8.
13 Chiller Optimization
B. G. LIPTÁK (1970, 1985, 1995, 2005)
INTRODUCTION
There are not many unit operations whose efficiency can be
doubled through optimization. The chillers are one of these few.
One unit of energy introduced at the chiller can remove 5, 6,
or even 7 units of heat energy from the process, if the chiller
controls are optimized. The previous section discussed the con-
ventional chiller controls, while this section is devoted to chiller
optimization.
Listed below are some of the performance character-
istics of optimized chillers as contrasted to conventional
ones:
• Coefficients of performance: 2.5–3.5 for conventional,
5–7 for optimized chillers.
• Operating cost reduction: 1.5% of yearly chiller operating
cost per °F reduction in chiller ∆T, 1% of total cooling
system operating cost per °F decrease in the cooling
tower’s approach (see Figure 8.13b for the definition of
‘‘approach’’). A 10°F reduction in cooling tower water
temperature cuts chiller operating costs by 30%.
• Pumping cost savings: 50–60 cents/gpm for each
psid.
• Operating cost distribution: 60% chiller compressor,
15% ea. cooling tower and chilled water pumps, 10%
cooling tower fans.
• Maximum turndown ratios: 8:1 for conventional, over
30:1 with hot gas bypass.
• Typical superheat at evaporator outlet: 9°F (5°C).
In order to fully optimize a cooling system, in addition
to the chiller, it is necessary to also optimize cooling tower
controls, the water distribution, and the pumping systems, as
will be discussed in the sections that follow.
© 2006 by Béla Lipták
Cooling tower
water
Condenser
Compressor
Evaporator
Chilled water
Flow sheet symbol
THE TOTAL COOLING SYSTEM
The piping layout of a conventional cooling system is illus-
trated in Figure 8.13a. In order to develop a completely
generalized method for controlling and optimizing such sys-
tems, the duplication of equipment will be disregarded, and
all chiller systems will be treated as if they were configured
as shown in Figure 8.13b. Here any number of cooling towers,
pumps, or chillers are represented by single units, because
variations in their numbers will not affect the overall optimi-
zation strategy.
In this generalized cooling system, the first heat transfer
step is when the cooling load from the process is transferred
to the chilled water. In the second step, the heat carried by
the chilled water is transferred to the refrigerant in the evap-
orator. The refrigerant takes the heat to the condenser, where
in the third step the heat is transferred to the cooling tower
water, so that it might finally be rejected to the ambient air.
This heat pump operation involves four heat transfer sub-
stances (chilled waster, freon or other refrigerant, cooling
tower water, and air) and four heat exchanger devices (pro-
cess heat exchanger, evaporator, condenser, cooling tower).
The total system operating cost is the sum of the costs of
circulating the four heat transfer substances (M1, M2, M3,
and M4 in Figure 8.13b).
In the traditional (unoptimized) control systems, such as
the one illustrated in Figure 8.12f, each of these four systems
were operated independently in an uncoordinated manner.
In addition, the conventional control systems did not vary
the speed of the four transportation devices (M1 to M4). By
operating them at constant speeds, they introduced more
energy than was needed for the circulation of refrigerant,
air, or water and therefore they had to waste that excess
energy.
1729
1730 Control and Optimization of Unit Operations

M1
SP = Twb + Ao Approach (A) = Tctws − Twb
Range (R) = Tctwr − Tctws
TIC
1
C Cooling
Air at Twb tower SP = Tctws + Ro
Cooling towers

CT bypass
T TIC
M2
Condenser 2
T 5000 GPM CTWP
Tctws Tctwr
Evaporator T FMT
200 HP Tc Freon

CWR
pump Condenser
200 HP Chiller 1 3400
Let-down
pump 2000 tons T GPM
valve M3
3750 GPM
3400 GPM Compressor
T FMT SP = max. TIC
150 HP 3
pump Tchwr Tchws
CWS

150 HP Chiller 2 Te
pump 1500 tons T 3750 GPM Evaporator
2570 GPM
FMT Process cooling
T TIC load
150 HP M4 SP = max.
4 (chilled water
pump CHWP users)
150 HP Chiller 3
pump 2 stage Chilled water return
2570 GPM 1500 tons T 3750 GPM
Abbreviations:
T FMT
150 HP Ao : Optimum approach
pump Ro : Optimum range
150 HP Chiller 4 Twb : Wet bulb temp.
pump 2 stage Tctws : Cooling tower water supply temp.
Tctwr : Cooling tower water return temp.
CHWS

2600 GPM 1500 tons

CHWR

Tchws : Chilled water supply temp.

Max 5220 GPM Tchwr : Chilled water return temp.
T min 1060 T CTWP: Cooling tower water pump
Building CHWP: Chilled water pump
load A Tc : Freon temperature in condenser
Max 4800 GPM Te : Freon temperature in evaporator
T min 0
T 1

Building FIG. 8.13b

load B In order to optimize a cooling system, the cooling tower(s), the
Max 380 GPM pumping stations, the chiller(s), and the process equipment should
min 0 be treated as an integrated single system.
2
T Building T
load C
return temperatures of chilled and cooling tower waters, and
FMT – Flow meter
the manipulated variables are the flow rates of chilled water,
End of line
T – Thermostat bypass valve
refrigerant, cooling tower water, and air.
If water temperatures are allowed to float in response to
FIG. 8.13a load and ambient temperature variations, the waste associated
A typical overall cooling system utilizes four heat transfer sub- with keeping them at arbitrarily selected fixed values is elim-
stances (CWS, CWR, CHWS, CHWR) and four heat exchange
inated, and the operating cost of the cooling system is dras-
devices (cooling towers, chiller condenser, chiller evaporator, pro-
cess heat exchangers).
1 tically reduced.
In order to control the total refrigeration system depicted
in Figure 8.13b, four control loops must be configured. In
OPTIMIZING MECHANICAL REFRIGERATION SYSTEMS these four loops the controlled (measured) variables are the
four water temperatures, and the four manipulated variables
Load-following optimization eliminates the waste resulting are the four motor speeds (M1 to M4) that drive the four
from constant speed operation by operating the aforemen- transportation devices. Table 8.13c lists the controlled and
tioned four systems as a coordinated single process, with the the manipulated variables in each of these loops and also
goal to keep the cost of operation at a minimum. In this indicates the optimization criteria for determining the set
control system, the controlled variables are the supply and points of the four controllers.

© 2006 by Béla Lipták

 8.13 Chiller Optimization 1731

both the chilled water supply and return temperatures (Tchws

TABLE 8.13c and Tchwr) are as high as the process will permit. This is true
Control Configuration and Optimization Criteria for the Four
under all load conditions, because the amount of work that
Loops Controlling a Chiller
the chiller compressor has to do is reduced as the suction
Optimization Criteria (Set pressure rises, and this is the case whenever the evaporator
Loop Controlled Manipulated Points of the Temperature temperature (Tchws) rises.
No. Variable Variable Controllers)
Therefore, the optimum set point of TIC-3 is the allowable
1 CT water Air flowrate Optimum approach (Ao) is maximum chilled water supply temperature. Consequently,
supply obtained by selected to keep the sum the minimum cost of operation will be achieved when the
temperature level of fan M1 + M2 + M3 to a chilled water supply temperature has been maximized, but it
(Tctws) operation minimum. Therefore, TIC can be maximized only to the point where it can still provide
(M1) set point becomes:
all the cooling that is required by any part of the process. The
Tctws = Twb + Ao
method of finding that maximum value will be discussed later.
2 CT water Rate of CT The optimum range (Ro) of The chilled water pumping cost (M4) depends only on the
return water CT is found as a function
∆T across the process users. The higher this ∆T, the less water
temperature pumping of the A0. Therefore, the
needs to be pumped to remove the required amount of heat.
(Tctwr) (M2) TIC set point becomes:
Tctwr = Tctws + Ro
As the chilled water supply temperature is set by the load,
this ∆T can be maximized only by maximizing the chilled
3 Chilled water Rate of chiller The optimum TIC set point
water return temperature (Tchwr). Therefore, the optimum set
supply compressor is the maximum Tchws
temperature operation temperature, which will
point for TIC-4 is the allowable maximum. By increasing this
(Tchws) (M3) satisfy all the process return temperature, the following benefits result:
loads
4 Chilled water Rate of The optimum TIC set point 1. The power consumed by M3 is lowered by 1.5% per
return chilled water is the maximum Tchwr °F rise.
temperature pumping temperature that will 2. The heat transfer efficiency of the evaporator is
(Tchwr) (M4) satisfy all the process improved by increasing the ∆T through it.
loads 3. Assuming that on the average the ∆T is 15°F, increasing
it by 1°F will lower M4 by about 6%.

Minimizing the Operating Cost
The yearly cost of operating the total cooling system can
typically be broken down as follows:
M1 (Fans) 10%
M2 (CT Pumps) 15%
M3 (Compressors) 60%
M4 (CH Pumps) 15%
TOTAL: 100%
These costs vary a great deal. The proportion of M1
increases in warm weather regions. M2 and M4 increase when
water transport lines are long and the proportion of M3 is
lowered as the maximum allowable chilled water temperature
rises. But regardless of these proportions in a particular instal-
lation, the goal of optimization is to find the minimum chilled
water and cooling tower water temperatures that will result
in meeting the needs of the process at minimum cost.
The overall control system can be reviewed in two steps:
First one can look at the chilled water side (the evaporator
side), then at the cooling tower water side (condenser side).
The Chilled Water Side The optimization of the lower
(evaporator) portion of Figure 8.13b is easily comprehended,
because the sum of M3 and M4 will be the minimum when
As M3 is about 60% and M4 about 15% of the total
operating cost, a 1°F decrease in ∆T will lower the yearly
cost of operation by about (1 × 60)/100 + (6 × 15)/100 = 1.5%.
The Cooling Tower Water Side Finding the optimum cool-
ing tower water temperatures is more complicated, because
changing the cooling tower water supply and return temperatures
will increase some costs, while it will also lower some others.
For example (as it is shown in Figure 8.17d), an increase in
the cooling tower water supply temperature (Tctws ) will increase
the tower’s approach and the cost of chiller operation but will
reduce the amount of work the cooling tower fans need to do.
Therefore, the optimum temperature in not necessarily
the minimum temperature (or approach) that the towers are
capable of providing, but rather the temperature that can meet
the particular load at the minimum total cost of operation of
all the equipment (M1 + M2 + M3). Figure 8.13d illustrates
that this is a function of the load and of the weather. When
the conditions are such that a lowering of the approach (or
Tctws) would require a large increase in fan operating costs
(M1″), the minimum point on the total cost curve ( A0′′ ) will
be above the minimum attainable approach ( A0′ ).
Inversely, if the approach can be reduced with a small
increase in fan operating costs (M1′), the optimum operating
point for the total system will correspond to the minimum
Tctws temperature and therefore to the minimum attainable
approach ( A0′ ). Whichever is the applicable total cost curve

© 2006 by Béla Lipták

1732 Control and Optimization of Unit Operations

$ Optimum approach (Ao)

Optimum Tctws = Twb + A0 from Fig. 8.13d

Total $″
10 Approach
Total $′ (5.6) °F(°C)

14
(7.8)
Operating cost ($)

