We know that eQUEST can edit personal weather data.
But the dry-bulb and wet-bulb temperature in weather data can only
enter integers.
Is it possible to have more precise temperature to decimal place?
Thank you very much.
This is not possible at present without changing the DOE-2.2 source code
to read a weather input file with decimal values.
When DOE-2 was first designed in the early 1980's, memory was a big
concern, so the weather data was reduced to integers and then packed,
which is why the DOE-2 *.BIN file is so small (146K). I have actually
developed a modified file format for *.BIN where I save an extra digit
of precision, i.e., temperatures to 0.1F instead of 1 F, but the source
code would also need to be changed slightly to read this extra
information. I've mentioned this to the developer of eQUEST/DOE-2.2 and
will be experimenting with
making this change to the source code. If and when it's proven to work
and gets incorporated into DOE-2.2, I'll let everyone know. I welcome
anyone who thinks this is a useful modification to send me an e-mail. It
might spur me on to do something!
Joe
Given that the time steps are an hour, and the
fact that weather data is averaged over an hour,
plus the fact that the building local will have
variations from the weather station local, would
an extra decimal point provide more useful information?
My attention on this issue was first raised about 15 years ago when I
was working with non-US weather data , i.e., the rest of the world, that
are all reported in 0.1 C. I've noticed since that US stations have also
moved to the use of metric units, i.e., 0.1 C for temperature. The DOE-2
weather format is still in integer F, which leads to three unfortunate
effects: (a) hourly records can be off by as much as 0.5 F, (b) clumping
of the temperature distribution, and (c) statistics such as degree-days
will be off by a percent or two compared to the original data. Now, one
can say that all this is immaterial in the bigger picture of things,
which has been the default attitude so far, but since it's really quite
simple to fix, why not get it right, i.e., doesn't it feel much better
to see the same temperatures in the DOE-2 outputs as in the original
weather data?
BTW, all the weather data that I've looked at are records of conditions
on the hour, not the average over the hour, except for solar radiation.
Since as you've raised the question of how significant would be adding extra precision to
the weather data in DOE-2, I was sent a copy of a recent paper by Annie-Claude Lachapelle
of the Univ. of Calgary given at eSim Canada 2012 on this exact topic, "DOE2 Dry-Bulb
Temperature Precision Level Impact on Sensible Economizer Performance". With the
author's permission, I've attached the paper with this post.
The dramatic effects of simply rounding off/ceiling the temperature data in
weather files is a perfect example of why we need to, as an industry, move
from deterministic energy predictions to stochastic energy predictions.
Weather and occupancy are inherently stochastic and when we couple
that stochasticity with our uncertainty in so many of the actual building
parameters (at least uncertainty as to their what their values will be when
installed), it seems to me that far more meaningful energy predictions
could be made using stochastic methods.
While you wouldn't necessarily want to use stochastic methods during much
of the design iteration, stochastic estimation should be a standard
procedure for comparing major iterations and for final energy predictions.
For all us researchers, there is plenty of work to be done in properly
quantifying the stochasticity of weather, occupancy and other stochastic
parameters and in developing uncertainty profiles for important parameters
that are not stochastic. There is also opportunities for industry to
develop the wrappers to take the info and create the DOE2 wrappers.
Am I alone on this or do others feel the same way?
Ralph T Muehleisen
PhD, PE, LEED AP, INCE Board Certified, FASA
Interesting discussion. I agree with you for the case of using modeling for energy usage predictions. I think a lot of us in the industry who are not researchers consciously don?t do energy usage predictions. We?re mostly using modeling for relative comparisons of energy usage where we?re looking at the model?s sensitivity to one or more energy reducing strategies. Or maybe that assumption is wrong? So, I agree when you say that you wouldn?t necessarily want to use stochastic methods in the design iterations. For the case of using modeling for predicting actual energy usage, the current modeling tools are inadequate.
