Food for thought....

18 posts / 0 new
Last post

Dear Forums,

I am busy preparing a short talk for the Fall ASHRAE Energy Modeling
Conference. The topic is "An Approach for Calibrating Existing Building
Energy Models to their Utility Consumption".

As part of the preparation, I will address the issue of how much difference
might result in energy conservation measure savings predictions if you use
actual weather data for the billing period versus TMY data.

To get a rough idea how much variation there might be, I looked at Degree
Days for a span of years. What a variation! (for the city I'm studying at

I am not yet sure how that affects total energy consumption - you'll have to
attend my presentation in Atlanta to find out J.

In the meantime, I am starting to think that existing building energy models
should use actual weather, not TMY data. Have any of you run similar
comparisons for existing building models?

James V Dirkes II, PE's picture
Joined: 2011-10-02
Reputation: 203

I would contact Zheng O'Neill. She was commissioned to do a similar study
for the Dept. of Defense as part of her work at United Technologies, which
she presented at IBPSA New York.

*Arpan Bakshi *

Arpan Bakshi3's picture
Joined: 2012-03-13
Reputation: 0


Actual weather data should be more important for calibrating a model to utility bills than for predicting ECM savings. Calibration usually corresponds to one year of utility data, and as you show below, weather can be highly variable year-to-year. (You could calibrate to an average of multiple years of data, and the variability would be reduced, making actual weather data less important.) For predicting savings however, you are concerned with performance over a period of five to twenty years (or whatever the expected lifetime of the measure is), so weather variability should have minimal influence.

It is an easy exercise to test the sensitivity of a particular model to weather by doing parametric runs using weather files from cities that are somewhat hotter or colder, say a few hundred miles north or south. This older study found 5% or less difference in general between using actual and TMY data (p. 58 of 128):
Of course, YMMV.

Best of luck with your presentation,

[Senior Energy Engineer 28Jun2012]

Bill Bishop's picture
Joined: 2012-02-25
Reputation: 7

Thanks to all who responded so far!

The two main points so far (without reading any of the suggested articles
and for all of you who have not yet had the pleasure of modeling an existing
building) are:

. Use actual weather data for calibrating an energy model to actual
energy use.

o This eliminates weather as an estimated variable and gives greater
confidence regarding your understanding of all the other variables.

. Use TMY data (or a generic weather pattern) for savings

o This eliminates the likelihood of predicting savings based on a
non-representative recent weather pattern.

I have lots of new reading as a result of this post and am looking forward
to it. Interesting that this is not really a new topic, as evidenced by the
1997 and 1998 publications referenced by some of you!

I've been thinking a fair amount lately about the huge and untapped savings
potential in existing buildings, but have not figured out the right sales
pitch for the owners yet. It seems that they have about 20 other factors
affecting their incentive for improving building operations. I'm not giving
up, however; there is too much good to be accomplished! (not to mention
challenge, learning and income J)

James V Dirkes II, PE's picture
Joined: 2011-10-02
Reputation: 203

Hi Jim,

Using actual local weather data (AMY vs. TMY) corresponding to the utility bills being calibrated against is standard practice in my experience. I started out using TMY weather and quickly ran into pitfalls as you're alluding to.


Nick-Caton's picture
Joined: 2011-09-30
Reputation: 805

EnergyPro can "normalize" the "predicted" results using historical
weather data. It seems to work.

Robert Wichert P.Eng. LEED AP BD&C

RobertWichert's picture
Joined: 2011-10-02
Reputation: 201

I've always thought it was a "no-brainer" to use actual weather data whenever you're
comparing simulation results to actual consumption data. Even with the earliest
degree-day software such as
PRISM (Princeton Scorekeeping Method) in the 1980's, it was stressed to use the degree days
from the period of record, and not the long-term average, so I'm not sure why this (using
year weather data) is such a revelation.

The variation in total energy consumption of course depends a lot on the building
Back in 1996, Dru Crawley and I wrote a paper on "Does it matter which weather data you
use in energy simulations?", for the ACEEE Summer Study on Energy Efficiency in Buildings
(it also appeared as two
separate ASHRAE papers at around the same time) where we took some prototypical building
models (Dru did commercial, I did residential) and ran them with various "typical year"
weather files and also 25 years of historical data in 10-12 US locations.

Joe Huang

Joe Huang's picture
Joined: 2011-09-30
Reputation: 406

Dear Joe,

No fair! You and Dru have been at the forefront energy modeling research
for most of my adult life, and have a big head start.