M3 A0 = Optimum approach 18
A0′ = Optimum at minimum (10.0)
A0″ = Optimum above minimum Ro
M1' = Fan cost increase small 22
(12.2)
M1" = Fan cost increase large
M2 = CT pumping cost 26
M3 = Chiller compressor cost (14.4) Ro = Optimum range, which is part of
M1"
$′ = Total cost w/expensive CT the set point for TIC-2 in Fig. 8.13b
$″ = Total cost when CT inexpensive Range °F (°C)
M1'
M2
FIG. 8.13e
The optimum range for a particular cooling tower is a function of
A the approach at which it is operating. Therefore, R0 can be found
A0′ A0″ Approach (Tctws − Twb)
on the basis of A0 .
FIG. 8.13d
The empirically determined operating costs are shown as a function
of the approach at two cooling loads. The optimum approach at of lowering approach will also increase (M1″). In that case,
each load is the one that corresponds to the minimum total cost of the optimum approach ( A0′ ) is at some intermediate value.
operation. The optimum return water temperature back to the cool-
ing tower (Tctwr) can be found by determining the optimum
range (Ro) for the particular set of operating conditions. The
($′ or $″), the Tctws temperature that corresponds to the min- relationship between operating approach and range is a func-
imum point becomes the set point for TIC-1 in Figure 8.13b. tion of the particular cooling tower design. Therefore, if one
The curves in Figure 8.13d can be based on the measure- has determined the optimum approach for a particular set of
ment of the actual total operating cost, on a projected cost conditions Ao (from Figure 8.13d), that value can be used to
based on past performance, or on some combination of the arrive at the optimum range (Ro) in Figure 8.13e.
two. The continuous storing and updating of the operating The set point for TIC-2 in Figure 8.13b then becomes the
cost history as a function of load, ambient conditions, and sum of the already-determined Tctws and the optimum range:
equipment configuration will not only provide the curves
Tctwr = Tctws + Ro 8.13(1)
required for optimization but can also be used to signal the
need for maintenance.
Chilled Water Supply Temperature Optimization
In Figure 8.13d the optimum Tctws is found by summing
M1, M2, and M3. In most installations the major cost element The yearly operating cost of a chiller is reduced by about 1.5%
is M3, which increases by about 1.5% of compressor oper- for each 1°F (0.6°C) reduction in the temperature difference
ating cost (or 1% of total cooling system cost) for each 1°F across this heat pump. In order to minimize this difference,
reduction in approach. TH must be minimized and TL maximized. Therefore, the opti-
The cost-vs.-approach curve is not necessarily a smooth mum value of Tchws is the maximum temperature that will still
one, because if constant-speed machines are stopped and satisfy all the loads. Figure 8.13f illustrates a method of con-
started or if positive-displacement compressors are loaded and tinuously finding and maintaining this maximum value.
unloaded, there will be steps in the M3 curve. This also is true It should be noted that an energy-efficient refrigeration
for constant speed fans (M1) or pumps (M2). The pumping system cannot be provided by instrumentation alone. Equip-
cost also increases as approach rises, but M2 is usually only a ment sizing and selection also play an important part in the
fraction of M3, unless the piping is very long or undersized. overall result. For example, the evaporator heat transfer area
The fan cost M1 tends to drop off as approach rises. If should be maximized so that TL is as close as possible to the
the load is low relative to the size of the cooling towers, the average chilled water temperature in the evaporator.
rate of increase in M1 with a reduction in approach will also Similarly, the refrigerant flow should be made to match
be small. In such cases (M1′) the optimum approach ( A0′ ) the load by adjusting the compressor motor speed and not by
corresponds to the safe lowest temperature that the tower can refrigerant throttling. TCV-1 and TCV-2 in Figure 8.12f in the
generate. If, on the other hand, the load is high, the fan cost previous section represent sources of energy waste, whereas

© 2006 by Béla Lipták

 8.13 Chiller Optimization 1733

Compressor This technique allows all the process users to obtain more
cooling (by further opening their chilled water supply valves)
M3
if needed, while the header temperature is continuously max-
imized. The VPC-1 set point of 90% is adjustable. Lowering
SIC it gives a wider safety margin, which might be required if
1
some of the processes that are being cooled are very critical.
On the other hand, increasing the set point will maximize
TIC SP HLL energy conservation at the expense of a reduced safety margin.
1 1 Integral
only An additional benefit of this load-following optimization
Feedback (ER) VPC Setpoint strategy is that because all chilled water valves in the plant
1 = 90% are opened as Tchws is maximized, valve cycling is reduced
RA
Figure 8.13g and pumping costs are lowered. This is so because when all
TT
1 chilled water valves are opened they require less pressure
TL = Tchws Brine or chilled water drop and therefore less discharge pressure from the pumps.
to other users Similarly, valve cycling is eliminated, because by increasing
Evaporator TY the opening of all user valves, they are moved away from the
1 2
unstable region near their closed position.
From all In order for the control system in Figure 8.13f to be stable,
FC other user it is necessary to use an integral-only controller for VPC-1,
TCV valves
1 with an integral time that is tenfold that of the integral setting
of TIC-1 (usually several minutes). This control mode selection
Typical TIC is needed to allow the optimization loop to be stable when the
process valve opening signal selected by TY-1 is either cycling or noisy.
cooler The high/low limits (HLL-1) on the set-point signal to
Abbreviations: TIC-1 guarantee that VPC-1 will not drive the chilled water
HLL : High and low limit temperature to unsafe or undesirable levels. Because these
SIC : Speed indicating controller limits can block the VPC-1 output from affecting Tchws, it is
SP : Set point necessary to protect VPC-1 against reset windup. This is done
VPC : Valve position controller through the external feedback signal shown in Figure 8.13f.
FIG. 8.13f
Maximizing the chilled water supply temperature by load-following Chilled Water Return Temperature Optimization
floating control is a means of continuously finding and maintaining
this temperature at its optimum value. The combined cost of operating the chilled water pumps and
the chiller compressor (M4 + M3) is a function of the tem-
perature drop across the evaporator (Tchwr − Tchws). Because
Figure 8.13f shows the energy-efficient technique of motor
an increase in this ∆T decreases compressor operating costs
speed control by eliminating TCV-2. This load-following mode
(suction pressure rises) while also decreasing pumping costs
of operation can be achieved by using variable-speed drives on
(the higher the ∆T the less water needs to be pumped), the
electric motors or by using steam turbine drives.
aim of this optimization strategy is to maximize this ∆T.
If several constant-speed motors are used, then all compres-
sors except one should be driven to the loading level at which
they are most efficient (TCV-1 and TCV-2 in Figure 8.12f fully Maximizing Tchwr This ∆T will be the maximum when the
open). The remaining one compressor should be used to match chilled water flow rate across the chilled water users is the
the required load by throttling. minimum. As Tchws is already controlled, this ∆T will be max-
imized when Tchwr is maximum. This goal can be reached by
Maximizing Tchws Figure 8.13f shows the proper technique evaluating the opening of the most-open chilled water valve in
of maximizing the chilled water supply temperature in a load- Figure 8.13f and, if even the most-open chilled water valve is
following, floating manner. The optimization control loop not yet fully open, making adjustments to further open it.
guarantees that all chilled water users in the plant will always The steps that can be taken include increasing the chilled
be satisfied while the chilled water temperature is maximized. water supply temperature (set point of TIC-1 in Figure 8.13f),
This is done by selecting (by TY-1) the most open chilled or increasing the temperature rise across the process users
water valve in the plant and comparing that opening with the by lowering the ∆P across the users (set point of PDIC-1 in
90% set point of the valve position controller, VPC-1. If even Figure 8.13g). Both methods can also be used simultaneously.
the most open valve is less than 90% open, the set point of Increasing the chilled water supply temperature reduces
TIC-1 is increased; if the valve opening exceeds 90%, the the yearly compressor operating cost (M3) by approximately
TIC-1 set point is decreased. 1.5% for each °F of temperature increase, whereas lowering

© 2006 by Béla Lipták

1734 Control and Optimization of Unit Operations
Compressor
M3 Most-open
chilled water
user valve signal
from Figure 8.13f
Tchws
Evaporator
Integral
only
PDIC SP HLL VPC
FSHL P1 1 1 2 SP = 95%
3 DA
PSH Feedback(ER)
1
SC
Start 1 in Figure 8.13f.
Stop
M4
FSL CHWP Tchwr
2
FIG. 8.13g
Adjusting the chilled water flow rate such that the most open user
valve is nearly full will keep the evaporator ∆T and chilled water
return temperature always at their maximum (optimum) values.
the ∆P across the users reduces the yearly pump operating
cost (M4) by approximately 50 cents/gpm for each psid.
The set points of the two valve position controllers (VPC-1
in Figures 8.13f and VPC-2 in 8.13g) will determine if these
adjustments are to occur in sequence or simultaneously. If
both set points are adjusted to be the same, simultaneous
action will result, while if one adjustment is economically
more advantageous or safer than the other, the set point of
the corresponding VPC should be set lower than the other.
This will result in sequencing, which means that the more
cost-effective or safer correction will take place before the
less effective one is started. In Figures 8.13f and 8.13g, it
was assumed that increasing Tchws is the more cost-effective
step. This is not always the case, and even when it is, there
might be process reasons that make it undesirable to allow
the floating of Tchws up or down.
If the settings are as shown, the system will function as
follows: If the most open valve is less than 90% open, VPC-2
in Figure 8.13g will lower the set point of PDIC-1, and VPC-
1 in Figure 8.13f will increase the set point of TIC-1. When the
opening of the most-open valve reaches 90%, VPC-1 in Figure
8.13f will slowly start lowering the TIC set point, while VPC-
2 in Figure 8.13g will continue to lower the set point of PDIC-1.
If VPC-1 in Figure 8.13f did not lower the TIC set point
sufficiently (or if the lowering was not fast enough) and the
most-open user valve continues to open, and at 95% the VPC-2
in Figure 8.13g will take fast corrective action by quickly rais-
© 2006 by Béla Lipták
ing the set point of PDIC-1. Thus, no user valve will ever be
allowed to open fully and go out of control as long as the pumps
and chillers are sized so as to be capable of meeting the load.
Pump Speed Control VPC-2 in Figure 8.13g is the cascade
master of PDIC-1, which guarantees that the pressure differ-
ence between the chilled water supply and return is always
high enough to motivate water flow through the users but never
so high as to exceed their pressure ratings. The high and low
limits are set on HLL-1, and VPC-2 is free to float this set point
within these limits to keep the operating cost at a minimum.
In order to protect against reset windup when the output
of VPC-2 reaches one of these limits, an external feedback
is provided from the PDIC-1 output signal back to VPC-2.
The VPC is an integral-only controller, which is tuned to be
Users much more responsive (shorter integral time) than is VPC-1
The high speed of response of the VPC in Figure 8.13g is
also important from a safety point of view. This is because
without it, in a conventional control system, the response can be
too slow. In such a case, when the demand for cooling suddenly
increases (because of some process difficulty), once the valve
(TCV-1 in Figure 8.13f) is fully open, the amount of cooling
provided cannot be increased faster than the rate at which the
chiller can lower the chilled water supply temperature. This does
take time, and therefore, if the cooling load evolves faster (run-
away reaction in a reactor), accidents can occur.
This is not the case when the control system shown in
Figure 8.13g is implemented, because as soon as the most-
open chilled water valve has opened to 95%, VPC-2 will
quickly speed up the chilled water pump (CHWP). Because
water is incompressible, this will immediately increase the
chilled water supply pressure in the main supply header,
therefore, the water supply pressure to TCV-1 in Figure 8.13f
will also rise, thereby immediately increasing the water flow
and therefore the amount of cooling of the process.
This practically instantaneous action, utilizing the
already-chilled water stored in the distribution headers, can
be fast enough to arrest a runaway reaction. Naturally, once
the chiller has had time to respond to the increase in load by
lowering the chilled water supply temperature, this emer-
gency action will no longer be needed and the water supply
pressure will be lowered back to its normal setting.
Multiple and Bypass Pumps When the chilled water pump
station consists of several pumps, only one of which is vari-
able-speed, additional pump increments are started when PSH-
1 in Figure 8.13g signals that the pump speed controller set
point is at its maximum. When the load is dropping, the excess
pump increments are stopped on the basis of flow, detected by
FSL-2. In order to eliminate cycling, the excess pump incre-
ment is turned off only when the actual total flow has dropped
to less than 90% of the capacity of the remaining pumps.
The load-following optimization loop in Figure 8.13f and
8.13g will float the total chilled water flow to achieve max-
imum overall economy. In order to maintain efficient heat
 8.13 Chiller Optimization 1735

