The basic required data for the definition of a PAN file are listed in How to establish the PAN files?
The PNom, STC values and temperature coefficients are always available in the datasheets.
The main unknown parameters are the Rshunt and the Rserie, as well as the exponential behavior of Rshunt. If not specified, the default
values of PVsyst are:
= Vmp / (0.2 * (Isc-Imp)). The 0.2 coefficient may be different for other technologies (0.33 for amorphous).
default is 5.5 for all technologies. It should be modified only if explicit measurements of Rshunt at different irradiances are available.
= Mult * Rshunt. In the present time, Mult = 4 for crystalline and 12 for all other technologies. Perhaps a value of 10 for crystalline modules would be more suitable (observed with many indoor low-light measurements) .
is determined according to the low-light performance resulting from the model, in order to obtain a relative efficiency of -3% at 200 W/m2.
NB: the effect of Rshunt and its irradiance behavior is rather marginal when we set the Rserie value according to low-light performance.
Submission of data by manufacturers
Now when receiving data from manufacturers, we use to check the following values:
- The first requirement is that the Pnom value is close to the Imp * Vmp product
. If the difference is greater than 0.2%, we use to redefine Imp = Pnom / Vmpp (therefore Imp is no more in conformity with the datasheet).
- All values Rshunt, RshExp, Rsh(0) and Rserie may be left blank, in this case they will be taken at their default value.
- If some of these values are explicitely specified (which is not advised without a very deep understanding of these parameters), we only accept them "as such" if the resulting low-light performance is lower than the default.
Otherwise we will require a full report of an independent institute, describing complete measurements at 200, 400, 600, 800 and 1000 W/m2 and 25°C.
Analysis of the low-light measurements
Attenuation filters uncertainties
When using a report, we have first to check/correct
the irradiance reference data.
The problem is indeed that the filters are never perfect. They may have inaccuracies either in irradiance and in spectral response.
Then, an error of 1 % on the irradiance means an error of 1% on the relative efficiency (remember that we analyze efficiency differences of fractions of %).
For avoiding this problem we can admit the hypothesis that the short circuit current is proportional to the irradiance (including spectral mismatch).
From experimental basis, this hypothesis is proposed by the Sandia National Laboratories (USA), as a direct measurement of the incident irradiance.
In the "theory", it is the basic hypothesis of the one-diode model.
When doing this the efficiency points - sometimes erratic - become very close to the one-diode model results.
Our techincal report (see below) includes all the file parameters suited for this measured Module.
It allows to adjust the Rserie for matching the measurements at best, and identifying namely the relative efficiency to be fixed in the model.
The report shows the accuracy of the modele for representing the STC measured points, and provides some additional information, namely about the Rshunt behaviour.
The STC parameters shown here are not the parameters of the database, as they correspond only to this specific module under measurements.
Setting in the database
The main result waited from this report is the low-light performance (quantized by the efficiency at 200 W/m2), which will allow to fix the Rserie value in the database.
In the database, we use to keep the values Rshunt and RshExp to their default values.
The last graph gives an information about the choice of Rsh(0), and also the pertinence of the RshExp parameter fixed at 5.5. Some manufacturers propose very high values for RshExp (more than 10 or 12), which boosts the global low-light efficiency when the value at 200 W/m2 is fixed. As we observe a good agreement with 5.5 for most of the crysrtalline modules, we will not accept anymore values higher than 5.5.
Now in the database, the STC basic values (Vmp, Imp, Voc, Isc) are those of the datasheets, different from the measured values shown here. Therefore the Rserie value will be different for getting the waited 200 W/m2 efficiency. In fact the Rserie has to be adjusted for each power class for getting this 200 W/m2 performance.
NB: We use to set the same low-light efficiency for each power class of the modules of a same model. Nothing ensures that this hypothesis is valid (it is probably not the case). However in spite of repeated requests to laboratories, we never got measurements of modules of different power classes nor saw any publication about this subject, so that we don't have any idea of the low-light performance as function of the power (quality of the module).
File comment: PVsyst technical report for parameter analysis
Low_Light_Report.png [ 79.77 KiB | Viewed 9775 times ]
Global fit methodology
Some laboratories perform the measurements proposed by the IEC 61863, i.e. at a specified set different irradiances and temperatures (i.e. 22 measurements at 4 temperatures fron 15 to 75°C).
And they perform a global fit
over all these points, using all the 5 parameters Rshunt, RshExp, Rsh(0), Rserie, muPmpp as variables.
Increasing the number of parameters in the fit allows obviously to get a better match at all the measured points. However this fit may be very unstable: intermediate values present hills and valleys, and a little variation of one of the points may lead to a completely different solution. Therefore:
- Using the STC basic values (of the datasheets) instead of the measured ones in the model may become irrelevant,
- The extrapolation of the measurements of one only module to several power classes becomes very difficult (they use to scale all the the data points and perform the fit again - what is the justification?).
- This extrapolation leads to very different low-light performances between different power classes,
- Some institutes use the same data set
(i.e. measurements of 3 or more modules of the same power), scale the "measured" values according to some algorithms (not described) as a function of the power class desired, and perform the 5-parameters fit on these values for each power class. This doesn't make much sense, as these data are obviously not representative of the real modules of each power class.
- Moreover, the Temperature behavior of the Pmpp issued from the global fit may be significantly different from the direct measurement at 1000 W/m2 and 25°C specified on the datasheets.
Therefore in a general way, we don't accept
the results of these multi-parameters fits for the database, as we don't have a mean for extending it to other modules. The methodology is too different from the standard parameter's choice used by all other manufacturers, and this may lead to irrelevant discrepancies between manufacturers in the simulation results.
Therefore when receiving such a report, we treat the original measured data as for other laboratories, using only the low-light data at 25°C, but applying stable and comprehensive methods for the extension to other power classes.