Ahmed:
?
The question is a bit perplexing.
?
If you spend enough time inputting rate details, time of use items, and taxes, you can literally come as close as you want. if your energy is off 5% annually, your costs should be similar.
?
John R. Aulbach, PE, CEM
On all the DOE modeling projects that we have done, of an existing
building, they have required the model to be within 1 to 2% or less of
the actual energy usage, based on an average two year history. Is that
what you are talking about?
How much effort was put into minimizing the error of your inputs to
justify that level of accuracy in your output? For example, did you
convert real weather data for the two years and use an average of the
simulations? Were you modeling small buildings so you could get a fairly
reasonable infiltration rate empirically? Were occupancy schedules
trended, and for how long? Were all the systems and controls working
correctly, with all sensors calibrated regularly?
1 to 2% seems to me to be fairly unreasonable. Unless you do an amazing
job verifying your inputs, in my opinion that level of precision doesn't
get you a better model. If you set everything up as best you can and it
comes in there, great! But a correctly set up model can be off by over
2% because of a couple "El Nino" years, a facilities guy locking a
humidity high limit to 50% for a summer, or any number of operational
factors. On the other side of it, we've tried to match a simple DOE2.2
model to a DOE2.1e model with limited success (I don't think we got
within 3%, although we didn't spend too much time with it). Who knows
what the difference would be with E+.
I know I always squirm when I'm asked to do existing building models,
and it may be irrational. Heck, if the DOE is asking for 1-2%, I
probably am being irrational. But it seems to me that the error from
assumptions could easily swing a model 1-2% (what would a 30% error on
your infiltration do to an otherwise correct building?). What are other
people's thoughts?
Much of the discussion on this so far has been about modelling existing
buildings.
If your situation is that you are modelling a new building and using
"default" (MNECB or A90.1) schedules, plug load values etc. Then the
actual building operation could be 50% higher than the model is
predicting. The difference will be in the occupancy, the schedules and
especially building operation and occupant behaviour. All of these are
hard to quantify at the design stage.
While I (and I am sure many others) have books to write regarding my
thoughts on the topic, I think I might suggest a simple step back: The
concept of predictively modeling to within a few percentage points of
accuracy is ridiculous. There are simply way to many things that could
not be known. HOWEVER: The concept of building a model for an existing
building, whose modeled energy consumption/costs fall within a 1-2% of
historical utility records, is entirely feasible - and a reasonable
requirement if the goal is to generate a model for predictive purposes.
How often such predictive models' accuracy is misinterpreted is
something I get depressed to think about.
I haven't done the DOE-modeling work that's being referenced with such
requirements, but I have done work for educational clients
(physics/building science departments) who wish to get these models set
up for ongoing study/tweaking purposes. The exercise is challenging,
and as close as energy modeling gets to "fun," when you can rest at ease
knowing the client is fully on board with what the model is and isn't.
It's fully rational to squirm and cringe when you have to make models
that you know will be mis-used and mis-understood, despite your best
efforts. Fortunate is the practicing energy modeler who gets to work
for fully educated clients and design teams all the time =).
Rather than drill David, I think it would be safe to assume those DOE
requirements exist to calibrate a model to a given degree to whatever
historical records are available, NOT to mandate a level of accuracy for
predictive purposes. If I'm wrong, I hope that work never crosses my
desk!
I should make it clear that my initial questions weren't meant to drill
David. I'm genuinely interested if the DOE projects required an
extremely high level of technical accuracy on the input side. (Although
they were also somewhat of a reaction to those requirements.) We have
this discussion in our office every so often, and I'm usually somewhat
on my own, but I'm still not convinced.
Would you mind elaborating a little more about why you feel this to be
true:
"HOWEVER: The concept of building a model for an existing building,
whose modeled energy consumption/costs fall within a 1-2% of historical
utility records, is entirely feasible - and a reasonable requirement if
the goal is to generate a model for predictive purposes."