transfer and the required turbulence within the evaporator, a
small local circulating pump (P1) is provided at the evapo-
rator. This pump is started and stopped by FSHL-3 in
Figure 8.13g, guaranteeing that the water velocity in the
evaporator tubes will never drop below the adjustable limit
of about 4 fps (1.2 m/s).
Cooling Tower Supply Temperature Optimization
As discussed in more detail in Section 8.17, minimizing the
temperature of the cooling tower water is one of the most
effective contributors to chiller optimization. Conventional
control systems in the past produced constant cooling tower
temperatures of 75°F (23.9°C) or higher.
It is an enemy of efficiency and therefore of optimization
to maintain a utility at a constant value. Each 10°F (5.6°C)
reduction in the cooling tower water temperature will reduce
the yearly operating cost of the compressor by approximately
15%. For example, if a compressor is operating at 50°F
(10°C) condenser water instead of 85°F (29.4°C), it will meet
the same load while consuming half as much power. Opera-
tion at condenser water temperatures of 50°F (10°C) or even
less can be quite practical during the winter months. Savings
3
exceeding 50% have been reported.
Minimizing Tctws As shown in Figure 8.13h, an optimiza-
tion control loop is required in order to maintain the cooling
tower water supply continuously at an economical minimum
temperature. This minimum temperature is a function of the
Total operating
cost ($/MBTU)
wet-bulb temperature of the atmospheric air. The cooling
tower cannot generate a water temperature that is as low as
the ambient wet bulb, but can approach it. The temperature
difference between Tctws and Twb is called the approach, as
was shown in Figure 8.13d.
Figure 8.13h illustrates the fact that as the approach and
the cooling tower supply temperature increases, the cost of
cooling tower fan operation drops, while the costs of pumping
and of compressor operation increase. Therefore, the total
operating cost curve has a minimum point that identifies the
optimum approach, which corresponds to the minimum cool-
ing tower temperature that will allow operation at an overall
minimum cost. This approach (∆T) automatically becomes
the set point of TDIC-1, and this optimum approach will
increase if the load on the cooling tower rises or if the ambient
wet bulb increases.
If the cooling tower fans are centrifugal units or if the
blade pitch is variable, the optimum approach is maintained
by continuous throttling. If the tower fans are two-speed or
single-speed units, the output of TDIC-1 will start and stop
the fan units incrementally in order to maintain the optimum
approach. In cases in which a large number of cooling tower
cells constitute the total system, it is also desirable to balance
the water flows to the various cells automatically as a function
of the operation of the associated fans.
The water flows to all cells whose fans are at high speed
should be controlled at equal and relatively high rates, while
the cells with fans operating at low speeds should receive
water at equal low flow rates. Cells with their fans off should
be supplied with water at equal minimum flow rates. Subjects
such as these and the cooling towers in general are discussed
in more detail in Sections 8.16 and 8.17.

Total cost Cooling Tower Return Temperature Optimization

(M1, M2 and M3)

Compressor
The optimum setting for Tctwr is based on the optimum
and approach (Ao) obtained from Figure 8.13h, using the charac-
pump costs teristic curve of the cooling tower depicted in Figure 8.13e.
(M2 and M3) This curve relates the approach to the range for the particular
tower design, and therefore once Ao is known, Ro can be
SC obtained from it.
Fan cost
1 M1
(M1) As shown in Figure 8.13i, the ∆T corresponding to this
Air
at Air
optimum Ro automatically becomes the set point of TDIC-1
Ao Twb at in the optimized control loop. This controller is the cascade
Approach (∆T) Twb master of PDIC-1, which guarantees that the pressure differ-
ence between the cooling tower supply and return water flows
Optimized is always high enough to provide the required flow through
set point the process users but never so high as to cause damage. The
to TDIC−1 SP TDIC Tctws
1 high and low limits are set on HLL-1. TDIC-1 floats the
PDIC-1 set point within these ∆P limits, to keep the operating
Approach (∆T ) = Tctws − Twb Twb cost at a minimum.
FIG. 8.13h
In order to protect against reset windup (when the output
Optimizing (minimizing) the cooling tower water supply tempera- of TDIC-1 reaches one of these limits), an external feedback
ture requires that the fan speed be so modulated as to keep the is provided from the output of PDIC-1 to TDIC-1.
approach (Tctws − Twb) at a value that corresponds to the minimum When the cooling tower water pump station consists of
cost of operation. several pumps, only one of which is variable-speed, additional

© 2006 by Béla Lipták

1736 Control and Optimization of Unit Operations

Total operating
cost ($/MBTU)

Total cost
(M1, M2 and M3)
Compressor
This is the range (Ro)
and pump costs
which corresponds to
(M2 and M3)
Optimized the optimum approach (Ao)
From set point To
cooling to TDIC-1 cooling
tower Ro = (Tctws −Tctwr) SP = ∆T
tower
Fan cost
Tctws (M1)
TDIC Tctwr
1
Ao
Feedback
(ER) Approach (∆T)
HLL
1

SP = ∆P
10 Approach
PSH PDIC
1 (5.6) °F (°C)
1
14
SP (7.8)
SC
Start 1 18
(10.0) Ro
Stop FSHL 22
Condenser 3 (12.2)
M2
26
FSL (14.4)
2
CTWP Range °F (°C)

P1

FIG. 8.13i
The cooling tower water flow rate is modulated to keep the ∆T of the chiller’s condenser (R0 , the range of the cooling tower) at the value
that corresponds to the optimum value of the approach.

pump increments are started when PSH-1 signals that the
pump speed controller set point is at its maximum. When the
load is dropping, the excess pump increments are stopped on
the basis of flow, which is detected by FSL-2. In order to
eliminate cycling, the excess pump increment is only turned
off when the actual total flow is less than 90% of the capacity
of the remaining pump(s).
The load-following optimization loop shown in Figure
8.13i will float the total cooling tower water flow to achieve
maximum overall economy. In order to maintain efficient heat
transfer and appropriate turbulence within the condenser, a
small local circulating pump (P1) is provided at the chiller’s
condenser. This pump is started and stopped by FSHL-3, guar-
anteeing that the water velocity in the condenser tubes will
never drop below the adjustable limit of 4 fps (1.2 m/s).
Heat Recovery Optimization
Figure 8.13j depicts the optimizing control loop required
when the heat pumped by a chiller is recovered in the form
of hot water and the hot water temperature is continuously
optimized in a load-following floating manner.
If, at a particular load level, it is sufficient to operate with
100°F (37.8°C) instead of 120°F (48.9°C) temperature hot
water, this technique will allow the chiller to meet the same
tonnage of refrigeration load at 30% lower operating cost.
The reason for this operating cost reduction is that the
required discharge pressure of the chiller compressor depends
on the hot water temperature in the split condenser.
The optimization control loop in Figure 8.13j guarantees
that all hot water users in the plant will always obtain enough
heat while the hot water temperature set point of TIC-1 is
minimized. In this system TY-1 selects the most open hot
water valve in the plant, and VPC-1 compares that with its
90% set point. If even the most open valve is less than 90%
open, the set point of TIC-1 is decreased, and if this opening
exceeds 90%, the TIC-1 set point is increased. This allows
all users to obtain more heat (by further opening their supply
valves) if needed, while the header temperature is continu-
ously optimized (minimized).

© 2006 by Béla Lipták

 8.13 Chiller Optimization 1737

M1

0 –50%
Cooling SP
FO TIC HLL
Air at Twb tower 1 1

TCV TCV
1A 1B
M2 Setpoint = 90%
50 –100% Feedback VPC
CTWP (ER) 1
Steam Integral
FC TT only
Condenser 1
HWP
To HW
Thws
FSHL users

Steam
M5 heater TY
1

From all
P1 FC other user
SC valves

TIC
Identical control
loop as shown in
Figure 8.13g for Typical
setting chilled process
water flowrate heater
Hot water return

FIG. 8.13j
When the heat pumped by a chiller is recovered as hot water, a load-following cascade master controller (VPC-1) can be used to set the
hot water temperature controller’s (TIC-1) set point.

Figure 8.13j also shows that an increasing demand for heat
will cause the TIC-1 output to rise. An increase in the heat load
of the plant will cause a decrease in the heat spill to the cooling
tower through TCV-1A when the TIC-1 output is between 0
and 50%. At 50% (0.6 bar) output, all the available cooling
load is being recovered and TCV-1A is fully closed. If the heat
load continues to rise (TIC-1 output signal rises over 50%), this
will result in the partial opening of the “pay heat” valve,
TCV-1B. In this mode of operation, the steam heat is used to
supplement the freely available recovered heat to meet the
prevailing heat load of the plant.
A local circulating pump, P1, is started by FSHL when-
ever the flow velocity is low. This prevents the formation of
deposit in the condenser tubes. P1 is a small 10–15 hp pump
that is operating only when the flow is low. The main cooling
tower pump (usually larger than 100 hp) is stopped when
TCV-1A is closed.
OPTIMIZATION BY OPERATING MODE SELECTION
The cost-effectiveness of heat recovery is a function not only
of the outdoor temperature, but also of the unit cost of energy
from the alternative heat source and the percentage of the
cooling load that can be used in the form of recovered heat.
When the outside air temperature is below 65°F (18°C) and
steam is available at $7/MMBTU while only half of the cooling
load is required in the form of hot water, it is more cost-effective
to operate the chiller in the regular cooling mode than in heat-
recovery. In that case, according to the data in Figure 8.13k, it
is cost-effective to use steam as the heat source.
Conversely, when the outdoor temperature is above 75°F
(23.9°C), the penalty for operating the split condenser at hot
water temperatures is no longer excessive; therefore, the plant
should automatically switch back to the heat-recovery mode of
operation. This cost-benefit analysis can be a continually used
and automatic element of the overall optimization scheme.
Artificial Heat Sources
In locations such as the southern United States, where there
is no alternative heat source, another problem can arise
because all the heating needs of the plant must be met by
recovered heat from heat pumps. It is possible that during
cold winter days there might not be enough recovered heat
to meet this load. Whenever the heat load exceeds the cooling
load and there is no alternative heat source available, an
artificial cooling load must be placed on the heat pump.
This artificial heat source can in some cases be the cooling
tower water itself. A direct heat exchanger between the cooling

© 2006 by Béla Lipták

1738 Control and Optimization of Unit Operations

Cooling tower To
$10 Cooling load met
water condenser
by refrigeration
$9
$/MMBTU (0.25 × 106 k calories)

machine
$8 $8
$8 50% Compressor
Optimum M
$7 Cost of pay heat Direct
50% exchanger
(purchase cost of steam)
$6 Optimum
$5.30 Chilled Evaporator
$5 water
$4.25 return Chilled water
$4 TIC supply
$3 Hot water bypass
$2 To cooling tower
$1 FIG. 8.13l
When the chiller compressor is on, the cooling tower water can be
35 40 45 50 55 60 65 70 75 80
(1.7) (4.4) (7.2) (10) (12.8) (15.6) (18.3) (21) (23.9)(26.7) used as an artificial heat source; with the compressor off, it can be
4
Available condenser water temperature, °F (°C)
used as a means of direct cooling.

