Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Diagnostics and Optimization of Reciprocating Chillers: Theory and Experiment

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Applied Thermal Engineering Vol. 17, No. 3, pp.

263-276, 1997
Copyright 0 1996 Elsevier Science Ltd
Pergamon Printed in Great Britain All rights reserved
PII: S1359-4311(%)00031-2 1359-4311/97 $17.00 + 0.00

DIAGNOSTICS AND OPTIMIZATION OF


RECIPROCATING CHILLERS: THEORY AND
EXPERIMENT

K. C. Ng,* H. T. Chua,* W. Ong,* S. S. Lee* and J. M. Gordont$§


*Department of Mechanical and Production Engineering, National University of Singapore, 10 Kent
Ridge Crescent, Singapore 0511; tCenter for Energy and Environmental Physics, Jacob Blaustein
Institute for Desert Research, Ben-Gurion University of the Negev, Sede Boqer Campus 84990,
Israel; and IThe Pearlstone Center for Aeronautical Engineering Studies, Department of Mechanical
Engineering, Ben-Gurion University of the Negev, Beersheva, 84105, Israel

(Received 14 June 1996)

Abstract-We develop a simple analytic diagnostic model for reciprocating chillers. With only a handful
of non-intrusive, in situ measurements, one can then ascertain quantitatively how chiller performance
changes with time or after a prescribed modification. We derive how reciprocating chillers can be
characterized by just three parameters with clear physical significance. We then verify the correspondence
between theory and reality with detailed experimental measurements. It is also demonstrated how this
model can be used to establish optimal operating conditions for reciprocating chillers, and to evaluate
potential improvements that stem from changes in operating conditions or the distribution of heat
exchanger inventory. Again, comparisons with actual performance data from commercial chillers are
provided. We give quantitative expression to the contribution to chiller efficiency of internal dissipation
from compression and throttling, and consequently reinforce the fact that endoreversible chiller models
are far off the mark. Copyright 0 1996 Elsevier Science Ltd.

Keywords-Thermodynamic modeling, reciprocating chiller, chiller diagnostics, chiller performance.

NOMENCLATURE
c coolant specific heat (kJ Km ’kg-‘)
c heat exchanger thermal conductance = mcE (kW K ‘)
COP coefficient of performance
E heat exchanger effectiveness
Q cycle-average heat transfer rate (kW)
R total thermal resistance of heat exchangers (K kW- ‘)
T temperature (K)
A& total internal entropy production (kW K-‘)

Subscripts and superscripts


camp compressor
cond condenser
eqv equivalent (heat leak)
evap evaporator
coolant inlet
;zak heat leak
out coolant outlet
total sum for both heat exchangers

1. INTRODUCTION

Say you need to perform a non-intrusive, in situ diagnostic analysis of an installed, operating
reciprocating chiller. Of interest could be quantifying and determining a source of performance
degradation, or simply examining how performance characteristics evolve with the months and
which chiller components are contributing toward those changes. In addition, this information
should be ascertained from just a handful of judiciously chosen measurements, rather than with
a large number of time-consuming test procedures. Our objective here is to develop a simple
analytic thermodynamic model that can address these issues for researchers and practising

§Author to whom correspondence should be addressed.

263
264 K. C. Ng et al.

engineers, and to validate the model against detailed experimental data from commercial
reciprocating chillers for a broad range of operating conditions.
Reciprocating chillers dominate the refrigeration and air-conditioning industries, particularly for
medium-size installations of cooling capacities up to around 1000 kW. Owing to their versatility
and durability, reciprocating chillers are subjected to operating conditions where the ranges of

coolant in Q cond coolant

leak
condenser Q cond

leak leak
Q cond Qcamp

x throttling
valve

t I I b loss
IQ evap

I- 7
coolant out Q evap coolant in

Fig. 1. Schematic diagram of chiller, showing the direction of flow of refrigerant, coolants and energy.
Diagnostics and optimization of reciprocating chillers 265

CHILLER CHARACTERISTIC PLOT (NON-QUANTITATIVE)

non-linear region:
dominated by external
(finite-rate heat transfer) losses

dominated by internal losses


e.g., fluid friction during
compression and throttling

point of maximum COP

l/(cooling rate)
Fig. 2. Characteristic chiller performance curve (non-quantitative).