So let me ask it this way. Let's say you've got a model, and it's coming
in about 6% low (both heating and electric), but to the best of your
knowledge everything is set up properly. You've got to start making some
educated guesses to close the gap. Do you assume the boiler needs a
tuning (drop your efficiency a few percent) and the AHU coils need to be
cleaned (bump up your static)? Or do you bump up your infiltration
slightly (because it's probably not a very tight building). Or did some
maintenance guy came along and switched on both the boiler pump manually
on when he got a cold complaint, and no one returned them to auto? Or
maybe you have data from a few years with abnormally warm summers and
cold winters. I could go on practically ad nauseum. There's just so many
things that can be wrong with an existing building, as all of you
probably know.
Whatever you choose, that guess you make may interact (or lack an
interaction) with some measures you're predicting. So, predicatively,
what makes a 1-2% off model better than a 5-6% off model unless you
really spend the time and find out whether your inputs are accurate?
I think it's important for anyone talking about modeling an existing building to qualify such/any levels of prediction with the amount of data that's used to calibrate the model. Utility bills and a simple walk through type audit (say a Level 2) will find you feeling pretty cheery if you're at +/- 10%. If you're at +/- 1-2%, it's most likely happenstance and you should trust that the model is dead-on accurate. If you've heavily instrumented the various systems throughout the building and your calibration finds you within 1-2% of actual consumption, you're doing your job well and have spent a fair bit of change on the effort.
As for new buildings and predictive modeling, this is where benchmarking and feedback to the modelers and design teams grows ever more important (and talking about 1-2% is nowhere near practical at this point). There are beginning to be more cases of litigation where large discrepancies exist, so be sure you know what you're promising and that you cya. It does seem that there will likely be a harder, faster push to focus on operational performance in terms of ratings/classifications as opposed to making the modeling better, but I do foresee the market trending towards the need for us to demonstrate that our models are getting better as time goes on if we want to keep market share.
In one of my learning course, the instructor mentioned that after they
used eQuest simulated an existing campus, they were not able to have the
simulation results to match utility bills close enough. It turned out
that the economizer damper was on all the time and the some meters were
set up wrong which leads the historical data was inaccurate.
That's a good example. I'm always amazed at the number of outside air
dampers that are broken/disconnected, have broken actuators, or crummy
seals. A significant percentage of existing buildings (and new ones for
that matter!) have that problem. It would be great if we had the time
and budget to test all the systems to find problems like this, but then
we're doing retrocommissioning on top of energy modeling. And even then,
our data is only as good as the meters and the people placing them (is
your mixed air temperature logger - with an error of 2-3% - just a
single logger ziptied right in the middle of the airstream, potentially
missing stratification?).
Sorry, I'm done ranting J. Just some pent up frustration from being
asked to do things that really don't make sense to me. I get the
heebie-jeebies even thinking about calibrating a model to an existing
building, let alone to within 1-2%. I'm just hoping someone will tell me
that I'm being too academic and that these things work themselves out in
the real world. The DOE must have good reason to request calibration
within 1-2% other than a warm fuzzy feeling when they get results,
right?
I think we're all in general agreement here - I've been tasked with
developing existing models to match historical consumptions to varying
deadbands of accuracy - but never with a prescriptive mandate. The
targeted degree of calibration for any of my projects is both selected
by myself and varies, and for a variety of reasons - primarily (1) time
allowed to develop the model, (2) the information available/obtainable,
(3) the perceived accuracy of that information.
I don't disagree: Tight levels of matched historical data may not
reflect a more sophisticated and accurate model, especially for
"predicitive" purposes, but until we get a time travel feature added to
the Wizards (come on dev team, get moving! ^_^) - what is the
alternative metric for modeling accuracy?
