Sunteți pe pagina 1din 24

Energy Conversion & Management 41 (2000) 677±700

www.elsevier.com/locate/enconman

Comparison of three transient methods for testing solar


¯at-plate collectors
J.K. Nayak a,*, E.H. Amer b, S.M. Deshpande a
a
Department of Mechanical Engineering, Indian Institute of Technology, Bombay, Powai, Mumbai 400 076, India
b
Department of Mechanical Engineering, Faculty of Engineering, Menou®a University, Egypt
Received 22 January 1999; accepted 17 August 1999

Abstract

This communication presents an extensive comparison of three transient methods (Perers, DSC and
Wijeysundera) for testing solar ¯at-plate collectors. Results from di€erent tests are compared with the
steady state values obtained on the basis of the ASHRAE 93-86 standard. Based on the experimental
results, it is seen that Perers' method yields parameters closer to the steady state values. The sensitivity
analysis shows that the results of Perers and DSC methods are greatly a€ected by measurement errors;
whereas the Wijeysundera method is less sensitive. Comparison of the prediction of outlet temperature
with measurements for the Perers and DSC methods show that the values based on the later method are
closer to the measurements. # 2000 Elsevier Science Ltd. All rights reserved.

Keywords: Transient methods; Dynamic testing; Flat-plate collectors

1. Introduction

Several dynamic methods have been reported in the literature for outdoor testing of solar
¯at-plate collectors. These methods have been developed because of the fact that the
conventional steady state methods are not always applicable due to the unavailability of
required ambient conditions, especially in tropical regions. The main feature of these methods
is that they enable one to test the collector under variable weather conditions. Thus, the total

* Corresponding author. Tel.: +91-22-578-2545; fax: +91-22-578-3480.


E-mail address: jknayak@me.iitb.ernet.in (J.K. Nayak).

0196-8904/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 9 6 - 8 9 0 4 ( 9 9 ) 0 0 1 4 2 - 9
678 J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700

Nomenclature

Ap absorber area (m2)


Cc e€ective heat capacity of collector per unit area (kJ mÿ2 Kÿ1)
Ct e€ective heat capacity of storage tank (kJ Kÿ1)
cp speci®c heat of ¯uid (J kgÿ1 Kÿ1)
E electric heater power (W)
FR collector heat removal factor
F' collector eciency factor
Ialb re¯ected component of radiation on collector plane (W mÿ2)
Ib beam radiation on collector plane (W mÿ2)
Id di€use radiation on collector plane (W mÿ2)
IT global radiation on collector plane (W mÿ2)
Ktab incidence angle modi®er for beam radiation
Ktad incidence angle modi®er for di€use radiation
m_ mass ¯ow rate (kg sÿ1)
N number of segments
qu rate of useful energy collected per unit area (W mÿ2)
Ta ambient temperature (8C)
T® ¯uid inlet temperature (8C)
Tfm ¯uid mean temperature (8C)
Tfo ¯uid outlet temperature (8C)
Ti temperature of water at inlet of pipe (8C)
To temperature of water at outlet of pipe (8C)
Ts temperature of water inside tank (8C)
Tsky sky temperature (8C)
t time (s)
UA product of heat loss coecient and e€ective area for pipes or tank (W Kÿ1)
UL overall collector heat loss coecient (W mÿ2 Kÿ1)
Up pipe loss coecient (W mÿ2 Kÿ1)
Usky equivalent radiative heat transfer coecient from ¯uid to sky (W mÿ2 Kÿ1)
Uv heat loss coecient due to wind speed (kJ mÿ3 Kÿ1)
U1, U2 ®rst and second order heat loss coecients (W mÿ2 Kÿ1), (W mÿ2 Kÿ2)
v wind speed (m sÿ1)
v_ volume ¯ow rate (lpm)
W Tfm ÿ Ta (K)
DTsky Tfm ÿ Tsky (K)
2 heat exchanger factor
Z thermal eciency
y angle of incidence (degree)
…ta†e e€ective transmittance absorptance product
J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700 679

Subscripts
i inlet pipe
n number of the segment
o outlet pipe
t tank

time available for outdoor testing is greatly increased. One of the earliest methods was
developed by Rogers [1], which enables one to obtain steady state collector parameters from
dynamic tests. Emery and Rogers [2] have implemented this method and the British Standard
Institution has adopted it as a standard for transient testing [3]. It however, requires the
handling of large data sets, and consequently, the amount of computation is heavy. Saunier
and Chungpaibulpatana [4] have developed a method, which does not require measurements of
the ¯uid ¯ow rate or temperature rise across the collector thus, reducing the parameters to be
controlled and measured. However, it incorporates inconvenient night time test. Also, the
results from di€erent tests di€er. In a later work, Chungpaibulpatana and Exell [5] have
modi®ed this method by eliminating the night time tests; but two additional tests have been
introduced, and the calculation procedure is quite tedious. Wang et al. [6] have proposed a
transient procedure, called the ®lter method. Though it incorporates simple experimental steps;
it requires familiarity with digital signal analysis and ®ltering. Souproun [7] has proposed a
method, which requires long periods of steady ambient conditions. This situation is rarely
available during varying weather. A relatively simple dynamic method has been developed by
Perers [8]. The distinguishing feature in it is that the incident angle modi®ers for beam, and
di€use solar radiation are considered as variables and, hence, are estimated in the method.
Muschaweck and Spirkl [9] have developed a method using the DSC (Dynamic Solar
Collector) model, which does not necessitate controlling the ¯ow rate or temperature during
testing under non-stationary conditions. Recently, this method has been modi®ed by Spirkl et
al. [10] for in-situ characterisation of collectors. Wijeysundera et al. [11] have developed a new
procedure (called the Wijeysundera method in this paper) that results in estimation of the
values of collector parameters from experimental tests conducted on a complete water heating
system. A detailed survey on transient testing of ¯at-plate collectors and experimental
evaluation of di€erent methods (Saunier and Chungpaibulpatana, Exell and
Chungpaibulpatana, Rogers and the ®lter method) have been reported by Amer et al. [12,13].
Schneiders' [14] has made very interesting theoretical studies and comparisons of one
stationary and ®ve dynamic models by applying them to a vacuum tube collector. The dynamic
models considered are: (i) three-node model of Kamminga [15], (ii) two-node model, a modi®ed
version of Kamminga's model, (iii) DSC model of Muschaweck and Spirkl [9], (iv) ``matched
¯ow collector'' (MFC) model of Isakson and Eriksson [16] and (v) 1-point model of Henning
[17]. The experimental investigations on the implementation of these models and their
parameters' estimations seem to be limited; but whatever has been reported, they refer to
vacuum tube collector.
This communication reports the results of extensive experimental investigations and
comparisons of three dynamic methods: Perers method [8], DSC method [9] and Wijeysundera
680 J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700