A basic question from someone who is not a developer of mathematical modeling methods: If you include probabilities in the model, would your end result be an energy usage prediction with a probability distribution wide that it wouldn?t be meaningful in the end?
These are my views in answer to recent posts on energy simulation and design program accuracy. Architectural design and simulation programs are based mainly on theory. Final decisions are subject to experience or statistical performance data using case studies.
Design programs are used to size & select systems & equipment and are based summer & winter design conditions. Design decisions are also based on first & maintenance costs, availability of parts, and maintenance staff at the location (particularly overseas). The performances of systems & plants decline with age. Scale builds up in piping and coils, filters and other equipment have to be cleaned frequently to maintain design conditions.
Energy program results are used to compare alternative design options and make decisions when selecting envelope, systems and plant and are based on 8760 hour data. They are based on the new theoretical performance of envelope, systems and plant. Energy program results are not used in design and therefore do not affect construction and operation.
The main reason for using energy programs should be to show energy code compliance and for sustainable buildings certification. Until 15 years ago they were rarely used. Experienced architects and engineers use the program to confirm the decisions they have already made. Example: glass properties & percent glass. They are not looking for surprises.
Both types of programs need user judgment & experience, besides theoretical knowledge, to check the results for input errors and interpret & apply the results. The best program to use is the one where you have a detailed understanding of the program from input to output and where you can detect inadvertent errors in the input by looking at the results.
There is no such thing as an accurate energy analysis program. Relative accuracy when comparing building components, loads, systems, equipment, plant with different Energy Savings Option is required and important. You cannot show energy savings by using different energy programs to compare the baseline with the proposed. Two people working with the same energy program on the same large commercial building will not come up with exactly the same results unless every detail of building is included in both models.
In the case of large commercial multi-use buildings and depending on the climatic location & season, more than 40% of the cooling energy can be internal ? lights, receptacle & people (includes ventilation). Internal cooling loads offsets winter heating loads that are based on weather. Two identical office bldgs located side by side will not have identical internal loads for any hour of any year. Occupancy and receptacle loads do not have to follow ASHRAE and other internal design standards & schedules after the building is occupied. The same is true for external loads which depend on the weather.
Few, if any, buildings are checked if the energy program predicted the actual energy consumption by the building during operation for each month of every year. So you can use any program accepted by code authorities and USGBC. One program (referring to major accepted programs) does not produce superior or better results compared to another.
Building energy codes and LEED certification have been around for over 10 years now. Is there statistical data comparing the results of different energy programs with measurement & verification of the building in operation? Is there statistical data on which energy program was used for code compliance and LEED certification?
Thank you Stormy! We need to seek intelligence here on earth before seeking artificial intelligence. There are so many variables that one cannot predict such as operational schedules, equipment load, occupancy, maintenance, or the complete lack of maintenance and training (especially in new construction) that a model on a complex building will NEVER match actual utility bills. I am SO tired of people raising unwarranted expectations of the results of models for customers and those who pay the actual utility bills. Here?s an excerpt from ASHRAE 90.1 that is often forgotten and/or overlooked by even those who wrote the document: Please see note 2 which is absolutely on the mark.
[cid:image001.png at 01CD6C27.A05E9590]
R I C H A R D E L L I S O N
BEAP, BEMP, BESA, CBEP, CEA, CEM, CMVP, CTAB, HBDP, LEED AP, PE
R I C H A R D E L L I S O N
BEAP, BEP, BEMP, BESA, CBEP, CDSM, CEA,
CEM, CMVP, CTAB, DGCP, HBDP, LEED AP, PE
Manager of Energy and Simulation
Southland Industries
22340 Dresden Street, Suite 177
Dulles, VA 20166
Office: 703.834.5570
Direct: 703.
