14 ASHRAE Jour nal ashr ae.

or g J ul y 2011
T
his is the frst of a series of articles discussing how to optimize the
design and control of chilled water plants. The series will summarize
ASHRAEs Self-Directed Learning (SDL) course called Fundamentals of De-
sign and Control of Central Chilled Water Plants and the research that was
performed to support its development (see sidebar [Page 20] for topics to
be discussed). The articles, and the SDL course upon which it is based, are
intended to provide techniques for plant design and control that require
little or no added engineering time compared to standard practice, but
at the same time result in signifcantly reduced plant life-cycle costs.
A procedure was developed to provide
near-optimum plant design for most chill-
er plants including the following steps:
1. Select chilled water distribution
system;
2. Select chilled water temperatures,
ow rate, and primary pipe sizes;
3. Select condenser water distribution
system;
4. Select condenser water tempera-
tures, ow rate, and primary pipe sizes;
5. Select cooling tower type, speed
control option, efciency, approach
temperature, and make cooling tower
selection;
6. Select chillers;
7. Finalize piping system design, cal-
culate pump head, and select pumps; and
8. Develop and optimize control se-
quences.
Each of these steps is discussed in
this series of ve articles. This article
discusses Step 1: Selecting chilled water
distribution system type.
Table 1 lists recommendations for
the life-cycle cost optimum distri-
bution system based on the size and
number of loads served and the num-
ber of chillers. Life-cycle cost opti-
mum is in quotes because these rec-
ommendations are generalizations that
should apply to the majority of typical
HVAC applications, but they may not
prove to be optimum for every appli-
cation and have not been rigorously
proven as the best choice. They are
based on the authors design and com-
missioning experience, the analysis
that was done in conjunction with de-
velopment of the ASHRAE SDL, work
done on an earlier chilled water plant
design manual,
1
and the prescriptive
About the Author
Steven T. Taylor, P.E., is a principal at Taylor
Engineering in Alameda, Calif.
By Steven T. Taylor, P.E., Fellow ASHRAE
Optimizing Design & Control
Of Chilled Water Plants
Part 1: Chilled Water Distribution System Selection
This article was published in ASHRAE Journal, July 2011. Copyright 2011 American Society of Heating, Refrigerating and Air-Conditioning
Engineers, Inc. Posted at www.ashrae.org. This article may not be copied and/or distributed electronically or in paper form without permission
of ASHRAE. For more information about ASHRAE Journal, visit www.ashrae.org.
Jul y 2011 ASHRAE Jour nal 15
requirements of ASHRAE/IESNA Standard 90.1.
2
The in-
tent is to allow designers to select the system that is most
often the best choice from a life-cycle cost perspective for
a given application without having to perform any lengthy
analyses.
Primary-Only Single Coil
With one or more chillers serving a single cooling coil (Fig-
ure 1), the simplest design strategy is to not use any control
valves at the coil. Instead, a constant-volume pump circulates
water between the chiller and the coil and supply air tempera-
ture is controlled by resetting the temperature of the chilled
water leaving the chiller. While constant chilled water ow
results in constant pump energy, chiller performance is im-
proved when the leaving chilled water temperature is reset
to be as high as possible. A variable frequency drive (VFD)
could also be added to the pump to make the system vari-
able ow, but that adds cost and complexity. VFDs are sel-
dom cost effective since pump power is typically small in a
single-coil plant because the chiller and coil are usually close-
coupled (physically close together), and it is more efcient to
increase chilled water temperature than to reduce pump speed
and pump energy. (Part 5 of this series will further discuss
the trade-off between resetting chilled water temperature and
pump energy.)
Many engineers are concerned about causing high space
humidity with this design because chilled water temperatures
must be aggressively reset to maintain supply air temperature
at setpoint under low load conditions. But, in fact, that is never
a concern because the leaving supply air condition is about the
* In all of the gures shown in this article, multiple chillers are shown in parallel. For most applications, the chillers could alternatively be piped
in series. This results in lower chiller energy use, partly offset by higher pump energy use. However, rst costs are almost always higher with
series piping due to larger piping and pumps and bypass piping typically provided to allow one chiller to operate while the other is down for
maintenance. Because of limited funding, we did not evaluate the life-cycle costs of series piping and so we have not included it in our recom-
mendations. This option will be evaluated for cost effectiveness in future versions of the ASHRAE SDL.
same regardless of chilled water temperature. For example, if
the supply air temperature leaving the coil is 55F (13C), the
air leaving the coil is close to saturated whether the chilled
water supply temperature is 42F (6C) or 50F (10C). It is
the supply air temperature setpoint that determines space hu-
midity conditions, not the chilled water temperature.