Operating costs of heat recovery chillers generating 105°F ambient conditions, the cooling tower water may be cold
(40.6°C) hot water with 50% and “optimum” loading of chiller.
enough to meet the cooling load of the process directly.
Operating costs of conventional mechanical chiller at various
condenser water temperatures and with 50% and “optimum” Alternatively, if the cooling tower water temperature is
loading of the chiller. below the chilled water temperature required by the process,
* the chillers can be operated in a free cooling or thermosiphon
Example
mode, which will be described in more detail later in con-
Temperature Heat nection with Figure 8.13o. In this case, the refrigerant circu-
of Condenser Cost Mechanical Recovery lation is driven by the temperature differential rather than by
Water Components Refrigeration Mode Mode the compressor, because the cooling tower water in the con-
Cost of Cooling $4.25 $8.00 denser is colder than the chilled water in the evaporator. In
65°F this thermosiphon mode of operation, the chiller capacity
(18.3°C) (0.5)(7.0) $3.50 $0.00 drops to about 10% of its rating.
Cost of Heating =
Total $7.75 $8.00
When the compressor of a chiller or heat pump is in
Cost of Cooling $5.30 $8.00 operation, the process is in the so-called mechanical refrig-
75°F eration mode (Figure 8.13m).
(23.9°C) (0.5)(7.0) $3.50 $0.00
Cost of Heating = When the process is cooled directly by the cooling tower
Total $8.80 $8.00
water, without the use of chillers or refrigeration machines,
* This example is based on the following assumptions: it is in the free cooling mode. Most modern cooling systems
a. The actual cooling load is 50% of chiller capacity (CL = 0.5 CAP). are provided with the capability of operating in several dif-
b. The heating load (the demand for hot water) is 50% of cooling load. ferent modes as a function of load, ambient conditions, and
(HL = 0.5 CL) utility costs.
The decision to switch from one operating mode to another
FIG. 8.13k is usually based on economic considerations. Such a control
The decision of operating in either the free cooling mode or in the system will automatically select the mode of operation that
heat recovery mode can be made automatically on the basis of cost- will allow the meeting of the load at the lowest total cost.
effectiveness.
Figure 8.13k illustrates the decision-making process by which
one can decide if the heat recovery or the mechanical refrig-
tower and chilled water streams is also advantageous in the eration mode will provide the better cost-effectiveness for the
case when there is no heat load but there is a small cooling prevailing conditions. The same kind of logic is applied when
load during the winter. At such times, the chiller can be the choice is between free cooling or mechanical refrigeration.
stopped and the cooling load can be met directly by the cool- Once it is decided to switch modes, the actual reconfig-
ing tower water (Figure 8.13l). uration of the associated piping and valve positions are
selected automatically. In larger, more complex cooling sys-
tems, there can be dozens of different modes of operation
OPTIMIZATION BY SYSTEM RECONFIGURATION (such a system was designed by the author for IBM corporate
headquarters at 590 Madison Avenue in New York). In this
In most plants, coolant can be provided from many sources, chapter, only four basic modes will be discussed.
so another approach to optimization is to reconfigure the The cooling system illustrated in Figure 8.13m is con-
system in response to changes in loads, ambient conditions, figured as a mechanical refrigeration system. The controls
and utility costs. For example, during some operating and for this configuration have been described in Figures 8.13g

© 2006 by Béla Lipták

 8.13 Chiller Optimization 1739

M1 and 8.13j. The heavily drawn pipelines show the flow paths
that are active in this mode of operation.
When the load drops off or when the ambient tempera-
Air Air ture decreases, it is possible to switch to one of the free-
at at cooling modes after the compressor has been stopped.
Twb Twb
Table 8.13n lists the equipment status and valve positions
for each of the operating modes that the system in Figure 8.13m
can support.
CTWP
B Indirect Free Cooling by Thermosiphon
A V7
Free cooling can be direct or indirect. When indirect free
Filter
cooling is used, the compressor is off and the refrigerant
A transports the heat to the cooling tower water from the chilled
Condenser water by evaporating in the “condenser” and condensing in
V1
B
Compr.
the “evaporator.” This is mode 2 in Table 8.13n. In this indirect
V2 (On) V3 free-cooling configuration the compressor is off and the heat
A is transferred from the evaporator to the condenser through
V6
Evaporator the natural migration of refrigerant vapors (Figure 8.13o).
B
Opening the refrigerant migration valves (V2 and V3)
A equalizes the evaporator and condenser pressures. Because
V5 the condenser is at a lower temperature, the refrigerant that
B is vaporized in the evaporator is recondensed and returns to
B Heat the evaporator by gravity flow. The cooling capacity of a
exchanger chiller is about 10% of full load in this mode of operation.
A V4
The cooling system is operated in this mode when the
load is low and the cooling tower water temperature is about
B 10°F (5.6°C) below the required chilled water temperature.
V8 This usually means 40–45°F (4.4–7.2°C) cooling tower and
A
50–60°F (10–15.6°C) chilled water temperatures. Such high
Process
CHWP load
chilled water temperatures are not unrealistic in the winter,
because the air tends to be dry and dehumidification is not
required.
FIG. 8.13m
This mechanical refrigeration system can be automatically recon-
This (mode 2 in Table 8.13n) can also be implemented
figured to operate in any of the three “free cooling” configurations, without controls. Under manual controls both sets of pumps
which are listed in Table 8.13n. are operated and the temperatures are allowed to float as a
function of load and ambient conditions.

TABLE 8.13n
Equipment Status in Four Operating Modes of a Refrigeration System
Mode 1 Mode 2 Mode 3 Mode 4
(Mechanical (Vapor migration based (Indirect free cooling by the (Direct free cooling
Equipment refrigeration) indirect free cooling) use of heat exchanger) with full filtering)

Compressor On Off Off Off

Cooling-tower pumps On On On On
Chilled-water pumps On On On Off
Valve V1 A A B B
Valve V2 and 3 Closed Open Closed Closed
Valve V4 A A B A
Valve V5 A A B A
Valve V6 A B A A
Valve V7 A A A B
Valve V8 A A A B

© 2006 by Béla Lipták

1740 Control and Optimization of Unit Operations

FC under normal conditions. Therefore, under these conditions,

M1 the PDIC-08 set point is kept at its allowable minimum. This
in turn results in minimizing the chilled water pumping cost
so long as all the users are satisfied. If any of the user valves
From Air Air Tctwr would open to over 95%, VPC-07 will quickly increase the
Fig.8.13d at at
DA Twb Twb pumping rate to guarantee sufficient coolant to that valve.
or 8.13h
TDIC Tctws Tctws TDIC When this occurs, the system is using up the cooling capac-
SP 1
01 02
Twb DA
SP 2 ity stored in the circulated chilled water. Once this is exhausted,
FC
CTWP ∗ control will be lost. Therefore, it is advisable to switch to a
B different mode of operation when this condition is detected.
A V7 RA
PDIC
03
Filter Indirect Free Cooling by Heat Exchanger
SP
A
Condenser Another method of indirect free cooling is to transfer the heat
Refrigerant
V 1 migration Set directly between the cooling tower water and the chilled water
B V 2 Compr. V3 DA VPC at
valves 04 through a heat exchanger (mode 3 in Table 8.13n). This allows
(Off ) 95%
DA TIC FC FC the complete bypassing of the refrigeration machine. This
05 SP A
V6 system has no heat transfer capacity limitation, because any
Evaporator
B cooling load can be handled as long as the plate-type heat
RA VPC Set at exchanger is large enough to handle it. Therefore, the main
06 90% V5
A advantage of mode 3 over mode 2 is that it is not restricted to
B loads of 10% or less of chiller capacity. Its main disadvantage
B Heat is the need for an additional major piece of equipment, namely
V4
exchanger the heat exchanger.
A Indirect free cooling using the plate-type heat exchanger
can be operated manually, without any controls. In that case,
B the chilled water temperature will float as a function of the
V8 process load and the ambient temperature, while the pump
A
stations operate at full capacity.
Process
CHWP load
DA
VPC Set at Optimized Operation If it is desired to optimize the opera-
FC RA 07 95%
PDIC SP tion of mode 3, the control system described in Figure 8.13p
08
can be implemented. Here, on the cooling tower, TDIC-01
* From Figure 8.31i.
maintains the optimum approach and TDIC-02 maintains the
optimum range. TDIC-02 can be overridden by PDIC-03 or
FIG. 8.13o by TIC-05. The pumping rate of cooling tower water circu-
In the free-cooling mode (mode 2 in Table 8.13o), the compressor lation will therefore be set by the highest of the three con-
is off. The refrigerant evaporates in the “condenser” and freely troller outputs. The purpose of PDIC-03 is to guarantee the
migrates to the evaporator, where it condenses, because the con- minimum pressure differential required for the plate-type
denser is colder than the evaporator.
exchanger, while TIC-05 can override TDIC-02 and PDIC-
03 if more cooling is required.
Optimized Operation To optimize this operation, the con- The set point of TIC-05 is set by VPC-06 to prevent the
trol strategy shown in Figure 8.13o has to be implemented. most-open user valve from exceeding a 90% opening. Under
Here TDIC-01 maintains the optimum approach and TDIC- normal conditions, the pumping rate of chilled water is kept
02 maintains the optimum range, unless these controls are at a minimum by PDIC-08, which maintains the minimum
overridden by VPC-04 when the refrigerant migration valves ∆P required across the process load. When VPC-06 is unable
approach their full openings, signaling that increased heat to keep the most-open user valve at a 90% opening and this
transfer is needed at the condenser in order to meet the load. opening rises to 95%, VPC-07 will start raising the set point
The migration valves V2 and V3 are throttled by TIC-05. of PDIC-08, thereby increasing the pumping rate.
The set point of TIC-05 is maximized by VPC-06 by allowing This is only a temporary cure, because the added cooling
all user valves to open until the most-open valve reaches 90% capacity is available only at the expense of heating up the
opening. The rate of chilled water pumping is minimized by stored chilled water in the pipe distribution system. There-
VPC-07, which modulates the set point of PDIC-08. fore, it is advisable to detect if this condition occurs for more
As VPC-06 normally keeps the most-open user valve at than a minute or so and, when it does, automatically switch
90% opening, therefore the set point of VPC-07 is not reached the system to a cooling mode that can handle the increased

© 2006 by Béla Lipták

 8.13 Chiller Optimization 1741

FC FC
M1 M1

From Air Air Tctwr Tctwr

From Air Air
Fig. 8.13d at at Fig. 8.13d at at
or 8.13h DA T Twb
wb or 8.13h DA Twb Twb
SP 1 TDIC TDIC TDIC TDIC
01 Tctws Tctws 02 SP 1
DA 01 Tctws Tctws 02
Twd FC SP 2 Twb DA SP 2
∗ FC
CTWP ∗
CTWP
B B DA
TIC
A V7
A V7 05
SP
Filter Filter
A
Condenser A
Condenser
V1
B V2 V3 B V1
Compr. Compr.
(off ) V2 (Off ) V3
A
V6 A
V6
Evaporator
B Evaporator
RA B
PDIC RA
03 PDIC
V5 A
03
DA V5 A
SP TIC
B
05 B
B Heat
exchanger B Heat
V4 exchanger
A V4
A
RA
VPC Set at
B 06 90 %
B RA
V8 VPC
A V8
A 02
Process Set at
CHWP load 90 % Process
CHWP load
FC DA
RA
PDIC SP VPC Set at
08 07 95% * From Figure 8.13i.

