This article was published in ASHRAE Journal, June 2012. Copyright 2012 ASHRAE. 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.

Optimizing Design & Control

Of Chilled Water Plants
Part 5: Optimized Control Sequences
By Steven T. Taylor, P.E., Fellow ASHRAE

his is the last of a series of articles discussing how to optimize the

design and control of chilled water plants. The series summarizes

ASHRAEs Self Directed Learning (SDL) course called Fundamentals of

8. Develop and optimize control sequences.

Each of these steps is discussed in this
series of five articles. This article discusses step 8.

Typical Chiller Plant

Figure 1 is a typical primary-only

Design and Control of Central Chilled Water Plants and the research variable flow chilled water plant. The
that was performed to support its development. 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
significantly reduced plant life-cycle costs.
A procedure was developed to provide
near-optimum plant design for most chiller plants including the following steps:
1. Select chilled water distribution
system.
2. Select chilled water temperatures,
flow rate, and primary pipe sizes.
3. Select condenser water distribution
system.
56

ASHRAE Journal

4. Select condenser water temperatures, flow rate, and primary pipe sizes.
5. Select cooling tower type, speed
control option, eff iciency, approach
temperature, and make cooling tower
selection.
6. Select chillers.
7. Finalize piping system design, calculate pump head, and select pumps.

plant has two of each major component (chillers, towers, condenser water
pumps, and chilled water pumps) each
sized for 50% of the load. This plant design is very common and was used as
the basis of the simulations and optimization for this series of articles and the
SDL course upon which it is based.
Note that the condenser water (CW)
pumps in Figure 1 do not have variable
speed drives (VSDs). Sequences for
variable speed CW pumps are also addressed in this article but, as discussed
in Part 21 of this series and in more detail below, VSDs on condenser water
About the Author
Steven T. Taylor, P.E., is a principal at Taylor
Engineering in Alameda, Calif.

a s h r a e . o r g

June 2012

Modeling the Plant

The plant in Figure 1, serving a typical office building, was
modeled with all permutations of the following design
variables:
Weather Oakland, Calif., Albuquerque, N.M., Chicago,
Atlanta, Miami, Las Vegas
CHWST Reset by valve position from 42F to 57F
Chillers
Two styles (two stage and open-drive)
Efficiency at 0.35, 0.5, and 0.65 kW/ton at AHRI
conditions
Towers
Approach: 3F, 6F, 9F, and 12F
Tower Range: 9F, 12F, and 15F
Efficiency: 50, 70, and 90 gpm/hp
Condenser water pumps With and without VSDs

VSD

VSD

Cooling
Tower 1

Cooling
Tower 2

VSD
Chiller 1

VSD

VSD
Chiller 2

VSD

The control equation coefficients were determined from

each run, then these coefficients were themselves regressed against various design parameters and weather
indicators. The results are shown below. The development
of these regressions is ongoing to include more weather
sites and chiller variations.
1. Condenser water temperature control. Control CW
return temperature to the setpoint determined from Equation 1a:

CWRT = CHWST + A PLR + B

(1a)

A = 63 + 0.0053 CDD65 0.0087 WBDD55 + 1.67

WB + 0.52 APPROACH 0.029 GPM/HP

(2)

C = 1.35 - 1.27E05 CDD65 + 1.36 NPLV 0.0212

WB 0.012 APPROACH + 0.0765 RANGE
D = 0.147 + 7.04E06 CDD65 0.124 NPLV + 0.0038
WB + 0.00133 APPROACH + 0.00217 RANGE

3. Chiller Staging. Use one chiller when PLR is less than

SPLR determined from Equation 3:

SPLR = E (CWRT CHWST) + F

(3)

E = 0.057 0.000569 WB 0.0645 IPLV 0.000233

APPROACH 0.000402 RANGE + 0.0399 KW/TON
F = 1.06 + 0.0145 WB + 2.16 IPLV + 0.0068
APPROACH + 0.0117 RANGE 1.33 KW/TON

These control sequences strictly apply to primary-only

plants with centrifugal chillers serving air handlers with
outdoor air economizers in a typical office building. It is not
known how well they apply to other applications.
June 2012

nus WB, F
CHWFR Chilled water flow ratio, actual flow divided
by total plant design flow