My statement that you quoted wasn't properly elaborated on my end, and I
apologize for writing in haste (end of the workday) - I don't mean to
suggest 1-2% is "good" and anything higher is "bad," or anything of that
nature. More directly: I simply mean to state that setting a target of
monthly utility consumption accuracy (either for yourself as the modeler
or for another), is entirely reasonable if the modeler is given the time
and resources to continue learning and investigating the building until
such "historical accuracy" is obtained. I personally, given no other
direction, consider +/-10% (for each utility and each modeled month) to
be a minimum before results are share-able.
I also want to emphasize: "Predictive modeling" as a concept is a
potential minefield, both for design teams and building owners if and
when it is (often) misunderstood. I'm no professional orator/writer, so
I don't know if there is better existing vocabulary for this, but I
forcefully insist on keeping the distinction between "predictive" and
"historical" accuracy understood by all parties and at all stages
whenever I am interacting with others over modeling projects.
To respond to your hypothetical - If I am in a project and I find the
monthly consumptions are generally ~6% off, where a target of 2% has
been chosen by myself or others, my first reaction would not be to
arbitrarily tweak unknown assumptions to make the number "work." When I
do this sort of project, I understand a myriad of assumptions have to be
made. In my mind, some assumptions are more of a "leap" than others, so
I keep a running list of those larger assumptions in a notepad at my
desk for future reference. Perhaps I never gathered information
regarding a specific roof retrofit's construction, or I haven't
investigated whether there are a separate set of setback conditions for
a given school season. When the time comes to calibrate the model,
after ensuring everything is entered as I intended, I then proceed to
investigate further (ask questions, make another site visit if
necessary) as required to resolve the items in my list to the best of my
ability. Nothing is altered purely for the purpose of "matching the
numbers," only in the name of improving the model's historical accuracy.
If a deadline should arrive someday where the targeted level of
calibration is not met (hasn't happened yet - but then I've set my own
targets...), I'll rest easy knowing I've developed the most accurate
model I can with the time and resources available to me. I also be
ready to discuss at length the major assumptions made and actions taken
to address those unknowns.
Off the record, I've approached +/- 2% within each monthly utility
record consumption one (1) time, and I had to become intimately familiar
with the project over months to get there. Much discussion with the
building maintenance crew, site visits, and poring over record documents
was required. I fully agree this is not a level of calibration that
ought to be required except when an exceptional level of time and
resources are dedicated to investigating the building (at least for a
modeler with only a few years experience). If you were to look at a
"predictive accuracy gained : time/money spent developing the model"
ratio, I'm sure for most projects you would find sharply diminishing
returns trying to approach such a deadband. If there is a metric to
measure "predictive accuracy" for any model better than calibrating to
historical data, I'd love to learn about it, but to my knowledge it's
all we've got to work with.
That I'm writing such lengthy responses hopefully conveys - this is a
kinda tough topic to discuss! I'm not trying to make any legally
binding statements here, so please don't read between the lines - if
anything I'm writing is clear as mud, I'll be happy to elaborate, time
permitting =).
Its quite amusing. However - I'm not sure I agree with him. To truly get an proper analysis you need to really look at it by building function. CBECS really doesn't include high energy intensity buildings like labs in its database. I'm not saying that LEED buildings are actually performing brilliantly - I just think we need better less sensationalistic studies.
Of course, your fee reflects the amount of accuracy the client is
requiring you to have. Some clients might only expect 5-20% accuracy,
and of course you would provide them with a lower fee. However, in over
17 years of running hourly, climate history based, energy estimating
software, (not always eQUEST), I don't ever remember doing a study where
we were satisfied with being more than 5% off on our final base model.
If you are happy with being 25% or more off, then why even do a model,
you might as well just guess or do some kind of energy saving hand
calculation. The fun part is getting modeled gas usage and electrical
usage, month to month, to follow historical monthly trends and still be
close to the annual energy usage on both. When you get both the annual
energy usage and the monthly energy usage of gas and electric close to
historical usage, then that's when you feel good about your base model
to compare against ECM runs.