* From Figure 8.13i. FIG. 8.13q

FIG. 8.13p
Indirect free cooling is obtained by sending the chilled water that
is returning from the process to a heat exchanger in which it is
cooled by the cooling tower water (mode 3).
load. This can be mechanical refrigeration (mode 1) or free
cooling through interconnection (mode 4).
Direct Free Cooling
In direct free cooling, the cooling tower water is piped
directly to the process load, as shown in Figure 8.13q. This
method (mode 4) is the most cost-effective mode of cooling,
because both the compressor and the chilled water pump
station are off. In the winter, this mode of operation can
handle high process loads, as it is limited only by the size of
the towers and their pumps.
The main disadvantage of direct free cooling is that it
brings potentially dirty cooling tower water to the process
users, causing plugging and build-up on the heat transfer
In the direct free-cooling mode, the cooling tower water is sent
directly to the process load.
surfaces. This problem can be solved either by full flow filter-
ing (also called strainer cycle), which is shown in Figure 8.13q,
or by the use of closed-circuit, evaporative cooling towers. In
such “noncontact” or “closed-loop” cooling towers, the water
has no opportunity to pick up contaminants from the air.
This configuration is also frequently operated manually,
without automatic controls. In that case, the cooling tower
water temperature floats as the load and ambient conditions
vary, and the fan and pumping rates are not optimized.
Optimized Operation If optimization is desired, the con-
trols shown in Figure 8.13q can be implemented. In this
figure, TDIC-01 serves to keep the cooling tower approach
at an optimum value, and TDIC-02 optimizes its range. The
range controller can be overridden by PDIC-03 or by TIC-
05 when either of these controllers require a higher pumping
rate than does TDIC-02, the range controller. The set point
of TIC-05 is optimized to keep the most-open user valve from
exceeding a 90% opening.

© 2006 by Béla Lipták

1742 Control and Optimization of Unit Operations

Operating Mode Reconfiguration If the mode changes are frequent or if the coolant capacity
of the piping headers is very small (insufficient to meet the
The number of possible modes of operation can be rather high, process load for even a few minutes), all systems must be
if the cooling system of the plant includes optional storage, heat switched while running. Dynamic switching requires more
recovery, and alternate types of motor drives. Switching from planning and a higher level of automation, because the auto-
one mode to another is not as simple as it might first appear, matic starting of certain pieces of equipment (such as chillers
because reconfiguration necessitates the stopping and starting driven by a steam turbines) requires a more comprehensive
of equipment and because the reconfiguration of control loops set of safety interlocks.
often requires the modification of pump discharge heads. Automatic operating mode reconfiguration is one of the
For example, when switching to mode 4, illustrated in most powerful tools of optimization, because it provides flex-
Figure 8.13q, the water circulation loop served by the cooling ibility to a previously rigid system. The mode reconfiguration
tower pumps becomes much longer. This in turn shifts the technique is not limited to cooling systems. It can be effective
operating point of the pumps and can lower their efficiency in any unit operation in which the system must adapt to
if the system is not carefully designed and evaluated for each changing conditions.
operating mode.
The control loop configuration requirements for all four
modes are tabulated in Table 8.13r. When the control loops RESPONSIVENESS AND PRIORITIZING
are automatically reconfigured and their manipulated vari-
ables are changed, it is important to also revise their tuning In some processes it is critical that the chillers be able to
constants, because the time constants of the loops are also respond to fast and drastic changes in the process cooling
changed. This should preferably be done automatically to load. In other processes, the plant’s cooling system simulta-
minimize the potential for human error. neously serves high- and low-priority users, and it is much
Whenever the process can tolerate an interruption of a more important to meet the needs of the high-priority users
few minutes in its cooling system, the switching from one than the others. The control systems described in the follow-
mode to another should be done in a stagnant state, while ing paragraphs are designed to improve responsiveness and
the equipment is turned off. This usually is acceptable for all to consider user priorities in chilled water distribution.
motors except for the pumps, which are serving the process
load. These pumps usually can be left running, and during More Responsive Control
switching they can utilize the coolant storage capacity of the
water distribution piping. Transferring to another mode while Feedforward anticipation can increase the responsiveness of
the pumps are running can be accomplished, because the the control system while also providing more precise temper-
slowly diverting three-way valves will never completely ature control. Such precision is desired when, for example, the
block the pump discharge but will only gradually change the goal of optimization is to maximize chilling without freezing.
destination of the water. When in a heat exchanger the heat transfer area is fixed
(such is the case in an evaporator with constant refrigerant
level), the rate of heat transfer (load) is a function of the
temperature difference between the process fluid and the
TABLE 8.13r boiling refrigerant. Therefore, as the load changes, the boiling
Listing of the Manipulated Variables That Are Assigned To
temperature (and pressure) must also change.
Be Throttled by the Various Controllers in the Four Modes
of Operation
Figure 8.13s shows the main components of such a
responsive control system. In this case the instantaneous load
Manipulated Variables Controlled by the Controller is calculated by multiplying the flow rate of the cooled pro-
Controller Mode 1 Mode 2 Mode 3 Mode 4
cess fluid (in FY-05) by the drop in temperature (TY-04) that
it should experience. Based on this flow rate and the desired
TDIC-01 Fan Fan Fan Fan process temperature (set point of TIC-01), FY-06 determines
TDIC-02 CTWP CTWP CTWP CTWP the desired refrigerant temperature. The value of K in this
PDIC-03 CTWP CTWP CTWP CTWP summing device is adjustable to reflect the actual slope of
load vs. refrigerant temperature. PY-07 serves to convert the
VPC-04 — PDIC-03 — —
Set point
desired boiling temperature to the corresponding vapor pres-
sure set point.
TIC-05 Compressor V2/3 CTWP CTWP
Dynamic compensation is provided by FY-03 to match
VPC-06 TIC-05 TIC-05 TIC-05 TIC-05 5
the response of process temperature to refrigerant pressure.
Set point Set point Set point Set point
A valve position controller can be added to detect the refrig-
VPC-07 PDIC-08 PDIC-08 PDIC-08 — erant control valve opening and to keep that valve over 90%
Set point Set point Set point
open by varying the speed of the compressor, in order to
PDIC-08 CHWP CHWP CHWP — minimize the operating cost of the compressor.

© 2006 by Béla Lipták

 8.13 Chiller Optimization 1743

Temperature Compressor
to pressure
SIC SP HLL
converter M TIC
Desired boiling 1 1 1
PY f (x)
temperature 07
of refrigerant
SP FY
PIC 2
∑ 07
−K Integral
FY
Actual 06 PT Integral Feed-
only
load + 07 only FIC Feedback back VPC Setpoint
Vapor 2 (ER) (ER) 1 = 90%
Lead/lag R/A RA
x Refrigeran
FY FY Brine or
03 05 TT
DP 1 chilled water
2 Tchws to other users
∑
TT + TY
04 Evaporator 04 Evaporator TY
− 1
FT #1
03
Process LT SP Desired From all
fluid 02 TT TIC process FC high FC
01 01 priority
temperature
valves
LIC High
02 TIC TIC
priority Low priority
process process
cooler cooler

Liquid refrigerant

FIG. 8.13s FIG. 8.13t

Increased responsiveness can be obtained by the feedforward The chilled-water supply temperature can be optimized by keeping
adjustment of refrigerant pressure. (Adapted from Reference 5.) the most-open high priority user valve at 90%. (Adapted from
Reference 2.)

Water Distribution Optimization cooling load. If valve-position-based optimization were

Figure 8.13f illustrates the ideal water distribution system,
in which the individual users are served by two-way valves
and the optimum supply temperature is found by keeping the
most-open valve at 90% opening. Although this is a very
effective control configuration, it is not always used when
some of the chilled water users are not critical.
If the number of user valves is very high and if the valves
are distributed over a large area, or if the loops are not properly
tuned and the valves have a tendency to cycle from fully
closed to fully open, the control shown in Figure 8.13f is not
practical. In such situations, the concept illustrated in Figure
8.13f can still be applied by selecting a few representative
user valves that are not cycling and by basing the optimization
on those.
If the chilled water user valves are grouped into high-
and low-priority categories or into other groups, the water
distribution and supply temperature can still be optimized.
Figure 8.13t illustrates a control system in which the high-
priority users are treated the same way as they were in
Figure 8.13f. The low-priority users, on the other hand, are
grouped together, and their demand is detected through the
measurement of total flow.
If all low-priority valves are the two-way type and if
supply temperature is constant, flow will vary directly with
applied, it would attempt to maximize valve opening, which
in turn would maximize flow by raising Tchws to the highest
acceptable level. This same goal is achieved by FIC-2 in
Figure 8.13t. As the flow drops off as a result of a reduction
in demand for cooling, FIC-2 will raise the supply tempera-
ture, which in turn will increase the total flow, as the valves
open up. Therefore, the total flow is kept constant, and load
variations result in supply temperature variations. VPC-1 and
FIC-2 should both be integral-only controllers, with time
constants of several minutes.
The control of total flow alone in Figure 8.13t can satisfy
the average user, but not all the users. Therefore, the critical
users cannot be treated in this manner, but must be protected
by valve position control. Such control is provided by VPC-1,
which overrides the flow controller whenever a high-priority
valve reaches 90% opening.
Three-Way Valves to Optimize HVAC Systems In many older
or air-conditioning-type applications, three-way control valves
are used. This is a highly undesirable practice, because the use
of three-way valves unnecessarily increases the required
amount of pumping and it also lowers return water tempera-
ture. Yet, these badly designed systems can also be optimized.
A control system that will do that is shown in Figure 8.13u.