CHWST Chilled water supply temperature (leaving
evaporator temperature), F

CWFR Condenser water flow ratio, actual flow divided by total plant design flow

CWRT Condenser water return temperature (leaving
condenser water temperature), F

CDD65 Cooling degree-days base 65F

DP Differential pressure, feet H2O

KW/TON Chiller efficiency at AHRI conditions, kW/ton

DT Temperature difference, F
GPM/HP Tower efficiency per ASHRAE Standard 90.1

IPLV Integrated part load value per AHRI 550/590,
kW/ton

NPLV Non-standard part load value per AHRI
550/590, kW/ton

RANGE Design tower entering minus leaving water
temperature, F

PLR Plant part load ratio, current load divided by
total plant design capacity

TOPP Theoretical optimum plant performance

WB Design wet-bulb temperature, ASHRAE 1%,
F
WBDD55 Wet-bulb cooling degree-days base 55F

2. Variable speed condenser water pumps. Control CW

flow ratio to the setpoint determined from Equation 2:
CWFR = C PLR + D

Variables

APPROACH Design tower leaving water temperature mi-

B = 18 0.0033 CDD65 + 0.0053 WBDD55 0.26

WB + 0.15 APPROACH 0.014 GPM/HP

Figure 1: Typical chilled water plant schematic.

ASHRAE Journal

57

pumps are usually not life-cycle cost effective for plants serving office building type loads.
Also note in Figure 1 that the cooling towers do not include
any isolation valves to shut off flow to allow one tower to operate alone. As discussed in Part 42 of this series, towers generally can be selected with nozzles and dams that allow half
flow from one CW pump while still providing full coverage of
fill and it is always most efficient to run as many tower cells
as possible. So whether one or two CW pumps are operating, both tower cells are enabled and fans are controlled to the
same speed.

Determining Optimal Control Sequences

www.info.hotims.com/41640-35

58

Chilled water plants have many characteristics that make

each plant unique so that generalized sequences of control that
maximize plant efficiency are not readily determined. Equipment and system variables that affect performance include:
Chillers. Each chiller has unique characteristics that affect
full-load and part-load efficiency such as compressor design,
evaporator and condenser heat transfer characteristics, unloading devices (such as variable speed drives, slide valves,
and inlet guide vanes), oil management systems, and internal
control logic.
Cooling towers. Tower efficiency (gpm/hp) varies significantly by almost an order of magnitude between a compact
centrifugal fan tower to an oversized propeller fan tower. Towers can also be selected for a wide range of approach temperatures.
Chilled and condenser water pumps. Pumps and piping
systems can be selected for a broad range of Ts and may or
may not include variable speed drives. Pump efficiency also
varies by pump type and size and pump head varies significantly depending on physical arrangement and pipe sizing
standards.
Chilled water distribution systems. Distribution system
arrangements, such as primary-secondary vs. primary-only
variable flow, significantly affect plant control logic.
Weather. Changes in outdoor air conditions affect loads and
the ability of cooling towers to reject energy.
Load profile. The size and consistency of loads will affect
optimum sequences. For instance, control sequences that are
optimum for an office building served by air-handling systems
with airside economizers may not be optimum for a data center served by systems without economizers.
With so many variables, no single control sequence will
maximize the plant efficiency of all plants in all climates for
all building types.
There are a number of papers3,4 on techniques to optimize
control sequences for chilled water plants. Almost all require
some level of computer modeling of the system and system
components, and the associated amount of engineering time
that most plant designers do not have. In writing this series of
articles and the SDL upon which it is based, significant modeling was performed in an effort to determine generalized control sequences that account for most of the variation in plant

A S H R A E J o u r n a l

June 2012

CHW Pump Control

Chilled water pump speed is typically controlled to maintain supply-to-return differential pressure (DP) at setpoint.
Standard 90.15 requires that the DP sensor(s) be located at the
most remote coil(s). This is because the lower the DP setpoint,
the lower the pump energy, as shown in Figure 2.* If the DP
setpoint is reset by valve position, as discussed further below,
pump energy can be close to the ideal curve in Figure 2 for
DP setpoint = 0.
Figure 3 shows the optimum number of CHW pumps as
a function of CHW flow ratio and as a function of pump
speed for the chilled water plant shown in Figure 1 based
on TOPP modeling. Unlike cooling towers, the optimum
sequence is not to run as many pumps as possible. This is
because the pumps all pump through the same circuit (other
than the pipes into and out of the each pump between headers) so there are not cube-law energy benefits for each
pump individually.

