COMPARISON OF TEST METHODS FOR EVALUATION

OF THERMAL PERFORMANCE OF PREHEAT

AND SOLAR-ONLY FACTORY MADE SYSTEM.
M.J. Carvalho
INETI - Department of Renewable Energies, Estrada do Pao do Lumiar 1649-038 Lisboa, Portugal,
Phone: +351 21 712 7193, Fax: +351 21 712 7195, Mjoao.Carvalho@mail.ineti.pt
D.J. Naron
Building and Systems - Renewable Energy, TNO Building and Construction Research, P.O.Box 49, Delft,
NL-2600 AA, The Netherlands, Tel: +31 15 2695249, Fax: +31 15 2695299, E-mail: d.naron@bouw.tno.nl
Abstract In the frame of the European Programme, SM&T, the project Bridging the gap: Research
and experimental validation on the DST performance test method for solar domestic water heaters
dedicated a work package to the comparison of DST test methods (ISO 9459.5) with CSTG test method
(ISO 9459.2). In the project, several European test laboratories tested different types of systems. The
systems have been selected in order to be as representative as possible of the European market. The
comparison of the two methods was based on long term performance prediction determined according to
each test method. In this paper the results of this comparison as well as Conversion Factors between the
two test methods are presented. Recommendations for a future revision of ISO 9459-2 are also presented.
1. INTRODUCTION
In the pre-Standard European documents for Solar
thermal systems and components, which are expected for
Formal Vote soon, two types of standards are included:
i) standards containing the definition of general
requirements;
ii) standards of test methods for evaluation of the general
requirements and also for evaluation of the thermal
performance of systems and components.
The European draft standard for Thermal solar systems
and components, prEN 12976 (1999), is dedicated to the
definition of general requirements and test methods for
Factory Made Systems, which are defined as solar
systems sold as products in large quantities as complete,
packaged and ready to install kits with one trade name.
The test methods considered to access thermal
performance of these systems are ISO 9459.2 (currently
designated by CSTG test method or Input-Output test
method) for preheat and solar-only systems, and ISO/DIS
9459.5 (currently designated by DST test method) for
solar plus supplementary systems. It is technically
possible to apply both test methods to preheat and solaronly systems, but it is necessary to verify the
comparability of results between the two methods.
CSTG test method was the result of the work developed
during the 80s by the Collector and Systems Testing
Group whose activities were co-ordinated by the Joint
Research Centre of Ispra (Aranovitch, E. et all, 1989).
This work included round robin tests of different systems
for validation of the test method. Analyses of these tests
and of its validation can be seen in reference (Bourges, et
all, 1991).
In the frame of the European Programme, SM&T, the
project Bridging the gap: Research and experimental

validation on the DST performance test method for solar

domestic water heaters (Naron, D.; 1999c) dedicated a
work package (WP2) to the comparison of the two test
methods. In this project, different types of Factory Made
Systems were tested by several European test
laboratories. The systems have been selected in order to
be as representative as possible of the European market.
The referred project had as its main objective the
validation of the DST test method (ISO 9459.5). A first
validation based on simulated test sequences was
performed (WP1) (Naron, D. et all, 1999a) as well as an
experimental validation (WP3) (Naron, D.;1999b). Some
of the tests performed in the frame of WP3 contributed
also to the comparison between the two test methods
(WP2) (Carvalho, M.J.; 1999). The systems were tested
between 1997 and 1999. The comparison between the
two test methods (DST and CSTG) was based on the long
term performance prediction determined considering the
test results and the calculation procedure associated with
each of the test methods.
The work developed within WP2 for comparison of the
two test methods showed the need to propose revision of
some aspects of CSTG test method. These aspects are
detailed in section 3 of this paper and the corresponding
recommendations for revision of ISO 9459.2 are listed. In
section 4 is given a brief description of the conditions for
parameter identification and Long Term Performance
Prediction (LTPP) calculations, for CSTG and DST tests
performed. Section 5 describes the results of the
comparison of the two test methods. In all situations
where there were doubts in relation to experimental test
results, simulated test sequences were obtained and
analysed to complement the comparison results. Section 5
also presents the determination of Conversion Factors.
These conversion factors will be proposed as an

2. SYSTEMS TESTED
The systems tested by the four laboratories involved in
WP2 are described in Table 1. Table 2 summarises the
tests performed by each laboratory involved in WP2 (1st
Laboratory) as well as the tests performed by the
laboratories participating in WP3 (2nd Laboratory) also
used to draw conclusions of the comparison between the
two test methods.