I think it all depends on the purpose of the simulation. Leaving aside the issue of design
sizing for the moment, use of energy simulations to predict, or more often, match observed
energy usage is typically concerned with annual or long-term energy usage. In that case,
stochastic variations around a mean doesn't provide much useful information, because the
concern is not variability, but bias, in the results. It reminds me of when I saw that the
uncertainty given for the modeled solar illuminance in the TMY2 weather files was less
than 3%. That seemed wrong to me, since I know that for any particular hour the
differences between modeled and measured solar could be quite large. However, when I read
the documentation, it made sense because the uncertainty indicates not the stochastic
variability, but the "mean bias error", i.e., systematic variations, between the modeled
and measured solar.
How does that relate to the value of stochastic modeling? If we're only concerned in
getting the annual totals to match, then whether or not we capture the stochastic
variations from hour to hour seems to be minor importance. I'm also having difficulties
in understand what
additional information is output from "stochastic modeling", except having error bars on
each hour, which undoubtedly will be large.
My last comment is that although weather patterns are stochastic, the data on the weather
files is quite deterministic, if we allow that the weather stations are doing a reasonably
good job in measuring temperatures, wind speeds, etc. The question of adding precision to
the DOE-2 weather files is not an issue of stochastic behavior, but simply that of
round-off errors.
I guess I should have started my discussion by differentiating building
design from analysis. I definitely see a place where actual energy
predictions and characterization of uncertainty is of prime importance -
retrofits.
While designing low energy new buildings is great, if we want to really
impact the building energy and carbon footprint we need to retrofit the
existing building stock as much as possible.
Since most owners look at this as an investment, we really need to given
them information like other investments and provide risk analysis. That
means probabilistic energy savings predictions and, when coupled
with probabilistic cost and future energy costs, a
true probabilistic return on investment and value added to the building.
Then owners and financiers can look at a retrofit just like any other
investment, and (I hope) pull the trigger on more retrofits than they are
currently doing.
I'm not sure of a way to generate real ROI probability without stochastic
analysis. The underlying model doesn't necessarily have to reflect the
true physics (or even full systems of the building) as long as it captures
the monthly energy use right and enough of the physics that when we make
changes to the building, the model correctly reflects the energy savings
and uncertainty in that savings.
Ralph T Muehleisen
PhD, PE, LEED AP, INCE Board Certified, FASA
This is not possible at present without changing the DOE-2.2 source code
to read a weather input file with decimal values.
When DOE-2 was first designed in the early 1980's, memory was a big
concern, so the weather data was reduced to integers and then packed,
which is why the DOE-2 *.BIN file is so small (146K). I have actually
developed a modified file format for *.BIN where I save an extra digit
of precision, i.e., temperatures to 0.1F instead of 1 F, but the source
code would also need to be changed slightly to read this extra
information. I've mentioned this to the developer of eQUEST/DOE-2.2 and
will be experimenting with
making this change to the source code. If and when it's proven to work
and gets incorporated into DOE-2.2, I'll let everyone know. I welcome
anyone who thinks this is a useful modification to send me an e-mail. It
might spur me on to do something!
Joe
Joe
Given that the time steps are an hour, and the
fact that weather data is averaged over an hour,
plus the fact that the building local will have
variations from the weather station local, would
an extra decimal point provide more useful information?
>> Christopher Jones, P.Eng.
Chris,
My attention on this issue was first raised about 15 years ago when I
was working with non-US weather data , i.e., the rest of the world, that
are all reported in 0.1 C. I've noticed since that US stations have also
moved to the use of metric units, i.e., 0.1 C for temperature. The DOE-2
weather format is still in integer F, which leads to three unfortunate
effects: (a) hourly records can be off by as much as 0.5 F, (b) clumping
of the temperature distribution, and (c) statistics such as degree-days
will be off by a percent or two compared to the original data. Now, one
can say that all this is immaterial in the bigger picture of things,
which has been the default attitude so far, but since it's really quite
simple to fix, why not get it right, i.e., doesn't it feel much better
to see the same temperatures in the DOE-2 outputs as in the original
weather data?
BTW, all the weather data that I've looked at are records of conditions
on the hour, not the average over the hour, except for solar radiation.
