy Use Data
In the age of the internet, data is king. There is energy data all around us and seeking it out is the game. Look for information that can give you relatable information about built projects such as the energy use intensity, building use, and types of systems used. For years, the two data points between the Packard Foundation and the office I sat in in Oakland provided me a great baseline for heating energy in bay area offices. Packard was built with a great envelope and a heat pump and without ventilation heat recovery. The heating EUI was always around 6 kBtu/sf. The office I sat in was an old historic building with single pane windows, perimeter radiators and a condensing boiler and around 20 kBtu/sf. Both these buildings have capacity to be tuned, but they provided me with reference points to compare with models I worked on.
With the amount of publications on buildings there are lots of free data points, especially in California. The Net Zero Energy Case Studies published by PG&E catalogs eight buildings, many built by my old company, and shares the modeled and measured EUI. Shown below are the building total EUIs:
All of these buildings are good examples of the leading low energy building designs. Low energy offices or near net zero offices in California tend to be in the low 20s. While schools and other buildings used less frequently tend to be in the high teens. More valuable than whole building EUI can be the end-uses for lighting, heating, cooling, equipment.
Additional end uses are in the spreadsheet posted here: Measured EUI Real Buildings
Even the building you are in now has an energy foot print and some visible signs of the various systems running. Some good additional resources include:
- AIA 2030 Challenge – National Average EUI by Building Type https://2030ddx.aia.org/helps/National%20Avg%20EUI
- ASHRAE High Performance Buildings Magazine – Case Studies http://www.hpbmagazine.org/Case-Studies/
- New Buildings Institute – Northwest Zero Energy Watchlist https://newbuildings.org/resource/northwest-zero-energy-watchlist/
- New Buildings Institute – California Zero Energy Watchlist https://newbuildings.org/?s=CALIFORNIA+ZNE+WATCHLIST
- Department of Energy – Building Performance Database https://www.energy.gov/eere/buildings/building-performance-database
- UC Davis Campus Energy Dashboard – https://ceed.ucdavis.edu/
City Benchmarking Reports.
- All San Francisco Municipal Buildings – https://sfwater.org/modules/showdocument.aspx?documentid=6271
- City of Chicago Benchmarking – https://www.cityofchicago.org/content/dam/city/progs/env/EnergyBenchmark/2016_Chicago_Energy_Benchmarking_Report.pdf
- City of Philadelphia – http://visualization.phillybuildingbenchmarking.com/#!/
Check with Simple Math
Some end uses in energy modeling can be easily checked with a bit of simple math. This only works for a few energy end-uses like lighting, plug loads, fan power (to some extent) and with a few more steps, domestic hot water. In a majority of the energy models people run the schedules of operation are fixed. Equipment in energy models by space types (office, corridor, classroom) have a default fractional hourly schedule for weekdays and weekends. This fractional hourly schedule can be very powerful. Understanding how many Full Time Equivalent (FTE) hours in a schedule can help quickly estimate energy end-use. Below is a schedule of plug load equipment for a typical week. Assuming this week repeats for an entire year, this schedule adds up to 3,000 FTE hours.
If we knew the plug load density in an office was 1 watt / sf we could quickly estimate the energy:
1.0 W/sf x 3000 hrs/yr = 3000 W-hrs/sf-yr = 10.2 kBtu/sf
Once we start to have some limits on starting and stopping hours, we can start to look at energy budgets with a few assumptions. Lights and plug loads can be the easiest to ballpark first.
For most default energy modeling plug loads or lighting schedules, you’ll start to see that equipment generally operates roughly 2000-3500 hours/year. This is a pretty conservative number for some equipment, though it is a start. Daylighting can easily take the hours a year down to 1,500 in some buildings.
Lighting built to code in 2013 might be:
0.8 W/sf x 2500 hrs/yr = 2000 W-hrs/sf-yr => 6.8 kBtu/sf
And new lighting systems pushing the boundaries might be 0.3 W/sf.
0.3 W/sf x 2500 hrs/yr = 750 W-hrs/sf-yr => 2.6 kBtu/sf
The office uses about 8 kBtu/sf of plug loads and about 3 to 4 kBtu/sf of lights. It had some pretty solid daylighting controls, yet they controlled older linear fluorescence fixtures and our plug loads are fairly typical with a small server closet and printers.
I have put together a condensed PDF of the typical hours in schedules provided in Title 24 as a cheat sheet. Use these to become familiar with hours/year. These can be combined with any power and even used quickly with a fan or a pump, but make sure to consider the part-load curve of these systems. The affinity laws are a big deal.
Download the full PDF set of Title 24 condensed schedules here: Energy Modeling – Typical Schedules T24 FTE
Excel version for all those who want to dig deeper: Energy Modeling Schedule Tool – Appendix_5.4B_Schedules
Stress Test Key Assumptions
For some models, it can be too difficult to read the results alone and see if the model is providing good results. For larger models, consider stress testing key inputs. In the massive 8760 spreadsheet models I use to build of complex high-tech facilities, a way to evaluate the model was to change a key input from 0% or of 100%. In conventional energy modeling software, this can be hard to actually do. Some ideas to consider:
- If a particular system is intended to save energy, try replacing it with a conventional system. The easiest would be to run energy models with district cooling and district heating instead of actual equipment. By knowing the ton-hours or BTU-hours a model needs, you can estimate how efficient a chiller or boiler would need to be on-average.
- Remove key components and changing discrete inputs to default values. This is a technique from the days of trouble-shooting eQuest. In testing models, one way is to default key assumptions on power, size, or control to something more drastic or something more conventional and re-run the model.
The key goal of stress testing a model is to uncover blind spots you may not be able to see. Always question any special work-arounds or unique features of your modeling.
Ask if the Answer is Rational
People expect energy modelers to be self-accountable in today’s age of consulting. While we strive towards good quality answers, factors like time, money, and experience are hard to balance with every project. Even if a model is made for a specific purpose, it is always a good idea to think rationally about the numbers it estimates. If I have learned one thing, it pays off to put in a little more effort to give a robust estimate and answer my primary objective (LEED points for instance). So many clients will look to a model to provide multiple insights, even if this was these insights were not included in your scope of work.
Checking you work can ensure you do not over promising savings or performance. LEED and California energy code, Title 24, have a lot of special cases. Title 24 software is also an extremely complicated set of rules. A good question to ask yourself is “where are my energy savings coming from?”.
- Do I have a low pressure drop fan system?
- Does the building have a high efficiency chiller, or a warm chilled water setpoint?
- Are the windows and envelope performance vastly above a conventional building?
If you cannot answer or list out at least 3 key areas where savings are coming from, and the energy model still shows savings, you are not doing your job. It is imperative to be able to point to the key parameters, even if they are control sequences, lighting schedules or other small factors. With LEED, this becomes much more important. If savings are too good to be true, they will ask you for proof. After reporting 87% cooling savings for the Packard Foundation, with a design narrative and detailed report, LEED responded “Please justify.”
Try to check a model rationally and use this same line of questioning to peer review others models. Overall the goal of this post is to help energy modelers build confidence and skills quickly. The world of simulated energy modeling is quickly converging with real measurements and we need to be focused on being able to utilize both to be successful.