[Bldg-sim] eQuest Default f(PLR) Skepticism

[Robby Oylear]

Excuse me for reviving an old thread here, but I’ve been having a similar
problem to what Taylor had in this thread and I can’t really find the
answer in the replies he received.  I have been inputting actual chiller
performance data for a project that I’m on and the results come out that
the proposed building uses much more cooling energy than the baseline.
Throughout this project I’ve thought the baseline cooling energy to be too
low, especially now when compared to a VSD centrifugal chiller which
should have better part load efficiency the baseline out performs it.

To get a concept of what a real baseline chiller should look like I asked
my local rep to produce data for a fictitious ARI chiller which matches
the COP and IPLV required by 90.1.  I then compared the EIR-f(PLR) curves
of the ARI chiller to the eQUEST default curve and found them to be
drastically different.  The eQUEST default curve starts at 1, and like
Taylor said, decreases almost linearly as the part load decreases.  The
actual data that I received shows quite the opposite, which is the chiller
starts at 1.0 at 100% and at around 70% PLR it begins to increase up until
30% where it reaches a value of 1.5.  To me this seems quite more
realistic, where the chiller (without a VSD of course) would perform much
worse at low part loads.  eQUEST defaults actually perform best at part
loads of 10% or less.

The choice of going with the default curves or with actual data of a 90.1
baseline compliant chiller has a huge impact on the energy use of my
building because 40% of my hours are in the 10-15% part load range.

Anyone have any thoughts on this subject?


Robby Oylear

Project Engineer

direct: 206.788.4571

cell: 206.354.2721

 <http://www.rushingco.com/> www.rushingco.com


The original message is attached below:

A few more comments regarding boilers:

We based the default curve for atmospheric boilers on information received
from a Northern California manufacturer?s representative. The default
atmospheric boiler consumes 45% fuel at 40% part load, so efficiency does
deteriorate. At full load, the efficiency is 80%; calculated as (PLR=1.) /
(HIR=1.25).  At 40% load, the default curve yields an efficiency of 71%;
calculated as (PLR=0.40) / (HIR=1.25 * HIRfPLR=0.45).

We believe that a 71% efficiency at 40% load is reasonable for a modern
atmospheric boiler. The default was based on a packaged water-tube boiler
with internal convective recirculation, but performance for other types is
expected to be similar.

The default atmospheric boiler unloads to 40% part-load ratio, and then
starts to cycle on/off; the HIRfPLR curve is valid only for part-load
ratios
above 40%. The loss during the off-cycle is characterized by the ?standby
time?, which is the equivalent full-load time in hours required to keep
the
boiler hot, if the boiler has no load at all. The default is 0.027 hours,
meaning that, at no load, the boiler would have to run about 2
minutes/hour
at full load to offset the jacket and flue losses. The documentation
describes the use of the standby-time vs. the part-load curve in further
detail. At very low loads, efficiency is seriously degraded; by definition
the efficiency at 0% load is 0%. You can observe this if you set up an
hourly report for the boiler and observe it at very low loads.

Our experience to date is that manufacturer?s part-load data for boiler
performance is quite difficult to obtain. If any of you have
well-documented
data that deviates significantly from the defaults, we would appreciate
hearing from you. As eQUEST/DOE-2 supports a library of equipment, it
would
be possible to add performance data for various types of boilers to the
library.


For chillers, let me further clarify Kevin's comment regarding the
EIRf(PLR,dT) curve. As Kevin explained, the dT term need be developed only
for variable-speed centrifugal chillers; it is of quite limited value for
constant-speed centrifugals, and of extremely limited value for
positive-displacement machines.

The discharge/suction temperature differential is closely associated with
the discharge/suction pressure differential. If the cooling tower is
controlled to a fixed, high setpoint such as 85F, then the chiller
impeller
may not be able to slow down much at all, even at low loads; the majority
of
capacity modulation will then be via the inlet vanes rather than speed.
This
is because the maximum possible pressure rise across the impeller drops
off
as the square of the impeller speed. Chiller impellers are typically
closely
matched to the design pressure rise specified by the engineer. If the
required pressure rise does not drop off substantially as the load drops,
then the impeller must maintain speed to avoid surge.

Data commonly published by manufacturers (and implicit in the IPLV)
ASSUMES
the condensing temperature drops with load; thereby allowing the impeller
speed to drop off. That may not actually be the case in real life! Nor is
it
true in eQUEST unless you specify a low tower setpoint, or utilize a tower
reset scheme.