Laboratory
INETI
(Portugal)

CSTB
(France)

UWCC
(UK)

NCSR
(Greece)

Table 1 Systems tested

Type of system tested
2 thermosyphon systems (T2, T3) with
non selective flat plate collectors
Collector aperture area: 1.6 m2
Tank volume: 100 l
1 ICS system (I2), with double glazing,
cylindrical tank and mirror
Collector aperture area: 1.2 m2
Tank volume: 160 l
1 ICS system (I4), with vacuum tubes
collector.
Collector aperture area: 0.9 m2
Tank volume: 80 l
Forced circulation system (P2) with a
selective flat plat collector
Collector aperture area: 3.6 m2
Tank volume: 281 l
Forced circulation system (P3) with
evacuated tubes collector
Collector aperture area: 3 m2
Tank volume: 252 l

Table 2 Tests performed within WP2 (1st Laboratory)

and WP3 (2nd Laboratory
st
System
1 Lab
Test
2nd Lab
Test
P2
NCSR
CSTG
SP
DST
Greece
Sweden
DST
P3
NCSR
CSTG
TNO
DST
Greece
Holland
DST
T2
INETI
CSTG
ITW
DST
Portugal
Germany
DST
T3
INETI
CSTG
TNO
DST
Portugal
Holland
DST
I2
CSTB
CSTG
SPF
DST
France
Switzerland
DST
I4
UWCC
CSTG (indoor)
UK
CSTG (outdoor)
3. CSTG TEST METHOD.
It is not the objective of this section to give a detailed
description of the test method. This can be seen in the

text of the standard ISO 9459.2 and also in references

(Aranovitch, E. et all; 1989; Bourges et all; 1991). The
objective is to highlight the points where revision of ISO
9459.2 should be considered.
According to CSTG test method, system characterisation
is obtained by the determination of:
i)
Input-Output diagram
ii) Draw-off temperature profile
iii) Tank overnight heat losses coefficient
This information is needed in order to obtain Long Term
Performance Prediction (LTPP) of the system.
3.1. Input-Output Diagram
The daily behaviour of the system is characterised by the
linear model:
Q = a1 H + a2 (ta(av) tmains) + a3

(1)

The parameters are obtained by Least Squares Fitting of

the daily test results. Fig.1 shows an example of an InputOutput diagram obtained from test results of system T2.
20

15
Test result

Qext [MJ]

addendum to prEN12976-2. Final conclusions are given

in section 6.

Delta T = 0 [C]
Delta T = 5 [C]

10

Delta T = 10 [C]
Delta T = -5 [C]
Delta T = -10 [C]

0
0

10

15

20

25

30

Hcol [MJ/m2]

Fig.1 Input-Output Diagram of system T2 (Test

performed at INETI in 1998).
According to ISO 9459-2 (point 7.2.) the number of days
needed for the test are at least six, provided that at least
four days have approximately the same values of [ta(day)tmains] and irradiation is evenly spread over the range 8
MJ.m-2 to 25 MJ.m-2. Two additional days with values
of [ta(day)- tmains] at least 9K above or below the values of
[ta(day)- tmains] obtained for the first four days must be
obtained too. These criteria for determination of the
minimum test days, is not enough to obtain a good
characterisation of the system. It is common practice in
test laboratories to test for a longer period (see Fig.1) in
order to obtain representative parameters of the system. A
criteria suggested in reference (Bourges et all, 1991) is
based on the uncertainty of LTPP. This calculation
should be done after each test day and tests stopped when
an uncertainty lower then 5% is attained, e.g.. According
to Bourges et all (1991) the uncertainty associated to the
LTPP is determined based on the statistical uncertainty of
the parameters a1, a2 and a3 as a result of Least Squares
Fitting.

Table 3 - I-O parameters of system T2 considering

a) Q = a1 H + a2 (ta(av) tmains) + a3
a1
a2
a3
[m-2]
[MJ.C-1]
[MJ]
a
0.57
0.23
-0.32
0.01
0.03
0.19
a
0.02
0.06
0.40
t(,95%) a
b) Q = a1 H + a2 (ta(av) tmains)
a1
a2
[m-2]
[MJ.C-1]
a
0.551
0.21
0.005
0.03

0.010
0.06
t(,95%) a

3.2. Temperature draw-off profile

This profile, obtained during the extraction of three times
the tank volume of water characterises the behaviour of
the storage tank in relation to mixture of cold water from
mains with hot water in the tank (see Fig.2).
65
Ideal stratification
Well mixed

55

Partial stratification

T [C]

