[Equest-users] Chiller Curves (oh boy!)

Nick Caton ncaton at smithboucher.com
Wed Nov 3 12:40:09 PDT 2010


Yikes… looks like a slew of responses got out before I finished this one, but I have a lot to say =)…  For context you might want to check which email I’m responding to first, everyone ;).

 

John, thanks very much for the encouragement.  I think I can help further your thoughts:

 

1.       Getting a manufacturer vendor to cooperate is indeed tricky for all parties, and I’m not convinced their selection software really affords them the capability to find maximum loading points for varying condenser/evaporater temps in a time-efficient fashion… time will tell but I’ve got a very smart rep right now promising to see what he can do.

2.       I’d rest easier knowing it’s perfectly fine to normalize all three curves to the design conditions and have a PLR of 1.0 equal the design conditions… except for the DOE2 help entries explicitly telling us that for centrifugal chillers, “full load” means “maximum” (see the highlighted copy-paste job way down below and the entry for EIR-FPLR as well).  

 

That said, your call to better understand the default curves sounds a lot like a request, so perhaps I can scratch your back in return ;).  Here are the three curves eQuest pulls from its library when you specify a water cooled centrifugal chiller… visualized!  

 

Note that each curve is indeed normalized to (Z=1.0 at) ARI conditions.  I’ve followed each with my (novice) observations/commentary.

 

The EIR-FPLR&dT curve seems, well, kinda wacky to me!  I’ve conferred with a few other mechanical designers, and we seem to agree that an efficiency “sweet spot” for a centrifugal variable speed chiller should lie somewhere between zero and full loading conditions – this curve’s shape seems to demonstrate continually improving efficiencies as you approach zero loading.  Truthfully, there’s a very slight apex that’s hard to see (you can tell by the curving bands) just off the charted values shown between the 10% and 20% loading marks, but what chiller has an optimum efficiency at 15% loading?  I’d of this as a “normal” shape if the z-axis were on some scale of straight power (kW)… I’d expect that to continually drop off with the PLR in this fashion, but that z-axis is normalized EIR (kW/ton conceptually, if not in magnitude)… shouldn’t the default curve have a “sweet-spot” efficiency somewhere in the vicinity of 40-60% loading?

 

This EIR-FT curve is again normalized (1.0) at ARI conditions, and tells us that the chiller will be pretty much linearly more efficient as you increase the CHWT target temperature, and will exponentially increase in efficiency as the condenser water temperature drops.  Raising either has less of an inverse effect.  Nothing seems odd to me.

 

This curve is also normalized (1.0) at ARI conditions, and is telling us the chiller loses capacity when the condenser water either rises or falls from ARI conditions (20-30% drop from ARI, when CWT is at 100F or 50F, for example).  I’m too new at this to know whether that makes total sense, but my initial guess would be that colder condenser water would always be help improve capacity…  I’d infer from this curve that centrifugal chillers are designed and optimized around a specific condenser water temperature, and that hotter or colder condenser water temperatures, for the same CHWT, hurts your equipment’s potential capacity.  I suppose it might well be a function of the refrigerant’s thermodynamic properties?  The CHWT selected has a linear, but a relative to the CWT minor effect on the capacity of the chiller.  

 

3.       You’ve nailed my current approach on the head!  I’m just concerned that going against the explicit instructions in the DOE2 help entries (mentioned above) may have some significant impact on what I’m modeling…

4.       Fully agreeing.  And kudos for condensing paragraphs of fretting into a single line answer =).

 

Thanks again so much for everyone’s input so far.  I’d be curious to hear any other thoughts or second-opinions we as a group may have regarding:

A.      Why does the default part load curve above seems such an odd shape (at least to my eyes) for a centrifugal VSD chiller?  Are my chiller fundamentals off?

B.      To what end/purpose do the DOE2 help docs and the EDR guidelines emphasis normalizing the EIR-FPLR curve to maximum vs. design capacities at the rated conditions, when you can normalize and specify to ARI conditions and set DESIGN-PLR=1.0 as John and I are discussing?  I mean, I understand that you could get a “comprehensive” set of curves allowing you to apply the same chiller to other projects where you have a higher design capacity (or expect higher capacities will be required), but is there any inherent loss of accuracy if you build everything around data points encompassing all extremes present in the simulated model?