Joe
Chris, others,
Since as you've raised the question of how significant would be adding extra precision to
the weather data in DOE-2, I was sent a copy of a recent paper by Annie-Claude Lachapelle
of the Univ. of Calgary given at eSim Canada 2012 on this exact topic, "DOE2 Dry-Bulb
Temperature Precision Level Impact on Sensible Economizer Performance". With the
author's permission, I've attached the paper with this post.
Joe
Joe Huang
Chris, Joe, others,
The dramatic effects of simply rounding off/ceiling the temperature data in
weather files is a perfect example of why we need to, as an industry, move
from deterministic energy predictions to stochastic energy predictions.
Weather and occupancy are inherently stochastic and when we couple
that stochasticity with our uncertainty in so many of the actual building
parameters (at least uncertainty as to their what their values will be when
installed), it seems to me that far more meaningful energy predictions
could be made using stochastic methods.
While you wouldn't necessarily want to use stochastic methods during much
of the design iteration, stochastic estimation should be a standard
procedure for comparing major iterations and for final energy predictions.
For all us researchers, there is plenty of work to be done in properly
quantifying the stochasticity of weather, occupancy and other stochastic
parameters and in developing uncertainty profiles for important parameters
that are not stochastic. There is also opportunities for industry to
develop the wrappers to take the info and create the DOE2 wrappers.
Am I alone on this or do others feel the same way?
Ralph T Muehleisen
PhD, PE, LEED AP, INCE Board Certified, FASA
Ralph and all,
Interesting discussion. I agree with you for the case of using modeling for energy usage predictions. I think a lot of us in the industry who are not researchers consciously don?t do energy usage predictions. We?re mostly using modeling for relative comparisons of energy usage where we?re looking at the model?s sensitivity to one or more energy reducing strategies. Or maybe that assumption is wrong? So, I agree when you say that you wouldn?t necessarily want to use stochastic methods in the design iterations. For the case of using modeling for predicting actual energy usage, the current modeling tools are inadequate.
A basic question from someone who is not a developer of mathematical modeling methods: If you include probabilities in the model, would your end result be an energy usage prediction with a probability distribution wide that it wouldn?t be meaningful in the end?
Stormy Shanks
These are my views in answer to recent posts on energy simulation and design program accuracy. Architectural design and simulation programs are based mainly on theory. Final decisions are subject to experience or statistical performance data using case studies.
Design programs are used to size & select systems & equipment and are based summer & winter design conditions. Design decisions are also based on first & maintenance costs, availability of parts, and maintenance staff at the location (particularly overseas). The performances of systems & plants decline with age. Scale builds up in piping and coils, filters and other equipment have to be cleaned frequently to maintain design conditions.
Energy program results are used to compare alternative design options and make decisions when selecting envelope, systems and plant and are based on 8760 hour data. They are based on the new theoretical performance of envelope, systems and plant. Energy program results are not used in design and therefore do not affect construction and operation.
The main reason for using energy programs should be to show energy code compliance and for sustainable buildings certification. Until 15 years ago they were rarely used. Experienced architects and engineers use the program to confirm the decisions they have already made. Example: glass properties & percent glass. They are not looking for surprises.
Both types of programs need user judgment & experience, besides theoretical knowledge, to check the results for input errors and interpret & apply the results. The best program to use is the one where you have a detailed understanding of the program from input to output and where you can detect inadvertent errors in the input by looking at the results.
There is no such thing as an accurate energy analysis program. Relative accuracy when comparing building components, loads, systems, equipment, plant with different Energy Savings Option is required and important. You cannot show energy savings by using different energy programs to compare the baseline with the proposed. Two people working with the same energy program on the same large commercial building will not come up with exactly the same results unless every detail of building is included in both models.