Discussion of these aspects during the project took into

considerations two important question:
i) How to establish the minimum number of test days
needed to obtain statistically valid parameters?
ii) How to establish the validity of the linear model given
by equation (1)?
It was established that Weighted Least Squares
Regression (WLSR) (Press, W. et all, 1992) should be
used instead of simple Least Squares Regression (LSR).
This regression technique gives the necessary instruments
for parameter and model validation. It also allows for the
inclusion of measurement uncertainty in the parameter
determination.
The answer to the above questions is given by the
following recommendation:
For determination of the models parameters a1, a2, a3
the SM&T group recommends to use Weighted Least
Squares Regression in order to take into account the
uncertainty of the measurements of Q, ta(av.day) and tmainss.
The determined uncertainty of parameter a3, gives
guidance for the adoption of a simplified model
Q=a1H+a2(ta(av,day) tmains)
i.e.
if a3 < t(,95%) a3 The simplified model can be
adopted.
t(,95%) is the Students distribution for degrees of
freedom and 95% confidence level.
The uncertainty of parameter a2 (determined from either
the complete or simplified model) gives guidance to the
need for more test days for a good determination of this
parameter, i.e.,
if a2 < t(,95%) a2 more test days are needed,
resulting in a2 > t(,95%) a2 ; this will correspond to a
larger value range for (Ta(av.day)-Tmains)
An example of the application of this criterion is given in
Table 3 for system T2. In Table 3a) the parameters
considering the complete model are given. In this case a
simplified model should be adopted since a3 falls off the
interval t(,95%) a3. Table 3b) lists the parameters of
the simplified model (a3=0).

45
35
25
15
0

Vext/Vtank

Fig. 2 Draw-off Profile

According to ISO 9459-2 the draw-off profile
characterisation is done using two different functions,
f(V) and g(V). The first is obtained based on the normal
draw-off profile performed at the end of each test day and
the second is based on a separate test for determination of
a mixing draw-off profile (see point 7.7. of ISO 9459.2).
The functions f(V) and g(V) are given by the following
expressions:
Q
f ( V ) Fi = i
(2a)
Q
with: Q i = Vi w C pw [t di ( V) t main ]
g( V ) G i =

Q*i
Q

with: Q*i = Vi w C pw t *di ( V ) t main

(2b)

where t di and t *di are the average temperature of the

volume of water Vi obtained for the normal draw-off
profile and the mixing draw-off profile, respectively. Vi is
the total volume of water drawn off. According to ISO
9459.2, Fi and Gi can be given by discreet values for 0.1
Vstore intervals.
The characterisation of the draw-off profile can also be
done according to Bourges et all (1991), using the
function, where is the fraction of energy remaining in
the storage tank after a draw-off of a water volume, Vext:
Q
= 1 ext
(3)
Q
Qext is the energy extracted during the draw-off and Q is
the total energy in the storage tank before the draw-off.

Considering an ideal storage tank with a draw-off profile

as indicated by the solid line of Fig.2, the function of
such a storage tank will be given by the expression:
1 g 0 < g < 1
V
( g) =
with g = ext
(4)
g >1
Vs
0
A well-mixed store is characterised by a function given
by the expression:
( g) = e g
(5)
A storage tank which behaves as an ideal store for a
fraction of storage volume g extracted, and then mixes
hot and cold water for higher storage volume fractions
extracted can be represented by a function given by the
following functional form:
= ag + 1 for g g'
= ' e A '(g-g' )

for

g > g'

with :
g=

Vext
V '
, g' = ext , ' = (g')
Vstore
Vstore

(6)

and :
( '-1)
a=
g'
The use of this function shows that the storage tank
behaves as a stratified tank for a draw-off volume lower
then Vext and behaves as a well mixed storage tank for
draw-off volumes higher then Vext. As an example, this
function is represented in Fig.3 for a specific test day in
order to show that it (gamma_calc) gives a good
representation of values measured (gamma_data).
1
gamma_data

gamma

gamma_calc
0.5

0
0

0.5

1.5
g

2.5

Fig.3 Gamma function (Test day 12.03.98)

Draw-off functions f(V) and g(V) can be determined
based on the function by the following expressions:
Vext

f ( V)

f ( V )dV = 1

or

F = 1
i

i =1

(7a)

Vext

g( V )

g( V )dV = 1

or

= 1 (7b)

i =1

with n such that:

= Vext .

i =1

3.3. Store heat losses

The tank overnight heat losses is obtained in an
independent test where the system with a hot tank is left
to cool down during about 24 h. The collector is covered
during this period. The coefficient is determined using the
following expression:
V T Ta

U s = 4186 s ln i
(8)
t Tf Ta
This value should be given with an associated uncertainty
resulting from measurement errors. An uncertainty in
LTPP resulting from the uncertainty of the heat losses
coefficient can be determined (Bourges et all, 1991).
3.4. Long term performance prediction
ISO 9459-2 also specifies the use of a calculation method
for the determination of long term system performance
prediction. In point 9 of the standard a detailed
description of the method is indicated. A programme in
BASIC code is also given in Annex B of the standard.
The calculation procedure considered includes two load
patterns:
Load pattern 1: Load determined by the volume of daily
hot water consumption
Load pattern 2: Load determined by a minimum useful
temperature limit for the hot water
consumption. When the outlet store
temperature is lower then this minimum
value no water is extracted from the
store.
Based on the results of CSTG test method (ISO 9459.2) ,
the long term performance prediction of the system can
also be obtained using the method presented by Bourges
et all, (1991). In this case a load temperature and a load
volume are specified in order to obtain the energy needed
to meet the load VloadwCpw(tload tmains).
When toutlet<tload , during draw-off, a volume equal to
Vload is extracted
When toutlet>tload , the existence of a mixing valve
regulated for tload is imagined and the volume
necessary to meet the energy needed to load
VloadwCpw(tload tmains) is extracted
This is considered as a third possibility for load pattern:
Load pattern 3: Load determined by maximum energy
needed.
This load pattern is in agreement with DST calculation
procedure (ISO 9459.5) and also with the Reference
conditions of prEN 12976-2 (1997) Annex B.