100
90

DP Setpoint = Design Head

DP Setpoint = Head 0.75

80

DP Setpoint = Head2
DP Setpoint = Head3

70

Pump kW (%)

design parameters summarized above. The technique used to

determine optimized performance is described in a June 2007
ASHRAE Journal article.4
In brief, the technique involves developing calibrated simulation models of the plant and plant equipment that are run
against an annual hourly chilled water load profile with coincident weather data while parametrically modeling virtually
all of the potential modes of operation at each hour. The operating mode requiring the least amount of energy for each hour
is determined. The minimum hourly energy use summed for
the year is called the theoretical optimum plant performance
(TOPP). Since all modes of operation were simulated, the
plant performance cannot be better than the TOPP within the
accuracy of the component models.
TOPP modeling was performed for the chilled water plant
shown in Figure 1 for a wide range of plant design options for
tower range, approach, and efficiency; different chiller types
and chiller efficiency; and varying climates (see Modeling
the Plant, Page 57). The operating modes (e.g., number of
chillers, condenser water flow and pump speed, tower fan
speed and related condenser water temperatures) that result in
the TOPP for each plant design scenario were studied to see
how they relate to independent variables such as plant load
and weather (e.g., wet-bulb temperature) to find trends that
can be used to control the plant in real applications through
the direct digital control (DDC) system.
Ideally, equipment should be controlled as simply as possible; complex sequences are less likely to be sustained since
operators are more likely to disable them at the first sign of
perceived improper operation. The remainder of this article
discusses the TOPP modeling and the generalized sequences
that were developed from the analysis for the chilled water
plant shown in Figure 1 serving a typical office building.

DP Setpoint = 0 (Reset)

60
50
40
30
20
10
0
0 10 20 30 40 50 60 70 80 90
100
gpm (%)

Figure 2: Variable speed performance at varying DP setpoint.

Figure 3 clearly indicates that staging pumps off of flow
provides better optimization than staging off of pump speed.
As suggested by Figure 3, CHW pumps should be staged
as a function of CHW flow ratio (CHWFR = actual flow divided by total plant design flow) at a staging point of 47%,
i.e., one pump should operate when the CHWFR is below
47% and two pumps should operate when CHWFR is above
47%, with a time delay to prevent short cycling. The 47%
optimum staging point assumes DP setpoint is reset by valve
position; it will be somewhat higher at higher DP setpoints.
For very large pumps (>~100 hp [75 kW]), it may be worth
the effort to determine the actual pump operating point (flow
vs. head) and optimize staging based on pump efficiency determined by flow and pressure drop readings mapped to pump
curves duplicated mathematically in the DDC system.6 This
can allow pumps to operate closer to their design efficiency
as the system operating curve varies from the ideal parabolic
curve due to modulating valves and minimum differential
pressure setpoint. But the small potential energy savings is not
worth the effort for most chilled water plants.

Chilled Water Temperature and DP Setpoint Reset

Chillers are more efficient at higher leaving water temperatures so, in general, optimum efficiency is achieved when
the chilled water supply temperature (CHWST) setpoint is
as high as possible. (The impact of CHWST on CHW pump
energy is discussed below.) Where all zones are controlled
by the DDC system, the best reset strategy is based on valve

* The curves in this figure assume pressure drop varies with flow to the 1.8 power since flow is typically in the transitional region between turbulent and laminar flow.
They do not account for the impact of opening and closing control valves, which change system geometry and hence the system flow characteristics. The curves
do include reductions to the efficiency of motors and VSDs at low load.

