Evolving Design
Of Chiller Plants
By Thomas H. Durkin, RE., Member ASHRAE
turing the last 15 years, mechanical rooms have seen dramatic
Dei The rooms have become smaller with fewer pumps
and valves, the equipment has become more efficient, and more
attention is paid to intricacies of connections. The quest is for solu-
tions that are less expensive to build, less expensiv
2 to operate and
easier to maintain than traditional (or previous) arrangements. This
article covers the improvements in chiller plant design, the reasons
why each variation was made, the challenges and opportunities that
were presented, and the results achieved
The evolution is driven by a con- that they work efficiently. and the yoal
timious search for improvement, and ought to be that each new system works
the only way that can happen is if the better and more efficiently than previous
performance of projects is tracked
desivns,
and all the results and implications are ‘This article traces the ongoing de.
Understood. The standaed should never velopment of mechanical room designs
be that systems work. That must be a from 1988. A baseline installed cost and
prerequisite. The standard should be baseline operati
plant have been inflation-adjusted so the
schemes allow for direct comparisons,
For example, if the baseline chiller
(1988 scheme) was 240 tons (S44 KW), it
would cost $134,000 to install today (RS,
Means 2005 Mechanical Cost Data). 8
chiller forthe same size building per the
1990 design would be 192 tons (675 kW)
(80% load diversity) and cost $103,000 in
today’s pricing. The 1990 chiller portion
of the total cost would be 0.76 of baseline
Increases in controls result in a total cost
oF 0.90 of baseline.
A similar analysis for operating cost
beuins with an estimate of $35,000 per
year to operate the 1988 design at today’s
utility rates, based on a 100.000 ft (9290
m2) school in the Midwest. The pump
energy saved by the 1990
timated to lower annwal operat
250, oF 0.95 of the baseline
umption that runs through all
ion based on 12° (44°F to
[About the Author
Thomas H. Durkin, PE
& diector of eng
iB ShoulderS6*F) AT. The design AT has a direct impact on pumping ener
and pump pipe sizing
1988: Primary/Secondary Constant Speed Pumping
(Of Both Chiller Circuits
All coils were fitted with three-way valves and automatic
Flow control valves, aka flow limiters (Figiy 1).
+ Installed Cost: baseline of 00
+ Operating Cost: baseline of 1.00
+ Advantages: Simple control, sae: easy to understand
+ Disadvantages: Unable to take advantage of lond diversity in
building; all pumps, chillers. pipiny are sized for the “sum
‘othe peaks.” This means low AT is Hed
‘water is mixed (see siebar "Mixing Chilled Water’)
normal” and
1990; Constant-Speed Primary, Variable-Speed
Secondary Pumping For Chiller Circuits
All coils ate fitted with two-way valves and automatic flow
linvters (Figure 2) The intent was o take advantage of the part
load efficiency of a pump running at reduced speed. Theoreti-
ly. as.a pump slows to half speed and half flow {they work
ther), the enerey used i (12) oF full speed eneruy: The ac
tual energy saved is not that dramatic due to inefficiency of the
variable speed drive, Consiclering that a properly sized cooling,
system will operate at full load only about SO hours a year. vs
about 1.500 hours of cooling system operations (a school inthe
Midwest, the savings potential is considerable.
COfeven bigger value is that now, the central plant can be down
sized, In the 1988 version, chillers were sized forthe “sum ofthe
individual spaces” vs. the “building peak” in the 1992 version.
‘Contributing to this would be solar load, which affects the east,
side inthe morning, and the south and west sides inthe afternoon,
Thisis referred toas diversity in the load. Iti typically about 80”
meaning the building peak is 20% less than the sum of the
vial spnces. Lange buildings with larve assembly spaces, such as,
‘auditoriums and gyms, may put the diversity as low as 70%,
+ Installed Cost: 0.900 of baseline
*+ Operating Cost: 0.950 of baseline.
+ Payback: Immediate
+ Advantages: Saves energy; and downsized central plant
pumps. piping.
+ Disadvantages: Increased control complexity: and low A
is still a limiting factor (see sidebar “What 1s Low 47°).