The default EIRf(PLR,dT) curve for variable-speed centrifugals was
developed
using a manufacturer's proprietary software package that was loaned to us;
it password-expired after a short time. To our knowledge, these data
cannot
be developed from software available to the general engineering community.
(You need to be able to vary the chilled-water and condensing temperatures
independently of the part-load ratio. This is also why IPLV data is
worthless for an hourly simulation program; the "condenser relief" is
built
into the part-load performance.) Let me know if you are aware of any
centrifugal chiller manufacturers that make this data generally available.

This brings to mind another interesting point that perhaps others can
respond to. Over the years I have heard unsubstantiated rumors that a
multi-impeller chiller, such as a Trane, may have significantly different
part-load performance compared to a single-impeller chiller, such as a
Carrier. Does anybody have information regarding multi-impeller vs.
single-impeller part-load performance?

Regards,

Steven Gates
eQUEST Development Team

-----Original Message-----
From: BLDG-SIM at xxxxxxxx [mailto:BLDG-SIM at xxxxxxxx] On Behalf Of Kevin
Madison
Sent: Friday, October 05, 2007 8:45 AM
To: BLDG-SIM at xxxxxxxx
Cc: BLDG-SIM at xxxxxxxx
Subject: [BLDG-SIM] eQuest Default f(PLR) Skepticism

You are too generous Mike. I should heed my own advice on the
documentation. Thanks for clarifying this.

I also thought and consulted a colleague (thanks Steve) on Taylor's
chiller question:

For chillers, the second term in EIRf(PLR,dT) is important only for
variable-speed centrifugal chillers. It is of negligible importance in
other types of chillers. It is important in variable-speed, because the
temperature differential is strongly correlated with the required speed
of the impeller. If the dT is high enough, then the impeller may have to
run at full speed even at low loads. The DOE-2 default curves are just
that, defaults. Some folks find it necessary to create new curves to
reflect their specific equipment. Cautions offered
* verify the conditions (condenser/evaprator) for efficiencies at lower
part loads are the same as for above 50%
* make sure the points are normalized around the rating condition

Sorry for my hasty response.

