IPLV Curve Fit

12 posts / 0 new
Last post


I am trying to create a part load curve for an air-cooled screw chiller. I have the EER & COP data for 100%, 75%, 50%, and 25% loading.

I have read through the various posts in the archives and I am still unsure if I should input this data as 'raw data' or 'coefficients'

IPLV (or NPLV) = 0.01A+0.42B+0.45C+0.12D
A = COP or EER @ 100% Load
B = COP or EER @ 75% Load
C = COP or EER @ 50% Load
D = COP or EER @ 25% Load

It is my understanding that EIR-PLR curves give the EIR as a function of part load ratio.

SO, if I create a 'cubic' curve with input type 'curve coefficients', and input a=(1/COP-100%), b=(1/COP-75%), c=(1/COP-50%), and d=(1/COP-25%) and specify this curve as the EIR - f(part load ratio) curve, would that be accurate?

Additionally, this chiller has a condenser water temp of 95 degF and design conditions that differ from AHRI conditions (design EWT=58 degF and LWT=44 degF)
Do these values have direct correlation to the curve creation ? Or can the condenser and CHW temp simply be defined in the 'basic specifications' tab for the chiller, and the EWT defined via the CHW loop deltaT?

I know this topic has been widely discussed, but I am still uneasy about the actual inputs for the creation of the part load performance curve.

Any input, or reference to a previous post would be appreciated.

Thanks in advance,

Lyle Keck

Lyle Keck's picture
Joined: 2011-09-30
Reputation: 1


The short answer is no. IPLV cannot be used to create performance curves in

Part-load curves are a strange animal in eQUEST. You'll need cooperation
from the chiller manufacturer to get the correct data. You need to have
enough data to describe the way energy varies as a function of part-load,
chilled water supply temp, and outside air temperature. Take a look at the
attached spreadsheet which can be used to create curves in eQUEST. The data
in yellow should be entered by the manufacturer or manufacturer's rep. The
data in red should be entered into eQUEST.

If you look into the formulas, you'll note that EIR-FPLR is actually not an
EIR factor, it is based on the power draw of the equipment at part-load
compared to at full-load. This throws a lot of people off as it is not
described that way in the help documentation.

The interesting question that arises when working with custom part-load
curves is how accurate does the eQUEST default performance curve model the
required IPLV for the baseline per ASHRAE 90.1? In my experience the
default curve is approximately 10% worse than the IPLV requirement.

Hope this helps,

Robby Oylear, LEED AP BD+C

Robby Oylear's picture
Joined: 2011-09-30
Reputation: 202

Hi Lyle,

I think you need to review a little further before proceeding. Read
through the attached discussion from the bottom up. This discussion
diverges a bit (as performance curve topics usually do) but I think
you're a bit ahead of the curve so to speak (har har!) in recognizing
what the IPLV numbers really mean. Plugging in your IPLV values as
coefficients for a cubic equation (where f(x) = Ax^3+Bx^2+Cx+D) is
pretty much wildly different from the correct course of action, but
you're definitely in the ballpark's parking lot ;).

Ultimately, you (and others - listen up!) will probably have the best
shot of getting it right on the first go if you visualize what you're
doing in excel. That's what worked for me when I was climbing this

* Assemble your part load percentages and corresponding EER's in

* Stop and review what the curve in eQuest is looking up, and
from what variable(s). You can glean this from the title of the default
curve used... In this case, you want eQuest to determine EIR each hour
as a function of Part load (EIRfPLR).

* Knowing that, convert convert your EER values to to EIR

* Now, normalize each EIR value to the ARI value*. If the EER
at ARI is 9, and your 50% part load EER is 6, then the normalized vaule
is 6/9 = 0.67. This normalized number is actually a multiplier of sorts
that eQuest will apply against your specified ARI efficiency in the
chiller input window. If none of this is making sense, go back further
and review my "recommended reading" list in a thread called "Chiller
curves, oh boy!" from sometime last year in the archives.