1993: Remove Flow Limiters
‘This may not sound like much of a change, but the results
were pretty interesting, While desivning the final phase of a
multiyear project that began in 1990, the last addition encled
tup being larger than had been conceived three years earlier
Faced with modifying a recently installed chiller plant to solve
‘critical issue of pump capacity, 10% more flow was needed,
By removing the pressure drop (flow limiters) from the pip-
oop, the system curve shifted to the right far enough to
give the capacity needed (Figure 5). The same pump was now
delivering more flow
“This raised some interesting questions:
+ What do flow limiters really do?
Answer; They make system balanci
easy and control
valve selection less important. [Fa control valve were to
fail open, the flow limiter would limit the impact on the
rest ofthe system, maybe even fo the point of masking the
filed control valve
+ What happens when the Hlow limiters are removed”
Answer: On the building side, nothing happens as long as
the control valves are working correctly n the mechanical
room, additional pump capacity was found, and 3 psi (7 10)
of head was taken off the pumps, equaling a drop in pump
energy forthe life ofthat build
+ Anything else”
Answer: In many control schemes, the winter fail-safe
{reeze-protection operation for heating systems isto open
the control valves and count onthe flow limiters to balance
the flow: This means that the hot water pumps are runnin
at 100% speed. Thisalso means that the hot water pumps
‘weren't sized for the building peak load, but were sized
forthe larger sum of the individual spaces. And, it means
that the speed drives aren't being allowed to save money
since most of the operating hours are unoccupied when the
pumps are at full speed. A much better scheme is to limit
the freeze protection valve position to some small percent
calculated to prevent freeze-ups. This will allow for pump
‘eneray savings, and eliminate the need for low limiters
+ Installed Cost: 0.890 of baseline
+ Operating Cost: 0.943 of baseline.
+ Payback: Immediate
+ Advantages: Saves pump ene
+ Disadvantage
and saves opetating cost
me consuming: control
Balancing more
valve and actuator selection more critical
1996: Variable Primary Flow
‘This was the next logical step in the evolution of variable-
speed pump systems, The constant-speed chiller primary pumps
design purpose is to provide constant flow through the chiller to
censure safe operation. However, with modem chiller controls,
the need for constant flow through the chiller is not as critical
as a generation ago, Minimum flow will always be required
“The variable primary-flow design did away with the constant-
speed chiller pumps and added chiller isolation valves. A flow
‘meter in the chiller inlet line monitored flow. with a setpoint
‘minimum corresponding to the number of chillers needed. The
Flow meter would open a chiller minimum flow valve to ensure
sate operation (Figure 4).
From a first-cost standpoint, replacing two large pumps with
two isolation valves, adding some additional controls. and
increasing the size of the remaining pumps resulted in a net
savings. That savings is compounded by reducing the required
size of the mechanical room,rE hy — non
Pane Let ‘Seay Parmer Peter
~6— fm ee -o- ill
‘coal (corer vom | | Fem seen |
Figure 1: Prinary/secondary constant speed pumping (1988),
40.
no.
100
%
Figure 2: Variable speed secondary pumping (19%,
"Naw Selection
680 gpm at 83
BD
i ie
= i
3 ¢
gli z
a é
z 2
® original Stecon
mH G00 gor at 90
»
» »
SPH RED 10
0 °
a a a
Figure 3: Removing flow limiters (1993)
Froma pump energy standpoint at fall load, litte difference
exists from a primary’secondary scheme. At all part loads,
variable primary flow will use less enerey than any primary
secondary scheme.
From an operational standpoint, we now have a scheme
that makes low AT-a nonissue. We can “over-pump” the
Mixing Chilled Water
All primary/secondary systems have a rmx point atthe de-
‘coupler whenever the secondary flow exceeds the primary
Mining chiled water is almost never a good idea, erases the
‘temperature and increases pump energy. Chilled water supply
temperature has t0 match the coll selection temperature so.