coolant temperature at the evaporator inlet T&, and condenser inlet Tznd are typically 5-l 8°C and
2445°C respectively. Figure 1 is a schematic of a reciprocating chiller.
A simple universal thermodynamic model for chillers in general, primary of which is
reciprocating chillers, was developed in refs [14]. Offering both a predictive and diagnostic
capability, this approach characterizes chiller performance in terms of three lumped empirical
parameters. The shortcoming of the procedure is that because these parameters represent the
combined effects of several noticeably different irreversibilities, one cannot attach a clear physical
significance to them.
Here we submit an alternative model of the chiller characteristic performance curve, more in the
spirit of the irreversible thermodynamic model developed for chiller optimization in ref. [5]. The
chiller can still be characterized by only three parameters, but parameters the physical significance
of which is clear. Therefore, monitoring changes in these parameters provides a clear indication
of where performance degradation is occurring, and how much each key loss mechanism is
contributing. All this while retaining a non-intrusive experimental procedure.
Validation of such a model requires detailed component-by-component chiller performance data,
the type and accuracy of which are not provided by manufacturers and are rarely reported in the
literature. We establish the model’s validity from measurements in our own chiller laboratory and
from a publication [6] that reports enough information to test the model. These data sets pertain
to constant coolant flow rates in the chiller’s heat exchangers-a situation typical to many
installations.
As shown in refs [l-5] and as derived below, chiller performance is conveniently plotted as l/COP
against I/(cooling rate) (see Fig. 2) where the coefficient of performance (COP) is defined as the
cooling rate (at the evaporator) divided by the input power to the compressor. The external
irreversibilities of finite-rate heat transfer at the condenser and evaporator heat exchangers
dominate COP at relatively high cooling capacities; whereas internal irreversibilities linked to fluid
friction primarily in the compression and throttling stages are the bottleneck at low cooling rates.
The opposite trends of these loss mechanisms imply a point of maximum COP. One would imagine
that chiller manufacturers would aspire to constructions that achieve maximum cooling capacity
266 K. C. Ng et al.

in the vicinity of the maximum COP point. As demonstrated in ref. [5], empirical chiller technology
appears to have evolved this way.
Most finite-time thermodynamic models developed for chillers ignore internal losses, i.e. the
chiller is viewed as an endoreversible system [7-161. Using experimental data, Chua et al. [5]
demonstrated that these endoreversible models not only fail even to come close to predicting chiller
COP values, but they cannot capture the key feature of the existence of a maximum COP point
under realistic operating conditions. In none of these instances was quantitative comparison with
chiller performance data attempted. With the detailed experimental data presented here, we will
reinforce the shortcomings of endoreversible chiller models, and quantify the significance of
omitting internal losses in the modeling procedure.
In the second half of this article, we will illustrate quantitatively how our thermodynamic model
can be used to: (1) determine optimal chiller operating conditions; and (2) evaluate the potential
improvement in COP solely from modifications in cooling capacity or the distribution of heat
exchanger inventory.

2. REVIEW OF THE THERMODYNAMIC MODEL

Consider the steady-state operation of a cyclic chiller that operates with constant flow rates. In
the derivation below, refrigerant temperature at the evaporator i&, and condenser Z&, are
process-average values, and all energy and entropy flows are cyclic-average values. From the First
Law, the energy balance on the refrigerant is

+ Q% - Qevap- Q%p - W + Q&


Qconc, = 0, (1)
where Q = heat transfer rate; W = electrical power input to the compressor; and the superscript
‘leak’ denotes heat leaks to or from the environment. All energy flows are defined as positive.
From the Second Law, the entropy balance on the refrigerant is

where A& is the total internal entropy production due to fluid friction (e.g. non-isentropic
compression and throttling) and other internal dissipative losses. The external losses are
predominantly due to finite-rate heat transfer at the condenser and evaporator heat exchangers:

Qcond = (~~‘%nd(~ond - T&d) = Ccand(%d - T:nd) (3)

and

Qevap= bWevap(T:ap- LqJ = C’evap(T&


- L,) 3 (4)

where c = coolant specific heat; E = heat exchanger effectiveness; the superscript ‘in’ denotes the
coolant inlet temperature; temperatures without a superscript refer to the refrigerant; and C denotes
the heat exchanger’s effective thermal conductance. Should evaporator coolant outlet (rather than
inlet) temperature be the measured variable, then equation (4) should be modified to

Combining equations (l)-(4), and defining chiller COP as QcV.,/W, we express COP as a function
of cooling capacity and the more convenient coolant (rather than refrigerant) temperatures:

- 4 - F2
(Q$ Qwa,,
2+
l/Cop = _ 1+ (6)
- +F4
Qevap
Diagnostics and optimization of reciprocating chillers 261

where the chiller variables F,, F,, F, and F4 are given by

F, = ~~apCevap{Ccon~(Q~~
- Qk=&) - <Qk$ - QEp - Q%>A&}

- CevapCcon~
XLi~L~& - Ce,,,Q%O’&&vap + Q$; - Q& - Qktd
= ~~a,Cev.,,Cc,,,<Q~p- QL%p>- CevapCcandEndT:apAST
- CevapQ2;T2n&‘evap (7)
6 = A&(Qk; - Q& - Qk!iJ + Cc,,(Q~& - Qk;)