In the case of large commercial multi-use buildings and depending on the climatic location & season, more than 40% of the cooling energy can be internal ? lights, receptacle & people (includes ventilation). Internal cooling loads offsets winter heating loads that are based on weather. Two identical office bldgs located side by side will not have identical internal loads for any hour of any year. Occupancy and receptacle loads do not have to follow ASHRAE and other internal design standards & schedules after the building is occupied. The same is true for external loads which depend on the weather.
Few, if any, buildings are checked if the energy program predicted the actual energy consumption by the building during operation for each month of every year. So you can use any program accepted by code authorities and USGBC. One program (referring to major accepted programs) does not produce superior or better results compared to another.
Building energy codes and LEED certification have been around for over 10 years now. Is there statistical data comparing the results of different energy programs with measurement & verification of the building in operation? Is there statistical data on which energy program was used for code compliance and LEED certification?
Varkie Thomas
Thank you Stormy! We need to seek intelligence here on earth before seeking artificial intelligence. There are so many variables that one cannot predict such as operational schedules, equipment load, occupancy, maintenance, or the complete lack of maintenance and training (especially in new construction) that a model on a complex building will NEVER match actual utility bills. I am SO tired of people raising unwarranted expectations of the results of models for customers and those who pay the actual utility bills. Here?s an excerpt from ASHRAE 90.1 that is often forgotten and/or overlooked by even those who wrote the document: Please see note 2 which is absolutely on the mark.
[cid:image001.png at 01CD6C27.A05E9590]
R I C H A R D E L L I S O N
BEAP, BEMP, BESA, CBEP, CEA, CEM, CMVP, CTAB, HBDP, LEED AP, PE
Ralph, others,
I think it all depends on the purpose of the simulation. Leaving aside the issue of design
sizing for the moment, use of energy simulations to predict, or more often, match observed
energy usage is typically concerned with annual or long-term energy usage. In that case,
stochastic variations around a mean doesn't provide much useful information, because the
concern is not variability, but bias, in the results. It reminds me of when I saw that the
uncertainty given for the modeled solar illuminance in the TMY2 weather files was less
than 3%. That seemed wrong to me, since I know that for any particular hour the
differences between modeled and measured solar could be quite large. However, when I read
the documentation, it made sense because the uncertainty indicates not the stochastic
variability, but the "mean bias error", i.e., systematic variations, between the modeled
and measured solar.
How does that relate to the value of stochastic modeling? If we're only concerned in
getting the annual totals to match, then whether or not we capture the stochastic
variations from hour to hour seems to be minor importance. I'm also having difficulties
in understand what
additional information is output from "stochastic modeling", except having error bars on
each hour, which undoubtedly will be large.
My last comment is that although weather patterns are stochastic, the data on the weather
files is quite deterministic, if we allow that the weather stations are doing a reasonably
good job in measuring temperatures, wind speeds, etc. The question of adding precision to
the DOE-2 weather files is not an issue of stochastic behavior, but simply that of
round-off errors.
Joe
Joe Huang
Great conversation everyone.
I guess I should have started my discussion by differentiating building
design from analysis. I definitely see a place where actual energy
predictions and characterization of uncertainty is of prime importance -
retrofits.
While designing low energy new buildings is great, if we want to really
impact the building energy and carbon footprint we need to retrofit the
existing building stock as much as possible.
Since most owners look at this as an investment, we really need to given
them information like other investments and provide risk analysis. That
means probabilistic energy savings predictions and, when coupled
with probabilistic cost and future energy costs, a
true probabilistic return on investment and value added to the building.
Then owners and financiers can look at a retrofit just like any other
investment, and (I hope) pull the trigger on more retrofits than they are
currently doing.
I'm not sure of a way to generate real ROI probability without stochastic
analysis. The underlying model doesn't necessarily have to reflect the
true physics (or even full systems of the building) as long as it captures
the monthly energy use right and enough of the physics that when we make
changes to the building, the model correctly reflects the energy savings
and uncertainty in that savings.
Ralph T Muehleisen
PhD, PE, LEED AP, INCE Board Certified, FASA