cooling cos can dehumidify properly and so the coils can de-
velop the intended AT. Although for many hours in the year the
coldest water wor'tactually be required for either temperature
‘or humility contro, if he system can tolerate warmer water
the better way is to raise the chiller setpoint
Ona related note, raising chilled water supply temperature
Capaciey in US. gpm
chiller if we need to, and stage chillers on flow and load
Father than just flow. This isa tar ery from saying that low A
is solved
+ Installed Cost: 0.867 of baseline
+ Operating Cast: 0.937 of baseline
+ Payback: Immediate
a8 an operating strategy can increase efficiency, increase ca-
pacity or lower demand, During the summertime, when eais-
ing supply temperature would save kWh and demand, that's
when the coldest water is needed to dehumidily property
Allchillers will be more efficent at warmer evaporator water
outlet temperatures. For example, a 4°F (2°C) rise in supply
temperature will roughly equal a 3% efficiency increase, some
of which willbe given back by increased pumping energy for
all variable speed distribution systems, Caution: if you're rai
ing supply temperature to save demand charges, remember,
thac although chillers have a 3% lower kW/ton at 4°F (2°C)
higher evaporator temps, they aso have 696 greater capacity
and a net increase in total power consumption,Figure 4: Variable primary low, Version 11996)
+ Advantages: Save fist cost; save operating cost; low AT
a nonisste
+ Disadvantay
Energy” sidebar)
s: Increased control complexity (see “Ghost
1997: The Good Idea Needed Optimizing
‘Too many small things popped up that negatively impacted
reliability, control and the ability to fine tune the system. Flow
meters did not perform as intencled due 10 water quality issues,
piping layout, ancl instrument accuracy.
Realizing that the most accurate and well-documented orifice
inthe mechanical room was the chiller barrel the solution was
to install differential pressure transmitter across the evapora
tor barrel and write a simple algorithm to control the chiller
minimum flow valve (Figwve 3). This allowed more stable
control, more finite control at low flow and a small inerease in
pump efficiency.
+ Installed Cost: 0.864 of baseline
+ Operating Cost: 0.937 of baseline
+ Payback: Immediate
+ Advantage: Better reliability
+ Disadvantages: None
What Is Low AT?
Ie is a phenomenon that occurs in most chilled water
systems when the return water coming back to the cooling
plant is net as warm as itis designed to be. For example. if
the design called for chilled water supply to the system to
be 45°F (7°C), and the return to be 55°F (13°C), then any
time return water was 54°F (12°C) or cooler, the plant would
be suffering from low AT. Usually worse at part loads, its
frequently seen at full load also. It defies the fundamentals of
heat transfer, since theoretically a al part-load conditions,
the AT should be above design rather than below i.
Why is Low AT a Problem?
‘Te limits the capacity out of your chillers. Generally, the
most expensive piece of equipment in a building is the
chiller, In a primary-secondary pumping arrangement
with the primary pumps sized for the chiller capacity at
Figure 5: Variable primary flow, Version 2 (1997)
1999; Pressure-Independent Flow Control Valves (More
Optimizing)
When designing, a myriad of operating conditions need to be
consideted that could affect the system, such as full oad, part,
Toad, one pump out, startup, shutdown, ete. One particularly
vexing condition is rapid increases in load
I the chiller plant were operating with pu
minimum flows valve in control, a large ait-handling unit start=
ing could cause a sueden increase in demand for chilled water:
‘cause a chiller to trip on low flow ifthe minimum
Flow valve could nor open up fast enough and the pump speed
could not inerease fast enowzh
‘The initial solution was to slow down the rate at which
air-handling unit control valves open anc close to avoid sud=
den tlow variations. It was felt that slow acting valves would
eliminate most ofthe problems. However a better desizn would
eliminate all of the nuisance trips, to be as operator friendly
as possible
After much product research in various valve and actuator
styles, an interesting device with its origins in the oilfield i
dustry appeared most promising. A pressure-independent flow
control valve (see sidebar “Pressure independent”) became
the design AT, chiller planes willbe limited on flow rather
than load unless the AT is at design
+ Iewastes pump energy. Ina 10°F (5.5°C) AT design. ifthe
water is coming back at 7°F (4°C) AT, then the system
‘s pumping 30% more than it should, and theoretically
using 220% of the pump energy that it should.