The comparison between the two test methods are always

made considering the application of Load pattern 3 in the
CSTG LTPP calculations.
The necessary changes in ISO 9459-2 for the calculation
procedure to be in agreement with reference conditions
and DST calculation was proposed in reference
(Carvalho, M.J.; 1999) and is reproduced here:
Changes in point 9 of ISO 9459-2 (pages 21, 22, 23 and
24), Step3 (Energy drawn off):
Calculations for Day 1:
Qc(1) is calculated according to:

Table 5 LTPP according to CSTG test method (P2)

(Yearly energy delivered by the system GJ/year)
Calculation
Load
procedure
Stockholm
Davos
Pattern
according to:
ISO 9459.2
1
6.72
10.18
2
4.60
6.68
Using function
(eq.6 and 7a,7b)
3
6.39
9.96
(Bourges et all,
3
6.42
9.96
1991). Using
function (eq.6)

V'

Q c (1) = Q(1) f (V )dV

(9)

where V is determined by two conditions:

Q c (1) Vload w C pw ( t load t main )

V' Vload

(10)

Calculations for Day 2 and subsequent days:

Qc(2:part1) is calculated according to:
V'

Q c (2 : part1) = Q(2 : part1) f (V)dV

(11)

and Qc(2:part2) is calculated according to:

V'

Q c (2 : part 2) = Q(2 : part 2) g (V )dV

(12)

The value V is determined when the total energy

extracted is calculated according to:
Q c (2) = [Q c (2 : part1) + Q c (2 : part 2)]

(13)

verifying the following two conditions:

Q c (2) = Vload w C pw ( t load t main )
(14)

V' Vload
The calculations done according to this procedure give
lower values than with one load volume extracted every
day, but will give higher values than with the
consideration of a minimum load temperature.
Using the test results reported by NCSR for the P2
system, values obtained according to calculations
considering the three load patterns described, are given in
Table 5. From this Table it is possible to conclude that
the use of a functional form for characterisation the drawoff profile is possible and it allows for the determination
of the draw-off profile in smaller intervals than 0.1Vstore.
It can also be seen that calculations for Load pattern 3 are
identical using the modification of ISO 9459.2 procedure
as suggested above or the calculation procedure from
Bourges et all (1991).

These results support the following recommendations:

i) use of a functional form to characterise the draw-off
profile allowing calculation in smaller steps than 0.1
of the storage volume as is considered in the standard.
ii) simplification of the CSTG test method because the
test for the determination of mixing draw-off profile
does not give significant additional information on the
behaviour of the system.
Another aspect that should be considered in a future
revision of ISO 9459.2 is the need to report an
uncertainty boundary on the LTPP of the CSTG test
method. The calculation of this uncertainty is not
considered in the calculation procedure of the present
version of the standard but it can be determined,
according to reference (Bourges et all, 1991) if the crosscorrelation matrix associated to the determination of
Input-Output parameters is obtained.
Also the uncertainty associated with LTPP of DST test
method is not determined based on a Weighted
Regression, although this is considered as a
recommendation for future improvement of DST (Naron,
D.; 1999a)
4. CONDITIONS FOR PARAMETER
IDENTIFICATION AND LTPP
4.1. CSTG conditions.
Within the project, CSTG tests were performed according
to ISO 9459.2 and the common procedures in each test
laboratory. This means that the above recommendations
for revision of the test method (ISO 9459.2) were
established from the discussion and analyses of the test
results. Some Laboratories already used WLSR but others
just used LSR. It will be necessary for adoption of WLSR
to establish a common methodology for the calculation of
uncertainty of measured values, i.e., Q, H, Ta(av.day) and
Tmains. It is suggested that the recommendations of the
ISO Guide (1995) are followed.
The complete linear model (see equation (1)) was used. A
complete listing of the parameters obtained can be seen in
reference (Carvalho, M.J.; 1999), as well as, the listing of
all LTPP results obtained for each system. As already
referred, the comparison between the two test methods is

Table 6 - Results of LTPP comparison of

DST and CSTG test methods for System P2
Difference (%)
Tested at:
NCSR(CSTG)
and:
NCSR(DST)
SP(DST)
Stock.
-37.7
-17.6
Ath.
-33.2
-14.4

always made considering the application of Load pattern

3 in the CSTG LTPP calculations.
4.2. DST conditions
The systems were tested according to the DST test
method, i.e., ISO/DIS 9459-5. The DST sequences used
for the comparison with CSTG test performed under WP2
or WP3 include only three types of sequences :
Type Ssol_A
Type Ssol_B with auxiliary off
Type Sstore