- Ceva,{QEp- Q%p - Qki> - ~kKconc4Cevap


- AS,) = ThLondCevap (8)
6 = Ceva,{Q~~ - KkdCcond- AS,)} = - Ce&onJL (9)
F4 = Gap + Ccond- AS, = Gap + Ccond
. (10)
The approximations in equations (7)-(10) are based upon actual measured values of the individual
terms for reciprocating chillers. Invoking these approximate relations, and treating the chiller heat
leaks as approximately constant, we obtain the key chiller performance equation:

T&A& + QkW%, - T&J + RQwa


=1+ ,kidfl + A]. (11)
Q cvap %dQevap

It is convenient to plot equation (11) as l/COP against l/Qe_ (Fig. 2) since the regime of cooling
rates below the maximum-COP point is then linear, with a slope dominated by internal losses AS,.
Equation (11) indicates that the chiller can be characterized in terms of three parameters:

(a) total internal entropy production AS,;


(b) total heat exchanger thermal resistance:

R = (l/ccond) + (l/ceva,>; (12)


(c) an equivalent heat leak:

(13)

The particular combination of variables chosen to define Qeq,, Ieat in equation (13) results in Q$
not being small relative to chiller cooling capacity. However, the heat-leak term in equation (11)
is typically an order of magnitude less than the other terms: small but not negligible for accurate
modeling. We find that accurate and physically meaningful parameter determination requires
retention of the heat-leak term, rather than treating heat leaks as zero and simply adopting a
two-parameter model. Furthermore, while our measured heat leaks are relatively small, it should
be kept in mind that the chiller we studied was new. For installed chillers that have degraded, or
in chillers with poor insulation, heat leaks will of course be greater, and this will be reflected in
larger best-fit values of Q&!.
The expression Q$.! contains a slight dependence on coolant temperatures; but for properly
operating commercial reciprocating chillers, this dependence exerts a small influence on COP. In
fact, R and AS, are not rigorously constant either, with a mild temperature dependence. However,
as evidenced below from detailed experimental data, adopting constant effectively average values
for these three chiller parameters (AS,, R and Qet) results in chiller performance curves the
accuracy of which is better than experimental uncertainty.

3. CHILLER OPTIMIZATION

Before proceeding to the description and analysis of the experimental results, we would like to
offer some observations on the implications of our simple thermodynamic model for chiller
optimization. Namely, how do practical constraints affect the location and attainment of the point
of maximum COP. In Section 5, we will compare these predictions against experimental realities.
A common and realistic economic limitation in chiller production is fixed heat exchanger
inventory C (i.e. total mcE). Accordingly, consider imposing the constraint

(14)
268 K. C. Ng et al.

Then in determining the chiller’s maximum COP operating conditions, there are two degrees of
freedom, which we select as Qevapand Cevap.Two equations must then be solved simultaneously:
XOP/aQ,,,, = 0 (1%
and
acoP/ac,,., = 0 (16)
(applied to equation (1 l)), and where, in order to ensure that only the physically meaningful
solutions are filtered from all possible mathematical solutions, it is required that

(17)

For experimentally determined values of Tknd, T&,, Ctota,, AS,, Q$,, Q:;$ and Q:Zgp, we can
calculate the values of Qevapand Cevapthat maximize COP and then compare them to the
corresponding measured chiller performance. This comparison, and the consequences for optimal
chiller design, will be presented in Section 5.

4. EXPERIMENTAL DATA

Two independent sets of experimental data are considered. The first is from a commercial
reciprocating chiller tested in our chiller laboratory. A schematic of our test rig is shown in Fig. 3.
It is a water-to-water chiller with a nominal cooling capacity of 10.5 kW. The semi-hermetic
reciprocating compressor has a displacement volume of 26.8 m3h-‘, configured with a coiled-tube
condenser and a similar evaporator. The refrigerant is dichlorodifluoromethane, commonly called
R12. It flows in the inner pipe at the evaporator and in the outer tube at the condenser. A
thermostatic valve is used to control the degree of superheating at the evaporator outlet by
regulating refrigerant flow rate. We equipped the chiller with RTD temperature sensors, calibrated
to an accuracy of + 0.07 K for differential temperature, f 0.05 K for absolute tempera-
ture, + 3 kPa pressure gauge, and + 0.00014 1 s ’for the refrigerant flow meter. The electric power
consumption was measured to an accuracy off 0.01 kW.
Table 1 summarizes 30 sets of our experimental measurements. They span TL, from 8 to 18°C
and T$, from 23.9 to 35”C, in accordance with Air-conditioning Refrigeration Institute (ARI)
Standard 590-86 [17]. Water flow rates at the condenser and evaporator were 0.616 1 s-’ and

ChiIled Pressure Water


Water Return I Gauge I Retum

Valve Valve

Fig. 3. Schematic of our chiller test facility.