 

~Nick

 

 

 

NICK CATON, E.I.T.

PROJECT ENGINEER

25501 west valley parkway

olathe ks 66061

direct 913 344.0036

fax 913 345.0617

Check out our new web-site @ www.smithboucher.com 

 

From: John T. Forester [mailto:JohnTF at BVHis.com] 
Sent: Wednesday, November 03, 2010 11:49 AM
To: Nick Caton; equest-users at lists.onebuilding.org
Subject: RE: [Equest-users] Chiller Curves (oh boy!)

 

Nick,

 

I think you’re on the right track.  Below are some of my thoughts on your conclusions.  I hope the modeling community will set me straight if I’m wrong here.

 

1.	Getting multiple performance runs from vendors that show part-load performance independent of the CW and CHW temps can often be challenging.  Adding the “maximum capability” task to that for each of the conditions requires a pretty detailed understanding of the selection software.  I’d say if you can get PLR data for 3-4 different CHW temps while holding the CW temps constant at 85, 75, 65 (and sometimes lower) – you’re ahead of most modelers.  Working with the Mechanical Design Engineer and the vendor together has been successful for me in the past.
2.	Defining the chiller capabilities at the “maximum” may only come into play if you expect your model to overload the chiller above the specified design capacity (I’m thinking building additions or process loads). At this point, this data (or knowing what the default eQuest curves do in that range) would be useful.  Depending on your project, the time spent on developing curves for PLR >1.0 may not be justified.
3.	If you don’t have “max” data and don’t want eQuest to assume performance at a part-load ratio >1.0, you can set the DESIGN-PLR to 1.0.
4.	Either way, you want your curves to be normalized at whatever condition you specify (Design or Rated) and you want to enter those values on the Basic Specifications tab.

 

John

 

John T. Forester, P.E., LEED AP, Mechanical Design Engineer I BVH Integrated Services I 617.658.9008 tel I 617.244.3753 fax I One Gateway Center Suite 506, Newton MA 02458 I www.bvhis.com I Hartford ● New Haven ● Boston

<hr size=2 width="100%" align=center tabindex=-1> 

From: Nick Caton [mailto:ncaton at smithboucher.com] 
Sent: Wednesday, November 03, 2010 12:06 PM
To: John T. Forester; equest-users at lists.onebuilding.org
Subject: RE: [Equest-users] Chiller Curves (oh boy!)

 

John,

 

The design/max ratio is exactly what I’m discussing below when I say “DESIGN-PLR ratio,” so we’re definitely in the same ballpark =).

 

I’d like to apply/paraphrase your advice to a few conditions to be sure I’m getting it correctly:  

-          If we create all 3 custom curves, and normalize each to a point at maximum (not design) capacity, then the design-to-max ratio (DESIGN-PLR) should be set to 1.00.

-          If we create only the part load efficiency curve (EIR-FPLR or EIR-FPLR&dT), and wish to use the library defaults for EIR-FT and CAP-FT, then we should normalize this curve’s data points to ARI conditions (as that’s what the library curves are normalized to, per James’s email – I think I’ve read this somewhere also), specify an ARI  capacity, EIR and conditions on the basic specifications tab, and enter a DESIGN-PLR of [ARI capacity/maximum capacity (for the same conditions)].

-          If we create all 3 custom curves, and try to normalize each to either ARI or design conditions, then we should specify capacity, EIR, CHWT, CWT and condenser GPM corresponding to either the ARI or design conditions of that normalizing point.  In that case, we also specify a DESIGN-PLR using either the ARI or design capacity divided by the maximum capacity for the same conditions.

 

Profound (to me) Conclusion:  In No instance should we Ever attempt creating custom curves and NOT have at least one run from our manufacturer telling us what the maximum (not design) capacity is for the normalizing point.  This conclusion would only apply to centrifugal chillers only.

 

Does this all sound right?

 

~Nick

 

 

James, 

 

Yeah, if all the part load data you received held the same CHWT and CWT equal, you might be able to make your part-load curve if it could have been a quadratic EIR-FPLR curve (like a reciprocating chiller), but not a bi-quadratic EIR-FPLR&dT (as with my centrifugal VSD chiller).  You definitely could not approach generating custom EIR-FT or CAP-FT curves without varying condenser and chilled water temps.  That exact issue happened to me the first few times I tried to reign my chiller reps in =).  