* Using excel, plot these normalized efficiencies (y-axis)
against the part load ratios (x-axis) with a scatter chart.

* Right click and add a curve-fit to the series. In the dialog,
try making it third order (quadratic) and see if you get a good fit.
Check the option to display the curve equation.

* Voila, there are your four coefficients (A,B,C,D - see
eQuest's curve inputs to be sure you know which is which in the
quadratic format) for eQuest! You've also developed a visual check to
ensure the numbers make sense.

*Important: You MUST work from ARI-condition data points, and normalize
to the ARI condition, and specify your chiller's capacity/efficiency at
ARI conditions, if you are trying to only make PLR curves without doing
a full set of custom curves. The reasons why are complex, and I've
already nearly written a book on the matter in the archives if you're
itching to know why ^_^. If you want to specify your chiller around
design conditions, then it's all or nothing (library curves) with regard
to custom curve creation!

A general statement/consolation to everyone: We all have to crawl before
we can run! Make no mistake, coming to a true fundamental understanding
of custom curves in eQuest/DOE2, to the point you can manipulate and
even create them on your own, is no small feat. They should give out
medals! The best way to develop this skill set is to accept you will
probably screw it up a few times to start, look hard at your results,
and read up on the literature/advice that's out there in order to learn
what it is that you personally don't know. Once you come around to
understanding what you do know, and more importantly what you know that
you don't know... (Donald Rumsfeld would make a good energy modeler, lol
), it's all downhill
from there, and you can begin to ask the right questions of yourself and
others to bridge your personal gaps. Making mistakes is commendable,
provided you resolve to learn from them!



Nick-Caton's picture
Joined: 2011-09-30
Reputation: 805
Robby Oylear's picture
Joined: 2011-09-30
Reputation: 202

Hi Robby!

Let me back into this with some clarifications for everyone's benefit -
I might be learning something here too!

First, the procedure I'm describing below is specifically to guide
someone in developing an EIR-fPLR, not an EIR-fPLR& dT curve. In that
point, you and I diverged in how we're advising Lyle to proceed. The
discussion I attached previously delves into whether one curve type or
another is appropriate for a given situation... but to summarize a long
discussion: a curve incorporating dT is not always critical to modeling
behavior in line with the actual equipment.

Second, below I've copied a library EIR-fPLR curve per your suggestion,
and I've generated a visualization as well for discussion:

[Inserted picture: eQuest library curve input screen illustrating curve
coefficients for an EIR-fPLR curve]

[Inserted picture: Excel plot of that library curve's coefficients]

(x-axis is the PLR, y-axis is the "EIR factor")

I think we all agree, but let's collectively be clear on this point:
This curve does not represent "efficiency." It is a factor applied
against the EIR specified for the chiller at full load. As we can
observe, the EIR "correction factor" decreases as part load decreases,
and that agrees with our expectation that the chiller should draw less
energy per unit work (or be "more efficient") at part load.

Efficiency is "what you get out / what you put in," where measures like
kW/ton and EIR are conceptually more of an inverse: "what you put in /
what you get out."

I took a look at the linked discussion Robby, and while I was not privy
to the off-list discussion, I sense we may have a different
understanding here (or we might concur?)... Does all this sit agreeably
with you to this point? If anything I'm stating seems incorrect, I (and
others) would very much appreciate being corrected, preferably with
illustrations ^_^!