Parameter identification was obtained with the Insitu

Software package version 2.7. Detailed listing of the
results can be seen in reference (Carvalho, M.J.;1999).
The LTPP was also determined using the Insitu Software
Package. Listing of LTPP results can be found in
Carvalho, M.J. (1999).
4.3. LTPP conditions
The LTPP was determined in both cases according to the
reference conditions of Annex B of prEn 12976.2 (1999):
Desired temperature - 45oC
Cold water supply temperature: as in table B.4 of
CEN prEN 12976.2
Collector tilt angle: 450
Draw-off flow rate: 600 Lt/h
One draw-off of 100% load volume in the evening
(18 h solar time)
Calculations were made for four locations: Stockholm,
Brussels, Davos and Athens. For illustration only the
results of Stockholm and Athens will be presented.
5. LTPP COMPARISON BETWEEN DST AND
CSTG RESULTS
The difference between the results of the two test
methods is calculated according to:
Q DST Q CSTG
100 (%)
(15)
Q CSTG
taking into account that, for solar only or preheat systems,
the CSTG is a validated method as already referred in
section 1 of this paper.
Difference =

5.1. Forced Circulation Systems

5.1.1. Experimental test results
Two forced circulation systems P2 and P3 were tested for
comparison of both test methods. In Table 6 the results of
LTPP comparison for one of the pumped systems (P2) is
indicated.
Unacceptably high differences were observed when the
LTPP was determined based on the tests performed at
NCSR. These high differences can (again) be explained
by this problem in the DST testing rig. The system was
then transported to the Netherlands and tested at TNO.

The high differences shown between the results of the

two methods for the tests performed at NCSR were
explained by a problem in the DST testing rig1 that was
only detected after the transport of the systems to another
test laboratory. The comparison of DST results from SP
with CSTG results obtained at NCSR shows lower
differences (see Table 6) but still very high. At SP
laboratory there was a suspicion of malfunction of the
control unit of the system.
In Table 7 the results of LTPP comparison for the other
pumped systems (P3) are listed.
Table 7 Results of LTPP comparison of
DST and CSTG test methods for System P3.
Difference (%)
Tested at:
NCSR(CSTG)
and:
NCSR(DST)
TNO(DST)
Stock.
-31.2
4.5
Athens
-34.6
-3.0
The comparison of CSTG test results from NCSR with
the results obtained for DST test at TNO shows
differences lower then 5%. Using the results of CSTG at
NCSR and the results of the DST test at TNO, the
difference between the two test methods for different load
volumes was determined. It was possible to see that these
differences are within 5% for load volume equal or
greater the store volume but there is a tendency of
underestimation of the DST test method when the load
volumes are lower then the store volume.
5.1.2. Simulated test results
In order to complement the test results specially in the
case of system P2, simulation of an identical system was
performed. The system behaviour was simulated with a
simulation programme developed by TNO called ZBOIL.
With this programme it was possible to generate test
sequences of DST test method and test days of CSTG test
method. With these simulated data, parameter
identification for both test methods has been carried out.
From the simulation results it is possible to establish the
comparison between LTPP DST and LTPP CSTG results.
Different results can be generated. To illustrate the result
the following options were considered:
1

Existence of a leaking magnetic valve in the DST-testinstallation.

1.

DST LTPP obtained with parameter identification

based on the Athens test (DST_Ath. Test).
2. CSTG LTPP obtained with I-O parameter
identification based on the Athens test (CSTG_Ath.
test) and CSTG LTPP obtained with I-O parameter
identification based on the Stockholm test
(CSTG_Stock. Test)
The next table (Table 8) shows the percentage difference
in results according to Eq.15. In the case of simulated test
sequences differences are mostly within 5%. Other
options gave similar results.
Table 8 Comparison results for a simulated
pumped system (Difference (%))
DST_Athens test
CSTG_Ath. test
CSTG_Stock. test
Vload (l)
Ath.
Stock.
Ath.
Stock.
250
-0.8
1.1
1.3
5.9
300
-0.2
1.3
0.7
4.1
400
-0.1
4.3
1.6
7.0

5.1.3. Comments on forced circulation systems results.

Taking into account that the determination of
Correlation factors between the two test methods was
one of the objectives of the project, it was considered that
this determination should not be influenced by the test
results that are clearly not acceptable and which are either
explained by problems with the test rigs or by a
malfunction of the system.
Based on the experimental results obtained for system P3
(i.e., CSTG test performed by NCSR and DST test
performed by TNO) and on simulation results for a
pumped system identical to system P2 it is possible to
conclude that the difference between the LTPP results for
DST and CSTG test method is within 5% for load
volume equal or greater the store volume. When the load
volume is greater then the store volume the differences
are all within these values as well as for load volumes
lower then the store volume but within 80% of the store
volume. Greater differences are observed for lower load
volumes.
For determination of Conversion Factors between the two
methods the following equation was considered:
Q DST = (a a )Q CSTG

(17)

The parameter a was determined for different load

volumes. The values obtained showed that the parameter
depends on the ratio Vload/Vstore showing almost constant
values for Vload Vstore ,i.e., Vload/Vstore1.
For load volumes lower then the store volume it would be
advisable to obtain more test results in order to clarify a
real dependency of a in Vload/Vstore. This statement is
justified by the fact that for the other systems this
dependency is not clear.