Figure 1 shows a single chiller, but any number of chillers
can be used. When two chillers are used, this is a good applica-
tion for piping chillers in series
*
rather than in parallel.
Figure 1 also shows an optional storage tank. Chilled water
systems must have a sufcient volume of water in the pip-
ing system to prevent unstable temperature swings, possibly
Application
Coils/Loads
Served
Chillers Size of Coils/Loads Served Control Valves
Recommended
Distribution Type
1 One Any Any None
Figure 1
Primary-Only Single Coil
2 More Than One One Small (< ~100 gpm) 2-Way and 3-Way
Figure 2
Primary-Only Single Chiller
3
Few Coils Serving
Similar Loads
More Than One Small (< ~100 gpm) 3-Way
Figure 3
Primary-Only Multiple Chillers
Few Coils With Similar Loads
4
Many Coils Serving
Similar Loads or
Any Serving
Dissimilar Loads
More Than One Small (< ~100 gpm) 2-Way
Figure 4
Primary-Only
or
Figure 5
Primary-Secondary
5 More Than One Any Large Campus 2-Way
Figure 6
Primary Distributed Secondary
6 More Than One Any Large Coils (> ~100 gpm) None
Figure 7
Primary Coil Secondary
Table 1: Chilled water distribution system.
Figure 1: Primary-only single coil.
Chiller
CHW
Pump
Optional
Storage Tank
Supply Air Temperature
Supply Water
Temperature
Coil
16 ASHRAE Jour nal ashr ae. or g J ul y 2011
causing chiller short-cycling. This is
a potential problem with single-coil
systems since they are typically close-
coupled with only short piping runs. To
compensate for the small water volume
in piping, a small storage tank is often
required. The minimum water volume
should be veried with the chiller manu-
facturer.
Primary-only Single Chiller
Small chilled water plants com-
monly have a single chiller, typically
air-cooled. Single chiller plants do
not have to deal with flow and staging
problems common to multiple chiller
plants and thus can have a simple
distribution and control system. The
recommended design is shown in Fig-
ure 2. It is the simplest variable flow
primary-only system. Two-way valves
are installed at most coils with just
enough three-way valves installed
to maintain the minimum flow re-
quired by the chiller. This minimum
rate, which can be obtained from the
manufacturer, will vary with design
chilled water flow rate and the chiller
type, size, and manufacturer but is
typically 25% to 50% of the design
flow. A VFD is shown in Figure 2;
VFDs are typically cost effective
except on very small systems. Note
that Standard 90.1 requires VFDs on
chilled water pumps exceeding 5 hp
(3.7 kW). The VFD is controlled by
a differential pressure (DP) sensor
located near the most remote coil so
that the DP setpoint can be as low as
possible; this is also a requirement
of Standard 90.1. Locating the sen-
sor near the pump requires a high DP
setpoint and eliminates most of the
energy savings from the VFD.
The three-way valves should be lo-
cated near the chiller if the pump has
a VFD to minimize pump energy. Lo-
cating them remotely increases ow to
the extremes of the system, which in-
creases the pump pressure and power
required. The one exception to this rule
is that three-way valves must be located
in a manner that engages enough water
volume to maintain the minimum wa-
ter volume required to minimize short
cycling as discussed previously. There
is usually no benet to locating three-
way valves remotely to keep the system
cold so that chilled water is instantly
available to coils; it typically takes just
seconds or perhaps minutes for water
to travel from the chiller to the most re-
Figure 2: Primary-only single chiller.
Chiller
Supply Water
Temperature
CHW
Pump
3-Way Valve
Coil
DP Sensor
2-Way Valve
Chiller No. 1
Supply Water
Temperature
CHW
Pumps
3-Way
Valve
Coil
Chiller No. 2
Figure 3: Primary-only multiple-chill-
ers few coils with similar loads.
VFD
mote coil, and the load will not be lost in
that short time.