June 2012

ASHRAE Journal

59

Wh/ft2yr

Wh/ft2yr

Optimum Number of Pumps

Optimum Number of Pumps

position where the CHWST setpoint is

reset upwards until the valve controlling the coil that requires the coldest
2
water is wide open. This strategy ensures that no coil is starved; all are able
1
to maintain their desired supply air or

space temperature setpoints.

Valve position can also be used to
0
0 10 20 30 40 50 60 70 80 90
100
reset the DP setpoint used to control
Percent of Design Flow
pump speed. In fact, this is required by
Standard 90.1. The logic is similar to
CHWST setpoint reset: the DP setpoint
is reset upwards until the valve control2
ling the coil that requires the highest DP
is wide open.
So we have a dilemma: Valve position
1
can be used to reset either CHWST setpoint or DP setpoint, but not both inde0
pendently; it is not possible to know if
0 10 20 30 40 50 60 70 80 90
the valve is starved for lack of pressure
Percent of Pump Speed
or from lack of cold enough water. We
must decide which of the two setpoints Figure 3: TOPP CHW pump staging vs. CHW flow ratio and pump speed.
to favor.
Houston
While resetting CHWST setpoint upward reduces chiller
Fans
Pumps
Towers
Chiller
energy use, it will increase pump energy use in variable
4,000
flow systems. Higher chilled water temperature will cause
3,500
coils to require more chilled water for the same load, degrading CHW T and increasing flow and pump energy re3,000
quirements. Degrading T can also affect optimum chiller
2,500
staging; however, this is not generally an issue in primary2,000
only plants with variable speed chillers.7 Furthermore, our
simulations have shown that the positive impact of reset1,500
ting chilled water temperature on chiller efficiency is much
1,000
greater than the negative impact on pump energy even when
500
distribution losses are high.
0
Figure 4 shows a DOE2.2 simulation of a primary-only
CHWST Reset,
Fixed CHWST,
Reset CHWST
plant with variable speed chillers and CHW pumps with
Fixed DP
DP Reset
Then DP
high pump head (150 ft [450 kPa]) using three reset strategies based on valve position: reset of chilled water temperaOakland
ture alone; reset of differential pressure setpoint alone; and
4,000
a combination of the two that first resets chilled water tem3,500
perature then resets DP setpoint. The simulations were done
3,000
in several climate zones (Houston and Oakland results are
shown in the figure) and in all cases, resetting chilled water
2,500
temperature was a more efficient strategy than resetting DP
2,000
setpoint. Sequencing the two (resetting chilled water temper1,500
ature first then DP setpoint) was the best approach, although
1,000
only slightly better than resetting chilled water temperature
alone.
500
Contrary to conventional wisdom, the impact of reset on the dehumidification
capability of air handlers is quite small and should not be a concern. Space humidity is a function of the supply air humidity ratio, which in turn, is a function of
the coil leaving dry-bulb temperature setpoint. Regardless of CHWST, the air leaving a wet cooling coil is nearly saturated; lowering CHWST only slightly reduces
supply air humidity ratio. As long as the supply air temperature can be maintained at the desired setpoint, resetting CHWST will not impact space humidity.

60

ASHRAE Journal

0
CHWST Reset,
Fixed DP

Fixed CHWST,
DP Reset

Reset CHWST
Then DP

Figure 4: Plant energy with CHWST Setpoint Reset, CW DP

Setpoint Reset and a combination of the two.
a s h r a e . o r g

June 2012

Figure 5 shows how this sequenced reset strategy can be

implemented. The x-axis is a software point called CHW
Plant Reset that varies from 0% to 100% using trim and
respond logic.8 The coil valve controllers generate requests
for colder chilled water temperature or higher pump pressure
when the valve is full open. When valves are generating requests, CHW Plant Reset increases; when they are not, CHW
Plant Reset steadily decreases.
When CHW Plant Reset is 100%, the CHWST setpoint is at
Tmin (the design chilled water temperature) and the DP setpoint
is at DPmax (the design DP setpoint). As the load backs off, the
trim and respond logic reduces the CHW Plant Reset point. As
it does, chilled water temperature is increased first up to a maximum Tmax (equal to the lowest air handler supply air temperature setpoint less 2F [1C]), then DP setpoint is reduced down
to a minimum value DPmin (such as 3 psi [21 kPa]).
In practice, this logic seldom results in much reset of the DP
setpointthe CHWST reset is aggressive enough to starve the
coils firstso it is important to locate the DP sensor(s) at the
most remote coil(s) so that DPmax can be as low as possible to
minimize pump energy (Figure 2).

