If we are doing a simulation for an existing building and the actual
electricity consumption does not match with the simulation, how can we
adjust/calibrate it?
The snapshot shows the actual (blue graph) vs. simulation.
Depending on the type of building it is very difficult to get an
accurate match between the predicted and actual energy usage. There are
a range of reasons why this can be the case.
*
Weather. How does the weather file used relate to the weather
from the measured period? If it was hotter or colder, this will have
implications for the comparison between the two figures.
*
Usage. All building simulation tools assume a set period of
occupancy and resultant heat gain - unless you know the occupancy and
equipment usage patterns for each hour of the year, there will be
differences between the two figures. Are there any data centres (and
resultant high power consumption) that have been implemented in the
building but aren't on drawings?
*
Operation. Perhaps the biggest cause of potential discrepancy.
Is the actual building being operated in exactly the same way as input
to the software? Commissioning and operation are perhaps the most
important things to get right if buildings are going to produce the kind
of energy savings that us engineers are predicting.
*
Sub metering. Has the building been submetered in any way to
allow you to identify where the discrepancies lie? From the graph it
looks like there is additional consumption during warmer months that
isn't being picked up to the same extent by the simulation.
Sorry not to be of more help, but these are a few of the initial ideas
to consider.
- You'll want to simultaneously look at both your consumption
and demand results/curves. Together you can use the resulting
differences to determine where you should make adjustments.
- You should also be looking at both gas and electricity
consumption/demand "actual/modeled" curves if they apply - 4 total.
- Take note that if you are using only a single-year's worth of
utility consumption data to develop your "actual" curve, you're bound to
have differences that come from your model using TMY weather conditions
(averaged over decades). The likelihood of any one year being a "hot"
or "cold" year relative to the average is high.
- The individual loads broken out are your first clue - some
loads when increased will raise the "modeled" curve uniformly, depending
on how they're scheduled. Miscellaneous Loads, Lights and DHW for
example are generally a flat consumption every month, unless you're
incorporating daylighting controls or some other scheduling schemes. As
a result, raising/lowering those loads (in a way that makes sense and
better syncs with the existing building) will raise/lower the "modeled"
curve, without changing its shape.
- Loads that vary over time will affect the shape of the curve.
Space heating, fans and space cooling energy are generally the major
ones. Increasing/Decreasing these is not as simple as entering a
different W/SF. A few major things affect the overall quantity of
heating/cooling provided over the year - internal loads (particularly
solar/people), their associated schedules, equipment/fan efficiencies
(these act sorta like multipliers to whatever annual consumption shape
you arrive at), building envelope - equally important is building
infiltration, and the true control behavior (actual thermostat
setpoints, setback behavior, economizer operation, etc...)
- It's best to try to match curve shape (variable loads),
allowing your modeled consumption to ride uniformly above/below actual,
before "playing with" the constant-over-time components.
- Ultimately, you will have a balance of work where you are
first matching the existing project to the best of your ability,
thereafter tweaking assumptions for which you cannot account for with
true accuracy. If you commit yourself to this reality, you can define
when you are doing the former and carefully keep track of your
assumptions - this is good because you can sleep at night being able to
provide concrete answers regarding exactly which elements of your model
are true behavior.
I've been down this route (using eQuest to match an existing building).
It's quite do-able but as a forewarning may require you to become pretty
intimate with the existing building when all is said and done ;).
Your model results have very little energy for fans and cooling (and basically no heating) - I'm not sure what you would calibrate. The actual data seems to show a much more pronounced weather dependence.
Probably some rough hand-calcs to check against annual fan and cooling energy usage estimates might be a good place to verify your model inputs for the HVAC systems. I think you have a little bit of debugging of the inputs to go before thinking about "fine tuning" or calibrating.
Mo,
Depending on the type of building it is very difficult to get an
accurate match between the predicted and actual energy usage. There are
a range of reasons why this can be the case.
*
Weather. How does the weather file used relate to the weather
from the measured period? If it was hotter or colder, this will have
implications for the comparison between the two figures.
*
Usage. All building simulation tools assume a set period of
occupancy and resultant heat gain - unless you know the occupancy and
equipment usage patterns for each hour of the year, there will be
differences between the two figures. Are there any data centres (and
resultant high power consumption) that have been implemented in the
building but aren't on drawings?
*
Operation. Perhaps the biggest cause of potential discrepancy.
Is the actual building being operated in exactly the same way as input
to the software? Commissioning and operation are perhaps the most
important things to get right if buildings are going to produce the kind
of energy savings that us engineers are predicting.
*
Sub metering. Has the building been submetered in any way to
allow you to identify where the discrepancies lie? From the graph it
looks like there is additional consumption during warmer months that
isn't being picked up to the same extent by the simulation.
Sorry not to be of more help, but these are a few of the initial ideas
to consider.
Best of luck,
Dan.
Some quick thoughts:
- You'll want to simultaneously look at both your consumption
and demand results/curves. Together you can use the resulting
differences to determine where you should make adjustments.
- You should also be looking at both gas and electricity
consumption/demand "actual/modeled" curves if they apply - 4 total.
- Take note that if you are using only a single-year's worth of
utility consumption data to develop your "actual" curve, you're bound to
have differences that come from your model using TMY weather conditions
(averaged over decades). The likelihood of any one year being a "hot"
or "cold" year relative to the average is high.
- The individual loads broken out are your first clue - some
loads when increased will raise the "modeled" curve uniformly, depending
on how they're scheduled. Miscellaneous Loads, Lights and DHW for
example are generally a flat consumption every month, unless you're
incorporating daylighting controls or some other scheduling schemes. As
a result, raising/lowering those loads (in a way that makes sense and
better syncs with the existing building) will raise/lower the "modeled"
curve, without changing its shape.
- Loads that vary over time will affect the shape of the curve.
Space heating, fans and space cooling energy are generally the major
ones. Increasing/Decreasing these is not as simple as entering a
different W/SF. A few major things affect the overall quantity of
heating/cooling provided over the year - internal loads (particularly
solar/people), their associated schedules, equipment/fan efficiencies
(these act sorta like multipliers to whatever annual consumption shape
you arrive at), building envelope - equally important is building
infiltration, and the true control behavior (actual thermostat
setpoints, setback behavior, economizer operation, etc...)
- It's best to try to match curve shape (variable loads),
allowing your modeled consumption to ride uniformly above/below actual,
before "playing with" the constant-over-time components.
- Ultimately, you will have a balance of work where you are
first matching the existing project to the best of your ability,
thereafter tweaking assumptions for which you cannot account for with
true accuracy. If you commit yourself to this reality, you can define
when you are doing the former and carefully keep track of your
assumptions - this is good because you can sleep at night being able to
provide concrete answers regarding exactly which elements of your model
are true behavior.
I've been down this route (using eQuest to match an existing building).
It's quite do-able but as a forewarning may require you to become pretty
intimate with the existing building when all is said and done ;).
NICK CATON, E.I.T.
Your model results have very little energy for fans and cooling (and basically no heating) - I'm not sure what you would calibrate. The actual data seems to show a much more pronounced weather dependence.
Probably some rough hand-calcs to check against annual fan and cooling energy usage estimates might be a good place to verify your model inputs for the HVAC systems. I think you have a little bit of debugging of the inputs to go before thinking about "fine tuning" or calibrating.
David