My guess is that you spent a lot of time preparing the actual weather files
for the research, however. Unless I'm missing something, the ready
availability of high quality (e.g., no big hunks of missing data) actual
weather data has been pretty limited until recently. With folk like Weather
Analytics getting on board and making it pretty easy to get and inexpensive,
it becomes a lot faster and lower cost than trying to clean some of the NOAA
/ NCDC data, not to mention getting good data for sites not in or near a
major city.

Kudos for being way ahead of the industry curve (at least my own curve)!
It's getting easier to catch up!

p.s., Dru sent me that paper and I'll be reading it with interest very soon.

James V Dirkes II, PE's picture
Joined: 2011-10-02
Reputation: 203


Sorry if I came off as a little smug. You've actually put in words the main reason why I
think building energy modelers have steered away from using historical weather data, which
is simply that of convenience, i.e., there are a lot of "typical year" weather files
floating around, but getting hold of
a usable historical year weather file takes a little more work.

For the study that I mentioned, Dru and I did not have to do anything because the SANSOM
data set,
available from NCDC, contained 25 years of historical weather files for 239 cities (same
as TMY2s).

That was in 1996, and the availability of historical weather data has only increased
exponentially since
then, although the building energy modeling community seems curiously to not have followed.
The most notable change, in my view, is NCDC's decision to put the entire ISH (Integrated
Surface Hourly) data online in 2005, and then to make it free to all in 2011. This means
that the raw weather reports
from major stations around the world ( ~ 1,500 in the US) are available stretching back to

I've been working with the ISH for many years now, and am able to generate complete
weather files from any ISH file within seconds, and have been providing that as a service
to customers. Sometime later this
year, I'll be rolling out something on the Web, but for now, those interested can just
send me an e-mail.

The idea that the ISH is of questionable quality is, in my view, rather backwards. The
ISH is a repository of the weather reports by the "official" weather stations around the
world, so if you can't trust that, what can you trust?

Joe Huang

Joe Huang's picture
Joined: 2011-09-30
Reputation: 406

I have a different take on this. I think a TMY is good enough for doing
model calibration that is used for ECM savings analysis on a whole building
level and yearly basis - I think this is the type of project Jim is
alluding to. In this scheme of things, the most important factors are the
scheduling of the various equipment/loads and the control logic and
gathering relevant data and making sure that the bills actually represent
the building in question properly. If you are going for hourly calibrations
etc, or calibrating to the few weeks of trend data that is collected,
then may be actual weather data is useful, although not 100% convinced on
this - depends on how much data one is able to collect.

If we go for a retrofit of a typical commercial building which is
mostly internal loads dominated (unless the perimeter to sqft is a lot
more-I am sure you can find papers related to this), I am not sure if
weather related issues are as important. Also, while doing retrofit, you
have access to three or more years of utility bill. Averaging out the
utility bills (removing obviously outlier data) and calibrating to it is a
reasonable way of proceeding to decide the savings due to ECMs. This takes
care of the weather issues as well as all the operating/internal load
issues that are also changing over time. Now you are no longer dealing with
one year.

It obviously depends on the building type - OA dominated buildings get
calibrated better with actual data - even here the more important data is
the amount of OA coming into the building and the control logic.

Residential and small buildings are a different category - they will get
calibrated better with actual data.

I agree with Jim that only recently a easy/non-time consuming way of using
actual data has become available. Has anyone tried it (the files from
weather analytics or equivalent) - is it as easy as using TMY without
having to worry if the results discrepancy is due to issues in the weather


R B's picture
Joined: 2011-09-30
Reputation: 4

I agree that measured weather data is the best way to go, but it is also difficult to get everything you need for a complete weather file. However, there are middle of the road approaches that work very well. Typically, the most important variables are the temperature and RH, which are also the easiest to measure or get from local weather stations. An easy approach is to replace the temperature and RH (or dew point temperature) data in the TMY file with your measured data. If you don't have hourly measured data, you can also adjust the temperatures up or down based on daily or monthly measured data. I have done this quite successfully on several projects.

Michael Deru, PhD

Deru, Michael2's picture
Joined: 2011-10-02
Reputation: 0

I think it's difficult to "calibrate" using only annual whole building data. Most of the
calibration that I've been involved uses monthly billing data and even at that admittedly
coarse level, using the actual year weather data becomes important, since the degree days
by month can easily vary by 40% or more year to year.

But there's another thing that puzzles me - when there are so many unknowns in the input
data, isn't it a relief whenever it's possible to eliminate one of them, like the weather
data? Asked another way, if you had the weather data for the right time period, why would
you not use it?