No one said this was easy. That's why you earn ever dime of your fee.
Your first preliminary base model run, if you do a good job, might be
5-20% off historical usage. To even get that close normally requires an
extensive survey, documenting ever identifiable energy using item,
interviews regarding usage by the occupants and maybe even data logging
usage of some important items, to confirm actual usage. It can,
depending on the size and complexity of the modeled building, take
several days to a week or more and a hundred or more runs to just create
a good solid base model, that reflects historical usage. That's why
modeling a new building, without any historic energy usage data to
match, other than approximate BTU's per square foot, is a piece of cake.
Two situations:
?
1) Matching energy?use - accurate metering
?
2) Matching energy bills - need first accurate energy measurement..THEN you need to input the rates correctly, including taxes..
?
Don't confuse the two..
?
My FOUR cents !!!!!
Thanks David, that's good to hear the level of rigor there was high and expected by the client.
I guess my concern here, if I had to summarize it, is that being X% away from the energy bills risks conflating precision for accuracy. Now, if you're 1-2% away on month by month maybe you can start to feel like you're more accurate, but even that still seems potentially too large grained to me. Maybe that's na?ve - I certainly don't have the experience you, or many of the people on this board have. I think ultimately, the energy modeler just needs to understand that the limitations of the tool and the uncertainty involved with inputs affect accuracy at least as much as being close to energy bills. Most importantly, that all needs to be communicated and understood by the client so they don't think they are receiving something that they are not.
Not energy cost, but here is a famous study that compared actual EUI to
modeled EUI. Check out page 8:
http://www.usgbc.org/ShowFile.aspx?DocumentID=3930
Nathan Miller
Ahmed:
?
The question is a bit perplexing.
?
If you spend enough time inputting rate details, time of use items, and taxes, you can literally come as close as you want. if your energy is off 5% annually, your costs should be similar.
?
John R. Aulbach, PE, CEM
On all the DOE modeling projects that we have done, of an existing
building, they have required the model to be within 1 to 2% or less of
the actual energy usage, based on an average two year history. Is that
what you are talking about?
David A. Bastow
How much effort was put into minimizing the error of your inputs to
justify that level of accuracy in your output? For example, did you
convert real weather data for the two years and use an average of the
simulations? Were you modeling small buildings so you could get a fairly
reasonable infiltration rate empirically? Were occupancy schedules
trended, and for how long? Were all the systems and controls working
correctly, with all sensors calibrated regularly?
1 to 2% seems to me to be fairly unreasonable. Unless you do an amazing
job verifying your inputs, in my opinion that level of precision doesn't
get you a better model. If you set everything up as best you can and it
comes in there, great! But a correctly set up model can be off by over
2% because of a couple "El Nino" years, a facilities guy locking a
humidity high limit to 50% for a summer, or any number of operational
factors. On the other side of it, we've tried to match a simple DOE2.2
model to a DOE2.1e model with limited success (I don't think we got
within 3%, although we didn't spend too much time with it). Who knows
what the difference would be with E+.
I know I always squirm when I'm asked to do existing building models,
and it may be irrational. Heck, if the DOE is asking for 1-2%, I
probably am being irrational. But it seems to me that the error from
assumptions could easily swing a model 1-2% (what would a 30% error on
your infiltration do to an otherwise correct building?). What are other
people's thoughts?
Eric
Much of the discussion on this so far has been about modelling existing
buildings.
If your situation is that you are modelling a new building and using
"default" (MNECB or A90.1) schedules, plug load values etc. Then the
actual building operation could be 50% higher than the model is
predicting. The difference will be in the occupancy, the schedules and
especially building operation and occupant behaviour. All of these are
hard to quantify at the design stage.