method [11]. The methods are also evaluated by comparing their results with those obtained
based on the steady state ASHRAE 93-86 standard [18]. In order to appreciate the results, a
brief theoretical background of each method is described at the outset.

2. Theoretical basis and test procedures

2.1. Perers method

Perers method [8] is based on a lumped capacitance model. It considers separate incident
angle modi®ers for direct and di€use solar radiation. These modi®ers are variables in the
model and are estimated from experimental data. The energy balance equation for the useful
heat gain is expressed as:

qu ˆ F 0 …ta†e Ktab …y †Ib ‡ F 0 …ta†e Ktad Id ÿ F 0 U1 W ÿ F 0 U2 W 2 ÿ F 0 Uv Wv ÿ F 0 Usky DTsky

dTfm
ÿ Cc ÿ Up W …1†
dt
where the collector parameters are F 0 …ta†e , Ktab , Ktad , F 0 U1 , F 0 U2 , F 0 Uv , F 0 Usky , Ce and Up. In
the case of glazed collectors, the terms involving F 0 Uv and F 0 Usky are excluded from Eq. (1).
The value of Ktab can be obtained from the following equation:
 
1
Ktab …y † ˆ 1 ÿ bo ÿ1 …2†
cos y
where bo is a constant.
The heat gain can also be calculated using the relation
_ p …Tfo ÿ Tfi †=Ap
qu ˆ mc …3†

The sum of the squares of the residual errors between the predictions of qu from Eq. (1) and
the measured values of qu using Eq. (3) is obtained. The technique of multiple regression is
used for minimising this error, and the collector parameters are estimated.
The experimental procedure is the same as that for steady state testing except that it involves
long hours of measurements, including some hours of darkness. Measurements of ¯ow rate,
solar radiation (global and di€use in the plane of the collector) and temperatures (inlet, outlet
and ambient) are required to be made. The requirements of this method are that (a) the inlet
¯uid temperature should be controlled over the test run and (b) the ¯uid ¯ow should be
continuous. In order to reduce the random errors in measurements, the author recommends
that the sampling rate should be less than 6 s.

2.2. DSC method

Muschaweck and Spirkl [9] have developed a dynamic parameter identi®cation method for
testing of solar ¯at-plate collectors under transient conditions. It is based on a one-node model
J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700 681

with heat capacities of the absorber plate, risers and ¯uid lumped together and referenced to
the mean temperature of the ¯uid. According to the model, the collector absorber plate, risers
and ¯uid are divided into N-segments along the ¯ow direction. The energy balance for a
typical segment is given by:
Cc dTfm, n 1 0 ÿ 
ˆ F …ta†e IT ÿ F 0 UL Tfm, n ÿ Ta ÿ qu …4†
N dt N
where,
ÿ 
_ p Tfm, n ÿ Tfm, nÿ1
qu ˆ mc …5†

IT ˆ Ktab Ib ‡ Ktad Id ‡ Ktad Ialb …6†

and
ÿ 
UL ˆ U1 ‡ U2 Tfm, n ÿ Ta ‡ Uv v …7†

The method uses a dynamic system testing algorithm DST developed by Spirkl [19] for
parameter identi®cation. The schematic diagram of the DST algorithm is shown in Fig. 1. It
involves the following steps:
1. Choose an initial set of collector parameters for calculating the model prediction,
2. Find the di€erence (called residue) between the measurements and predictions,
3. Subject the residue to a low pass ®lter for eliminating errors such as due to noise,
4. Minimise the root mean square of the ®ltered residue for estimating the parameters along
with their standard errors. The minimisation is done using the Levenberg±Marquardt
Algorithm LMA [20].
The test procedure and the parameters required to be measured are similar to those of the
Perers method. Measurements are taken for several days under variable conditions. The
method does not prescribe controlling either the ¯ow rate or ¯uid inlet temperature.

Fig. 1. DST algorithm.


682 J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700

2.3. Wijeysundera method

Wijeysundera et al. [11] have proposed a test procedure to estimate the daily mean values of
the collector characteristics, FR UL and FR …ta†e from measurements of the variations of the
storage tank temperature of a complete system shown in Fig. 2. The method prescribes that (a)
the tank temperature be maintained in a fully mixed condition, (b) the experiment be
conducted for a long duration of time so that on the average, the dynamic behaviour of the
tank temperature remains signi®cant and that of the collector temperature becomes
insigni®cant.
Energy balance conditions for the complete system, collector and for the inlet pipe from the
storage tank to the collector are expressed as:
Closed system
dTs 
Ct ˆ Ap FR …ta†e IT ÿ UL …Tfi ÿ Ta † ÿ …UA †t …Ts ÿ Ta † ÿ …UA †i
dt
2i …Ts ÿ Ta † ÿ …UA †o 2o …Tfo ÿ Ta † …8†

Collector

_ p …Tfo ÿ Tfi † ˆ Ap FR …ta†e IT ÿ UL …Tfi ÿ Ta †
mc …9†
Inlet pipe from storage tank to the collector
_ p …Ts ÿ Tfi † ˆ …UA †i 2i …Ts ÿ Ta †
mc …10†

Fig. 2. Sketch of set up (Wijeysundera method).