Perhaps the question gets back to convenience, and this is where I say that building
modelers should get better acquainted with what's now available in historical weather
data. Just remember that every "typical year" weather file is made from many years of
historical weather data. So, at any time historical year data should be more plentiful
than "typical year" weather data. The only question is how to get it. Happily, now that
the ISH is online, there's very little need to search elsewhere, at least for the US.

If you want to see the situation for 2010, look at
which shows all the ISH stations for which I've already created simulation-ready weather

Joe Huang

Joe Huang's picture
Joined: 2011-09-30
Reputation: 406

My experience here in Canada is that the larger organizations with
existing big buildings have done a significant amount of work over
the past decade to reduce energy use. I know a number of firms that
specialize in energy management that are very busy with organizations
like the banks, large utilities, etc.

The one thing to remember when pitching energy improvement is that if
energy use costs $2.00 per square foot, employees cost $20 per square
foot. Owners like benefits that keep employees happy.

>> Christopher Jones, P.Eng.

Chris Jones's picture
Joined: 2011-09-30
Reputation: 0

The discussion has become two discussions -
weather data and processing. And calibrating to
TRY (what I call acutal weather data for the same
billing period) or TMY, tyipcal weather data.

Environment had a number of weather stations with
hourly data - some with full weather data and
some with only a few data channels. Some
stations were daytime hours only. One had to
request a period of data and pay a setup
fee. The setup fee was more than the data fee so
it paid to request a bunch of sites at the same time.
We have an old C-code processor that converts the
EC data into an old DOE2.1e format text format
for conversion to a .WTH binary weather
file. The processor could use a number of
different channels depending on the weather
site. The input file signalled what channels
were coming in the text file. The Hirsh
processor can convert .WTH files to .bin files.

Environment Canada now has hourly data available
for much fewer sites. Cut backs closed dozens of
sites across the country. There is a significant
difference in weather between places like
Revelstoke and the closest site with TMY
weather. With the availability of stations
smaller places, one could aggregate a weather
file that captured the differences in temperature and humidity at least.

The good thing to happen with Environment Canada
is that the WEB site makes hourly data available
on line for a large number of stations across the
country. You can download a month at a time to a
.csv file. We now have a spreadsheet with the
appropriate psychometric functions that converts
the EC data to TRY data. We just save the data
to a text file for the Hirsh weather processor.

We use these TRY files for M&V projects.

>> Christopher Jones, P.Eng.

Chris Jones's picture
Joined: 2011-09-30
Reputation: 0


I hate to jump in this late in the conversation, but here are a few more papers that you might find useful:

Song, S., Haberl, J. 2008. ?A Procedure for the Performance Evaluation of a New Commercial Building: Part I ? Calibrated As-built Simulation?, ASHRAE Transactions-Research, Vol. 114, Pt. 2, pp. 375-388 (June ).

Song, S., Haberl, J. 2008. ?A Procedure for the Performance Evaluation of a New Commercial Building: Part II ? Overall Methodology and Comparison of Results?, ASHRAE Transactions-Research, Vol. 114, Pt. 2, pp. 389 ? 403 (June).

Haberl, J., Bou-Saada, T. 1998. ?Procedures for Calibrating Hourly Simulation Models to Measured Building Energy and Environmental

Data,? ASME Journal of Solar Energy Engineering, Vol. 120, pp. 193 - 204 (August).

Haberl, J., Abbas, M. 1998. ?Development of Graphical Indices for Viewing Building Energy Data: Part 2,? ASME Journal of Solar Energy Engineering, Vol. 120, pp. 162 - 167 (August).

Haberl, J., Abbas, M. 1998. ?Development of Graphical Indices for Viewing Building Energy Data: Part 1,? ASME Journal of Solar Energy Engineering, Vol. 120, pp. 156 - 161 (August).

Haberl, J., Bronson, D., Hinchey, S., O'Neal, D. 1993. ?Graphical Tools to help Calibrate the DOE-2 Simulation Program to Non-weather Dependent Measured Loads,? ASHRAE Journal, Vol. 35, No. 1, pp. 27 - 32 (January).

Bronson, D., Hinchey, S., Haberl, J., O'Neal, D. 1992. ?A Procedure for Calibrating the DOE-2 Simulation Program to Non-Weather Dependent Loads,? ASHRAE Transactions-Research, Vol. 98, Pt. 1, pp. 636 - 652 (January).