The parameter value considering all load volume values

higher then the store volume is indicated in Table 9.
Table 9 - Parameter a value for pumped systems
Vload [l]
Vload/Vstore
a
a
1.004
0.004
Vstore
1.00
5.2. Thermosyphon systems
5.2.1. Experimental results
Only one thermosyphon system type has been tested,
however these were two identical systems. In the first test
laboratory the two systems were tested according to both
the CSTG and the DST test method.
Using the CSTG results of the test performed at INETI
the comparison between CSTG and DST for the two
systems is also determined in relation to the DST tests at
ITW, TNO and NCSR.
In Table 10 the results for Stockholm and Athens are
listed. From Table 10 it is possible to see that, for both
systems T2 and T3, the difference between the results of
CSTG test method and the different results obtained by
the DST test method (in different test laboratories) show
similar results.
In order to analyse the differences for load volumes lower
or higher then the store volume the LTPP with CSTG and
DST was done for different load volumes For system T2
the results are within 5 and 14% and for system T3 within
-2% and 6% showing a behaviour that does not show
dependence on the load volume. Considering for each test
comparison the average over all climates considered in
the LTPP, the results are well within 0 and 10%.
Table 10 Results of LTPP comparison of DST and
CSTG test methods
Difference (%) System T2
INETI(CSTG)
Tested at:
INETI(DST)
TNO(DST)
and:
Stock.
8.2
7.5
Ath.
5.8
5.4
Difference (%) System T3
Tested at:
and:
Stock.
Ath.

INETI(CSTG)
INETI(DST)

TNO(DST)

NCSR(DST)

2.0
2.8

2.7
3.7

-0.2
2.8

5.2.3. Comments on Thermosyphon systems results.

An average behaviour of the two systems was considered
to be characteristic for thermosyphon systems. An
average difference between the two test methods is found
to be within 0 and 10% and there is no dependency in
load volume. Based on the results of both systems and
considering Eq. 17, it was possible to determine a
conversion factor independent of the load volume.

Table 11 - Parameter a value for thermosyphon systems

a
a
1.056
0.004
5.3. ICS System
5.3.1. Experimental test results
The ICS system was tested at CSTB according to both the
CSTG and the DST test method. At SPF the system was
tested according to the DST method only. Table 12 show
the comparison results:
Table 12 Results of LTPP comparison of DST and
CSTG test methods for System I2
Difference (%) System T2
Tested at:
CSTB(CSTG)
and:
CSTB(DST)
SPF(DST)
Stock.
2.5
-3.3
Ath.
2.0
-1.9
It is possible to see that when the results of DST of the
two test laboratories are compared with CSTG the results
the differences are all within 5%. The differences were
also determined for different load volumes. It can be seen
that these differences are also within 5%.
5.3.2. Simulated test results
In order to confirm the experimental results, due to the
short test data obtained at CSTB for the CSTG test
method, simulation of an ICS system was used to obtain
test sequences of both methods.
The simulation programme used was developed by TNO
and it is referred here as ZICS. This system simulation
has been based on the characteristics of the ICS system
used for WP1-simulations (Naron, D. et all; 1999a). The
following characteristics were considered:
Collector Area = 2.2 m2

Collector efficiency curve: eta0=0.769;

Ca=1.623W/m2; Cb=0.091 W/K2 m2
IAM: equivalent to Ambrosetti equation r=0.47
Storage Volume = 186 litre
Us=5 W/K
Using the simulation program it was possible to generate
test sequences of DST test method and test days of CSTG
test method. With these simulated data, parameter
identification for both test methods has been carried out.
From the simulation results it is possible to establish the
comparison between LTPP DST and LTPP CSTG results.
A methodology similar to the one adopted for the forced
circulation systems was used. Table 13 shows the
percentage difference in results according to Eq.15.
The CSTG parameters of the simulated tests were
obtained considering, according to ISO 9459.2, that:
6 pm

H=

I
6 am

(18)