J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700 683

Eqs. (8)±(10) can be combined to get


dys
‡ lys …t† ˆ gIT …t† ‡ lya …t† …11†
dt
where
ys …t† ˆ Ts …t† ÿ Ts …0 † …12†

ya …t† ˆ Ta …t† ÿ Ts …0 † …13†

l1
gˆ …14†
Ct

…l2 ‡ l3 †
lˆ …15†
2
where
" #
…UA †o 2o
l1 ˆ Ap FR …ta†e 1 ÿ …16†
_ p
mc

" #" #
…UA †o 2o …UA †i 2i
l2 ˆ Ap FR UL 1ÿ 1ÿ …17†
_ p
mc _ p
mc

and
…UA †o 2o …UA †i 2i
l3 ˆ …UA †t ‡…UA †o 2o ‡…UA †i 2i ÿ …18†
_ p
mc
The parameters g and l include the characteristics of the system and are estimated by
minimizing the sum of the squares of the di€erence between measured and predicted tank
temperatures. It may be noted that the characteristic parameters of the collector are taken as
FR …ta†e and FR UL in this model, and they are similar in nature compared to the steady state
model. Thus, this model provides a method of estimating the collector parameters from the
daily variation of the storage tank temperature of a collector-storage system.
The method requires two calibration tests; one test is meant for the storage tank to estimate
its parameters, viz., Ct and …UA†t ; the other for the pipe losses to estimate …UA†i and …UA†o .

2.3.1. Tank calibration test


The e€ective thermal capacitance and the e€ective loss coecient of the tank are obtained
by conducting a calibration test. In this, water in the tank is heated to about 808C using a
constant input of electric power (E ). Water is continuously circulated through the tank by a
pump for achieving a uniform tank temperature. Then, heating is stopped and water is allowed
684 J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700

to cool due to the circulation. For these conditions, Eq. (11) is used as a governing equation in
which l ˆ …UA†t =Ct and gIT ˆ E=Ct are substituted and the tank parameters are estimated
using regression analysis.

2.3.2. Estimation of pipe heat loss coecient


The authors have adopted the procedure recommended by ASTM for testing of cold air
ducts [21] to obtain the pipe heat loss coecient.

2.3.3. Test procedure


According to the authors, the collector test procedure consists of exposing the collector to
variable weather conditions for a number of days from 9:00 am to 5:00 pm. Water is
circulated at a constant ¯ow rate through the system, even when there is no useful heat gain
from the collector. The measurements are taken every 15 s for radiation (incident on the
plane of the collector), ambient temperature, storage temperature, inlet and outlet ¯uid
temperatures and ¯ow rate. The temperature di€erence across the inlet pipe and outlet pipe
are also recorded.

Fig. 3. Sketch of the test rig.


J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700 685

3. Experimental evaluation

3.1. Test rig

In order to perform the comparative study, an experimental rig was designed, fabricated and
assembled. It has facilities to conduct both steady state and transient methods, since the rig
can be operated in di€erent modes depending on the test procedure. A schematic sketch of the
test rig is shown in Fig. 3. The set up has been equipped with suitable instruments for various
measurements and for controlling the system operation. All measuring transducers are
connected to a PC 486. The sensors ®xed in the rig for measurements of temperature,
temperature di€erence, ¯ow rate and solar insolation enable measurements with accuracies of
0.1, 0.018C, 0.05 lpm and 2% of reading, respectively.
Tests have been conducted at the Solar Energy Laboratory, IIT, Bombay (19807' N, 72851 '
E), on a ¯at-plate collector commercially available in the market. It is single glazed and
selectively coated, and the gross area of the collector is 2.06 m2. Experiments have been
conducted as per the sequences described in the last section for each method. Thus, all the
methods have been implemented and compared on a common basis. The sampling rate of
measurements has been used as 2 s, and depending on the prescription of a method, data sets
of appropriate intervals are extracted.

3.2. Perers method

The experimental set up is similar to that of steady state tests. Testing has been conducted
for 7 days with 9±12 h a day, which also includes some hours of darkness. The inlet
temperature of water is kept constant for a day's test and increased in steps of 58C from one
day to the other till the inlet temperature is around 708C. Flow rate is also controlled within
20.05 lpm. As per the method, data recorded at intervals of 6 s are used for measuring data.
In the present calculations, second order temperature dependent loss has been neglected,
since the temperature level of operation was not very high. Also, the inlet and outlet pipes are
well insulated, so the heat loss through the pipes is neglected. Further, terms involving F 0 Uv
and F 0 Usky are excluded from Eq. (1), since a glazed collector is used for experimentation.
Thus, the characteristic parameters considered are F 0 …ta†e , F 0 U1 , Cc, bo and Ktad . U1 , therefore,
refers to the overall loss coecient of the collector. The experimental measurements made over
a week have been used for calculations.

3.3. DSC method

For our evaluation purpose, the same data sets as those used for the Perers method are used
and the calculations have been performed using the DST algorithm [22]. It may be mentioned
that the authors have not recommended any de®nite sampling rate for measurements. In our
investigation, data recorded at intervals of 2 s are used for calculations. The loss term having
explicit dependence on wind speed has been neglected in our calculation, since the wind speed
is not so high at our experimental site. Also not considered is the second order temperature
dependent heat loss, as explained in the previous section. Thus, the parameters estimated are
686 J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700

F 0 …ta†e , F 0 U1 , and Cc. U1 , thus, corresponds to the overall loss coecient of the collector. The
calculation has been performed using any one-day's data or a number of days' data combined
together.