Haberl, J., Bou-Saada, T., Soebarto, V., Reddy, T. 1998. ?Use of Calibrated Simulation for the Evaluation of Residential Energy Conservation Options of Two Habitat for Humanity Houses in Houston, Texas,? Proceedings of the 11th Symposium on Improving Building Systems in Hot and Humid Climates, Texas Building Energy Institute, Ft. Worth, Texas, pp. 1-15 (June).

In addition, there is ASHRAE research project 1404 by Abushakra and Reddy that is about to finish that has good advice on the use of short term data to predict long term energy use. Copies of the final report should be available shortly from Mike Vaughn.

Jeff S. Haberl, Ph.D.,P.E., FASHRAE

Jeff Haberl2's picture
Joined: 2011-10-02
Reputation: 0

Classification: UNCLASSIFIED
Caveats: NONE

Many building control systems measure the outside air temperature, can the building's own measured outside air temperature be used? If you assume that the building can also measure wet-bulb temperature this still leaves cloud cover & wind speed (I'm not sure what other factors are involved)

Maybe a person could use the build's measurements and splice in any remaining variables from actual annual data and if there are still gaps, fill in with average data.

This brings up a big question; is there a right or wrong way to do it? Any set standards?

Jeurek's picture
Joined: 2010-10-07
Reputation: 0


On a related note - Thanks to Brian Fountain, a very intriguing (free) piece of energy simulation software called Sim-EB (available here: which focuses on modeling existing commercial buildings recently came to my attention. While quite powerful, the Sim-EB software provides a number of unique features that I think are appropriate to this discussion:

1) Given 'monthly' utility consumption and demand, and user-defined input parameter constraints, Sim-EB provides an interface for supporting 'auto-calibration' of DOE2.1e models.

2) Given user defined building descriptions of elements common across the engines, (form, fabric, HVAC, controls, etc), provides a means for using either DOE2.1e, DOE2.2 or EnergyPlus to perform calculaitons

3) Allows for easy import of user edited .inp files or .idf into the Sim-EB software. These files could have been created elsewhere, and just use Sim-EB for calibration or analysis, results review, etc.

4) Has a user website (,) which automatically creates and emails to users a user defined hourly 'AMY' model files (for Quebec related cities) to support model calibration.

5) Give a flexible, uniform data series (short term data logger data, 15 minute energy data, etc.), Sim-EB provides a flexible clustering tool for observing 'day-type load profiles' from within a larger data set. One can cluster model output as well, as Sim-EB provides a good visualization tool for comparing (day-type load profile) clusters of model output to clusters of metered data.

#2) lets people see, even for a shoebox model, the differences in the loads/systems/plants between DOE2.1, DOE2.2 and EPlus calculation methodologies. Of course, there are a great deal of 'switches' between the three different simulation engines that the software automatically sets, but one can export the input files and study the switch settings if they need to.

#5) above supports a case of real field data informing model input, i.e. the generation of hourly schedules given actual, real data!

This software was created and is distributed by the good folks at Hydro Quebec. As such, it supports both English and Metric units and the menu trees can be switched between French and English (but the help is in French only). I've use the Google translator tool and the example files to learn how to use the tool successfully. There are great pdf reports distributed along with the software which describe the clustering algorithms, the auto-calibration routines, etc.

Finally - if one renames the Sim-Eb .swdf file format weather files with a .zip extension, you can see how to replace them with non-Canadian files. It's a bit of a pain, but possible.

While it is certainly no panacea, I personally think this tool is a demonstration of how simulation software could be used to 'marry' simulation input/output with field data collection to provide 'better' calibrated simulations for examining retrofit potential. Recognizing the 'perfect is not the enemy of the good', this software, IMHO, forms a good compromise towards using energy models to create actionable results.

All the Best,

Chris Balbach, PE, CEM, CMVP, BEAP, BESA, BEMP

Chris Balbach's picture
Joined: 2011-09-30
Reputation: 1

In the IBPSA Building Energy Modeling Workshop, we have a few slides on
creating custom weather files, which I have copied below.

If your site has data available, it is not as painful as you use to be to
create a weather file based on actual weather data. See example below. I
have used the weather processing tool that comes with Energy-10 to modify a
TMY files monthly average temperature to match the actual weather year. Of
course it doesn't adjust for solar. But often you don't have the solar data
for the exact site anyway.

Also, RMI has commissioned a software developer to create a visual weather
processing tool to facilitate making custom weather files. It's called
Elements. It's hard to know when it will be realized but a reasonable guess
is Fall 2012. We're alpha testing it now.


[image: Inline image 1]

[image: Inline image 2]
[image: Inline image 3]

Ellen Franconi2's picture
Joined: 2011-10-02
Reputation: 0