Table I. Summary of allexperimental
measurements,
including
experimental
uncertainties,
forthe nominal 10.5kW capacity
water-cooled
reciprocating
chiller
testedin our laboratory

cd (“CL T& (“C), (kiy (kc) I/COP, Q% W’). QZ:p Wh AS, (kW Km'), Q& WV, (mew,, (kW (mc&., (kW
f 0.05 + 0.05 f 0.2 * 0.; 0.01 +0.006 + 0.4 k 0.3 ~0.0003 f 0.01 Km'), f 0.04 Km'), + 0.05
24.09 8.02 13.8 9.8 3.66 0.373 -0.173 0.289 0.00531 0.127 0.840 0.608
23.89 10.00 14.5 10.5 3.71 0.353 0.017 0.358 0.00537 0.082 0.884 0.622
23.81 12.37 15.3 II.4 3.81 0.336 0.043 0.318 0.00539 0.110 0.918 0.643
23.94 13.96 15.6 II.7 3.84 0.329 0.075 0.306 0.00528 0.133 0.905 0.639
23.90 15.99 16.3 12.4 3.90 0.315 0.053 0.191 0.00529 0.118 0.907 0.646
23.88 18.00 16.8 12.9 3.92 0.302 0.044 0.114 0.00520 0.135 0.917 0.649
26.75 8.00 13.7 9.8 3.74 0.380 -0.035 0.227 0.00526 0.152 0.869 0.618
26.75 10.00 14.4 10.5 3.81 0.364 -0.010 0.238 0.00539 0.122 0.900 0.640
26.75 12.41 15.3 II.3 3.94 0.348 0.006 0.230 0.00555 0.130 0.939 0.672
26.67 14.00 15.9 II.9 3.98 0.335 -0.003 0.187 0.00557 0.117 0.953 0.689
26.74 16.01 16.8 12.7 4.10 0.322 0.057 0.156 0.00578 0.100 0.992 0.716
26.67 17.99 17.6 13.4 4.14 0.308 0.105 0.196 0.00576 0.076 1.007 0.734
29.49 8.00 13.3 9.4 3.85 0.408 0.059 0.261 0.00537 0.170 0.817 0.602
29.38 10.00 14.1 IO.1 3.91 0.388 -0.005 0.213 0.00536 0.140 0.853 0.595
29.43 12.39 14.7 10.7 3.98 0.373 0.006 0.219 0.00511 0.195 0.838 0.594
29.43 13.98 15.6 II.4 4.09 0.357 0.081 0.189 0.00567 0.087 0.884 0.659
29.39 15.97 16.3 12.1 4.17 0.345 0.075 0.187 0.00566 0.099 0.907 0.665
29.41 17.99 17.1 12.9 4.22 0.328 0.143 0.147 0.00573 0.034 0.949 0.668
32.19 7.98 13.2 9.2 3.94 0.428 0.014 0.223 0.00543 0.159 0.794 0.581
32.24 10.00 14.1 9.9 4.04 0.406 -0.013 0.212 0.00556 0.147 0.831 0.631
32.22 12.40 15.0 10.8 4.16 0.386 -0.108 0.162 0.00551 0.191 0.885 0.644
32.18 13.99 15.6 II.3 4.25 0.375 -0.025 0.162 0.00566 0.183 0.895 0.673
32.22 16.01 16.4 12.1 4.32 0.357 0.000 0.140 0.00578 0.122 0.922 0.695
32.26 18.02 17.2 12.8 4.43 0.347 0.026 0.190 0.00579 0.131 0.928 0.711
34.99 8.01 13.0 8.9 4.01 0.450 -0.174 0.148 0.00527 0.215 0.722 0.577
34.88 10.00 13.7 9.5 4.10 0.429 -0.048 0.178 0.00540 0.175 0.750 0.603
35.01 12.40 14.8 10.4 4.26 0.408 -0.075 0.172 0.00565 0.145 0.818 0.633
34.97 13.99 15.4 II.0 4.36 0.396 -0.042 0.172 0.00575 0.152 0.852 0.648
35.05 15.98 16.2 II.7 4.45 0.379 -0.093 0.088 0.00581 0.150 0.895 0.667
34.99 17.99 16.9 12.5 4.55 0.366 0.045 0.099 0.00590 0.154 0.901 0.702
K. C. Ng et al.