+ The piping/pumping syseem may be unable to meet the
building toad
Possible Causes of Low AT
* Dirty coll, airside or water-side:
+ Poor coil selection;
+ Changing design conditions:
+ Laminar flow on waterside:
+ Coll unequal air distribution;
+ Three-way valves: |
+ System DP above valve shutoff head
+ Piping configuration, ete.the standard for chiller minimum ow protection, and a big,
drop in spurious trips was seen. The piping schematie did not
change trom the 1997 version, but operation became much
more stable:
* Installed cost: 0.865 of baseline (increase vs. 1997)
+ Operating cost: 0.937 of baseline
+ Payback: Never, but, highly worthwhile
+ Advantage: Better reliability
+ Disadvantage: Slight price increase
2001 and 2002: Two Changes in Two Different Steps,
Both Involving Cooling Coil Control Valves
First, the pump speed differential pressure transmitter (DPT)
\was relocated om across the terminal unit branch main (where it
had been from 1999) to across only the most remote contol valve,
‘The theory is that best control and most energ
seen by monitoring as close to the control device as possible. And,
additional energy savings is possible since, in the original scheme
the DPT setpoint was typically around 25 ft (75 kPa) (coil 10
[30 kPa}: controt valve = 10 ft (30 kPa}; al valves’strainer pipe
511 [15 kPa), Inthe new location, the DPT setpoint would only
need to be 10 f(30 kPa), meaning that the variable speed pumps,
‘would spend more hours running at lower speed.
This was an example of what Albert Einstein called a
scheme: an improvement that made things worse” Without
the dampening effect ofthe coil, piping and the mains, even small,
‘changes in valve position resulted in wild varationsin differential
pressure and constant speed swings at the pumps. As with the
chiller minimum flow valve, slowing down the valve response
slowed the DPT swings but could not eliminate them,
Conventional contrat valves could not control accurately or
respond quickly enough to allow reasonable contol, hence the
second improvement The success that was seen with the pressure
indepenslent control iller minimum flow
ils Figure 4), Inaddition, wth the control valves actu-
ally controlling correctly, cooling coil A was now at ar above
design, occasionally significantly above des
‘Two things were immediately evident: the pumps could be
accurately and smoothly controlled at a much lower setpoint.
hhence more energy savings: and low AT no longer existed
Al previous efforts were about making low AT'a non-issue,
this change solved it. With PICVs on all the cooling coils, the
system finally obeyed the fundamentals of heat transfer. At all,
part-Load conditions, return water is consistently above design,
‘with no change in room comtort level,
From afirst-cost standpoint, PICV are about 50% more than
‘conventional control valves. First-cost is akvays a concern, as
‘owners and construction manayers look for value engineering
items. On a 100,000 ft (9290 m*) building, PICV's will add
about $20,000 to the cost, which is 0.162% ofthe total project,
cost ancl 0.7% of the HVAC package. These certainly are not
ignificant percentages.
From our energy savings point, the ghost energy was elimi:
nated (see “Ghost Energy” sidebar). and control was much,
smoother. However. calculated payback was 12 years, prompt-
ing a reevaluation of other parameters to lessen the first-cost,
impact, One suggestion was that since low AT was no longer an
issue, piping mains could be downsized. Building load diversi
was showing up as much greater flow diversity. Resizing and
recaeulating the mains is worth abou hal the premium forthe
PICY, bringing the payback to
“Serendipity” is find oul that you weren't
looking for. So, solving the low A was serendipity. However,
it presented another concern. With the PICVs on all our cool-
ing coils, the chilled water AY’ was now consistently at 16°F
(0°C) and occasionally in the 20°F ange (11°C) (12°F (7°C]
design). Most chiller manufacturers will say that, when used
in a variable primary flow system, that chiller minimum flow
should be 1.5 gpnvton (0.03 Lis per kW). That equates to
a chiller AV of 16°F (9°C), which means that at many load
conditions. considerable water is being bypassed when the
chiller was heavily loaded. The chiller industry may need t0
better define minimum flow so a chiller can run fully loaded
with almost any A. without introducing another kind of ehost
‘energy (unnecessary bypas
3 water}
Pressure-Independent Control Valves
The flow-to-position response of valves (called valve au-
thority) is characterized by ASHRAE as either quick opening,
linear or equal percentage. Almost all valves used for tem:
perature control systems are specified as equal percentage
Chapter 42 of the ASHRAE Handbook—HVAC Systems and
Equipment shows how flow in different valve styles reacts
to changes in position.