© 2006 by Béla Lipták

1744 Control and Optimization of Unit Operations

Compressor percentage of water diverted as at full load. The benefit of

SIC TIC SP HLL this optimization strategy is in reducing the temperature dif-
M
1 1 1
ferential across the chiller, which in turn lowers the operating
Σ
cost by 15% for each 10°F (5.5°C).
+ The setting of G determines the percentage of coolant
To TDY VPY
2 1 diverted by the average valve. If G is set too low, some valves
−G
Integral will be 100% open to the chilled water supply and therefore
Feed-
back VPC only Set point out of control. If G is set too high, the chilled water supply
(ER) 1 = 90% temperature will be lower than needed, and compressor
RA
Brine or
energy will be wasted.
TT chilled water In order to protect the high-priority users from running
1
Tchws to other users out of coolant, when G is set low, a valve-position-based
override is used in Figure 8.13u. Whenever a high-priority
TY
Evaporator 1 valve is more than 90% open, VPC-1 will override the set
point of TIC-1 and will lower it until all high-priority valves
Water From all
flow const. high priorty are less than 90% open.
TDT valves
therefore 2
∆T reflects Set Set
load at at
High
T
Low
T
OPTIMIZING STORAGE AND LOAD ALLOCATION
priority TIC o priority TIC o
process process
cooler cooler The total cost of cooling can also be reduced by making the
coolant when it is less expensive to make it and placing it in
Tchwr storage until it is needed. When a variety of equipment is
available to meet the cooling load of the plant, the total cost
FIG. 8.13u of operation can also be minimized by always using the most
The optimization of chilled water supply temperature on the basis efficient combination of equipment. These are the topics of
of keeping the most open high priority three-way user valve 90% the next paragraphs.
open. (Adapted from Reference 4.)
Optimized Storage
When three-way valves are used, the total flow is rela- If storage tanks are available, it is cost-effective to generate
tively constant, and it is the return water temperature that the daily brine or chilled water needs of the plant at night,
reflects the variations in the cooling load. Therefore, the when it is the least expensive to do so, because ambient
control system in Figure 8.13u determines the set point for temperatures are low and night-time electricity is less expen-
TIC-1 by relating the temperature rise (Tchwr − Tchws) and the sive in some areas.
6
5
desired space temperature set points (To) as follows: When demand is low, operating costs can be reduced by
TIC-1 setpoint = To − G(Tchwr − Tchws) 8.13(2) operating the chillers part of the time at peak efficiency rather
than continuously at partial loading. Efficiency tends to be
If the thermostats are set at To = 80°F (27°C) and if, at lower at partial loads because of losses caused by friction drop
full load, Tchws = 50°F (10°C) and Tchwr = 65°F(18°C), the across suction dampers, prerotation vanes, or steam governors.
value of G can be determined as follows: Cycling is practical if the storage capacity of the distri-
G = (To − setpoint)/(Tchwr − Tchws ) = (80 − 50)(65 − 50) = 2 bution headers is enough to avoid frequent stops and starts.
When operation is to be intermittent, data such as the amount
8.13(3)
of heat to be removed and the characteristics of the available
TDY-2 in Figure 8.13u therefore calculates the desired chillers are needed to determine the most economical oper-
set point for TIC-1 at any load. For example, if the load drops ating strategy.
to 50%, the space temperature To will begin to fall, and the When the chiller is cycled, the thermal capacity of the
thermostats will divert more coolant into the return line. This chilled water distribution system is often used to absorb the
will reduce Tchwr from 65°F (18°C) to 57.5°F (14°C). In load while the chiller is off. For example, if the pipe distri-
conventional systems, this then would become the new steady bution network has a volume of 100,000 gallons (378,000 l),
state. With the control system shown in Figure 8.13u, TDY-2 this represents a thermal capacity of approximately 1 million
BTUs for each °F of temperature rise (1.9 × 10 J/°C).
6
will respond by revising the TIC-1 set point as follows:
So, if one can allow the chilled water temperature to
set point = 80 − 2 (57.5 − 50) − 65°F 8.13(4)
float 5°F (2.8°C) (such as from 40°F to 45°F, or from 4.4°C
As Tchws is slowly increasing, the space thermostats will to 7.2°C) before the chiller is restarted, this represents the
divert less and less water until they come to rest at the same equivalent of approximately 400 tons (1405 kW) of thermal

© 2006 by Béla Lipták


capacity. If the load happens to be 200 tons (704 kW), the
chiller can be turned off for 2 hours at a time. If the load is
1000 tons (3,514 kW), the chiller will be off for only 24
minutes. This illustrates the natural load-following, time-
proportioning nature of this scheme.
If the chiller needs to be off for a longer period than the
thermal capacity of the distribution system can provide, three
options are available:
1. Storage tanks can be added to increase the water volume.
2. A second chiller can be started (not the one that was
just stopped).
3. The load can be distributed among several chillers of
different sizes, keeping some in continuous operation
while cycling others.
Optimized Load Allocation
Continuous measurement of the actual efficiency ($/ton) of
each chiller can enable the plant to meet all loads with the
most efficient combination of machines. In plants with mul-
tiple refrigerant sources, the cost per ton of cooling can be
calculated from direct measurements, and it can be used to
select the most efficient combination of units to meet any
existing or anticipated loads. The differences between the
cost per ton characteristics and the efficiency-vs.-load char-
acteristics of the individual coolant sources can be combined
and used as in Figure 8.13v.
In simple load allocation systems, only the starting and
stopping of the chillers is optimized. In such systems, when
the load is increasing, the most efficient idle chiller is started,
and when the load is dropping, the least efficient one is stopped.
In more sophisticated systems, the load distribution between
operating chillers is also optimized. In such systems, the real-
time efficiency of each chiller is calculated. This determines
the incremental cost for the next load change for each chiller.
If the load increases, the incremental increase is sent to the
set point of the most cost-effective chiller. If the load decreases,
the incremental decrease is sent to the least cost-effective unit
(Figure 8.13w). Software packages are readily available that
0.10
Cost ($/ton)
0.08 Steam
0.06 Electric
0.04
0 20 40 60 80 100
% Load
FIG. 8.13v
The unit cost of refrigeration is a function of both the load and of
the characteristics of each chiller. Installed systems should be used
6
as the basis of obtaining the operating cost curves.
© 2006 by Béla Lipták
8.13 Chiller Optimization 1745
Total demand
for cooling
Select least Decrease Nature Increase Select most
cost-effective of cost-effective
chiller change
chiller
No
change
To chiller To chiller
loading controls loading controls
FIG. 8.13w
Optimized, computer-based load allocation will direct load
increases to the most cost-effective chiller, while sending the load
decreases to the least cost-effective chiller.
have proven capabilities for continuous load balancing through
7
the predictions of efficiencies and operating costs.
With the strategy described in Figure 8.13w, the most effi-
cient chiller will either reach its maximum loading or will enter
a region of decreasing efficiency and will no longer be the most
efficient. When the loading limit is reached on one chiller, or
when a chiller is put on manual, the computer will select another
unit as the most efficient one for future load increases.
The least efficient chiller will accept all decreasing load
signals until its minimum limit is reached. Its load will not be
increased unless all other chillers are at their maximum load or
when it is in manual. As shown in Figure 8.13v, some chillers
can have high efficiency at normal loads while being less effi-
cient at low load. Such units are usually not allowed to shut
down, but are given a greater share of the load by the control
system.
If all chillers are identical, some will be driven to the
capacity corresponding to maximum efficiency while others
will be shut down by this strategy, and only one chiller will
be placed at an intermediate load.
In starting up chiller stations, the optimization system
“knows” how many BTUs need to be removed before start-up
and the size and efficiency of the available chillers. Therefore,
the length of the “pull-down” period can be minimized and the
energy cost of this operation can also be optimized.
OPTIMIZATION OF SPECIAL CHILLERS
In the following paragraphs the optimization of some special
chiller configurations will be discussed. These will include
the commercial chillers that utilize a number of positive dis-
placement refrigerant compressors. Also discussed will be
the multistage refrigeration units that are needed to generate
very low temperature coolants and the multiple-user systems.
Optimized Commercial Chillers
In supermarkets, warehouses, and other commercial facili-
ties, there usually are a large number of loads (zones, each
served by an evaporator coil), and there are a number of
1746 Control and Optimization of Unit Operations

Compr. M1
1
Suction pressure (P) 1 HP
Incr.
∆T1

Condenser
Compressor
Option*

HP
load levels Status
Compr.
(HP) M1 M2 M3 M4 #2 M2
PSH SET PT. 2
1 HP A × 2 HP
2 HP A × #1
3 HP A × ×
× PSL SET PT.† Compr. M3
B
Decr. 3
A × × 3 HP
4 HP #3 HP
B × #4
A × ×
5 HP Time ∆T2
B × × Compr. M6
A × × × 4
6 HP 4 HP
B × ×
A × × ×
7 HP
B × ×
8 HP A × × × Receiver
9 HP A × × ×
10 HP A × × × ×

∆T2 = Seconds
Decr. operating HP TD P
PSL PT
2

Increase. operating HP TD SP2 = 0.95 SP1

PSH P
1
SP1 = A + 0.9P
∆T1 = Minutes
+ 0.9P
0.95 x Σ x 0.9

Refrigerant liquid “A”

Refrigerant vapor

<
Balancing Balancing Balancing Balancing
valve valve valve valve

R/A R/A R/A R/A

Exp. Exp. Exp. Exp.
TC Int. TC Int. TC Int. TC Int.
VA. VA. VA. VA.

S TS S TS S TS S TS

Coil defroster Coil defroster Coil defroster Coil defroster

Cooling zone #1 Cooling zone #2 Cooling zone #3 Cooling zone #4

FIG. 8.13x
Here a control system is shown that will minimize the energy cost, while equalizing the operating times of multiple chiller compressors,
which are serving an unlimited number of evaporator coils.8

positive-displacement refrigerant compressors, which are and to select the operating compressors in such a way that
turned on or off to follow their loads. The goals of optimi- the total run time of each machine is about the same.
zation are to minimize the cost of operation (to meet any Figure 8.13x illustrates such a control system, which
cooling load with a minimum of compressor horsepower) incorporates some of the ideas contained in the patents listed