Table 13 Comparison results for an ICS simulated

system (Diferrence (%))
DST_Ath. Test
CSTG_Ath. test
CSTG_Stock.test
Vload (l)
Ath.
Stock.
Ath.
Stock.
140
10.6
13.8
11.7
15.7
170
9.8
15.2
10.7
17.4
200
8.8
14.2
10.4
18.4
250
7.6
11.8
9.8
15.7
This means that IAM (Incidence Angle Modifier) effect
will affect CSTG results. In the case of DST there is a
recommendation from WP1 (Naron, D. et all; 1999a) to
use radiation values for LTPP which are already
corrected for IAM effect if the parameter r of Ambrosetti
formula is higher the 0.4 and this was done for the system
simulated. In order to see if differences between CSTG
and DST can be reduced, simulations for generation of
the CSTG results were repeated considering that the daily
radiation incident on the collector is:
6 pm

H=

I IAM

(18)

6 am

The I-O parameters obtained are given in the following

table (Table 14) for both cases.
Table 14 I-O parameters for the ICS simulated system
Without IAM
Test
a1a1
a2a2
a3a3
location
Stockholm 1.310.02
0.150.06 -0.590.51
Athens
1.450.03
0.150.03 -1.840.36
With IAM
a1a1
a2a2
a3a3
Stockholm 1.520.01
0.160.01
0.10.06
Athens
1.530.01
0.160.01
-0.020.1
This table shows that the new parameters are much closer
for both test locations, which is understandable since
IAM influence will be different in each of these locations
due to the different latitude values. LTPP calculations
were repeated for CSTG and in the next table the new
comparison values can be seen.
Table 15 Comparison results between the two test
methods considering IAM correction in CSTG test
(Diferrence (%)).
DST_Ath. test
CSTG_Ath. test
CSTG_Stock. test
Vload (l)
Athens
Stock.
Athens
Stock.
140
1.2
1.7
1.2
1.7
170
-0.1
0.8
-0.1
1.6
200
-1.4
-0.9
-1.4
-0.9
250
-3.9
-4.3
-3.9
-4.3

Now differences are within 5% as was the case for the

experimental results. These results show that IAM
introduction in CSTG should also be considered and
shows lower differences between CSTG and DST results.
5.3.3. Comments on ICS systems results.
One important conclusion of the ICS simulated results is
the need to determine the CSTG parameters using
radiation data corrected for the IAM effect when the
collector has a strong IAM effect (e.g. Ambrosetti
equation with r>0.4). Based on the results of the system
tested and of the system simulated the parameter a (Eq.
17) was determined. Since no dependence on Load
volume was observed only one a value was determined.
Table 16 - Parameter a values for for ICS systems
a
a
1.022
0.004
6. FINAL CONCLUSIONS
During 1997 and 1998, five systems with different
characteristics (two pumped systems, two thermosyphon
systems and one ICS system) were tested in three
laboratories. Each laboratory analysed the data from tests
using two test methodologies, currently designated as the
CSTG and DST test methods, described in the ISO
Standard 9459, part 2 and part 5., respectively. The Long
Term Performance Prediction LTPP was calculated
based on the results of the tests. The LTPP calculations
were based on the identification of parameters as
considered in the present version of ISO standards, ISO
9459.2 and ISO 9459.5. Recommendations for exclusion
of parameters are not considered in this analyses although
recommendations are given for future consideration of
use of simplified models based on the statistical
significance of the parameters. Reference conditions of
prEN12976-2 (1997) were followed as closely as
possible.
The conclusions on the analyses of test results for the
different systems are:
i) Pumped systems
It is not possible to draw conclusions on the
comparison of the two methods based on results
which show high differences (P2: 20-40%; P3: 3040%) which were due to problems in the test rig or
due to malfunction of the system.
For clarification of the comparison results of the two
test methods, a simulation of a P2 equivalent system
was used. The differences observed for the simulation
results are within 5% when a load volume equal or
greater the store volume is considered., as well as, for
load volumes lower but within 80% of the store
volume. For lower load volumes higher differences
were observed.

In the case of system P3 it is possible to compare the

results of CSTG for the 1st test Laboratory (NCSR)
with the results of DST in the 2nd laboratory (TNO).
In this case the differences observed are within 5%
when a load volume equal to the store volume is
considered.
ii) Thermosyphon systems
It is possible to say that for thermosyphon systems, on
the average, the results are well within 0 and 10%.
Only test results were used for the determination of a
Correlation Factor
iii) ICS systems
From all the test results presented it is possible to
conclude that the differences between the test
methods, DST and CSTG, for ICS systems, are within
5% when load volume is equal to store volume.
Simulations were performed which pointed out the
importance of IAM in the comparison of the two test
methods when the parameter r of the Ambrosetti
equation is higher then 0.4. In this case the
identification of parameters for the CSTG test must be
made considering the radiation corrected by the IAM
effect.
Considering the claimed uncertainty of LTPP for each
method:
i) for CSTG: 5% for clear climates and 10% for
cloudy climates (Bourges et all, 1991)
ii) for DST: 5% for most cases and 10% for extreme
cases (Naron, D. et all, 1999a; Naron, D., 1999b)
Differences of the order of 7% in the best case and
14% in the worst case would be expected. The results
obtained are within these limits.
For all systems there is a slight tendency of
overestimation of DST in relation to CSTG test method.
A relation between the LTPP of the test methods can be
determined and is given by:
Q DST = (a a )Q CSTG
The parameter a has the following values:
Table 32 - Parameter a values for different load volumes
Type of system
Condition
a
a
Forced circulation Vload Vstore
1.004
0.004
Thermosyphon
all Vload
1.037
0.003
system
ICS system
all Vload
1.056
0.004
In the case of Forced circulation systems, when
Vload < 0.8 Vstore, the parameter a values determined
are higher indicating a stronger tendency for
overestimation with the DST test method. This result may
need further research.
The parameter a determined is based on LTPP results
considering the present version of ISO Standards, ISO
9459-2 and ISO 9459-5, i.e., without any model