3.4. Wijeysundera method

The tank calibration test has been conducted by heating water inside the tank to about 808C
using a constant known input of electric power (6.9 kW). The water is then allowed to cool
with the circulating pump on.
Tank temperatures are recorded at regular intervals of 60 s, both, during the heating and
cooling periods. With these experimental data, the e€ective thermal capacitance and loss
coecient-area product of the tank are determined by regression analysis.
The facilities for conducting testing of cold air ducts, as recommended by ASTM are not
available in the Solar Energy Laboratory, IIT Bombay. So, the pipe loss coecients are
determined by a separate test in which ¯ow rate, inlet and outlet temperatures for both the
pipes are recorded. Also measured are the storage tank temperature (Ts) and ambient
temperature. With the help of these measurements, the UA product (product of heat loss
coecient and surface area) of the pipes is calculated from the following standard expressions:
 
_ p …To ÿ Ti †
mc …Ti ÿ Ta †
UA ˆ   ln …19†
…Ti ÿ Ta † ÿ …To ÿ Ta † …To ÿ Ta †

and
  
UA UA
2ˆ 1 ÿ exp …20†
_ p
mc _ p
mc

This test has been conducted at four di€erent temperatures of ¯uid, and the average of all the
results is used for our calculation purpose.

Table 1
Collector parameters from Perers method

Test no No. of days F 0 …ta†e bo F 0 …ta†e Ktad F 0 UL (W mÿ2 Kÿ1) Cc (kJ mÿ2 Kÿ1)

1 All 0.76 (0.001) 0.119 (0.000) 0.64 (0.000) 3.34 (0.001) 12.20 (0.29)
2 1, 2, 3, 4 0.77 (0.001) 0.108 (0.000) 0.70 (0.000) 3.81 (0.003) 17.05 (0.40)
3 3, 4, 5, 6 0.78 (0.000) 0.170 (0.000) 0.73 (0.000) 3.76 (0.003) 13.70 (0.45)
4 1, 2, 3 0.76 (0.000) 0.107 (0.000) 0.65 (0.000) 3.39 (0.004) 13.32 (0.51)
5 4, 5, 6 0.79 (0.000) 0.127 (0.000) 0.73 (0.001) 4.11 (0.004) 13.12 (0.45)
6 1 0.76 (0.000) 0.113 (0.001) 0.66 (0.001) 3.33 (0.016) 12.67 (0.95)
7 2 0.80 (0.000) 0.137 (0.000) 0.69 (0.001) 4.09 (0.007) 13.68 (0.90)
8 3 0.79 (0.000) 0.128 (0.001) 0.68 (0.000) 3.95 (0.012) 16.65 (2.35)
9 4 0.76 (0.003) 0.143 (0.000) 0.65 (0.001) 3.41 (0.009) 13.21 (1.74)
10 7 0.78 (0.001) 0.138 (0.000) 0.73 (0.004) 3.95 (0.019) 16.40 (1.92)
Average 0.78 0.129 0.69 3.71 14.2
J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700 687

Collector testing has been conducted by exposing the complete system to actual weather
conditions from 9:00 am to 5:00 pm, and the variation of the storage tank temperature has
been measured. The pump is kept on throughout the test period. Measurements are done every
15 s for IT, Ta, Ts, T®, m_ and temperature di€erences across each of the collector, inlet pipe
and outlet pipe.
The daily steady state parameters are obtained by minimising the sum of the squares of the
di€erence between measured and predicted tank temperatures using the Levenberg±Marquardt
Algorithm.

4. Results and discussion

The collector parameters obtained on the basis of Perers method are presented in Table 1.
The values within parentheses represent the standard errors in the estimation of the respective
parameters. To calculate the parameters, computations have been performed using data sets
formed by combining several days' data together. In a second step, the data of individual days
have been used independently for calculations. The number of days used to obtain each set of
results is mentioned in the table. The table also presents the average values of the parameters
from all the tests conducted.
It is seen that the variation of F 0 …ta†e from one set to the other is relatively small. Signi®cant
variations are seen in the results for the remaining parameters, especially bo and Cc. Thus, it
appears that the estimation of the collector parameters by this method does not seem to be
very good. It may be noted that the calculations are sensitive to the averaging period. The
values reported in the table correspond to the averaging period of 30 s; but if a higher
averaging period is taken, the calculations did not converge satisfactorily. Secondly, if the data
sets include measured values of qu less than 50 W/m2 or IT less than 100 W/m2, then the
method yields unrealistic values for the parameters. Such records are, therefore, to be excluded
before calculations are performed. Consequently, there is no need to collect data during
darkness, as suggested by the author.

Table 2
Collector parameters from DSC method

Test no No. of days F 0 …ta†e F 0 UL (W mÿ2 Kÿ1) Cc (kJ mÿ2 Kÿ1)

1 All 0.71 (0.003) 4.19 (0.17) 16.38 (0.3)


2 1, 2, 3, 4, 5, 7 0.72 (0.002) 4.32 (0.21) 16.41 (0.4)
3 1, 2, 3 0.69 (0.021) 3.20 (0.41) 17.14 (0.5)
4 4, 5, 6 0.69 (0.031) 3.61 (0.04) 17.55 (0.6)
5 1 0.70 (0.008) 3.72 (0.02) 15.94 (0.3)
6 2 0.71 (0.006) 4.24 (0.01) 16.53 (1.2)
7 3 0.72 (0.001) 4.15 (0.01) 16.79 (0.2)
8 4 0.69 (0.025) 3.43 (0.04) 19.05 (1.3)
9 6 0.69 (0.011) 3.74 (0.04) 10.77 (2.3)
Average 0.70 3.84 16.28
688 J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700