For plants with more consistent loads that do not vary with weather, such as
those serving data centers and those located in consistently humid climates
such as Miami, correlation of load with CWRT/CHWST temperature difference
is poor. For these plants, optimum CWST vs. wet-bulb temperature was found
to have better correlation. But for office buildings in general, the correlations in
Figure 7 were more consistent.

June 2012

DPmax

Tmax

DP
Setpoint

DPmin

DP
Setpoint

CHWST
Setpoint

CHWST
Setpoint

Tmin
0 50% 100%
CHW Plant Reset

Figure 5: CHWST Setpoint and CW DP Setpoint Reset sequenced off of CHW valves.

Tower Fan Speed Control

A common approach to controlling cooling towers is to
reset condenser water supply temperature based on outdoor
air wet-bulb temperature. But our simulations seldom indicated a good fit; as shown in Figure 6, the correlation was
fairly good in Miami but not in Oakland and most other
climates.
For plants serving typical office buildings, good correlations were found in all TOPP simulations between plant part
load ratio (PLR, actual plant load divided by total plant design

www.info.hotims.com/41640-25

ASHRAE Journal

63

Oakland

CWST

90
85
y=0.9133x+12.997
R2=0.927
80
75
70
65
60
55
50
45
40
35
30
30 35 40 45 50 55 60 65 70 75 80 85 90

90
85
y=0.1223x+58.823
R2=0.0119
80
75
70
65
60
55
50
45
40
35
30
30 35 40 45 50 55 60 65 70 75 80 85 90

CWST

Miami

OAWB

OAWB

Figure 6: TOPP optimum condenser water supply temperature vs. wet-bulb temperature.
Miami
y=47.729x+11.656
R2=0.7848

CWRTCHWST

CWRTCHWST

80
75
70
65
60
55
50
45
40
35
30
25
20
15
10
5
0

Atlanta

0 10 20 30 40 50 60 70 80 90
100

80
75
70
65
60
55
50
45
40
35
30
25
20
15
10
5
0

y=58.332x+4.0753
R2=0.9417

0 10 20 30 40 50 60 70 80 90
100

Percent Total Design Chiller Capacity

Chicago

y=44.87x+4.2464
R2=0.9341

CWRTCHWST

CWRTCHWST

Oakland
80
75
70
65
60
55
50
45
40
35
30
25
20
15
10
5
0

Percent Total Design Chiller Capacity

0 10 20 30 40 50 60 70 80 90
100

80
75
70
65
60
55
50
45
40
35
30
25
20
15
10
5
0

Percent Total Design Chiller Capacity

y=57.434x+3.8301
R2=0.9475

0 10 20 30 40 50 60 70 80 90
100
Percent Total Design Chiller Capacity

Figure 7: TOPP [CWRT-CHWST] vs. plant load ratio.

capacity) and the difference between the condenser temperature return temperature (CWRT, leaving the condenser) and
the CHWST. Examples are shown in Figure 7. The CWRTCHWST difference is a direct indicator of the refrigerant lift
(the condenser and evaporator leaving water temperatures are
64

ASHRAE Journal

determined by the condenser and evaporator temperatures),

which drives chiller efficiency.
The data in Figure 7 can be fit to a straight line:

CWRT CHWST = A PLR + B

a s h r a e . o r g

(1)

June 2012

A and B are coefficients that vary with climate and plant design
(see Modeling the Plant, Page 57). Equation 1 can be solved
for the optimum CWRT setpoint given the current CHWST:

CWRT = CHWST + A PLR + B

(1a)