 

This time, I convinced my rep to give me multiple part load runs holding the CHWT constant and varying the CWT incrementally.  This let me build the bi-quadratic EIR-FPLR&dT curve as I had at least three different dT’s represented in my part load data points.  I plotted the 3D curve in excel to check my work and darned if the generated coefficients seem to be really accurate =)!   It’s currently looking like a bittersweet revelation however – the library curve for a water-cooled centrifugal VSD chiller (see attached visualization) seems a LOT more generous (more efficient) at low part loads than the one I’ve generated which matches my rep’s data…  I might share a visual of my custom curve for comparison once I’m dead-sure it’s accurate – I’m trying to clarify a few things with my rep right now.

 

~Nick

 



 

NICK CATON, E.I.T.

PROJECT ENGINEER

25501 west valley parkway

olathe ks 66061

direct 913 344.0036

fax 913 345.0617

Check out our new web-site @ www.smithboucher.com 

 

From: John T. Forester [mailto:JohnTF at BVHis.com] 
Sent: Wednesday, November 03, 2010 10:29 AM
To: Nick Caton; equest-users at lists.onebuilding.org
Subject: RE: [Equest-users] Chiller Curves (oh boy!)

 

Nick,

 

When defining a centrifugal chiller in eQuest, one of the items on the Basic Specifications tab under the Design vs. Rated Conditions is a Design/Max Cap ratio.  By default, this is 92% for a water cooled unit.  I believe this gets at the discussion in the help pages that talks about maximum capacity versus design capacity and how the chiller vendor will spec a piece of equipment.  Typically, vendors don’t often get asked (or provide) what the “Maximum” capacity of a spec’d unit is.  Therefore the performance data that they provide are at “design conditions.”

 

If you change the chiller type to a reciprocating chiller, this “Design/Max Cap” ratio is disabled and the default specified condition changes from “Design Conditions” to “Rated Conditions.”  This suggests that there is little “extra” capacity when a selection is done for that type of chiller.

 

If you do get “maximum capacity” data and create curves from that data, you will want to change the Design/Max Cap ratio to 1.0 so eQuest knows that there isn’t any spare capacity at the chiller.  Also if the data points you are using to normalize your curves are different than the design conditions for your energy model, you will want to change the “Chiller Specified at” value to “Rated Conditions” and enter the rated conditions for CHW temp, CW temp and CW gpm/ton to match your normalized curves.

 

Hope this helps,

 

John

 

John T. Forester, P.E., LEED AP, Mechanical Design Engineer I BVH Integrated Services I 617.658.9008 tel I 617.244.3753 fax I One Gateway Center Suite 506, Newton MA 02458 I www.bvhis.com I Hartford ● New Haven ● Boston

<hr size=2 width="100%" align=center tabindex=-1> 

From: equest-users-bounces at lists.onebuilding.org [mailto:equest-users-bounces at lists.onebuilding.org] On Behalf Of Nick Caton
Sent: Wednesday, November 03, 2010 11:03 AM
To: Carol Gardner; equest-users at lists.onebuilding.org
Subject: Re: [Equest-users] Chiller Curves (oh boy!)

 

Thanks for the response Carol!

 

That 120% load case is what I’m getting at – let me try to explain a little further:

 

In the DOE2 help files, the vocabulary for centrifugal chillers is “design capacity” and “maximum capacity,” where “design” means the capacity at the rated or designed conditions (at which you define / specify your chiller), and “maximum” means the capacity the chiller is really capable of under the same conditions if it runs balls-out (maximum power to the refrigerant drive).

 

The help file excerpt I copied below with the red line is pretty explicitly telling us to normalize the part load values to the maximum capacity for centrifugal chillers.  I’ve highlighted a second line for clarity.  The EDR guidelines I linked below are saying you can instead normalize to the design capacity for the EIR-PLR curve if that’s all your field measurements or manufacturer rep can provide.

 

I’m asking – are both approaches right?  

 

My first and second questions are kinda tied together… How would choosing to normalize to either the maximum or design conditions affect how we should handle the DESIGN-PLR ratio, if at all?

 

 

~Nick.

 



 

NICK CATON, E.I.T.