simplification based on statistical significance of the

parameters.
When a revision of these two standards is undertaken for
inclusion of the recommendations listed in section 3 of
this paper, as well as, in the other main reports of this
project (Naron, D. et all, 1999a; Naron, D., 1999b), a new
evaluation of the parameter a is made.
As referred along the analyses made in section 5., these
results are only possible following reference conditions of
prEn12976-2 and also taking care that:
i. the LTPP of CSTG test method follows the above
referred Load Pattern 3
ii. if the information available on the collector IAM
corresponds to Ambrosetti equation with r > 0.4, the
parameter identification in the CSTG test method
should be done considering the daily radiation values
corrected for the IAM effect.
iii. the LTPP of the DST test method maintains the
auxiliary off.
iv. the LTPP of the DST test method uses a 100% drawoff at 18 h solar noon.
It is also important to notice that a criterion for
acceptance/rejection of test results should be adopted
based in a Figure of Merit to be developed in future for
DST test method and using WLSR for CSTG test
method.

REFERENCES
Bourges, B.; A. Rabl; B. Leide; M.J. Carvalho; M.
Collares Pereira (1991) Accuracy of the European solar
water heater test procedure. Part 1: Measurement errors
and parameter estimates. Part 2: Prediction of long-term
performance, Solar Energy, vol.47, pag.1-25
Carvalho, M.J. (1999) Final Report WP2: Comparison
with CSTG test Methods, INETI, Portugal
Dynamic System Testing, Program Manual, version 2.7
InSitu Scientific Software 1997
ISO 9459.2 (1995) Solar heating Domestic water
heating systems Part 2: Outdoor test methods for
system performance characterisation and yearly
performance prediction of solar only systems
ISO/DIS 9459 .5 (1996) Solar heating Domestic water
heating systems Part 5: System Performance
characterisation by means of whole-system tests and
computer simulation
ISO Guide (1995), Guide to the expression of uncertainty
in measurements

NOMENCLATURE

Naron, D.J.; B. van der Ree (1999a), Bridging the Gap,

Final Report of Work Package 1: Definition of Scope,
TNO, Delft, The Netherlands.

cpw
f(V)
g(V)
H

Naron,D.J. (1999b) Bridging the Gap; TNO work done

in Work Package 3: Experimental Validation, TNO,
Delft, the Netherlands

specific heat capacity of water [J/(kg.K)]

normalised draw-off temperature profile
normalised mixing draw-off temperature profile
daily solar irradiation (radiance exposure) in the
collector aperture [MJ/m2]
Q
useful energy extracted from the system [MJ]
Qc
energy contained in a volume of water Vc [MJ]
ta
ambient or surrounding air temperature [C]
td
water temperature of load drawn off [C]
tmains cold water supply temperature [C]
Us
storage tank heat loss coefficient [W/K]
Vc
volume of daily hot water consumption [l]
Vs
fluid capacity of the store [l]
w
density of water [kg/m3]

standard deviation of parameter models

QDST Yearly energy delivered by the system determined
according to DST test method [MJ/year]
QCSTG Yearly energy delivered by the system determined
according to DST test method
Subscripts
(av) average (mean) value of parameter

Naron, D. (1999c), 'Bridging the Gap' Final Report of

Work Package 3: Experimental Validation, TNO, Delft,
The Netherlands
prEn 12976 (1999) Thermal solar systems and
components. Factory made systems part 2: Test
methods
Press, W. ; S. Teukolsky; W. Vetterling; B. Flannery,
(1992) Numerical Recipes, Cambridge University Press
Recommendations for performance and durability tests of
Solar Collectors and Water Heating Systems (Part A),
Edited by E. Aranovitch, D. Gilliaer, W.B. Gillett,
J.E.Bates (1989)

ACKNOWLEGMENTS

SDHW Simulation Programme ZBOIL - version 4.5,

TNO, Delft, The Netherlands, 1995.

The authors want to thank Dr. Richard Marshall (UWCC)

for their suggestions during elaboration of this work.

ICS Simulation Programme ZICS version 1.3, TNO,

Delft, The Netherlands, 1994.