3.0-
4’
, /%
k
A
/
&
experimental error = 2.2%
P
2.5-
/Y model prediction RMS error = 0.894
/A’
.A’
/
/

2.0 //‘I I I I I I 1 I I I I I I
2.0 2.5 3.0 3.5
COP (measured)
Fig. 4. Comparison between model-predicted COP and corresponding experimentally measured values for
the nominal IO.5 kW capacity chiller tested in our laboratory.

0.461 1 s - ‘, respectively, and were maintained to within f 1%. Specific entropy and enthalpy
values were computed with computational procedures recommended by ASHRAE [ 181,based upon
the simultaneous temperature and pressure measurements. All uncertainty analyses were performed
according to ASHRAE Standard 413-75 [19]. Our root-mean-square (RMS) error for determining
COP turns out to be 2.2%.
The second data set was taken from ref. [6], with a total of 60 experimental points for a chiller
of 70.4 kW nominal cooling capacity. These measurements cover a similar range of evaporator and
condenser inlet coolant temperatures. Detailed measurements reported are not as extensive as those
we performed, so that experimental validation of our thermodynamic model is more limited with
this data set than with our own laboratory measurements. The RMS experimental error for
determining COP with this second data set is 5%. The noticeably higher RMS error probably stems
from non-steady-state measurements and/or poorer instrument accuracy.

5. TESTING THE DIAGNOSTIC POWER OF THE THERMODYNAMIC MODEL

Two exercises are performed here. First, we need to ascertain how accurately the simple
three-parameter model of equation (11) can fit actual chiller performance data. This will establish
the degree to which the model can be employed as a predictive and diagnostic tool.
Second, in order to demonstrate that the model parameters truly correspond to the physical
variables which they purport to represent, we will compare the best-fit values of these three
parameters from strictly statistical regression fits against their corresponding experimentally
Diagnostics and optimization of reciprocating chillers 271

measured values. This will provide convincing evidence that the accurate fits are neither fortuitous
nor for parameters the physical significance of which is misidentified.
For the two experimental data sets delineated in Section 4, we regressed for the three chiller
characteristic parameters, to obtain:
AS, = 0.00555 kW K- ‘,
R = 2.505 K kW-‘,
Q$ = 4.38 kW,
for our 10.5 kW capacity chiller and
AS, = 0.0366 kW K- ‘,
R = 0.127 K kW-‘,
Q&! = 26.1 kW,
for the 70.4 kW capacity chiller of Leverenz and Bergen [6]
How well do these best-fit values account for chiller COP? The answer is provided by Figs 4
and 5. For our 10.5 kW capacity chiller, the RMS error in predicting COP is 0.9%, which is well
below the experimental uncertainty of 2.2%. For the 70.4 kW capacity chiller of Leverenz and
Bergen [6], the RMS error in predicted COP is 1.9%, which is comfortably less than the
experimental error of 5%.
Although the data reported in ref. [6] are not sufficiently detailed to permit experimental
determination of these three parameters at each chiller operating point, the data measured in our

4.5

//
//
///
1 /

4.0-
1 P
/
/A
A

d
3.5-
A a
A4

3.0-
yx
A t
experimental error = 5%

model prediction RMS error = 1.924

2.5 1’1 ’ ’ ’ 1 ’ ’ ’ ’ 1 ’ ’ ’ ’ 1 1 ’ 1 1 1
2.5 3.0 3.5 4.0 4.5
COP (measured)
Fig. 5. Comparison between model-predicted COP and corresponding experimentally measured values for
the nominal 70.4 kW capacity chiller described in ref. [6].
272 K. C. Ng et al.

Table 2. Exoerimentallv determined values of the three chiller oarameters at each chiller oueratinr! point

TLd°C) T&W) R (K kW-‘) AS, (kW K-‘) QZ WV


24.09 8.02 2.84 0.00531 2.51
23.89 10.00 2.74 0.00537 2.03
23.81 12.37 2.64 0.00539 3.06
23.94 13.96 2.67 0.00528 4.13
23.90 15.99 2.65 0.00529 4.50
23.88 18.00 2.63 0.00520 6.80
26.75 8.00 2.77 0.00526 2.51
26.75 10.00 2.67 0.00539 2.30
26.75 12.41 2.55 0.00555 2.82
26.67 14.00 2.50 0.00557 2.84
26.74 16.01 2.40 0.00578 2.85
26.67 Il.99 2.36 0.00576 2.75
29.49 8.00 2.89 0.00537 2.49
29.38 10.00 2.85 0.00536 2.26
29.43 12.39 2.88 0.00511 3.49
29.43 13.98 2.65 0.00567 1.81
29.39 15.97 2.61 0.00566 2.32
29.41 Il.99 2.55 0.00573 1.01
32.19 7.98 2.98 0.00543 2.07
32.24 10.00 2.79 0.00556 2.08
32.22 12.40 2.68 0.00551 2.91
32.18 13.99 2.60 0.00566 3.05
32.22 16.01 2.52 0.00578 2.32
32.26 18.02 2.48 0.00579 2.87
34.99 8.01 3.12 0.00527 2.39
34.88 10.00 2.99 0.00540 2.17
35.01 12.40 2.80 0.00565 2.00
34.97 13.99 2.72 0.00575 2.25
35.05 15.98 2.62 0.00581 2.36
34.99 17.99 2.53 0.00590 2.74
Corresponding model values obtained by regression: R = 2.505 K kW -‘, AS, = 0.00555 kW K-‘, Q$,! = 4.38 kW.