PROJECT ENGINEER

25501 west valley parkway

olathe ks 66061

direct 913 344.0036

fax 913 345.0617

Check out our new web-site @ www.smithboucher.com 

 

From: Carol Gardner [mailto:cmg750 at gmail.com] 
Sent: Tuesday, November 02, 2010 10:04 PM
To: Nick Caton
Cc: equest-users at lists.onebuilding.org
Subject: Re: [Equest-users] Chiller Curves (oh boy!)

 

Let me take a crack at this. If by design capacity you mean the chiller running at 100% load, you would create the curve(s) by normalizing around your ARI design conditions i.e. the PLR curve would be 1.0 at this point, call it ARI Cap and the other points would be 90% Cap/ARI Cap, 80% Cap/ARI Cap, etc. The same would go for your temp curves. If, however, your chiller is operating at 120%, or some such other level, I would normalize the curve around the ARI design conditions of the chiller at 120%. I had to do this for a VRV hp that was selected at the 120% design condition.

I find this from the DOE2 manual the most helpful:

  

Volume 2: Dictionary > HVAC Components > CURVE-FIT > INPUT-TYPE = DATA


INDEPENDENT-2


Used for all curves having two independent variables. A list of up to twenty values of the second independent variable. The number of values should be the same as for DEPENDENT.

Example 1: defining a curve by inputting a set of data points.

A packaged system (PZS) has cooling performance significantly different from that used in the default  model. The manufacturer lists the data shown in Table 46, for cooling capacity, at 2000 cfm design air flow rate, as a function of outside dry-bulb temperature and entering wet-bulb temperature.

Table 46  Cooling capacity (kBtu/hr) vs. temperature

Outside
Dry-bulb

Entering Wet-bulb

72F

67F

62F

85F

69

65

60

95F

68

63 (ARI)

57

105F

65

60

53

115F

62

55

49

 

In this example the independent variables are the entering wet-bulb temperature and the outside dry-bulb temperature. Because there are two independent variables and they have units of temperature, we input a curve of TYPE BI-QUADRATIC-T using the given data points. The dependent variable is not the cooling capacity listed in the table but rather the cooling capacity divided by the cooling capacity at the ARI rating point (95 F outside dry-bulb and 67 F entering wet-bulb). In other words, the capacities should be normalized to the ARI rating point., as shown in Table 47

Table 47  Normalized capacity vs. temperature

Outside
Dry-bulb

Entering Wet-bulb

72F

67F

62F

85F

1.095

1.032

0.952

95F

1.079

1.0 (ARI)

0.905

105F

1.032

0.952

0.841

115F

0.984

0.873

0.778

 

The CURVE-FIT input will look like the following:

CAP-CURVE-1 = CURVE-FIT
TYPE               = BI-QUADRATIC-T
INPUT-TYPE         = DATA
DEPENDENT          = (1.000,1.079,0.905,1.032,0.952,0.841,
                      0.984,0.873,0.778,1.095,1.032,0.952) ..
IN-TEMP1           = (   67,   72,   62,   72,   67,   62,
                         72,   67,   62,   72,   67,   62) ..
IN-TEMP2           = (   95,   95,   95,  105,  105,  105,
                        115,  115,  115,   85    85,   85) ..

Example 2: Defining a curve by inputting coefficients

We want a furnace to have a constant efficiency as a function of part load. To do this we must replace the default FURNACE-HIR-FPLR with a curve that will give a constant efficiency. The curve TYPE is QUADRATIC in the part load ratio (PLR). PLR correction curves are always multiplied by the unit capacity, not the load, to obtain the energy (fuel or electricity) use. Thus the curve we want is: 0.0 + 1.0*PLR + 0.0*PLR*PLR. The input will look like:

New-Furnace-HIR-fPLR = CURVE-FIT
TYPE               = QUADRATIC
INPUT-TYPE         = COEFFICIENTS
COEFFICIENTS       = (0.0,1.0,0.0) ..

Then in the SYSTEM command we include:

   FURNACE-HIR-FPLR = New-Furnace-HIR-fPLR

 

On Tue, Nov 2, 2010 at 3:21 PM, Nick Caton <ncaton at smithboucher.com> wrote:

Hi everyone!