Primary-Only Multiple Chillers:
Few Coils With Similar Loads
When systems have multiple chillers,
chiller staging can be a problem when
ow and load do not track, and they
generally do not when three-way valves
are used. Consider the system shown
in Figure 3. When the system operates
near full load, performance is satisfac-
tory since both chillers and pumps are
operating. However, the system can
have problems during part-load condi-
tions depending how coil loads vary.
For example, suppose the system shown
in Figure 3 had two equally sized chill-
ers and served two equally sized coils,
each serving a hotel ballroom. If there
were functions in both rooms and both
rooms were above 50% load, the system
operates well; both chillers with their as-
sociated pumps will run and each func-
tion space will receive its design ow.
But when only one of the two function
spaces is occupied and the other is va-
cant, the system, as a whole, will be
below half load, so in theory only one
chiller and pump could satisfy the load.
However, the coil serving the unoccu-
pied room will still use its design ow,
bypassing it around the coil to the re-
turn. If the plant operates with only one
chiller and pump, it has sufcient chiller
capacity to meet the load, but it cannot
meet the ow demands; the coil serv-
ing the occupied meeting room will be
starved of ow. To avoid this problem,
both pumps will have to operate, so both
chillers will have to operate at or below
50% load. This problem is one of the
reasons designers have migrated to vari-
able ow designs, discussed later.
But this system can work well as long
as all coil loads tend to vary in the same
proportion, as they might if all coils
served similar occupancies (e.g., all
serve ofces on the same schedule). For
instance, if the coils served are below
half load and only one chiller and pump
are operating, all coils will be capable of
meeting their loads. The system is thus a
quasi-variable-ow system in that pumps
and chillers can be staged. Also, because
Advertisement formerly in this space.
18 ASHRAE Jour nal ashr ae. or g J ul y 2011
the loads vary similarly, chilled water
temperature may be reset aggressively,
which allows the plant to be about as
efcient as one of the true variable ow
systems discussed later. This system is a
reasonable choice for small applications
with only a few coils serving similar
loads; it is simple and inexpensive and
avoids all of the complexities of vari-
able ow systems. Also, small systems
like this are typically close-coupled, so
there is not much pump energy to save.
Note that Standard 90.1 only allows this
approach for systems with three coils or
fewer or a total chilled water pump sys-
tem power of 10 hp (10.7 kW) and less.
Primary-Only Variable Flow and
Primary-Secondary Systems
Application 4 in Table 1 is probably
the most common. It applies to systems
serving many small coils (or a few coils
with dissimilar loads) and more than
one chiller. In this case, either of two
systems is recommended, primary-only
variable ow (Figure 4) or primary-sec-
ondary (Figure 5). Both systems have
plusses and minuses, discussed in detail
in an article by Taylor
3
and summarized
in Table 2.
Primary-only systems always cost less
and take up less space than primary-
secondary systems, and with variable
speed drives, primary-only systems also
always use less pump energy than tradi-
tional primary-secondary systems. The pump energy savings
are due to:
Reduced system head as a result of the elimination of the
extra set of pumps and related piping and devices (shut-off
valves, strainers, suction diffusers, check valves, etc.).
More efcient pumps. The primary pumps in the primary-
secondary system will be inherently less efcient due to their
high ow and low head. This can be partially mitigated by
using larger pumps running at lower speed, but at an increase
in rst costs.
Variable ow through the evaporator, which allows ow to
drop below design ow down to some minimum ow rate pre-
scribed by the chiller manufacturer. VFDs can be added to the
primary pumps of a primary-secondary system and controlled
to track secondary ow down to the chiller minimum rate, but
at an increase in rst costs and control complexity.
The lower energy costs and lower rst costs of the prima-
ry-only system often make it an easy choice versus primary-
secondary, but the system does have two signicant disad-
vantages.
Figure 4 (left): Primary-only variable fow. Figure 5 (right): Primary-secondary.
DP Sensor
2-Way Valve
Chiller 1
Primary
Pumps
Flow Meter
Coil
Chiller 2
Bypass
VFD
VFD
DP Sensor
2-Way Valve
Chiller 1
Primary
Pumps
Coil
Chiller 2
V
F
D
V
F
D
Common Leg (Decoupler)
Secondary Pumps
1. Bypass Control Complexities
A bypass valve (Figure 4) is required to ensure that mini-
mum ow rates are maintained through operating chillers. The
valve must be automatically controlled by ow, typically using
a ow meter in the primary circuit (as shown in Figure 4) or
differential pressure sensors across chillers correlated to ow.
