You are on page 1of 6

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/262826975

Field performance of PV modules quality


control process

Conference Paper · September 2008


DOI: 10.4229/23rdEUPVSEC2008-4AV.3.33

CITATIONS READS

3 67

3 authors, including:

Rodrigo Moretón Eduardo Lorenzo Pigueiras


Universidad Politécnica de Madrid Universidad Politécnica de Madrid
20 PUBLICATIONS 113 CITATIONS 115 PUBLICATIONS 1,558 CITATIONS

SEE PROFILE SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Rodrigo Moretón
letting you access and read them immediately. Retrieved on: 22 June 2016
FIELD PERFORMANCE OF PV MODULES QUALITY CONTROL PROCESS

R. Moretón, E. Lorenzo, F. Martínez-Moreno


Instituto de Energía Solar, Universidad Politécnica de Madrid 28040, Madrid, Spain.
email: rodrigo.moreton@ies-def.upm.es

ABSTRACT: In the current photovoltaic market, the importance of module quality control process is growing
although its implementation in large size plants is not evident and it can also produces delays in the setting up of
the power plant. In this article, the authors describe a simple module testing process aimed to be applied in the
field. It allows more than 1MW per day testing rate and a repeatability better than 1,5%. The process has been
tested in the facilities of the Instituto de Energía Solar of the Universidad Politécnica de Madrid, and it has been
applied in the field in a 12 MW photovoltaic plant in Spain. The results show different values from what was
expected if attending to manufacturer data, in particular a higher dispersion in power values. This make the
authors think that classifying the modules on the basis of their flash-report power has no relevance.

Keywords: PV modules, Quality control process, flash-report.

1. INTRODUCTION

Independent measures show repeatedly the existence of modules with electric characteristics below the
supposed values in the current photovoltaic market [1]. For the particular case of the Spanish market, dominated
by megawatts-size power plants, photovoltaic developers usually require independent tests of certain sets of
modules in certified laboratories (CIEMAT, CENER; IES-UPM). This fact entails transporting the modules to
the laboratories and subjecting them to standardized measure processes. In practice, this procedure can only be
applied to a little set of modules (typically 20 for each project), and it represents a real delay risk.

The main objective of this work has been to develop and test a new photovoltaic modules quality control
procedure for testing under real conditions, based on the previously gained experience of the IES-UPM [2]. This
procedure is aimed to be applied directly in the field by the installation companies. So, the number of analyzed
modules increases significantly without affecting their installation rates. In essence, the purpose is to measure
up to 200 modules per day, which is equivalent to a modules installation rate of more than 1MW per day.

2. PROCEDURE

The basic idea is to analyze modules by comparing each of them with a reference module of the same
technology. This reference module has been previously calibrated by a certified institution. The output power,
measured by an I-V curve tracer, is the control parameter. Both modules are tested almost simultaneously.
Then, the measurements are converted into Standard Test Conditions (STC) using [3], [4], [5]:

1 − γ (TRM − 25º C )
*
* PRM G
PAM = PAM × × RM ×
PRM G AM 1 − γ (T AM − 25º C )

where * refers to STC, MR and MA to the reference and the analyzed module respectively, and γ is the
variation coefficient of the power with the temperature, with a typical value of 0,5%/ºC; G is the irradiance and
T the cell temperature.

The irradiance is obtained from the short-circuit current of a second reference module, measured with the
voltage dropped in a calibrated shunt according to:

I shunt G *
G = Vexp × ×
Vshunt I *

where Vexp is the experimental voltage measured in the shunt, Ishunt and Vshunt are respectively the
calibration values of the shunt current and voltage, and I* refers to the calibration value of the reference
module’s short-circuit current.
Cell temperature is obtained from the open-circuit voltage of the reference module using:

*
* VOC − VOC
TC = TC +
NS × β

where NS is number of series connected cells, and β is the variation coefficient of the open-circuit voltage
with the temperature, with a typical value for monocrystalline silicon of 2,3 mV/ºC.