laboratory are. The measured parameter values are summarized in Table 2. The experimental
uncertainty for AS, is 0.0003 kW K-‘. The relative uncertainty in the heat exchanger parameter
R is 5%, with a maximum absolute uncertainty of 0.16 K kW - ‘. For the heat leak parameter Qg:,
the absolute uncertainty is 0.4 kW, with a maximum absolute uncertainty of 0.5 kW. The relevant
comparison here is between the regressed values of the three model parameters noted above and
the experimental values listed in Table 2.
Also keep in mind that the dominant contributions to COP are from the AS, and R terms, with
contributions of a few percent at most from the heat leak term-a point that is consistent with
earlier findings on the small-to-negligible contributions of heat leaks to properly operating chillers
of this type [5]. The combination of highly accurate fits for COP and of confirmation of the physical
significance of the formulated parameters attests to the robustness of the diagnostic model.

6. THE RELATION OF MODEL PREDICTIONS TO CHILLER OPTIMIZATION

Commercially produced reciprocating chillers appear to have evolved such that their maximum
cooling capacity operation, and the division of their heat exchanger inventory between evaporator
and condenser, are at, or close to, maximum COP conditions [5]. Examination of optimal chiller
design here is restricted to our own data set, rather than to that of ref. [6], too, due to a dearth
of detailed pressure, temperature and refrigerant flow rate measurements for the latter.
Consider two chiller rating points for this discussion: (1) a typical rating point for normal chiller
operation; and (2) test conditions where the throttle valve at the inlet to the compressor is partially
shut, such that refrigerant flow rate is reduced to about 75% of the nominal flow rate. The latter
reduces internal dissipative losses.
Measured and computed results are listed in Table 3 and Table 4 for both rating conditions.
Both tests have the same Z$,, = 18.O”C, but a slightly different TEndof 29.4 and 30.4”C. The
reduced refrigerant flow rate results in lower heat exchanger thermal conductances (and hence an
increased value of R) due to reduced heat exchanger effectiveness. It also lessens internal dissipation
AS,. Indeed, based on pressure and temperature measurements, witness the change in AS, from
the nominal normal rating value of 0.00573 to 0.00203 kW K- ‘.
A highly unrealistic, theoretical extreme one could consider is that of vanishing AS-the
Diagnostics and optimization of reciprocating chillers 273

Table 3. Comparison of measured chiller variables with model predictions of maximum-COP conditions, for normal rating and throttled
(reduced refrigerant flow rate) configurations. Model results for the endoreversible limit of vanishing internal losses are also listed.
Exncrimentallv measured chiller variables
Chiller variable Normal rating test Rating test with throttled flow rate
To., (“C) 29.4 30.4
X, (“C) 18.0 18.0
Refrigerant flow rate (kg ss’) 0.1009 0.0766
‘Go,.,(kW K - ‘) 1.619 1.347
A& (kW K - ‘) 0.00573 0.00203