The ow meter is more costly but is more easily adapted into
plant load (Btu) calculations, which will be necessary for op-
timum chiller staging (discussed in Part 5).
Selecting the bypass control valve and tuning the control
loop is sometimes difcult because of the widely ranging
differential pressure across the valve caused by its location
near the pumps. The valve must be large enough to bypass the
minimum chiller ow through it with a pressure drop as low as
the differential pressure setpoint used to control chilled water
pump VSDs. This is because if only a few valves are open
in the system, the pressure at the DP sensor location will be
what is available at the plant as well since there is little pres-
sure drop between the two points due to the low ow. But this
makes the valve oversized for other ow scenarios that can
Advantages of Primary-Only Disadvantages of Primary-Only
Lower First Costs
Less Plant Space Required
Reduced Pump Peak Power
Lower Pump Annual Energy Use
Complexity of Bypass Control
Complexity of Staging Chillers
Table 2: Advantages and disadvantages of primary-only vs. primary-secondary systems.
Advertisement formerly in this space.
20 ASHRAE Jour nal ashr ae. or g J ul y 2011
occur, so tuning can be difcult. If the control loop is unstable,
cold chilled water supply can be fed back into the return inter-
mittently and cause chillers to cycle off due to low load or cold
supply water temperatures. But if the loop is too slow, it may
not respond quickly enough to sudden changes in ow (e.g.,
when a large number of air-handling units shut off at the same
time), causing insufcient ow through the chillers, causing
them to trip on low ow or low temperature.
Complex control systems are prone to failure, so at some
point in the life of the plant, one can expect the bypass con-
trol to fail. A failure of the bypass system can cause nuisance
chiller trips, which generally require a manual reset. If an op-
erator is not present to reset the chiller, the plant can be out of
service for some time.
2. Staging Control Complexities
When one or more chillers are operating and another chiller
is started by abruptly opening its isolation valve (or starting its
pump for dedicated pumps), ow through the operating chill-
ers will abruptly drop. The reason for this is simple: ow is
determined by the demand of the chilled water coils as con-
trolled by their control valves. Starting another chiller will not
create an increase in required ow, so ow will be split among
the active machines. If this occurs suddenly, the drop in ow
will cause operating chillers to trip.
To stage the chillers without a trip, active chillers must rst
be temporarily unloaded (demand-limited or setpoint raised),
then ow must be slowly increased through the new chiller
by slowly opening its isolation valve. Then, all chillers can be
allowed to ramp up to the required load together. During the
staging sequence, chilled water temperatures will rise some-
what. This is seldom a problem in comfort applications, but
may be an issue for some industrial applications.
Given these considerations, primary-only systems are most
appropriate for:
Plants with many chillers (more than three) and with fairly
high base loads, as might be expected in an industrial or data
center application. For these plants, the need for bypass is
minimal or nil due to the high base loads, and ow uctuations
during staging are small due to the large number of chillers.
Plants where design engineers and future on-site opera-
tors understand the complexity of the controls and the need to
maintain them.
The primary-secondary system may be a better choice for
buildings where fail-safe operation is essential or on-site op-
erating staff is unsophisticated or nonexistent.
Primary Distributed Secondary
For plants serving groups of large loads such as buildings in
a college campus, terminals in an airport, etc., the primary-dis-
tributed secondary system (Figure 6) is usually the best solu-
tion. The secondary pumps at the central plant are deleted and
variable speed pumps are added at each building. The build-
ing pumps are controlled by differential pressure sensors at the
most remote coil in each building. Building pump heads are
This series of articles will summarize the upcoming
Self-Directed Learning (SDL) course called Funda-
mentals of Design and Control of Central Chilled Water
Plants and the research that was performed to support its
development. The series will include ve segments:
Chilled water distribution system selection. This
article will discuss distribution system options, such as
primary-secondary and primary-only pumping, and pro-
vide a simple application matrix to assist in selecting the
best system for the most common applications.
Condenser water distribution system selection. This
article will discuss piping arrangements for chiller-con-
densers and cooling towers, including the use of variable
speed condenser water pumps and water-side economiz-
ers.