This setpoint must be bounded by the minimum CWRTCHWST difference at low load prescribed by the chiller
manufacturer. This minimum (9F [5C] for the chiller in
Figure 7) is a function primarily of the chillers oil management design and can range from only a few degrees for oilfree chillers (e.g., those with magnetic or ceramic bearings)
to as high as about 20F [11C]. The lower this minimum is,
the lower annual chiller plant energy will be, particularly in
mild climates.
So near-optimum tower performance can be achieved by
controlling tower fan speed based on condenser water return
temperature to the setpoint determined from Equation 1a.
Controlling tower fan speed based on return temperature rather than supply temperature is non-conventional but it makes
sense because it is the temperature leaving the condenser that
determines chiller lift, not the entering (supply) water temperature. Chiller efficiency is not sensitive to entering chilled
or condenser water temperature.

www.info.hotims.com/41640-65

66

Condenser Water Pump Control

No good correlations were found for control of VSDs on
condenser water pumps. Optimum condenser water pump
speed and flow were plotted against various parameters such
as PLR, wet-bulb temperature, chiller percent power, and lift
with no consistent relationships. The best correlation was
flow vs. PLR as shown in Figure 8, but the correlations were
seldom strong (R2 typically less than 0.85 and some as low
as 0.5). The correlations were significantly weaker for pump
speed than for flow so a condenser water flow meter should be
added if one is not already part of the design.
The curve fit can be expressed as follows

CWFR = C PLR + D

(2)

where CWFR is the ratio of desired CW flow setpoint to the

design CW flow. The CW flow setpoint is then calculated as:

CWFSP = CWFR CWDF

(2a)

where CWDF is the design CW flow rate for the plant

(both pumps). This setpoint must be bounded by the minimum required CW flow rate obtained from the chiller manufacturer. The minimum flow from most manufacturers cor-

www.info.hotims.com/41640-30

A S H R A E J o u r n a l

June 2012

Miami

Atlanta

90

80

80

70

70

60

60

50

50

40
30
20

y=0.8431x+0.3286
R2=0.8284

CW gpm (%)

100

90

CW gpm (%)

100

40
30

y=0.9833x+0.2667
R2=0.8356

20

10

10

100
10 0 10 20 30 40 50 60 70 80 90

100
10 0 10 20 30 40 50 60 70 80 90

20

20

30

30
Percent Total Plant Design Capacity

Percent Total Plant Design Capacity

Chicago

Oakland

90

80

80

70

70

60

60

50

50

40
30
20
10

y=1.335x+0.1784
R2=0.8488

CW gpm (%)

100

90

CW gpm (%)

100

40
30
20

y=1.0327x+0.2503
R2=0.8333

10

100
10 0 10 20 30 40 50 60 70 80 90

0
100
10 0 10 20 30 40 50 60 70 80 90

20

20

30

30
Percent Total Plant Design Capacity

Percent Total Plant Design Capacity

Figure 8: TOPP CW flow vs. plant load ratio.

68

ASHRAE Journal

3
Optimum Number of Chillers

relates to the onset of laminar flow and will be on the order

of 40% to 70% of design flow depending on the number of
tubes, number of passes, and tube design (e.g., smooth vs.
enhanced). Higher rates are reputed to discourage fouling
of condenser tubes but to the authors knowledge, no studies
have been done to support that notion.9 Once the flow rate
is determined, CW pump speed is modulated to maintain the
CW flow at setpoint.
When C and D coefficients determined for specific plants
were fed back into the energy model, actual performance
ranged from 101% to 110% of the TOPP. With this less than
optimum performance, VSDs were found to be marginally
life-cycle cost effective in dry climates (Albuquerque, N.M.)
and not cost effective elsewhere. This performance gets
worse when C and D are determined from the regression
equations based on plant design (see Modeling the Plant),
rather than from actual plant performance modeling (e.g.,
Figure 8). In some cases, particularly in humid climates, the
CW pump control logic caused energy use to increase vs.
constant speed CW pumps. Therefore, VSDs on CW pumps
are recommended only on plants in dry climates and only if

0
0 10 20 30 40 50 60 70 80 90
100
Percent Total Design Chiller Capacity

Figure 9: TOPP variable speed chiller staging vs. plant load

ratio (Albuquerque).
C and D coefficients are based on TOPP simulations of the
actual plant, not from the equations list in Modeling the
Plant.
a s h r a e . o r g

June 2012

Speed
100%
90%

Two Chillers

One Chiller

Surge Region
Refrigerant Lift

Optimum staging for variable speed CW pumps was found

to correlate very well to CW flow with 60% of the total design flow as the staging point, i.e., one pump should operate
when the CWFR is below 60% and two pumps should operate
when CWFR is above 60%, with a time delay to prevent short
cycling.
Optimum staging for constant speed pumps was found to
vary with both CWRT-CHWST difference and with PLR, but
with fairly weak correlations and relatively small energy impact regardless of logic. For simplicity, constant speed CW
pumps should simply be staged with the chillers.