endoreversible limit. We say ‘unrealistic’ since either the cooling capacity would be orders of
magnitude too small at the desired level of internal dissipation, or the predictions of COP at actual
cooling rates would be gross overestimates. As an illustration of the latter point, Table 3 and
Table 4 also lists the COP value arrived at by adopting the endoreversible assumption of AS, = 0
(while retaining the actual heat leak contributions which, in some endoreversible models, are
excluded too). The endoreversible COP predictions are around a factor of two larger than actual
chiller COPS-an indication that about 50% of the physics of the problem has been omitted.
Furthermore, endoreversible models predict that maximum COP is reached only in the limit of
vanishing cooling rate. Hence, they miss the key qualitative and practical feature of a maximum
COP point at considerable cooling capacity (in fact, the cooling capacity for which commercial
chillers are invariably designed), and the fundamental competition between internal and external
losses in dictating actual chiller performance. A considerable number of journal articles have
analyzed endoreversible chiller models [7-161, regrettably without any attempt to compare
predictions against actual performance or experimental data.
Figures 6 and 7 are plots of our model’s predictions for l/COP against l/Qcvaprfor normal rating
and for reduced-flow-rate conditions, respectively. Both Figs 6 and 7 show for two values of Cevap:
the value used in the actual installed chiller, and the value that we calculate would correspond to
globally maximum COP at that particular refrigerant flow rate. Each graph also includes the
corresponding experimental point. One would expect that at nominal normal rating (Fig. 6), the
actual design operating point would be close to the theoretical optimum-an expectation that is
seen to be borne out to within 1%. The empirical wisdom embodied in chiller design is also attested
to by the relatively broad optimum with respect to both cooling capacity and heat exchanger
thermal conductance. The lower dotted curves in Figs 6 and 7 correspond to the nominal
endoreversible limit of vanishing internal losses (AS, = 0) at the actual installed value of Cevap.
When chillers are driven under throttled conditions (reduced refrigerant flow rate), the actual
operating point, and the maximum-COP point for that particular reduced flow rate, can be
markedly different. This is to be expected since manufacturers will invariably aim to attain
maximum COP at maximum rated cooling capacity rather than at reduced or part-load capacities.
This point is graphically clear from Fig. 7. The actual chiller COP under these conditions is 2.56.
Were the chiller optimized for these reduced-flow conditions, its maximum COP would be 3.46.
Understandably, however, chillers are typically optimized for full-flow maximum-rated cooling
capacity.
The nominally optimal COP can be higher at lower flow rates because throttling reduces internal
losses. This occurs because throttling significantly reduces refrigerant flow rate, and total entropy
production (not to be confused with entropy production per unit mass) decreases with refrigerant
flow rate. (The relative contribution of heat leaks increases slightly under throttled conditions but
still has only a small effect on COP.)

Table 4. Comparison of measured chiller variables with model predictions of maximum-COP conditions, for normal rating and throttled
(reduced refrigerant flow rate) configurations. Model results for the endoreversible limit of vanishing internal losses are also listed.Measured
chiller variables and model predictions for maximum-COP conditions
Normal rating test Rating test with throttled flow rate
Chiller variable Measured Model prediction Endoreversible value* Measured Model prediction Endoreversible value*
Q..., WV 12.9 12.6 9.43 9.09
C,, (kW K - ‘) 0.668 0.761 0.237 0.646
COP 3.05 3.09 5.83 2.56 3.46 4.67
*Since the endoreversible model predicts maximum COP at zero cooling rate, this column simply evaluates COP at the actual chiller
cooling rate and evaporator thermal conductance, for the model equation with A& = 0.

ATE 1713-C
274 K. C. Ng et al.

NORMAL OPERATING CONDITIONS

globally optimal operation (broken curve)

. .
endoreversible model ___. - . _

0.0 1 , I 1 I
I I I I I
I I I I I

0.00 0.05 0.10 0.15


l/(cooling rate) (kW-‘)
Fig. 6. Characteristic chiller plot of l/COP against l/Q...,. The upper two curves are model predictions,
at normal rated capacity conditions with T:“,, = 29.4”C and T:“,,, = 18.O”C, for two values of C,,,,: solid
curve = at actual installed C,,,, = 0.668 kW K- ‘; broken curve = at C,,,, = 0.761 kW K- ‘, where the
model predicts a global maximum for COP. The point indicated by a solid triangle represents measured
chiller performance. Note the near coincidence of the predicted global optimum and actual chiller design
operation. The lower dotted curve corresponds to the nominal endoreversible limit of vanishing internal
losses (A& = 0), at the actual installed C,,,, value of 0.668 kW K-‘.

7. SUMMARY

To a large extent, chiller diagnostics and optimization can be performed with a simple analytic
thermodynamic model in which chiller performance is characterized by three physically meaningful
parameters. Using two independent sources of experimental data for commercial reciprocating
chillers, we have demonstrated: (1) good predictive accuracy for COP; and (2) that the three
parameters regressed from actual performance data agree well with independent experimental
determination of the actual physical variables to which they are purported to correspond.
Observations based upon the model and the experimental data reveal that external (heat
exchanger) and internal (fluid friction and throttling) dissipative losses each contribute very roughly
on a Xl-50 basis to chiller COP, with a comparatively small contribution from heat leaks. Here
we refer to the magnitudes of the individual terms in the governing chiller performance equation
(11). Our thermodynamic model partitions the contributions into separate heat exchanger, internal
dissipation and heat leak terms. Endoreversible chiller models miss around 50% of the physics of
the problem and emerge with predictions of COP that are about a factor of two in excess of reality.
Furthermore, endoreversible models fail to account for the existence of a maximum-COP point
at considerable cooling capacities, and the associated consequences for optimal chiller design.
Our thermodynamic model accounts accurately for the optimal (maximum-COP) configurations
of commercial reciprocating chillers, i.e. the cooling capacities and the division of typically fixed
heat exchanger inventory (total mcE value) at which they are designed to operate. We have
observed the evolved empirical wisdom in the chiller industry that the maximum-COP point
corresponds to normal rated capacity. We also have quantified the extent to which COP must be
compromised under throttled (reduced refrigerant flow rate) conditions, and what the nominally
optimal chiller configuration would be under such conditions.
Diagnostics and optimization of reciprocating chillers 275