Pipe sizing and optimizing T. This article will dis-
cuss how to size piping using life-cycle costs then how to
use pipe sizing to drive the selection of chilled water and
condenser water temperature differences (Ts).
Chillers and cooling tower selection. This article will
address how to select chillers using performance bids and
how to select cooling tower type, control devices, tower
efciency, and wet-bulb approach.
Optimized control sequences. The series will con-
clude with a discussion of how to optimally control chilled
water plants, focusing on all-variable speed plants.
The intent of the SDL (and these articles) is to provide
simple yet accurate advice to help designers and opera-
tors of chilled water plants to optimize life-cycle costs
without having to perform rigorous and expensive life-
cycle cost analyses for every plant. In preparing the SDL,
a signicant amount of simulation, cost estimating, and
life-cycle cost analysis was performed on the most com-
mon water-cooled plant congurations to determine how
best to design and control them. The result is a set of im-
proved design parameters and techniques that will pro-
vide much higher performing chilled water plants than
common rules-of-thumb and standard practice.
Central Chilled Water Plants Series
sized for the pressure drop of the loop from the plant, to the
building, through the buildings coils, then back to the plant
through the common leg. Therefore, each pump has a different
head customized for the building.
The advantages of this design compared to conventional
primary-secondary and primary-secondary-tertiary systems
include:
Overall pump horsepower is reduced. With the conven-
tional system, secondary pump head must be sized for the
most remote building (say 100 ft [299 kPa]) while the distrib-
uted building pumps close to the central plant can have much
smaller heads (say 50 ft [150 kPa]).
Advertisement formerly in this space.
22 ASHRAE Jour nal ashr ae. or g J ul y 2011
The system is self-balancing via the
speed controls on the secondary pumps.
There is no need to throttle pressure at
close buildings and ow self-adjusts
over time as additional buildings are
connected to the system.
Overpressurization of control valves
located near the central plant is elimi-
nated. With large, high head second-
ary systems, these valves must operate
against excess differential pressure,
which can reduce controllability and
may even force ow through the valve if
it does not have sufcient shut-off head.
Pump energy is reduced because of
the custom pump heads and the more
precise control of the variable speed
drives. With a conventional primary-
secondary system, the secondary pumps
are typically controlled to maintain dif-
ferential pressure at the entry to the most
remote building. Therefore, the setpoint
must be higher than for the distributed
pump system, which is controlled by
differential pressure at the most remote
coil in each building. At part load, the
pumps therefore can operate at slower
speeds.
With primary-secondary-tertiary sys-
tems, the tertiary pumps are generally
piped with a bridge and two-way control
valve. Control of the bridge is always
difcult and, if done incorrectly, often
the cause of degrading DT.
4
With this
Figure 6: Primary-distributed secondary.
Building A Building B
CHW
Supply
CHW
Supply
Primary
Pumps
Chiller N
Chiller 2
Chiller 1
Common Leg
Secondary
Pump
VFD VFD
distributed pumping system, bridge con-
nections are eliminated.
The system will be less expensive,
more energy efcient, and have lower
maintenance costs than primary-second-
ary-tertiary systems.
Disadvantages include:
Expansion tank pressurization may
have to be increased to maintain posi-
tive suction pressure at building pumps
if the pumps are located at the top of
campus buildings. This has only a mi-
nor cost impact to the expansion tank.
Figure 7: Primary-coil secondary.
VFD
VFD
Chiller 1
Primary
Pumps
Coil
Chiller 2
Coil
Pump
AHU 1
(Large Coil)
AHU 2
(Large Coil)
AHU 3
(Large Coil)
VFD VFD VFD
Chiller 1
Chiller 2
Common Leg
VFD VFD
Figure 8: Hybrid primary-coil secondary and primary-secondary system.
Primary-distributed secondary sys-
tems will usually cost more than con-
ventional primary-secondary systems
because there are more pumps and
space is required to house them in each
building.
Advertisement formerly in this space.
24 ASHRAE Jour nal J ul y 2011
Primary Coil Secondary
For plants serving large individual air-handling systems,
using distributed variable speed driven coil secondary pumps
(Figure 7, Page 22) is usually the best solution.