© 2006 by Béla Lipták


in Reference 8. In Figure 8.13x each of the cooling zones can
be a separate room, freezer, or other cooled area, which is
provided with its own temperature controls (TC). Each of the
zones is cooled by an evaporator coil, which is kept cold by
its boiling liquid refrigerant.
The liquid refrigerant is admitted into each coil by an
expansion valve, which controls the superheat of the evapo-
rated vapors by detecting both the pressure and the temper-
ature of those vapors. The expansion valve admits as much
refrigerant liquid as is required to keep the evaporator super-
heat (the temperature above the boiling point at the operating
pressure of the refrigerant) constant.
A manual balancing valve is also provided for each zone.
These valves are located in the refrigerant vapor lines leaving
the zones. If all zones are to be kept at the same temperature,
all balancing valves can be kept fully open. This will mini-
mize the pressure drop on the suction side of the compressors
and thereby will maximize their efficiency.
If some of the zone temperatures are to be controlled (by
their TCs) at a higher temperature than the others, the bal-
ancing valves on these higher temperature zones must be
throttled to increase the vaporization pressure and thereby
the boiling temperature of the refrigerant in the coils of those
zones.
Defrosting and Sequencing Controls If the zones are oper-
ated below or near to the freezing temperature (below 35°F,
or 2°C), it is also necessary to provide defrosting controls
because of the ice build-up on the cold evaporator coils. The
defrosters are controlled by coil surface temperature detector
switches (TS), one in each zone (TS). These switches are
continuously detecting the coil surface temperatures, and
when that temperature drops below some low limit for some
preset time period, the switch first shuts off the refrigerant
liquid supply by closing a solenoid valve.
Once all refrigerant has been vaporized, the switch turns
on a coil defroster heater for a short time, to remove the ice.
During the defrosting cycle, the zone temperature controller
(TC) is disabled by setting its measurement and set point to
be equal and thereby causing its output signal to stay constant.
The nonoptimized method of turning the refrigerant com-
pressors on and off is by sequential controls, where each
added compressor unit is turned on when the suction pressure
to the overall compressor station has dropped to an even
lower level. This approach is inefficient, because the reaching
of these lower levels of suction pressure requires added com-
pressor horsepower.
Optimized Controls The control system shown in Figure
8.13x eliminates the waste of added compressor horsepower
by turning added compressors on (PSH) or turning unneces-
sary compressors off (PSL) at the pressure settings of PSH and
PSL. This is achieved by both pressure switches actuating a
time delay (TD), and whenever the pressure switch is actu-
ated for the time period set on the time delay, it changes the
© 2006 by Béla Lipták
8.13 Chiller Optimization 1747
operating horsepower level of the system. TD-1 increases it
while TD-2 decreases the horsepower level.
TD-1 is set in minutes, while TD-2 is set in seconds.
Therefore, a response to a load increase is relatively slow
(minutes), while the response to a load decrease is fast (sec-
onds), which serves energy conservation.
Further optimization is obtained by pushing the control
gap between the pressure switches (PSH and PSL) as high
as the loads will allow. This continuous floating of the con-
trolling suction pressure gap is achieved by feedforward
action based on zone temperature control. As any one of the
zone temperatures rises above the zone thermostat (TC, a
reverse-acting, integral-only controller) set point, the output
signal of that TC drops. A low-signal selector selects the
lowest of the TC outputs (A) and thereby identifies the zone
with the highest cooling load (the critical zone, which, if
satisfied, will result in all others being satisfied).
The set point of PSH is obtained by taking 90% of the
suction pressure of the compressor station (0.9P) and adding
to it the output signal (A) of the critical TC. The set point of
PSL is 95% of the PSH set point. Both the 90% and the 95%
values are adjustable and can be varied from one installation
to the next.
The energy consumption of the system can be minimized
and its operation optimized by maximizing these set points.
This is because whenever all the zone temperatures are on
set point (or below), the suction pressure of the operating
compressor station is increased, and therefore the amount of
work the compressors need to do is reduced.
The graph of suction pressure (P) vs. time in Figure 8.13x
illustrates the operation of the system. As the suction pressure
rises, at point #1 the set point of PSH is reached and the time
delay TD-1 is actuated. At point #2 the time delay that was
set on TD-1 has run out, which increases the operating horse-
power level by one unit. As a result, the suction pressure
starts to drop.
If at this point the operating compressor capacity exceeds
the cooling load, the suction pressure will continue to drop,
and when it reaches the set point of PSL (point #3), TD-2
will be energized. After the time delay of TD-2 has run out
(point #4), the operating horsepower level of the system is
reduced by one unit.
Minimum Maintenance In addition to energy optimization,
the system is also optimized from a maintenance point of
view by keeping all compressor running times approximately
the same. This is accomplished by the use of the horsepower
table in the top left of Figure 8.13x. Let us assume that the
system is operating at the 4 hp level when PSH calls for an
increase to 5 hp. The 5 hp level can be met by either operating
compressors 1 and 4 (option A) or by operating compressors
2 and 3 (option B). In this situation the least used compressor
pair will be started.
If at some time later PSL calls for reducing the operating
level to 4 hp, there again are two options. Option A is to run
compressors 1 and 3, while option B is to run compressor 4.
1748 Control and Optimization of Unit Operations

Here, the choice is made in such a manner that the most used M
Envelope algorithm
compressor will be turned off. The net result is a continuous
redistributes
balancing of running times between machines. According to load between
Reference 8, the total energy savings obtained by this control Air at Cooling stages to:
Twb tower
system amounts to about 20%. 1) Minimize cost
2) Maximize
Condenser rangeability
Multistage Chillers CTWP
Tctws Tctwr
It is not practical to obtain a compression ratio outside the Tc
Propane
range of 3:1 to 8:1 when using the compressors usually refrigerant
installed in the process industry. This places a limitation on
the minimum temperature that a single-stage refrigeration Letdown
valve
unit can provide. For example, in order to maintain the evap- Compressor
#2 M
orator at −80°F (−62°C) while the condenser is operating at #2
86°F (30°C), the required compression ratio for Freon-12 is Evaporator
PT PIC PY
calculated as 02
02 02
Ethylene
×
refrigerant presure at 86°°F 108 refrigerant Speed
compression ratio = = = 37 #2 PY
refrigerant presure at −80°F 2.9 ratio 01 SP
Letdown Compressor
valve M
8.13(5) #1
#1
For a single-stage machine, such a compression ratio is Evaporator PT PIC
not practical, and therefore, a multistage system is required 01 01
to provide it. Figure 8.13y illustrates a multistage refrigera- SP
tion system, with ethylene as the lower and propane as the #1
TT TIC
01 01
higher temperature refrigerant. In order to minimize the total
operating cost, the loading of the two stages must be coor- Process
cooling
dinated to balance the total work between the stages. load
The higher the interstage temperature (evaporator #2), Chilled process stream
the more work is done by stage #1 and the less work remains
for stage #2. If the compression ratios (and therefore the FIG. 8.13y
temperature differences) of the two stages are the same, the The operation of multistage refrigeration units can be optimized by
work will be nearly equally distributed. If true equality is to automatically balancing the work load between the stages. (Adapted
be achieved, the differences in the properties of the refriger- from Reference 5.)
ants and the added load on the higher stage caused by the
work that is introduced by the lower stage compressor must
also be considered. As long as both stages are far from surge, the algorithm
redistributes the load to minimize total operating cost. When
one of the stages is nearing surge, the envelope algorithm
Optimized Controls One of the goals of optimization is to increases its loading to keep the system out of surge. There-
maximize the rangeability of the multistage refrigeration fore, at high loads the goal of load distribution between the
units. This rangeability will be maximum when the two stages stages is to minimize the total operating cost, and at low loads
approach their surge limits at the same time. This can be the goal is to keep the system out of surge.
guaranteed by redistributing the load between the stages in As ambient temperatures drop, the cascaded stages will
such a manner that they maintain equal distance from their operate against a lower total temperature difference, and
respective surge lines as the load drops. therefore, each of the stages will require less work. As the
In Figure 8.13y, TIC-01, the temperature controller condenser temperature drops, suction pressure of compressor
of the chilled utility supply to the process, sets the speed of #2 will also drop. This in turn will cause the temperature in
the first stage compressor #1 by modifying the set point of evaporator #2 to drop, which will result in the lowering of
its suction pressure controller PIC-01. The speed of com- the suction pressure of compressor #1.
pressor #2 is set in speed ratio to that of compressor #1 by As the suction pressures drop, PIC-01 and PIC-02 will
PY-01. The set point of PY-01 is optimized by an envelope reduce the speeds of their respective compressors, thereby
algorithm that attempts to minimize the operating cost and reducing the total work introduced by the heat pumps. Using
to maximize rangeability, as it is moving the PY-01 set point these controls, this multistage refrigeration system will
to its optimum. respond to ambient variations in a flexible and efficient manner.

© 2006 by Béla Lipták

 8.13 Chiller Optimization 1749

Optimized Control of Multiple Users M

When the same compressor serves several evaporators in

parallel, either the evaporator temperature or the refrigerant Air at Cooling
Twb
level in the evaporators can be varied to follow the load. tower
Figure 8.13z shows the controls for a configuration, where
the evaporator levels (their wetted heat transfer areas) are
Condenser
kept constant, and therefore the load variations result in CTWP Tctws Tctwr
changes in the evaporator temperature. In the control system
shown, TY-03 selects the most open TCV valve opening and Tc
Refrigerant
keeps that valve from opening to more than 90% by speeding
up the compressor when the valve opening reaches 90%.
Set at LC M
The resulting reduction in compressor suction pressure Compressor
90% 03
lowers the evaporator temperature and increases the rate of
heat transfer. Thus, no user is ever allowed to run out of
coolant, and the compressor operating costs are kept to a
minimum. This type of load-following optimization system, LY
03
where compressor speed varies with the requirements of the Evaporator Evaporator
#1 #2
most heavily loaded user, is very sensitive to dynamic upsets.
Therefore, VPC-03 must be tuned for slow adjustment of the LT LT
01 TT 02 TT
compressor speed to avoid upsetting the other users that are 01 02
not selected for control.
If one of the user valves is consistently more open than TIC TIC
the others, it might be possible to repipe the other loads to 01 02

FIG. 8.13aa
Air at Cooling The vapor-side control valves on multiple users can be eliminated
Twb by automatically adjusting the compressor speed, so as to keep the
tower
highest level in the evaporators at 90%. (Adapted from Reference 5.)

Condenser the interstage of the compressor. This might provide a better

CTWP Tctws Tctwr balance between users.
Tc
Refrigerant
Eliminating the Vapor-Side Valves Figure 8.13aa illustrates
a control system in which the vapor-side control valves have
set at M Compressor been eliminated. This advantage of the illustrated configura-
90%
VPC tion must be balanced against a number of disadvantages. In
03
the illustrated system, the heat transfer area is modulated and
is exposed to boiling, and the user with the highest load will
TY have the highest refrigerant level.
03 LY-03 selects the most flooded evaporator for control by
LC-03, which, when the level rises above 90%, increases
Evaporator TCV Evaporator TCV
01 02 the speed of the compressor and lowers it when the level is
#1 #2
below 90%.
LT LT This method of load following has some drawbacks. One
01 TT TIC 02 TT TIC
01 01 LIC 02 02
problem is that the heat transfer surfaces within the evaporators
LIC
01 02 are not fully utilized. Another difficulty is that the relationship
between level and heat transfer area can be confused by foam-
5
ing, which keeps the tubes wet even when the level has dropped.
For these reasons, tuning of such control system can be difficult,
FIG. 8.13z necessitating the sacrifice of responsiveness for stability.
When a chiller is serving a number of users, the speed of the chiller Therefore, if responsiveness is critical, it is better to keep
can be automatically modulated so as to keep the temperature the refrigerant level constant and manipulate the evaporating
control valve of the most demanding user at 90% opening. pressure, as shown in Figure 8.13z.