The parameters calculated on the basis of the DSC method are shown in Table 2, and the
standard errors are mentioned within parentheses. As pointed out earlier, the same set of data
as used for the Perers method is used in this calculation. Also, like the Perers method,
calculations have been performed by combining several days' data. It is seen that the values of
F 0 …ta†e do not vary much from one set to the other; but a large variation is noticed in the
values of Cc. This behaviour is similar in nature to those reported by the authors. It may be
mentioned that the authors have a sampling rate of 5 min in their method. In the present case,
even if the sampling rate has been reduced to 2 s, the variation in the values of Cc is
signi®cant. The calculations do not yield realistic results for the data having radiation level less
than 75 W/m2. Hence, the data collected for dark hours has not been used.
In order to estimate the parameters on the basis of Wijeysundera method, two calibration
tests are conducted. Table 3 presents the results of tank heat capacity and loss term. The pipe
loss terms are also presented in the same table. These values have been used for estimating
collector parameters.
As far as the collector test is concerned, each days experimental observations constituted one
single test, and hence, we get daily mean values of steady state parameters FR …ta†e and FR UL .
These results, along with the corresponding standard errors, are presented in Table 4 for a
week's experiments. It may be noted that there are not much variations of the values of
parameters from one set of experiments to the other.

4.1. Comparison of results with ASHRAE 93-86 standard

The test set up required for the Perers and DSC methods is the same as that for the steady
state ASHRAE 93-86 standard [18]. Wijeysundera's method requires an additional
arrangement to measure the variation in the tank temperature. The experimental conditions
required are less stringent as compared to the steady state test procedures. The Perers method
needs control over inlet temperature, whereas the DSC method does not prescribe controlling
either the ¯ow rate or ¯uid inlet temperature. In the case of the Wijeysundera method, a steady
¯ow rate is required. In addition to collector testing, this method requires two additional tests,
viz., tank calibration test and pipe loss coecient estimation test. Thus, the Perers and DSC
methods involve simpler test procedures and do not require any additional tests, unlike the
Wijeysundera method.
The steady state results obtained according to the ASHRAE 93-86 standard are taken as a
reference for comparing results of the di€erent tests. The thermal eciency curve obtained by
this standard is shown in Fig. 4. The collector parameters estimated from it are presented in

Table 3
Results of calibration tests

Tank parameters Pipe loss terms

Ct (kJ Kÿ1) …UA†t (WKÿ1) …UA†I (WKÿ1) …UA†o (WKÿ1)

1003.91 5.74 6.12 6.01


J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700 689

Table 4
Collector parameters from Wijeysundera method

Test no FR …ta†e FR UL (W mÿ2 Kÿ1)

1 0.71 (0.051) 3.21 (0.29)


2 0.70 (0.049) 3.17 (0.27)
3 0.71 (0.056) 3.14 (0.31)
4 0.68 (0.059) 3.06 (0.28)
5 0.69 (0.048) 3.10 (0.31)
6 0.68 (0.051) 3.15 (0.27)
7 0.70 (0.047) 3.16 (0.30)
Average 0.70 3.14

Table 5, and the corresponding correlation coecient is also mentioned in the table. The
standard errors are given within the parentheses. The constant ``bo'' appearing in the
expression for Ktab has been estimated for steady state case and is presented in Table 5. As
pointed out in previous sections, the characteristic parameters of the collector are F 0 …ta†e ,
F 0 U1 , Cc, bo and Ktad in the case of the Perers method, and those for the DSC method are
F 0 …ta†e , F 0 U1 and Cc. In view of the simpli®cations adopted in our calculations, U1 is nothing
but UL. Hence, for direct comparison with the results of the ASHRAE standard, the steady
state values of FR …ta†e and FR UL (Table 5) are used to calculate F 0 …ta†e and F 0 UL , and these
values too are presented in Table 5. The total heat capacity of the absorber plate, risers and
¯uid inside have been theoretically calculated from the physical dimensions of the collector, as
provided by the manufacturers. This value has been reported in the table for comparison
purpose.

Fig. 4. Eciency curve of the collector (ASHRAE standard).


690 J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700

Table 5
Steady state test results

FR …ta†e FR UL (W mÿ2 Kÿ1) R2 bo F 0 …ta†e F 0 UL (W mÿ2 Kÿ1) Theoretical (kJ mÿ2Kÿ1

0.75 3.61 0.975 0.142 0.77 3.68 14.4


(3.91  10ÿ5) (0.001)

The following conclusions can be drawn from the comparison of results presented in Tables
1±5.
1. The results presented in Tables 1 and 2 show that there are variations in the values of
parameters from one test to another. In order to make a quantitative representation of their
variations, individual results have been compared with the corresponding average values.
Table 6 gives the percentage deviation from the average values for each method. In the
Perers method, the maximum variation observed for F 0 …ta†e , F 0 UL and Cc are, respectively,

Table 6
Percentage deviations from the average value

Method Test no DF 0 …ta†e DF 0 UL DCc Dbo DF 0 …ta†e Ktad DFR …ta†e DFR UL
(%) (%) (%) (%) (%) (%) (%)