The advantages of this design compared to a conventional
primary-secondary system include:
Connected pump motor hp is reduced. This is due in part be-
cause of the customized heads for each pump but also because the
control valve is eliminated. Two-way control valves are typically
selected for a wide-open pressure drop of 4 to 5 psi (27.5 kPa to
34.4 kPa), about 10 ft (29.9 kPa). This is a substantial savings.
The system is self-balancing. There is no need for bal-
ancing valves of any kind nor are there any advantages to
self-balancing designs such as reverse-return arrangements.
Pump energy is significantly lower with this design.
This is due mostly to the reduced pump heads but also
because there is no need to maintain
a minimum differential pressure in
the system as there is with conven-
tional secondary pumps. Because of
this minimum DP and because of the
throttling caused by partially closed
control valves, conventional second-
ary pumps will not follow the theo-
retical parabolic system curve. Hence,
pump efficiency will generally get
worse, particularly at low load. With
the variable speed coil pump design,
there are no control valves or mini-
mum DP, so pump efficiency will be
nearly constant.
Control of large control valves
is inherently slow due to the size and
slow responsiveness of the valve. With
the coil pump design, ow can be con-
trolled almost instantaneously with the
VFD, so control is precise. There is
also no fear of over-pressurizing con-
trol valves, which reduces their con-
trollability.
Because of the eliminated control
valves and lower pump HP, this system
will generally have lower costs than a
conventional primary-secondary sys-
tem. It is usually a little more expensive
than a primary-only system.
Control valves can be thought of as
brakes on a car while pumps are the
car engines; from an energy perspec-
tive, it never makes sense to press both
the brake and the accelerator pedals at
the same time, but that is effectively
what systems with control valves do.
So, this system is actually ideal from
a pumping perspective: it has no
brakes.
Unfortunately, there are a few dis-
advantages of this system. First, all
coils must have a pump. If a coil were
connected to the secondary circuit
without a pump, flow through the coil
will be backwards from the return to
the supply. For a building that has a
Advertisement formerly in this space.
Jul y 2011 ASHRAE Jour nal 25
mixture of small coils and large coils, pumps for the small
coils will most likely have to be expensive multistage
pumps.
Another disadvantage is the increased exposure to
equipment failure. A control valve is extremely reliable
the pump and VFD in this design are more likely to fail.
Duplex pumps could be used to improve redundancy, but
the cost is prohibitive in most situations. Our philosophy
is to provide the same level of redun-
dancy as the rest of the system served.
For instance, if the air handler has
only a single fan, then it makes sense
to provide only a single pump. For
more critical applications, redundant
pumps, or an alternative distribution
system design, should be considered.
For this design to be energy effi-
cient, coils must be large due to the
inherent inefficiency of small pumps,
particularly low flow/high head
pumps. For instance, a typical pump
at 60 ft of head (179 kPa) will have
an efficiency on the order 30% at 20
gpm (1.3 L/s), 50% at 50 gpm (3.2
L/s), 60% at 100 gpm (6.3 L/s), and
70% at 200 gpm (12.6 L/s). That is
why this system is recommended only
for coils with flows greater than 100
gpm (6.3 L/s) in Table 1. This flow
limit is obviously a rough rule-of-
thumb since efficiency will vary over
a range, not drop abruptly below 100
gpm (6.3 L/s).
If a project includes both small and
large coils, a hybrid system of both dis-
tributed coil pumps and conventional
secondary pumps to serve small coils is
possible. See Figure 8, Page 22, for an
example hybrid plant.
Summary
This article is the rst in a series of
ve that summarize chilled water plant
design techniques intended to help en-
gineers optimize plant design and con-
trol with little or no added engineering
effort. In this article, a simple look-up
table is provided to allow designers to
select a near optimum chilled wa-
ter distribution system based on their
application without having to do any
rigorous life-cycle cost or system
analysis. Next month, condenser water
distribution system selection will be
addressed.
References
1. Energy Design Resources. 1999. CoolTools Chilled Water
Plant Design Guide.
2. ASHRAE Standard 90.1-2007, Energy Standard for Buildings
Except Low-Rise Residential Buildings.
3. Taylor, S. 2002. Primary-only vs. primary-secondary variable
ow systems. ASHRAE J ournal 44(2).
4. Taylor, S. 2002. Degrading chilled water plant delta-T: causes
and mitigation. ASHRAE Transactions 108(1).
Advertisement formerly in this space.