3. FIELD IMPLEMENTATION

A first experiment was conducted at the IES-UPM facilities to test the new procedure and reduce the
possible error sources [6], [7]. Six modules were measured daily as described before, from May to December
2007. The main result achieved was the repeatability of the STC power values obtained with this method, better
than 1,5% for G≥700 W/m2. For the southern and central regions of Spain, this irradiance condition is satisfied
during 4 hours per day for at least the 80% of the days.

During the next step, the measure process was applied in a 12 MW photovoltaic plant under construction.
Each day, between 150 and 200 photovoltaic modules were measured on average, and the corresponding power
results were compared with the data included in the flash-report given by the manufacturer. This process was
performed between January and April 2008. More than 2800 modules were tested, reaching 280 modules in a
one single day. It is important to remark that the repeatability obtained during the application in a real situation
has remained under the range of the 2%.

4. RESULTS

Fig. 1 shows the comparison between the power values obtained after the measuring process under real
conditions (PRS) and those given by the supplier (PFR).

240

235
y = 1,0042x
230

225

220
P* RS (W)

215

210

205

200

195

190
200 205 210 215 220 225 230
P* FR (W)

Figure 1: Comparison between real sun power (P*RS) and flash-report power (P*FR)

Tolerance, established at 3%, is delimited for the two red lines. The mean value of modules power (the
better adjust of a function of the type P*RS = a*P*FR; black line in the chart) is slightly superior to the expected
one. However, it is possible to notice a much higher dispersion value when measuring in the field than when
testing in a simulator. Thus, there are units out of tolerance when testing under real sun conditions that are not
detected when measuring with flash.
This fact can be observed better if the power distribution of the tested modules is considered separately.

1,8

1,6
Average power = 219,82 W
1,4
σ = 1,49 %
Number of modules (%)

1,2

1,0

0,8

0,6

0,4

0,2

0,0
200 205 210 215 220 225 230 235
Power (W)

1,2

1,0
Average power = 219,87 W
Number of modules (%)

σ = 2,57 %
0,8

0,6

0,4

0,2

0,0
200 205 210 215 220 225 230 235
Power (W)

Figure 2: Power distribution according to flash-report data (a) and real sun measures data (b).

Fig. 2 represents the laboratory measures power distribution and the outdoor tests power distribution
respectively. As it can be observed, although the power average is the same in both cases, the power dispersion
in the real case (σRS=2,57%) is much higher than the one resulted from the laboratory values (σFR=1,49%). So,
the dispersion losses related to the associations of modules in the field can be higher than the expected ones.

Trying to explain this result, the internal functioning of solar cells has been reviewed. In general, when a
solar cell is illuminated with a short light pulse it has a different behavior than when working under continuous
light. Particularly, the control mechanisms of the internal recombination process of the current carriers can be
very different, if the light pulse length is shorter than the denominated life time of these carriers. This is the
main reason that explains the obtained result and so, although supposing perfectly calibrated instruments, the
STC power value of a single module obtained in a solar simulator, P*FR, can differ from that one reached in a
real condition situation, P*RS [8].
Fig. 3 represents the percentage difference distribution between both values of power [100*(PFR- PRS)/ PRS]
corresponding to the 2817 modules tested during the process.

2,5

Average deviation = -0,43 %


σ = 2,18 %
2,0
Number of modules (%)

1,5

1,0

0,5

0,0
-8 -6 -4 -2 0 2 4 6 8
Deviation regarding P* RS (%)

Figure 3: Distribution of P*FR deviation regarding P*RS

The data distribution fits approximately a Gaussian, typical of random phenomena, with an average value
of E= -0,43% and a standard distribution of σE = 2,18%. The mean value indicates the difference between the
calibration of the manufacturer solar simulator and the calibration of the flash used for the characterization of
both the reference modules used in the outdoor measures (calibration laboratory: CIEMAT). In this case, the
difference of 0,43% is really small, as it is understood considering the calibration differences between
independent laboratories, that can reach values up to 3%.