While I (and I am sure many others) have books to write regarding my
thoughts on the topic, I think I might suggest a simple step back: The
concept of predictively modeling to within a few percentage points of
accuracy is ridiculous. There are simply way to many things that could
not be known. HOWEVER: The concept of building a model for an existing
building, whose modeled energy consumption/costs fall within a 1-2% of
historical utility records, is entirely feasible - and a reasonable
requirement if the goal is to generate a model for predictive purposes.
How often such predictive models' accuracy is misinterpreted is
something I get depressed to think about.
I haven't done the DOE-modeling work that's being referenced with such
requirements, but I have done work for educational clients
(physics/building science departments) who wish to get these models set
up for ongoing study/tweaking purposes. The exercise is challenging,
and as close as energy modeling gets to "fun," when you can rest at ease
knowing the client is fully on board with what the model is and isn't.
It's fully rational to squirm and cringe when you have to make models
that you know will be mis-used and mis-understood, despite your best
efforts. Fortunate is the practicing energy modeler who gets to work
for fully educated clients and design teams all the time =).
Rather than drill David, I think it would be safe to assume those DOE
requirements exist to calibrate a model to a given degree to whatever
historical records are available, NOT to mandate a level of accuracy for
predictive purposes. If I'm wrong, I hope that work never crosses my
desk!
NICK CATON, E.I.T.
Nick,
I should make it clear that my initial questions weren't meant to drill
David. I'm genuinely interested if the DOE projects required an
extremely high level of technical accuracy on the input side. (Although
they were also somewhat of a reaction to those requirements.) We have
this discussion in our office every so often, and I'm usually somewhat
on my own, but I'm still not convinced.
Would you mind elaborating a little more about why you feel this to be
true:
"HOWEVER: The concept of building a model for an existing building,
whose modeled energy consumption/costs fall within a 1-2% of historical
utility records, is entirely feasible - and a reasonable requirement if
the goal is to generate a model for predictive purposes."
So let me ask it this way. Let's say you've got a model, and it's coming
in about 6% low (both heating and electric), but to the best of your
knowledge everything is set up properly. You've got to start making some
educated guesses to close the gap. Do you assume the boiler needs a
tuning (drop your efficiency a few percent) and the AHU coils need to be
cleaned (bump up your static)? Or do you bump up your infiltration
slightly (because it's probably not a very tight building). Or did some
maintenance guy came along and switched on both the boiler pump manually
on when he got a cold complaint, and no one returned them to auto? Or
maybe you have data from a few years with abnormally warm summers and
cold winters. I could go on practically ad nauseum. There's just so many
things that can be wrong with an existing building, as all of you
probably know.
Whatever you choose, that guess you make may interact (or lack an
interaction) with some measures you're predicting. So, predicatively,
what makes a 1-2% off model better than a 5-6% off model unless you
really spend the time and find out whether your inputs are accurate?
Cheers,
Eric
Eric/Nick,
I think it's important for anyone talking about modeling an existing building to qualify such/any levels of prediction with the amount of data that's used to calibrate the model. Utility bills and a simple walk through type audit (say a Level 2) will find you feeling pretty cheery if you're at +/- 10%. If you're at +/- 1-2%, it's most likely happenstance and you should trust that the model is dead-on accurate. If you've heavily instrumented the various systems throughout the building and your calibration finds you within 1-2% of actual consumption, you're doing your job well and have spent a fair bit of change on the effort.
As for new buildings and predictive modeling, this is where benchmarking and feedback to the modelers and design teams grows ever more important (and talking about 1-2% is nowhere near practical at this point). There are beginning to be more cases of litigation where large discrepancies exist, so be sure you know what you're promising and that you cya. It does seem that there will likely be a harder, faster push to focus on operational performance in terms of ratings/classifications as opposed to making the modeling better, but I do foresee the market trending towards the need for us to demonstrate that our models are getting better as time goes on if we want to keep market share.
My two cents...
Paul Erickson LEED(r) AP
In one of my learning course, the instructor mentioned that after they
used eQuest simulated an existing campus, they were not able to have the
simulation results to match utility bills close enough. It turned out
that the economizer damper was on all the time and the some meters were
set up wrong which leads the historical data was inaccurate.