Perers method 1 2.6 10.0 14.1 7.8 7.2 ± ±


2 1.3 2.7 20.1 16.3 1.4 ± ±
3 0.0 1.3 3.5 31.8 5.8 ± ±
4 2.6 8.6 6.2 17.1 5.8 ± ±
5 1.3 10.8 7.6 1.6 5.8 ± ±
6 2.6 10.2 10.8 12.4 4.3 ± ±
7 2.6 10.2 3.7 6.2 0 ± ±
8 1.3 6.5 17.3 0.8 1.4 ± ±
9 2.6 8.1 7.0 10.9 5.8 ± ±
10 0.0 6.5 15.5 7.0 5.8 ± ±
DSC method 1 1.4 9.1 0.6 ± ± ± ±
2 2.9 12.5 0.8 ± ± ± ±
3 1.4 16.7 5.3 ± ± ± ±
4 1.4 6.0 7.8 ± ± ± ±
5 0.0 3.1 2.1 ± ± ± ±
6 1.4 10.4 1.5 ± ± ± ±
7 2.9 8.1 3.2 ± ± ± ±
8 1.4 10.7 17.0 ± ± ± ±
9 1.4 2.6 33.8 ± ± ± ±
Wijeysunder a method 1 ± ± ± ± ± 1.4 2.2
2 ± ± ± ± ± 0.0 1.0
3 ± ± ± ± ± 1.4 0.0
4 ± ± ± ± ± 2.9 2.5
5 ± ± ± ± ± 1.4 1.3
6 ± ± ± ± ± 2.9 0.3
7 ± ± ± ± ± 0.0 0.6
J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700 691

about 2.6, 10.8 and 20.1%. The corresponding values for the DSC method are about 2.9,
16.7 and 33.8%, respectively. Similarly, large ¯uctuations of the order of 31.8 and 7.2%,
respectively, for the values of bo and F 0 …ta†e Ktad compared to their average values have been
observed in the Perers method. Thus, the values of the characteristic parameters, except
F 0 …ta†e , vary signi®cantly from one test to the other. In the case of the Wijeysundera method
(Table 4), the maximum variations observed for FR …ta†e and FR UL are, respectively, 2.9 and
2.5%. Clearly, the repeatability of results in the Wijeysundera method are quite good.
2. The individual methods have also been compared with the steady state values of the
ASHRAE standard. The maximum deviations of F 0 …ta†e and F 0 UL obtained on the basis of
the Perers method are, respectively, 3.9 and 11.7%. The corresponding values for the DSC
method are, respectively, 10.4 and 17.4%, and the values of F 0 …ta†e are consistently lower
than those of the steady state case. Similarly, the maximum deviations of FR …ta†e and FR UL
obtained for the Wijeysundera method are 9.3 and 15.2%, respectively, and the values for
both parameters are consistently lower than the steady state values. So, the values obtained
on the basis of the Perers method are close to the steady state parameters.
3. The values for the collector heat capacity obtained on the basis of the Perers and DSC
methods are compared with the theoretically calculated value (Table 5). The maximum
deviations of about 18.4 and 32.3% are observed for these two methods. Hence, the
estimation seems to deviate signi®cantly from the theoretical value.

4.2. Sensitivity analysis

In order to determine the e€ect of uncertainties in measurements on the values of estimated


parameters, a sensitivity analysis has been performed. All measurements of temperatures, ¯ow
and solar radiation are assumed to be changing by 20.58C, 210% and 210 W/m2, respectively.
For both the Perers and DSC methods, sensitivity analysis has been performed for a data
sequence which combines all the days' measurements. The results for the Perers method are
shown in Table 7. It can be concluded that the results obtained are sensitive to all the
measured values. This is especially true with the parameter F 0 …ta†o bo . Since bo is calculated
from F 0 …ta†o bo , there is a propagation of error in its estimation, and it is found to be as high

Table 7
Sensitivity analysis for Perers method

Parameter Variation DF 0 …ta†e (%) DF 0 …ta†e bo (%) DF 0 …ta†e Ktad (%) DF 0 UL (%) DCc (%)

v_ 2 10% 2 10 3 49 2 11 2 15 2 10
Tfi 2 0.58C 33 2 13 3 13 25 36
Tfo 2 0.58C 23 3 93 2 15 21 24
Ta 2 0.58C 3 0.2 3 44 2 0.5 25 2 0.1
IT 2 10 W/m2 3 0.7 3 48 31 26 3 0.9
Id, T 2 10 W/m2 3 0.3 3 42 3 0.2 34 3 0.3
All ± 2 10 2 14 27 2 10 2 16
692 J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700

as 98% for the case when the Tfo variation is ÿ0.58C. Thus, it can be said that the method
requires accurate measurements of data to obtain reliable results.
Table 8 presents the sensitivity results for the DSC method. It can be seen that the values of
the collector parameters are sensitive to the measurements of ¯ow rate and inlet and outlet
temperatures. There are, however, no signi®cant e€ects of errors in the measurement of
ambient temperature and solar radiation (total as well as di€use) on the tilted plane.
Sensitivity analysis is also performed for the tank calibration and pipe loss coecient
estimation tests along with the collector test for the Wijeysundera method, and the respective
results are presented in Tables 9±11. It can be seen from Table 9 that when both Ts or Ta are
increased or decreased, there is no signi®cant change in both Ct and …UA†t . The results of
Table 10 show that the ¯ow rate should be measured accurately. The change in ambient
temperature does not a€ect the values of the parameters signi®cantly. Table 11 presents the
results for the collector parameters. The calculations have been performed without considering
any changes in the values of Ct, …UA†t , …UA†i and …UA†o as obtained from the calibration tests.
Only the measurements made during the collector testing are considered. The results of Table
11 correspond to the values reported in the ®rst row of Table 4. Thus, it can be seen that the
results obtained by the Wijeysundera method are not very sensitive to the measurements. In
order to see the e€ect of propagation of errors in the prediction of the collector parameters,
the changes reported in the calibration parameters (Tables 9 and 10) are considered, and the
sensitivity calculations are repeated with the results shown in Table 12. It is seen that the
method estimates fairly stable values for FR …ta†e and FR UL . Thus, the method is less sensitive
to the parameters obtained from the calibration test.