Our syntax might be getting in the way... so let me touch on that as
well: When I say this library curve is "normalized," it is because the
derived equation returns a correction factor, not an actual EIR figure.
The library curve above is telling us that at 100% loading, the EIR
should be 1.00 x [specified EIR at 100%]. If you were build a curve as
I describe below without first dividing the EIR data points by the 100%
figure (so the factor at 100% is 1.0), you will have coefficients that
will produce actual EIR's for any given PLR, which would be multiplied
for each hour against the specified EIR... The results could either be
wildly off from reality, or could also produce deceptively sensible
behavior as well. I suppose a curve created this way might be made to
work correctly if you make a point to specify the chiller's EIR as 1,
but I'm not entirely sure something else wouldn't be affected... =)

I do want to support that Robby is correct to assert one cannot build a
performance curve from an IPLV rating alone, but one can work towards a
simple fPLR curve if that's desirable and appropriate for the model at
hand, if you do have the values for each part load point as Lyle is
describing (you cannot extrapolate these from the IPLV number).

A few extra points: I forgot to mention is that EIR-fPLR curves can be
entered as either Quadratic (2nd order) or Cubic (3rd order), so in
excel it would be a good exercise to try out both with your curve-fit
and see what shape fits your data/expectations best. Also, I may have
somewhat overstated the importance of specifying a chiller at ARI
conditions when using EIR-fPLR curves... but suffice to say - the
library curves are built around ARI conditions, so specifying your
chiller at ARI conditions is always a safe bet if you aren't planning to
make a full set of curves around your design conditions.

Game on!



Nick-Caton's picture
Joined: 2011-09-30
Reputation: 805
cmg750's picture
Joined: 2010-10-05
Reputation: 0

Robby, Nick, and Carol,

Thanks for all of the input regarding this issue. I am working on obtaining some temperature specific data from the manufacturer.

I understand this is one of the more complicated procedures to tackle, but I hope revisiting this issue will benefit all of the archive users.

Lyle Keck

Lyle Keck's picture
Joined: 2011-09-30
Reputation: 1


The link below has a great section on creating DOE2 chiller curves.


Paul Diglio

Paul Diglio's picture
Joined: 2011-09-30
Reputation: 400
Robby Oylear's picture
Joined: 2011-09-30
Reputation: 202
Paul Diglio's picture
Joined: 2011-09-30
Reputation: 400

Hello All,

I think it might be helpful to the community if we had available a spreadsheet that has been reviewed and approved by the community. I attached the spreadsheet I use to create custom chiller curves which could be used for this purpose and an image of the spreadsheet with some notes that might be helpful to the following post on eSimForums:


The spreadsheet contains the information required from a manufacturer, the calculations used to create custom curves, and the curve information in a format that can be pasted into an inp file. The spreadsheet is based on the attached document written by Steve Gates which I received from a post by Carol Gardner on this list. Feel free to look the spreadsheet over, verify the formulas match the equations in Steve Gates's document, and vote on whether or not you approve of the spreadsheet using the poll at the top of the screen. If you find an error, please modify and attach the modified spreadsheet in a reply to the post with a description of the error and the changes you made.

In response to some comments made regarding normalizing the capacity and EIRfPLR, I refer to the attached document which I find much less confusing than the EDR document. It states the following for capacity:
?Calculate the normalized capacity for each set of temperatures as:
CapfT = CapOp / CapNom (4)
As defined above, CapNom is the capacity at the rated conditions and is constant, CapOp varies with the off-rated chilled water and condenser temperatures.?

I take this to mean that the capacity is normalized to the rated capacity for the efficiencies you are using, which may or may not be the ARI rated capacity.

The following definition is given for EIRfPLR:
?EIRfPLR = PwrPLR / PwrOp
PwrOp = the full-load power for the given operating capacity
PwrPLR = fraction of full-load power consumed at the given part load ratio?

My interpretation of this is that EIRfPLR is referenced to the full load power at the current lift conditions (given operating capacity), and is a ratio of power, not efficiency.


Jeremy McClanathan, P.E., BEMP, LEED? AP BD+C

Jeremy McClanathan's picture
Joined: 2011-09-30
Reputation: 200

The following link takes you to a publication by Taylor Engineering on two methods they?ve used for calculating performance curves for different chillers. Its very accurate and easy to use.


Travis Rhodes

Travis Rhodes's picture
Joined: 2011-09-30
Reputation: 0