Lan Li, PE
Lan,
That's a good example. I'm always amazed at the number of outside air
dampers that are broken/disconnected, have broken actuators, or crummy
seals. A significant percentage of existing buildings (and new ones for
that matter!) have that problem. It would be great if we had the time
and budget to test all the systems to find problems like this, but then
we're doing retrocommissioning on top of energy modeling. And even then,
our data is only as good as the meters and the people placing them (is
your mixed air temperature logger - with an error of 2-3% - just a
single logger ziptied right in the middle of the airstream, potentially
missing stratification?).
Sorry, I'm done ranting J. Just some pent up frustration from being
asked to do things that really don't make sense to me. I get the
heebie-jeebies even thinking about calibrating a model to an existing
building, let alone to within 1-2%. I'm just hoping someone will tell me
that I'm being too academic and that these things work themselves out in
the real world. The DOE must have good reason to request calibration
within 1-2% other than a warm fuzzy feeling when they get results,
right?
Eric
Eric and everyone,
I think we're all in general agreement here - I've been tasked with
developing existing models to match historical consumptions to varying
deadbands of accuracy - but never with a prescriptive mandate. The
targeted degree of calibration for any of my projects is both selected
by myself and varies, and for a variety of reasons - primarily (1) time
allowed to develop the model, (2) the information available/obtainable,
(3) the perceived accuracy of that information.
I don't disagree: Tight levels of matched historical data may not
reflect a more sophisticated and accurate model, especially for
"predicitive" purposes, but until we get a time travel feature added to
the Wizards (come on dev team, get moving! ^_^) - what is the
alternative metric for modeling accuracy?
My statement that you quoted wasn't properly elaborated on my end, and I
apologize for writing in haste (end of the workday) - I don't mean to
suggest 1-2% is "good" and anything higher is "bad," or anything of that
nature. More directly: I simply mean to state that setting a target of
monthly utility consumption accuracy (either for yourself as the modeler
or for another), is entirely reasonable if the modeler is given the time
and resources to continue learning and investigating the building until
such "historical accuracy" is obtained. I personally, given no other
direction, consider +/-10% (for each utility and each modeled month) to
be a minimum before results are share-able.
I also want to emphasize: "Predictive modeling" as a concept is a
potential minefield, both for design teams and building owners if and
when it is (often) misunderstood. I'm no professional orator/writer, so
I don't know if there is better existing vocabulary for this, but I
forcefully insist on keeping the distinction between "predictive" and
"historical" accuracy understood by all parties and at all stages
whenever I am interacting with others over modeling projects.
To respond to your hypothetical - If I am in a project and I find the
monthly consumptions are generally ~6% off, where a target of 2% has
been chosen by myself or others, my first reaction would not be to
arbitrarily tweak unknown assumptions to make the number "work." When I
do this sort of project, I understand a myriad of assumptions have to be
made. In my mind, some assumptions are more of a "leap" than others, so
I keep a running list of those larger assumptions in a notepad at my
desk for future reference. Perhaps I never gathered information
regarding a specific roof retrofit's construction, or I haven't
investigated whether there are a separate set of setback conditions for
a given school season. When the time comes to calibrate the model,
after ensuring everything is entered as I intended, I then proceed to
investigate further (ask questions, make another site visit if
necessary) as required to resolve the items in my list to the best of my
ability. Nothing is altered purely for the purpose of "matching the
numbers," only in the name of improving the model's historical accuracy.
If a deadline should arrive someday where the targeted level of
calibration is not met (hasn't happened yet - but then I've set my own
targets...), I'll rest easy knowing I've developed the most accurate
model I can with the time and resources available to me. I also be
ready to discuss at length the major assumptions made and actions taken
to address those unknowns.