80%

Two Chillers

70%

One Chiller

60%

Chiller Staging
Figure 9 (Page 68) shows the optimum number of chillers
that should be run plotted against plant load for variable speed
centrifugal chillers. The graph shows that it is often optimum
to operate two chillers as low as 25% of overall plant load.
This result may seem somewhat counterintuitive; conventional wisdom is to run as few chillers as possible. That is true for
fixed speed chillers, but not for variable speed chillers, which
are more efficient at low loads when condenser water temperatures are low.
Figure 9 shows that staging chillers based on load alone
will not optimize performance since there is a fairly wide
range where either one or two chillers should be operated.

70

Load

Figure 10: Possible surge problem staging by load only.

There is also another problem with staging based on load
alone: it can cause the chillers to operate in surge. This can
be seen in Figure 10, which schematically shows centrifugal chiller load vs. lift, defined as the difference between
condenser and evaporator refrigerant temperature. If two
chillers are operated when the refrigerant lift is high (red
line), the chillers will operate in the surge region. To avoid

www.info.hotims.com/41640-3

A S H R A E J o u r n a l

D63898asjl_1206885.indd 1

June 2012

5/10/12 8:28:54

SPLR = E (CWRT CHWST) + F

(3)

Oakland
80
70
60
CWRTCHWST

surge, the chiller controllers will speed up the compressors and throttle inlet guide vanes to control capacity. This
reduces chiller efficiency so that it would then be more efficient to operate one chiller rather than two. But if the lift
is low (green line in Figure 10), the chillers would not be
in surge so operating two chillers would be more efficient
than operating one. So in addition to load, chiller staging
must take chiller lift into account. (This consideration applies only to centrifugal chillers; surge does not occur with
positive displacement chillers such as those with screw
compressors.)
Figure 11 shows the optimum number of operating chillers (blue dots indicate one chiller while red dots indicate two
chillers) for example TOPP simulations. For all plant design
options and for all climate zones, good correlations were
found for the optimum staging point described by a straight
line:

50
40
30
20
10
0

0 10 20 30 40 50 60 70 80 90
100
Percent Total Design Chiller Capacity

Albuquerque
80
70
60
CWRTCHWST

where SPLR is the staging PLR and E and F are coefficients

that vary with climate and plant design (see Modeling the
Plant). If the actual measured PLR is less than SPLR, one
chiller should operate; if the PLR is larger than SPLR then
two chillers should operate, with a time delay to prevent short
cycling.
Note that the number of operating chilled water pumps and
the number of operating chillers may not match. The pumps
must respond to the flow and pressure requirements of the system, not to the load, and thus are staged independently from
chillers.
Primary-only variable flow plants like this also will require
soft staging and minimum flow control. These sequences
and why they are needed are discussed in more detail in the
SDL and in Reference 10.

June 2012

40
30
20
10
0
0 10 20 30 40 50 60 70 80 90
100
Percent Total Design Chiller Capacity

Chicago

Example

80
70
60
CWRTCHWST

The TOPP model results for an Oakland plant were plotted

per Figures 7, 8, and 11 and the following slopes and intercepts were determined from curve-fits:

A = 47, B = 5.2

C = 1.3, D = 0.13

E = 0.009, F = 0.21
Figure 12 shows the theoretical optimum performance for
both variable speed (VS) constant speed (CS) CW pumps
compared to our proposed real sequences using the coefficients listed above. Despite their simplicity, our sequences
resulted in only about 1% higher energy use than the TOPP.
Variable speed drives on the CW pumps saved 3% vs. constant
speed pumps, but this was not enough savings to make them
cost effective at a 15 scalar ratio (simple payback period) for
this plant. Also shown in the figure for comparison is plant
performance using the AHRI 550/590 condenser water relief
curve to reset condenser water temperature (4% higher energy
use than our sequences) and performance assuming CWST