 

I think I have finally wrapped my mind completely around custom chiller performance curves for a centrifugal VSD chiller.  I’ve got a few specific questions now that I’m on the other side of the fence:

 

1.       Is it necessary for the data points of a part load efficiency curve (EIR-FPLR&dT in my case) to originate from data with a 1.0 (100%) PLR ratio corresponding to a maximum vs. a design load capacity?  From what I gather in the EDR reference <http://www.energydesignresources.com/Portals/0/documents/DesignGuidelines/EDR_DesignGuidelines_%20HVAC_Simulation.pdf>  (re: “Method 2” on PDF page 32/65), this curve can be generated using part-load readings assuming a design capacity at the 100% loading mark… but the DOE2 help entry for “EIR-FPLR” seems to suggest otherwise (copied below – see highlighted line).

2.       If the above part load efficiency curve is created based on data where the 100% loading point corresponds to the maximum (not design) capacity, should “DESIGN-PLR” (the ratio of design to maximum capacity) be set to 1.00 and the capacity of the chiller be specified at its maximum (not design) for the design/rated conditions?  As I write this question it sounds like I’m chasing my tail – someone straighten me out =)!

3.       When you veterans finish a project with sets of custom performance curves, do you have any suggestions for a naming scheme for future reference/re-use?  I’m currently thinking to keep the curves grouped in an .inp snippet I for importing along with an equipment cutsheet… but I’m certain I’ll forget the all the details as quickly as humanly possible when this project is behind me…

 

 ~Nick

 



 

NICK CATON, E.I.T.

PROJECT ENGINEER

25501 west valley parkway

olathe ks 66061

direct 913 344.0036

fax 913 345.0617

Check out our new web-site @ www.smithboucher.com <http://www.smithboucher.com>  

EIR-FPLR

Takes the U-name of a curve that adjusts the electric input ratio as a function of

· The part load ratio (PLR) –  The PLR is defined as the ratio of the hourly load to the hourly capacity;  Load / Caphour

· The evaporator/condenser dT -  The temperature differential between the condenser and leaving chilled-water. The  meaning of the condenser temperature varies according to condenser type. 

For most chillers, the dT has a relatively small effect on part-load performance. However, for variable-speed centrifugal chillers, the effect of dT is as important as the PLR.  This is because the pressure rise across the impeller is proportional to the square of the impeller’s speed. Unless some form on condenser temperature relief is employed to reduce the temperature (and pressure) differential across the chiller at part load, the performance of a variable-speed chiller may not be significantly different than that of a constant-speed chiller.

To model power consumption as a function of the PLR only, use a CURVE-FIT of TYPE = QUADRATIC or CUBIC.  To model as a function of both PLR and dT, use a BI-QUADRATIC-RATIO&DT curve. The curve must be normalized to 1.0 at full load and the rated temperature differential. 

Note that, for centrifugal chillers, ‘full load’ is defined as the ‘maximum capacity’, not the ‘design capacity’. Refer to the DESIGN-PLR keyword for more information.

 


_______________________________________________
Equest-users mailing list
http://lists.onebuilding.org/listinfo.cgi/equest-users-onebuilding.org
To unsubscribe from this mailing list send  a blank message to EQUEST-USERS-UNSUBSCRIBE at ONEBUILDING.ORG




-- 
Carol Gardner PE

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.onebuilding.org/pipermail/equest-users-onebuilding.org/attachments/20101103/5b6d97d0/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/jpeg
Size: 1459 bytes
Desc: image001.jpg
URL: <http://lists.onebuilding.org/pipermail/equest-users-onebuilding.org/attachments/20101103/5b6d97d0/attachment.jpeg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/png
Size: 28596 bytes
Desc: image002.png
URL: <http://lists.onebuilding.org/pipermail/equest-users-onebuilding.org/attachments/20101103/5b6d97d0/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/png
Size: 26884 bytes
Desc: image003.png
URL: <http://lists.onebuilding.org/pipermail/equest-users-onebuilding.org/attachments/20101103/5b6d97d0/attachment-0001.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/png
Size: 26225 bytes
Desc: image004.png
URL: <http://lists.onebuilding.org/pipermail/equest-users-onebuilding.org/attachments/20101103/5b6d97d0/attachment-0002.png>


More information about the Equest-users mailing list