Off the record, I've approached +/- 2% within each monthly utility
record consumption one (1) time, and I had to become intimately familiar
with the project over months to get there. Much discussion with the
building maintenance crew, site visits, and poring over record documents
was required. I fully agree this is not a level of calibration that
ought to be required except when an exceptional level of time and
resources are dedicated to investigating the building (at least for a
modeler with only a few years experience). If you were to look at a
"predictive accuracy gained : time/money spent developing the model"
ratio, I'm sure for most projects you would find sharply diminishing
returns trying to approach such a deadband. If there is a metric to
measure "predictive accuracy" for any model better than calibrating to
historical data, I'd love to learn about it, but to my knowledge it's
all we've got to work with.
That I'm writing such lengthy responses hopefully conveys - this is a
kinda tough topic to discuss! I'm not trying to make any legally
binding statements here, so please don't read between the lines - if
anything I'm writing is clear as mud, I'll be happy to elaborate, time
permitting =).
Thanks,
NICK CATON, E.I.T.
I little video about LEED energy savings. My favorite part it is at 4:52 ~
5:20. This guy is my new hero.
http://www.youtube.com/watch?v=mvCP3s7Xq48
Lan,
This is why commissioning is one of the best things you can do for a project.
Commissioning is very very cost effective.
Almost every piece of equipment is tested in the factory before being
shipped.
The building is a piece of equipment with many parts, it must be tested
(Commissioned).
Its quite amusing. However - I'm not sure I agree with him. To truly get an proper analysis you need to really look at it by building function. CBECS really doesn't include high energy intensity buildings like labs in its database. I'm not saying that LEED buildings are actually performing brilliantly - I just think we need better less sensationalistic studies.
Vikram Sami, LEED AP
Of course, your fee reflects the amount of accuracy the client is
requiring you to have. Some clients might only expect 5-20% accuracy,
and of course you would provide them with a lower fee. However, in over
17 years of running hourly, climate history based, energy estimating
software, (not always eQUEST), I don't ever remember doing a study where
we were satisfied with being more than 5% off on our final base model.
If you are happy with being 25% or more off, then why even do a model,
you might as well just guess or do some kind of energy saving hand
calculation. The fun part is getting modeled gas usage and electrical
usage, month to month, to follow historical monthly trends and still be
close to the annual energy usage on both. When you get both the annual
energy usage and the monthly energy usage of gas and electric close to
historical usage, then that's when you feel good about your base model
to compare against ECM runs.
No one said this was easy. That's why you earn ever dime of your fee.
Your first preliminary base model run, if you do a good job, might be
5-20% off historical usage. To even get that close normally requires an
extensive survey, documenting ever identifiable energy using item,
interviews regarding usage by the occupants and maybe even data logging
usage of some important items, to confirm actual usage. It can,
depending on the size and complexity of the modeled building, take
several days to a week or more and a hundred or more runs to just create
a good solid base model, that reflects historical usage. That's why
modeling a new building, without any historic energy usage data to
match, other than approximate BTU's per square foot, is a piece of cake.
David A. Bastow
Two situations:
?
1) Matching energy?use - accurate metering
?
2) Matching energy bills - need first accurate energy measurement..THEN you need to input the rates correctly, including taxes..
?
Don't confuse the two..
?
My FOUR cents !!!!!
Thanks David, that's good to hear the level of rigor there was high and expected by the client.
I guess my concern here, if I had to summarize it, is that being X% away from the energy bills risks conflating precision for accuracy. Now, if you're 1-2% away on month by month maybe you can start to feel like you're more accurate, but even that still seems potentially too large grained to me. Maybe that's na?ve - I certainly don't have the experience you, or many of the people on this board have. I think ultimately, the energy modeler just needs to understand that the limitations of the tool and the uncertainty involved with inputs affect accuracy at least as much as being close to energy bills. Most importantly, that all needs to be communicated and understood by the client so they don't think they are receiving something that they are not.
Good discussion, thanks!
Eric