50

50
40
30
20
10
0 0 10 20 30 40 50 60 70 80 90
100
Percent Total Design Chiller Capacity
One Chiller

Two Chillers

Figure 11: Optimum staging vs. (CWRT-CHWST) and plant

part load ratio.
ASHRAE Journal

71

setpoint is fixed at the design temperature (16% higher than

our sequences).

+20%

This article is the last in a series of five that summarize

chilled water plant design techniques intended to help engineers optimize plant design and control with little or no added
engineering effort. In this article, optimized control logic was
addressed. The logic is very simple and easily programmed
into any DDC system controlling the plant. With these sequences properly implemented, chiller plants can perform
within a few percent of their theoretical optimum.

CHW Plant Energy Use (kWh/yr)

180,000

Summary

+3%

160,000

+1%

+4%

Real vs

Real CS

+8%

140,000
120,000
100,000
80,000
60,000
40,000
20,000

References
1. Taylor, S. 2011. Optimizing design and control of chilled water plants
part 2: condenser water system design. ASHRAE Journal 53(9):2636.
2. Taylor, S. 2012. Optimizing design and control of chilled water
plants part 4: chiller and cooling tower selection. ASHRAE Journal
54(3):6070.
3. Hartman, T. 2005. Designing efficient systems with the equal
marginal performance principle. ASHRAE Journal 47(7):6470.
4. Hydeman, M., G. Zhou. 2007. Optimizing chilled water plant
control. ASHRAE Journal 49(6):4554.

www.info.hotims.com/41640-28

72

200,000

0
TOPP vs TOPP CS

CWP kWh

CHWP kWh

Tower kWh

AHRI
Constant
Reset CS CWST CS

Chiller kWh

Figure 12: TOPP vs. real sequences for both constant speed
and variable speed CW pumps.

www.info.hotims.com/41640-4

A S H R A E J o u r n a l

June 2012

Central Chilled Water Plants Series

This series of articles summarizes the upcoming Self Directed
Learning (SDL) course Fundamentals of Design and Control of
Central Chilled Water Plants and the research that was performed
to support its development. The series includes five segments.
Part 1: Chilled Water Distribution System Selection (July 2011),
Part 2: Condenser Water System Design (September 2011),
Part 3: Pipe Sizing and Optimizing DT (December 2012), and
Part 4: Chiller & Cooling Tower Selection (March 2012).
Optimized control sequences. The series concluded
with a discussion of how to optimally control chilled water
plants, focusing on all-variable speed plants.

5. ANSI/ASHRAE Standard 90.1-2010, Energy Standard for Buildings Except Low-Rise Residential Buildings.
6. Rishel, J.B. 2001. Wire-to-water efficiency of pumping systems.
ASHRAE Journal 43(4):4046.
7. Taylor, S. 2002. Degrading chilled water plant T: causes and
mitigation. ASHRAE Transactions 108(1):641653.
8. Taylor, S. 2007. Increasing efficiency with VAV system static

The intent of the SDL (and these articles) is to provide

simple yet accurate advice to help designers and operators
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 significant amount of simulation, cost estimating, and life-cycle cost analysis was
performed on the most common water-cooled plant configurations to determine how best to design and control
them. The result is a set of improved design parameters
and techniques that will provide much higher performing chilled water plants than common rules-of-thumb and
standard practice.

pressure setpoint reset. ASHRAE Journal 49(6): 2432.

9. Li, W., R. Webb. 2001. Fouling characteristics of internal helical-rib roughness tubes using low-velocity cooling
tower water. International Journal of Heat and Mass Transfer
45(8):16851691.
10. Taylor, S. 2002. Primary-only vs. primary-secondary variable
flow systems. ASHRAE Journal 44(2):2529.

www.info.hotims.com/41640-68

74

A S H R A E J o u r n a l

June 2012