<< Click to Display Table of Contents >> Navigation:  Validations > Validations on old versions of PVsyst

The seven tested grid-connected installations

We have carried out detailed validations of the programme, using the data from 7 grid-connected systems, whose detailed characteristics appear on table 1. These systems have been chosen, on the one hand for the quality of gathering data, and on the other hand as an attempt to represent a variety of different situations: sizes of 0.5 to 100 kWc, fields in sheds or integration (roofing, facade), types of collectors (Si-mono, Si-poly, or amorphous), types of climates (plain, mountain, etc.), possible near shadings.

The validations have to be carried out carefully using measured data, with an hourly or sub-hourly time step, over sufficient periods (one year). According to the measurement's parameters available, we have tested the various stages of the simulation, giving special attention to those which involve the most delicate physical models.

 

 

Table 1. - Summary of the 7 tested installations:comparisons  between  simulation and measurements.
(A positive MBE indicates that the simulation overestimates the measured values)

 

Incident irradiation models

The first delicate stage is the treatment of the incident irradiation in the collector plane: it involves models for the estimation of the diffuse irradiation (from the global irradiation) and for transposition. These models have been otherwise studied [Perez et al], but proved to be the weakest link in the comparison process, with differences reaching more than ten percent for some of the data used.

For the SIG installation at Geneva, the only site where measurements of the horizontal diffuse irradiation is available, the deviation does not exceed 2 to 3%.

But other sites show differences going up to more than 10%. It should be emphasised that these results are highly dependent on the quality of the instruments used for the irradiance measurements (especially the calibration and sometimes the positioning). We supposed that the standard deviation in hourly values is a good indication of the performances of the model itself, and in this respect, the installation of Marzili (MBE=1% and hourly RMSE < 7.8%) confirms their validity. The mediocre results of the LESO can be explained by the considerable distance (several hundred meters) between the horizontal solarimeter and the measurement in the collector plane.

 

Fig 1a. - Measurement-Simulation Comparison for Incident Irradiation,
Marzili installation, daily values

 

Fig 1b. - Measurement-Simulation Comparison for Incident Irradiation,
Marzili installation, hourly values

 

Array temperature model

The second model is the estimation of the temperature of  the PV field using values of ambient temperature and irradiation. This temperature only acts as an auxiliary parameter in the calculation of the electrical production of the field, and its specification is not critical. Our model, resulting from a simple thermal balance, gives remarkable results. By using the default value suggested by the programme (k=29W/m²k) for all the collectors without back-covering, and by adapting this value for the integrated installations, we obtain, in all cases, errors lower than 1°C, with hourly dispersions of 2 to 4°C at the most.

 

Fig  2. Comparison  of the measured temperature with respect to the model

Marzili Installation.

Array output calculation

The electrical output, measured at the collector array terminals (DC energy), is calculated by the simulation on the basis of the incident irradiation (given the shading and non normal incidence corrections), the temperature of the modules, and the collector model (operating at MPP), keeping in mind the ohmic losses of the wiring and the module's mismatch. The excellent results obtained especially in LESO-sheds, the SIG or at the EIV, show that these models work perfectly, at least for silicium crystalline modules.

 

Fig. 3a. - Simulation-Measurement Comparison for the output of the PV field
(LESO-sheds installation, daily values)

 

Fig. 3b. - Simulation-Measurement Comparison for the output of the PV field
(LESO-sheds installation, hourly values)

 

Amorphous tandem technology

The LESO-USSC installation, equipped with a:H-SI tandem amorphous collectors, has shown one of the limits of the collector's model used.  Table 1 shows that the global results of the field are clearly underestimated by the simulation. Fig. 4 shows that the values with strong beam are very well reproduced (MBE=0.5 %), while the points of pure diffuse irradiation are clearly under-evaluated. This is attributed to the fact that this type of collector is more sensitive to diffuse irradiation, which cannot be reproduced by our model which does not take the spectral distribution of the incident irradiation into account.

 

Fig 4a. - Comparisons for the amorphous collectors of the LESO-USSC, strong beam

 

Fig 4a. - Comparisons for the amorphous collectors of the LESO-USSC,  purely diffuse irradiation

Inverter modelling

Inverter's operation modelling does not give any special problems, except the exact determination of the power efficiency profile: the programme offers a specific tool to superpose the characteristic of an existing model (present in the component library) on the measured input/output data of the inverter. The exact profile corresponding to the data can then be adjusted interactively by the user, and it can be assumed that the residual errors of the simulation are not imputable to the calculation model of PVsyst, but only to the inaccuracy of the parametrisation of this inverter efficiency.

 

Fig 5a. - Inverter response, with standard available inverter specification  (Marzili).

 

Fig 5b. - Inverter response, after manual adjustment  (Marzili).

 

Conclusion (1996):

With the exception of amorphous collectors, we can assess that the PVsyst software is capable of simulating a large number of different grid-connected systems with excellent accuracy. By independently testing each of the algorithms, we have tried to identify the uncertainties related to measurement and parameter's determination, and those inherent to the modelling. Finally, it can be stated that the accuracy of the global results of the simulation is of the order of 2 to 3% (MBE).

However, the specificity of the PVsyst programme will rather be the detailed and comparative study of special disturbing effects, and in this use, the quality of the different models allow us to hope for much higher precision.

Stand-alone systems, and particularly the battery model, have not been validated up to now. We hope to find properly measured data, recorded in hourly values, to further analyse the software accuracy.