4.3. Prediction of collector performance

In order to examine the capability of the methods for predicting the collector performance,
calculations have been made for the useful heat gain rate and compared with the experimental
measurements. Since the Wijeysundera method estimates only steady state parameters, it is not
expected to predict the dynamic behaviour of the collector and hence has been excluded from
this comparison. In order to calculate the heat gain, Eqs. (1) and (4) are recast in terms of
outlet temperatures and solved numerically. Thus, Eqs. (1) and (4) respectively become

Table 8
Sensitivity analysis for DSC method

Parameter Variation DF 0 …ta†e (%) DF 0 UL (%) DCc (%)

v_ 2 10% 2 10 +0.3 0
Tfi 2 0.58C 33 2 12 2 0.2
Tfo 2 0.58C 25 3 17 32
Ta 2 0.58C 3 0.2 2 0.4 0
IT 2 10 W/m2 31 24 2 0.3
Id, T 2 10 W/m2 2 0.4 22 3 0.6
All ± 29 24 2 0.4
J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700 693

Table 9
Sensitivity analysis for tank calibration test

Parameter Variation (8C) DCt (%) DUAt (%)

Ts 3 0.5 3 0.1 2 3.1


Ta 3 0.5 3 0.1 2 3.1
All ± 0.0 +0.4

ÿ 
Tfo …t ‡ Dt † ˆ Tfo …t† ‡ Q1 …t† ÿ P1 …t†  Tfo …t† Dt …21†

ÿ 
Tfm, n …t ‡ Dt † ˆ Tfm, n …t† ‡ Q2 …t† ÿ P2 …t†  Tfm, n …t† Dt …22†

where
   
0 0 Tfi Cc dTfi 2
Q1 …t† ˆ mc
_ p Tfi ‡ F …ta†e …Ktab Ib ‡ Ktad Id † ‡ F UL Ta ÿ ÿ …23†
2 2 dt Cc
 
_ p ‡ F 0 UL
2mc
P1 …t† ˆ …24†
Cc
Similarly,
 
_ p Tfm, nÿ1 ‡ F 0 …ta†e IT ‡ F 0 UL Ta
Nmc
Q2 …t† ˆ …25†
Cc
and
 
_ p ‡ F 0 UL
Nmc
P2 …t† ˆ …26†
Cc
n varies from 0 to 31 so that Tfm, o …t† ˆ Tfi …t† and Tfm, 31 …t† ˆ Tfo …t†.
The initial conditions are:
Tfo …0 † ˆ Tfi …0 † …27†
and

Table 10
Sensitivity analysis for pipe loss test

Parameter Variation D…UA†i (%) D…UA†o (%)

Ta 2 0.58C 21 2 0.7
v_ 2 10% 27.5 2 10
All ± 27.5 2 10
694 J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700

Table 11
Sensitivity analysis for Wijeysundera method

Parameter Variation DFR …ta†e (%) DFR UL (%)

Ts 2 0.58C 31 24
Ta 2 0.58C 21 34
v_ 2 10% 0.0 2 0.1
IT 2 10% 2 0.8 3 0.2
All ± 3 0.4 22

Table 12
Error propagation for Wijeysundera method

Parameter Variation (%) DFR …ta†e (%) DFR UL (%)

DCt 0 ÿ0.2 +0.7


D…UA†t +0.4
D…UA†i 2 7.5 32 2 0.7
D…UA†o 2 10.4
All ± 21 3 0.6

Tfm, n …0 † ˆ Tfi …0 † …28†

Figs. 5 and 6 show the results of such calculations for two di€erent days. While part (b) of
each ®gure presents the variation of the useful heat gain rate as predicted by the two methods
with time, the corresponding weather data are plotted in part (a). It may be mentioned that the
experimental data set considered for this comparison is not used in estimation of the collector
parameters. Though the calculations have been performed for the complete time duration
corresponding to the experimental conditions, only a part of the results are shown in Figs. 5(b)
and 6(b) for the sake of clarity. It is seen that the DSC method predicts results close to the
measured values as compared to the Perers method. In order to get a clear idea, the frequency
of occurrences of data points deviating from measurements in di€erent percentage ranges have
been calculated. The range has been chosen as 2%. Plots of such frequency of occurrences for
di€erent ranges of deviations are shown in Figs. 7 and 8. It can be seen that the percentage of
data points showing deviations within 6% are about 60 to 80% in the case of the DSC
method, and the corresponding ®gure for the Perers method is about 35±46%. Thus, the DSC
method predicts better than the Perers method. This may be attributed to the modelling of the
individual methods. In the case the of Perers method, a mean ¯uid temperature is assumed for
the entire collector. Consequently, the outlet temperature starts changing as soon as one of the
input variables is changed. This is not true in the case of the DSC method, which splits the
collector into a number of segments along the ¯ow direction.
The e€ects of measurement errors in the estimation of the collector parameters have been
discussed in the previous section. The variations in the values of the collector parameters due
to these e€ects would cause variations in the predictions of heat gain. In order to study this
e€ect, the collector parameters F 0 …ta†e , F 0 UL and Cc are increased by 10, 15 and 16%,
J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700 695

Fig. 5. (a) Experimental measurements (20 February 1998), (b) Comparison of prediction of heat gain rate with
measurements (20 February 1998).
696 J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700

Fig. 6. (a) Experimental measurements (21 February 1998), (b) Comparison of prediction of heat gain rate with
measurements (21 February 1998).
J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700 697

Fig. 7. Frequency distribution for deviations in predicted heat gain rate (20 February 1998).

Fig. 8. Frequency distribution for deviations in predicted heat gain rate (21 February 1998).
698 J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700

Fig. 9. E€ect of measurement errors on the prediction of heat gain (Perers method).