THROTTLED OPERATING CONDITIONS

\
0.6

0.5 model predictions for


actual operation (solid measure4l operating point
6 0.4 globally optimal operation

u
2 \
\
0.3 --___----

3 . - .
. . __
. . - - . _
- _. _
endoreversible model

0.0 7 I 1 1 I
I I I I I
I I I I I

0.00 0.05 0.10 0.15


l/(cooling rate) (kW- ‘)
Fig. 7. As in Fig. 6, but throttled to 75% of the nominal refrigerant flow rate so as to reduce internal
dissipative losses, with T&, = 30.4”C and 7’& = 18.O”C. The uppermost (solid) curve and measured
operating point pertain to the actual installed chiller evaporator thermal conductance for this reduced flow
rate, &.r = 0.237 kW K- ‘. The middle (broken) curve was generated for Cc..r = 0.646 kW K- ‘, where
the model predicts a global maximum for COP at this reduced flow rate.

Acknowledgement-J. M. Gordon gratefully acknowledges the generous hospitality of the Department of Mechanical and
Production Engineering at the National University of Singapore during part of the period of this research.

8. REFERENCES

1. J. M. Gordon and K. C. Ng, Thermodynamic modelling of reciprocating chillers. J. Appl. Phys. 75, 2769-2774
(1994).
2. J. M. Gordon and K. C. Ng, A general thermodynamic model for absorption chillers. Heaf Recouery System & CHP
15, 73-83 (1995).
3. J. M. Gordon and K. C. Ng, Predictive and diagnostic aspects of a universal thermodynamic model for chillers. Znt.
J. Heat Muss Transfer 38, 807-818 (1995).
4. J. M. Gordon, K. C. Ng and H. T. Chua, Centrifugal chillers: thermodynamic modelling and a diagnostic case study.
Int. J. Refrig. 18, 253-257 (1995).
5. H. T. Chua, K. C. Ng and J. M. Gordon, Experimental study of the fundamental properties of reciprocating chillers
and its relation to thermodynamic modelling and chiller design. In?. J. Heat Muss Transfer (in press).
6. D. J. Leverenz and N. E. Bergan, Development and validation of a reciprocating chiller model for hourly energy analysis
programs. ASHRAE Trans. 89(1A), 156174 (1983).
7. Z. Yan and J. Chen, An optimal endoreversible three-heat-source refrigerator. J. Appl. Phys. 65, l-4 (1989).
8. J. Chen and Y. Zan, Unified description of endoreversible cycles. Phys. Rev. A 39, 4140-4147 (1989).
9. D. C. Agrawal and V. J. Menon, Performance of a Camot refrigerator at maximum cooling power. J. Phys. A 23,
5319-5326 (1990).
10. Z. Yan and J. Chen, A class of irreversible Carnot refrigeration cycles with a general heat transfer law. J. Phys. D
23, 136141 (1990).
Il. A. Bejan, A theory of heat transfer: irreversible refrigeration plant. Inr. J. Heat Mass Transfer 32, 1631-1639
(1989).
12. C. Wu, Cooling capacity optimization of a geothermal absorption refrigeration cycle. Int. J. Ambient Energy 13,
136-141 (1992).
13. S. A. Klein, Design considerations for refrigeration cycles. Znf. J. Refrig. 15, 181-185 (1992).
14. C. Wu, Specific heating load of an endoreversible Carnot heat pump. J. Ambient Energy 14, 25-28 (1993).
15. C. Wu, Specific heating load of an endoreversible Carnot heat pump. Int. J. Ambient Energy 14, 77-82 (1993).
16. N. E. Wijeysundera, Analysis of ideal absorption cycle with external irreversibilities. Energy 20, 123-130
(1995).
276 K. C. Ng et al.

17. Air-conditioning Refrigeration Institute, AR1 Standard 590, Standard for reciprocating water-chilling packages. ARI,
Arlington, VA (1986).
18. ASHRAE, Thermodynamic Properties of Refrigeranfs. The American Society of Heating, Refrigerating and
Air-conditioning Engineers, Inc., Washington (1986).
19. ASHRAE, Standard 413-75, Standard measurement guide: engineering analysis of experimental data. The American
Society of Heating, Refrigerating and Air-conditioning Engineers, Inc., Washington (1975).

You might also like