Considering the real sun estimation as reference, the error associated to the measure of a single module is ±
4,36%. Strictly speaking, the width of the 95% good decision probability interval is ±2*σE. This error margin is
intrinsic to the fact of measuring with flash and, therefore, it is additional to any other possible calibration error.
As this error is significantly larger than the width of the intervals in which the classification processes pretend
to differentiate the modules, it can be concluded that the classifications made on the basis of flash-report data
result irrelevant. In fact, modules are currently in the market with a tolerance of ±3%, so if a 3 class
classification is performed, it should distinguish between three ±1% tolerance intervals, [-3,-1), [-1,+1) and
[+1,+3], which obviously can not be done when each module measurement error is larger.

To achieve a positive feedback of the modules classification, in terms of improving the total efficiency of
the PV generators constituted with them, the modules order according to flash-report values (for example, from
higher to lower power) must be sensibly similar to the modules order when classifying them according to the
real sun values. Fig. 4 shows a very different reality.
3000

2500

Order according to P * RS 2000

1500

1000

500

0
0 500 1000 1500 2000 2500 3000
Order according to P* FR

Figure 4: Order according to P*RS vs order according to P*FR

Each point in the chart is associated with a singular module, and it represents the position in which this
module is in the flash-report power values and the real sun power values classifications. If both classifications
matched, points should be situated on a Y=X line.

Anyway, the number of modules needed to be tested under real conditions to check the manufacturer flash
calibration state with a better than 1% precision is at least of 19 (σE,19 = σE/√19; here σE,19 = 0,5%, so 2σE,19 =
1%). In addition, this data could be understood as the minimum number of modules needed to be measured to
control a set of modules from the same fabrication batch.

5. CONCLUSSIONS

Through this work, it has been developed a photovoltaic modules measure method, which is very useful in
the module quality control processes. It is a simple and direct application in the field test site procedure that
guarantees precision and repeatability, and it also allows a high testing rate. After the control period, it has been
observed a higher dispersion value in outdoor measures than in the laboratory ones. Therefore, laboratory tests
do not result much representative in comparison with the real behavior of the modules under real outdoor
conditions. So the modules classifications made according to these indoor-tested values are not so indicative.

6. REFERENCES

[1] Fraunhofer ISE annual report 2007. Achievements and results. Fraunhofer Institut Solare Energiesysteme.
[2] Zilles R., Lorenzo E., López-Bravo I., Llorens J.D. (1995). Power rating and photovoltaic modules
measurements. Proceeings of the International Solar Energy Society (ISES). Solar World Congress, Harare.
[3] Caamaño E., Lorenzo E., Lastres C. (2002). Crystalline silicon photovoltaic modules: characterization in the
field of rural electrification. Progress In Photovoltaics: Research Applications 10. 481.
[4] Caamaño E., Lorenzo E., Zilles R. (1999). Quality control ofwide collections of PV modules: lesson learned
from the IES experience. Progress In Photovoltaics: Research and Applications 7. 137.
[5] King D. L., Boyson W.E., Kratochvil J.A. (2004). Photovoltaic array performance model. Sandia National
Laboratories, Report SAND 2004-3535.
[6] Zdanowicz T. Common falls and error sources occurring during outdoor monitoring of PV modules. 20th
European Photovoltaic Solar Energy Conference.
[7] Betts T., Bliss M., Gottschalg R., Infield D. (2005). Consideration of error sources for outdoor performance
testing of photovoltaic modules. 20th European Photovoltaic Solar Energy Conference.
[8] Fabero F., Vela N. , Alonso-Abella M., Chenlo F. (2005). Characterization of recent commercial
technologies of PV modules based on outdoor and indoor I-V curve measurements. 20th European Photovoltaic
Solar Energy Conference.

You might also like