Fig. 10. E€ect of measurement errors on the prediction of heat gain (DSC method).
J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700 699

respectively (Table 7), for the Perers method and 10, 17 and 2%, respectively (Table 8), for the
DSC method. The results are shown in Figs. 9 and 10 for the Perers and DSC methods,
respectively. It can be seen that both the Perers and DSC methods are sensitive to the errors in
the prediction of F 0 …ta†e , the changes in the value of average deviation being 13.3 and 7.9%,
respectively. It can further be seen that there is an insigni®cant e€ect on the prediction of heat
gain for both the methods due to the errors in estimation of F 0 UL and Cc. Thus, the Perers
and DSC methods are sensitive to the errors in the measurements, leading to relatively large
error propagation in predicting heat gain rate.

5. Conclusions

A detailed study of the three transient test procedures (Perers, DSC and Wijeysundera) has
been conducted. All the methods have been implemented and extensive comparisons have been
made on a common basis. Their results are also compared with the steady state values based
on the ASHRAE standard. A detailed sensitivity analysis has been performed for all of them.
The predictions of the methods, namely Perers and DSC, have been compared with
experimental measurements. The conclusions can be summarised as follows:
1. Test set up and procedures for both the Perers and DSC methods are simpler as compared
to the Wijeysundera method, which requires two additional calibration tests. Perers method
requires that the inlet ¯uid temperature should be controlled over the test run. Control of
neither ¯ow rate nor ¯uid inlet temperature is needed in the DSC method, but the
Wijeysundera method prescribes control of ¯ow rate.
2. For computation of results, Perers method requires any spreadsheet, while the DSC method
requires a special program and knowledge of digital ®ltering. Both the DSC and
Wijeysundera methods use subroutines of the Levenberg±Marquardt Algorithm for
optimisation purpose.
3. There are large variations in the estimation of F 0 UL and Cc from one set of results to
another for both, the Perers and DSC methods. Thus, the repeatability of results is not quite
good. However, in the case of the Wijeysundera method, the results are repeatable.
4. The Perers method yields F 0 …ta†e close to the steady state value (within 4%), whereas the
DSC method underpredicts it (the maximum deviation being about 10%). The values of
F 0 UL di€er signi®cantly from the steady state values. However, in the case of the
Wijeysundera method, both FR …ta†e and FR UL are underpredicted.
5. Similar behaviour is seen in the case of estimation of Cc from both Perers and DSC
methods, leading to large variations.
6. Sensitivity analysis shows that the results based on Perers method are quite sensitive to
errors in the measurements of all variables. Especially, the value of F 0 …ta†e bo (in turn bo) is
very sensitive. In the case of the DSC method, the ¯ow rate and inlet and outlet
temperatures are a€ected signi®cantly, compared to other measurements. The results of the
Wijeysundera method are not sensitive to the errors in measurements.
7. The Wijeysundera method estimates steady state collector parameters from daily tests of a
complete system. It is, thus, not expected to predict the dynamic behaviour of a collector.
700 J.K. Nayak et al. / Energy Conversion & Management 41 (2000) 677±700

Compared to Perers method, the prediction of heat gain rate by the DSC method is closer
to the measurements.
8. In the DSC method, a skip time is required to be given for obtaining reliable results,
otherwise the calculation yields a zero value for Cc. There is not a clear guideline about
selecting this value. The collector parameters are found to depend signi®cantly on the choice
of skip time.

Acknowledgements

The ®nancial assistance received from the Solar Energy Centre, Ministry of Non-
Conventional Energy Sources, Government of India, for the work is gratefully acknowledged.

References

[1] Rogers BA. In: Hall DO, Mortan J, editors. Solar world forum, vol. 1. Oxford: Pergamon Press, 1981. p. 898.
[2] Emery M, Rogers BA. Solar Energy 1984;33:117.
[3] British Standard, BS 6757, 1986.
[4] Saunier GY, Chungpaibulpatana S. In: Szokolay SV, editor. Solar World Congress, vol. 2. Oxford: Pergamon
Press, 1983. p. 910.
[5] Chungpaibulpatana S, Exell RHB. Solar and Wind Technology 1988;5:411.
[6] Wang XA, Xu YF, Meng XY. Solar Energy 1987;38:125.
[7] Souproun AV. ASME Journal of Solar Energy Engineering 1992;2:1149.
[8] Perers B. Solar Energy 1993;50:517.
[9] Muschaweck J, Spirkl W. Solar Energy Materials and Solar Cells 1993;30:95.
[10] Spirkl W, Muschaweck J, Kronthaler P, Schoelkopf W, Spehr J. Solar Energy 1997;61:147.
[11] Wijeysundera NE, Hawlader MNA, Foong KY. ASME Journal of Energy Engineering 1996;118:30.
[12] Amer EH, Nayak JK, Sharma GK. Energy Conversion and Management 1998;39:285.
[13] Amer EH, Nayak JK, Sharma GK. Solar Energy 1997;60:229.
[14] Schnieders J. Solar Energy 1997;61:179.
[15] Kamminga W. Int J Heat and Mass Transfer 1985;28:1393.
[16] Isakson, P., Eriksson, L.O. Nordic Solar Energy R & D Meeting, BorlaÈnge, 1991.
[17] Henning HM. In: Regenerierung von Adsorbention mit solar erzeugter Prozesswaerme. Duesseldorf: VDI-
Verlag, 1994. p. 45.
[18] ASHRAE 93-86 standard, ASHRAE Inc. New York, 1986.
[19] Spirkl W. ASME Journal of Solar Energy Engineering 1990;112:98.
[20] Marquardt D. SIAM Journal of Applied Mathematics 1963;11:431.
[21] ASTM, 1984 Annual Book of ASTM standards, ASTM C1003-83, 1984.
[22] Insitu Scienti®c Software, Klein and Partners, Dynamic System Testing Program Manual, Version 2.4a, 1994.

S-ar putea să vă placă și