Kevin Madison


Michael Tillou wrote:
> Actually Kevin didn't get the boiler hourly energy equation quite
> right. The actual equation is:
> Hourly Boiler Energy = DesignCapacity * HIR * HIRf(plr)
> The part of this that Kevin didn't explain is that the boiler
> HIRf(PLR) curve includes the PLR which explains why the curve is
> nearly linear. The value that the performance curve returns is actually
> (HIRadj) * PLR
> HIRadj = the multiplier that indicates how the full load HIR changes
> with respect to part load. If the boiler efficiency at a given part
> load goes down, HIRadj > 1. If the boiler efficiency goes up at a
> given part load HIRadj<1. HIRadj is really the ratio HIR-partload over
> HIR-fullload.
> PLR = hourly load on the boiler / total capacity of the boiler
> To create a curve that describes boiler HIR at various part loads you
> will need to divide the performance curve output at each part load
> point by the part load value and then multiply by the full load HIR.
> Taylor - You should double check your "custom" chiller curves I'm
> pretty sure from what you describe they are not correct. The Vol6 -New
> Features user manual does a good job describing how the chiller curves
> work. I suggest you review this. You can use the Excel function
> "LINEST" to create the necessary coefficients for a bi-quadratic curve
> from manufacturers chiller data. Typically you will need to request
> data for a specific chiller from the chiller rep. The hardest data to
> get is the chiller capacity data at various CHW/CW temperatures.
> Remember total chiller capacity is different than the rated 100% part
> load point, most chillers can provide 10-20% extra capacity.
> I have had good experience creating custom chiller curves for DOE2.2
> and I think the default curves in eQuest are representative of the
> various chiller types. Obviously if you are evaluating a specific
> chiller you should try to create custom curves.
> Mike
> ------------------------------------------------------------------------
> *From:* BLDG-SIM at xxxxxxxx [mailto:BLDG-SIM at xxxxxxxx] *On Behalf Of
> *Kevin Madison
> *Sent:* Thursday, October 04, 2007 9:46 PM
> *To:* BLDG-SIM at xxxxxxxx
> *Subject:* [BLDG-SIM] eQuest Default f(PLR) Skepticism
>
> Perhaps it would help to clarify how DOE-2.2 (the simulation engine
> behind eQUEST) calculates hourly energy input for boilers and chillers.
>
> For boilers, the hourly energy input is:
> Hourly Energy = Cap(hour) * HIR * HIRf(plr)
>
> So while the HIRf(plr) may increase as part load decreases, which is
> not uncommon for standard atmospheric boilers, the energy use will
> certainly decrease with plr because the required output of the boiler
> for the hour decreases.
>
> For chillers, DOE-2 uses the following relationship to calculate the
> electricity input to the chiller each hour:
>
> Caphour = Capacity * CAPf(t1,t2)
> PLR = Load / Caphour
> dT = t2 ? t1
> Elechour = Caphour * EIR * EIRf(t1,t2) * EIRf(PLR,dT) / 3413 Btu/kW
>
> where
>
> Caphour hourly capacity, Btuh (this is dependent on condenser and
> evaporator conditions for that hour)
> Capacity rated capacity, Btuh
> CAPf(t1,t2) correction to capacity for temperatures, curve CAP-FT
> t1 leaving chilled-water temperature, °F
> t2 condenser temperature, °F
> PLR Part load ratio
> Load Hourly load, Btuh
> dT Temperature differential across chiller, °F
> Elechour electric input to the chiller, kW
> EIR rated electric input ratio
> EIRf(t1,t2) correction to EIR for temperatures, curve EIR-FT
> EIRf(PLR,dT) correction to EIR for part-load ratio and dT, curve
EIR-FPLR
>
> Again, the primary factor affecting chiller energy use is the cooling
> capacity needed for that hour. Just because you don't have access to
> the dual function information doesn't mean you shouldn't be accounting
> for it in the simulation. Chiller performance is dependent on all
> operating conditions including load, condenser conditions and
> evaporator conditions.
>
> For a more complete discussion on these simulation concepts, refer to
> the DOE-2 documentation included with the eQUEST installation. Look in
> Dictionary:HVAC Components:Boiler:Boiler Energy Consumption and
> Dictionary:HVAC Components:Chiller:Chiller Energy Consumption.
>
> Kevin Madison
> Madison Engineering PS
> Seattle WA
> USA
>
> Taylor Keep wrote:
>>
>> eQuest models boiler and chiller plants with default part load curves
>> that I think may be incorrect. As I understand it, the f(PLR) curves
>> are used as a direct multiplier on the HIR for boilers and EIR for
>> chillers, with full load (1.0 PLR) corresponding to a 1.0 multiplier.
>> If this is true, the f(PLR) curve should increase at part load for
>> atmospheric boilers (atmospheric boilers become somewhat less
>> efficient at part load). The default atmospheric boiler curve
>> decreases almost linearly down to zero! I am having a tough time
>> wrapping my head around this.
>>
>> On the chiller side, the default f(PLR) is a bi-quadratic function
>> using both dT and PLR as variables, so it is f(PLR,dT). Since I never
>> have this dual function information in my general chiller selections,
>> I have been using a standard f(PLR) function quoted at a fixed dT
>> from the manufacturer. The curve I get from a McQuay 400-ton chiller
>> selection is quadratic, with decreasing EIR down to 50% load and
>> increasing EIR below 50% load. I seriously doubt that the eQuest
>> default corresponds with this entry because changing the function
>> produces a huge change in performance.
>>
>> Do any of you have any thoughts or suggestions about the accuracy of
>> default f(PLR) curves? Should I scrap my "improved," real curves -
>> they are drastically changing the model performance?!?!
>>
>> Taylor
>>
>>
>> ________________________________________________________
>> Taylor Keep
>> Mechanical, LEED® AP
>> _ _
>> Arup
>> 901 Market Street Suite 260
>> San Francisco, CA 94103
>> tel: 415 946 0279
>> fax: 415 957 9096
>> taylor.keep at xxxxxxxx
>> _www.arup.com_ <file://www.arup.com>
>>
>>
>> ____________________________________________________________
>> Electronic mail messages entering and leaving Arup  business
>> systems are scanned for acceptability of content and viruses
>>
>> ======================================================
>> You received this e-mail because you are subscribed
>> to the BLDG-SIM at xxxxxxxx mailing list.  To unsubscribe
>> from this mailing list send a blank message to
>> BLDG-SIM-UNSUBSCRIBE at xxxxxxxx
>>
> ======================================================
> You received this e-mail because you are subscribed
> to the BLDG-SIM at xxxxxxxx mailing list.  To unsubscribe
> from this mailing list send a blank message to
> BLDG-SIM-UNSUBSCRIBE at xxxxxxxx
>
> ======================================================
> You received this e-mail because you are subscribed
> to the BLDG-SIM at xxxxxxxx mailing list.  To unsubscribe
> from this mailing list send a blank message to
> BLDG-SIM-UNSUBSCRIBE at xxxxxxxx
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.onebuilding.org/pipermail/bldg-sim-onebuilding.org/attachments/20080204/511284d4/attachment.htm>