© 2006 by Béla Lipták

1750 Control and Optimization of Unit Operations
RETROFIT OPTIMIZATION
In new plants it is easy to install variable-speed pumps, to
provide the thermal capacity required for chiller cycling, or
to locate chillers and cooling towers at the same elevation
and near each other so that the cooling tower water pumping
costs will be minimized. In existing plants, one has to accept
the inherent design limitations and the system must be opti-
mized without major changes to the equipment. Even with
3
those limitations, one can still produce savings up to 50%,
but certain precautions are needed.
In optimizing existing chillers, it is important to give
careful consideration to the constraints of low evaporator
temperature, economizer flooding, steam governor rangeabil-
ity, surge, and piping/valving limitations.
Surge Protection
A detailed discussion of surge controls is given in this chapter
under Compressor Controls. Therefore, here it should suffice
to just mention that surge occurs at low loads when not
enough refrigerant is circulated. A surge condition can cause
violent vibration and damage.
Old chillers might not have automatic surge controls and
have only vibration sensors for shutdown. If the chillers will
operate at low loads, it is necessary to add an antisurge
control loop. Surge protection is always provided at the
expense of efficiency. To bring the machine out of surge, the
refrigerant flow must be artificially increased if there is no
real load on the machine. The only way to provide this
increase in flow is to add artificial and wasteful loads (for
example, hot gas or hot water bypasses).
Therefore, it is much more economical either to cycle a
large chiller or to operate a small one than to meet low loads
by running the machine near its surge limit.
Low Evaporator Temperature
Low temperatures can occur in the evaporator when an old
chiller is optimized—for example, when one that has been
designed for operation at 75°F (23.9°C) condenser water is
run in the winter using 45 or 50°F (7.2 or 10°C) condenser
water. This phenomenon is exactly the opposite of surge,
because it occurs when refrigerant is being vaporized at an
excessively high rate. Such vaporization occurs because the
chiller is able to pump twice the tonnage for which it was
designed as a result of the low compressor discharge pressure.
In such a situation, the evaporator heat transfer area
becomes the limiting factor. Furthermore, the only way to
increase heat flow is to increase the temperature differential
across the evaporator tubes. This shows up as a gradual
lowering of refrigerant temperature in the evaporator until it
reaches 32°F (0°C), at which point the machine shuts down
to protect against ice formation.
There are two ways to prevent this phenomenon from
occurring in existing chillers. The first is to increase the
© 2006 by Béla Lipták
Condenser
Cooling tower water
Freon
Compressor
Set at
Freon TT TRC 33°F
Evaporator Chilled water
FIG. 8.13bb
Protection against evaporator freeze-up can be provided by the evap-
4
orator temperature controller throttling the cooling tower water flow.
evaporator heat transfer area (a major equipment modifica-
tion). The second is to prevent the refrigerant temperature in
the evaporator from dropping below 33°F (0.6°C) by not
allowing the cooling tower water to cool the condenser to its
own temperature. The latter solution requires only the addi-
tion of a temperature control loop (Figure 8.13bb). This pre-
vents the chiller from taking full advantage of the available
cold water from the cooling tower by throttling its flow rate,
thereby causing its temperature to increase.
Economizer and Steam Governor
In existing chillers, the sizing of the economizer level control
valves (LCV-1 and LCV-2 in Figure 8.13cc) is often based on
an assumption. This assumption is that the refrigerant vapor
pressure in the condenser (P3) is constant and corresponds to
a condenser water temperature of 75 or 85°F (23.9 or 29.4°C).
Naturally, when such units are operated with 45 or 50°F (7.2
or 10°C) condenser water, P3 is much reduced, as is the
available pressure differential across LCV-1 and LCV-2.
If this occurs—when the freon circulation rate is high
—the control valves will be unable to provide the necessary
flow rate, and flooding of the economizer will occur, because
the flow is higher and the ∆P is lower than was the basis of
valve sizing. The solution is to install larger valves, preferably
external, and throttle them using controllers with proportional
and integral control modes.
The proportional control mode alone cannot maintain the
set point as load changes. The addition of the integral mode
is necessary to eliminate the offset. This is important in
machines that were not initially designed for optimized, low-
temperature condenser water operation, because otherwise
the liquid refrigerant that overflows from the flooded evapo-
rator can damage the compressor.
To optimize a steam-turbine-driven compressor, one must
be able to modulate its rotational velocity over a reasonably
wide range. This is not possible with old, existing machines,
if they are provided with quick-opening steam governor
valves. When such governors are throttled, a slight increase
in lift from the fully closed position results in a substantial
 8.13 Chiller Optimization 1751

P3 100 85F
Cooling Based on constant:
water 90 1. Condenser water flow rate

Entering condenser
water temperatures
2. Cooler water flow rate
Condenser 80 3. Leaving chilled water temperature 75F
Liquid
70

Percent design BHP

P3 LCV−1 65F
Vapor 60
Flash gas to
50 55F
2nd stage
P2 40
LCV−2
30

20
P1 Economizer 10
P2
P3
M 0
Vapor 10 20 30 40 50 60 70 80 90 100
Percent design load
Liquid Speed varied with entering condenser water
Compressor temperature
Speed varied with changes in load

FIG. 8.13ee
Evaporator
The actual power consumption of a variable-speed compressor is
P1
reduced by lowering either condenser water temperature or the
Chilled water
loading (speed). (Adapted from Reference 9.)

FIG. 8.13cc
by welding two rings with V-notches to the seats of the existing
To protect against economizer flooding, the internal valves (LCV-1
steam governor valves. This is shown in Figure 8.13dd.
and LCV-2) can be supplemented with larger external valves.

steam flow and therefore a substantial rotational velocity.

CONCLUSIONS
Consequently, the valves become unstable and noisy.
The characteristics of the steam governor valves can be
The goals of chiller optimization include:
changed from quick-opening to linear at minimal cost. The linear
characteristics and the desired wide rangeability can be obtained
• Minimizing the temperature difference across the chiller
by minimizing the cooling tower water temperature and
C B maximizing the chilled water temperature
Part B • Minimizing pumping costs by transporting only as
much water as is required to meet the load and by
Part A
using variable-speed pumping
Quick • Minimizing cooling tower operating costs by approach
opening
plug optimization
A • Operating chillers at the load at which their efficiency
is the highest and by meeting partial loads with part-
Part A Part B time operation at maximum efficiency
• Making coolant when least expensive, if night storage
A + 0.03" B + 0.03" is available
• Making load allocation on the basis of maximizing
c + 0.1"
operating efficiency
30° 30°
0.2" 0.1" 0.2" • Initiating maintenance if a drop in efficiency is
0.5" c + 0.5" detected

FIG. 8.13dd
As shown in Figure 8.13ee, when using variable-speed
A double-seated steam governor valve can be rebuilt to provide compressors, either a reduction in speed or a reduction in
the required rangeability for optimized variable-speed service, if condenser water temperature will reduce the power consump-
two notched rings are installed in the valve seat to modify its tion. For example, when the design load is 50% and the
4
characteristics. condenser water temperature is 55°F, the power consumption

© 2006 by Béla Lipták

1752 Control and Optimization of Unit Operations
can be less than 20%. This compares with 53% if guide vanes
are used on a constant-speed machine operating with 85°F
condenser water.
The operating costs of chiller stations that operate
throughout the year and are located in the northern regions
3
can be cut in half if the potentials of all of the optimization
strategies that were discussed in this section are fully
exploited. The payback period on investment for chiller opti-
mization controls is usually under 1 year.
With the availability of inexpensive solid-state sensors
and microprocessors, it is possible to incorporate all the opti-
mization strategies in a single chiller controller. Such multi-
variable control software makes it possible to replace the
uncoordinated control of flows and temperatures with opti-
mized load following that always corresponds to the lowest
possible cost of operation.
References
1. Lipták, B. G., “Cutting in Half the Cost of Cooling,” Control, November
2003.
2. Lipták, B. G., “Optimizing Controls for Chillers and Heat Pumps,”
Chemical Engineering, October 17, 1983.
3. Romita, E., “A Direct Digital Control for Refrigeration Plant Optimi-
zation,” ASHRAE Transaction, Vol. 83, Part 1, 1977.
4. Lipták, B. G., “Optimizing Plant Chiller Systems,” InTech, September
1977.
5. Shinskey, F. G., Energy Conservation Through Control, New York:
Academic Press, 1978.
6. Lipták, B. G., “Save Energy by Optimizing Your Boilers, Chillers, and
Pumps,” InTech, March 1981.
7. Zimmer, H., “Chiller Control Using On-Line Allocation for Energy
Conservation,” Paper No. 76-522, Advances in Instrumentation, Part
1, ISA/76 International Conference, Houston, TX, 1976.
8. Alsenz, R. H., U.S. Patent Numbers 4,612,776 (dated September 23,
1986) and 4,628,700 (dated December 16, 1986).
9. Carrier Corporation, “Centrifugal Liquid Chillers,” Application Data,
17-IXA, 1975.
Bibliography
ASHRAE Handbook and Product Directory, Systems Volume, Chapter 7,
New York: ASHRAE, Inc.,
© 2006 by Béla Lipták
Backus, A. O., “Energy Savings through Improved Control of Heat Pump
Setback,” ASHRAE Journal, February 1982.
Bower, J. R., “The Economics of Variable-Speed Pumping with Speed Chang-
ing Drives,” Pumpworld, Mountainside, NJ: Worthington Pump.
Brownell, D. L., “Handbook of Applied Thermal Design,” London: Taylor
and Francis, March 1999.
Carrier System Design Manual, Part 7, Syracuse, NY: Carrier Corporation.
Chiller Control, Installation and Operation Manual, Moeller Refrigeration
Products, April 2002.
CIBSE, Building Control Systems, Butterworth-Heinemann, August 2000.
Cooper, K. W., and Erth, R. A., “Centrifugal Water Chilling Systems: Focus
on Off-Design Performance,” Heating/Piping/Air Conditioning, Janu-
ary 1978, p. 63.
Cooper, K. W., and Erth, R. A., “Utilizing Water Chillers Efficiently,” Inter-
national Conference on Energy Use Management, Tucson, AZ, Vol. 1,
New York: Pergamon Press, 1977, p. 233.
Edwards, T. C., “A New Air Conditioning Refrigeration and Heatpump
Cycle,” ASHRAE Transactions, Vol. 84, Part 2, 1978.
Frayne, C., Cooling Water Treatment, Chemical Publishing, January 2001.
Hartman, T., “Ultra-Efficient Cooling with Demand-Based Control,” HPAC
Engineering, December 2001.
Lipták, B. G., “Optimizing Plant Chiller Systems,” InTech, September 1977.
Lipták, B. G., “Optimizing Controls for Chillers and Heat Pumps,” Chemical
Engineering, October 17, 1983.
Lipták, B. G., “Envelope Optimization,” Control, November 1998.
Lipták, B. G., “Cutting in Half the Cost of Cooling,” Control, November 2003.
Luyben, W. L., Process Modeling, Simulation, and Control, New York:
McGraw-Hill, 1989
McMillan, G., “Good Tuning: A Pocket Guide,” ISA, June 2000.
Null, H. R., “Heat Pumps in Distillation,” Chemical Engineering Progress,
July 1976.
Polimeros, G., “There Is Energy for Use in Refrigeration Waste Heat,”
Specifying Engineer, January 1980.
Rishel, J. B., “Cut Energy Costs with Improved Controls for Pumping
Systems,” Instruments and Control Systems, August 1976.
Roper, M. A., Energy Efficient Chiller Control, BSRIA Ltd., September
2000.
Schwedler, M., “Multiple Chiller System Design and Control,” TRANE,
2001.
Shinskey, F. G., “Energy Conservation,” Instruments and Control Systems,
January 1977.
Stair, W. S., “Air Conditioning System,” U.S. Patent Numbers 2,715,514
and 2,715,515, issued August 16, 1955.
Turpin, J., “How to Control Chillers,” ES: Engineering Systems, February
2001.
Werden, R. G., “Heat Pump Gets by with Less Energy by Making Ice
Summer and Winter,” Architectural Record, November 1976.
Zimmer, H., “Chiller Control for Energy Conservation,” 1976 ISA Confer-
ence, Reprint No. 76-522, Research Triangle Park, NC: Instrument
Society of America, 1976.