Iealso has achartillustratinga phenomenon called authority
distortion. The chart shows that any valve will only perform
as advertised ifthe pressure differential across the valve stays
‘constant, which almost never happens. From fullload to part
load. pressure diferential across a control valve can vary sig
nificantly. For example, 50% valve position may equal 10%
Slow at design differential, or it may equal 0% flow at part
load (authority distortion = 0.1. see $42.9, Figure 19)
‘Apressure-independent control valve automatically and im-
‘mediately responds co variations in pressure, Iisa relatively
recent innovation for hydronic applications. A perfect analogy |
isthe variable-air-volume (VAV)box. Years ago, when VAV was. |
inits infancy (a great idea but inconsistent space control), the
invention that made VAV the smart choice for most modern
buildings was making the VAV boxes pressure independent
The addition ofa flow ringin the box inlet meant thae the box:
ould maintain constant arflow into the space regardless of
anything that was happening on the upstream side. A pressure
independent control valve has a sinilar device atthe inlet.cae 1s vot Tecan
en, [omer tin “OED | Gage
es 7 Fora
i a iS Moore cnr
Figure 6 Version 3 with PICT 200
+ Installed Cost: 0.872 of baseline (increase)
+ Operating Cost: 0.900 of baseline
+ Payback: Six years.
+ Advantage: Better control; ener
host eneruy and low AZ.
+ Disadvantage: Slight price increase
2003: Dedicated Heat Recovery Chillers (DHRC) Became
‘Standard on All Four-Pipe Projects
Heat recovery chillers are not intended to replace the main cool-
iny plant only to replace summertime boiler operation, Theyarean
environmentally conscious way of controlling humidity. decreasin
greenhouse wases. and providing domestic water heat
‘The ideal location is where it will see the warmest enterin
evaporator temperature, which is before the chiller minimum,
flow bypass (Figure 7). Typical sizing for a 100,000 tt (9290
m2) school might be a 200-ton (703 kW) main cooling plant
and a 30-t0n (106 kW) DHRC
Even though itis called a chiller, its summertime function is
tomake hot water, and cooling isa by-product. That reverses in
the winter when it makes chilled water withthe rejected heat
as a by-product. This is distinctly different than a heat pump,
which has reversing valves on the refrigerant side, Ina DHRC,
the evaporator is alays the evaporator; the condenser is always
the condenser. The control parameters differ trom summer
(condenser control) to winter (evaporator control)
Figure 7: Dedicated heat recovery chillers (2003)
Consider the operating economics of heat recoxery chillers
ipplied to any concurrent heating cooling load (the heating
‘may include domestic water heating or swimming pool water
heating). The utility rates shown are typical of central Indiana
before Hurricane Katrina,
Option 1: Run boilers and ehillers.
Chiller: Air-cooled setew at 1.25 KWiton (0.36 KW'KW).
Electricity at $0.07/kWh = $0.729/100 MBtu
Boiler: Condensing boiler at 92% efficient gas at S085
therm = $0.924/100 MBtu,
‘Total Option 1 = S16:
100 MBIu of cooling.
Option 2: It’s cool enough to run economizers,
still a heating load.
Chiller: Of
Cooling = $0.00
Condensing boiler at 92% efficient gas at $0.85
therm,
Heating
‘Total Option 2 = $0.924 for 100 MBtu of heating and
100 MBtu of cooling.
0.924/100 MBnu
n 3: Heat recovery chiller.
Chiller: Modular scroll at 12 kWiton (0.34 kW/kW)
(55°F [13°C] Evap EAT, 120°F [49°C] Cond LWT); at
Opt
A condensing boiler is one that is built to withstand the
corrosive conditions of cool flue gases. It operates at cem-
peratures significantly cooler and at efficiencies significantly
higher than would be found in conventional boiler systems.
The most important factor of boiler efficiency is enter-
ing water temperature (EWT), see Figure 6, p. 27.4, 2004
ASHRAE Handbook—HVAC Systems and Equipment, Most
Conventional boilers have arninimium EWT low limit of 140°F
(60°C) to prevent condensation on the fireside. This is one
of the key factors that has pushed HVAC engineers ¢o the
“standard” 180°F (82°C) hot water systems
Condensing boilers are, by defirition, meant to condense,
so that the latent heat from the flue gas is sent into the hy
Condensing Boilers and Heat-Recovery Chillers