Validations

Currently the only validation of the simulation results were done in the context of the PVsyst software development (1996, version 2)

We are not aware of any third party validation measurements. The reason for this is that only very few people perform this kind of validation, and if they do, the results are not made public.

Validating the simulation results is not an easy task, because:

- The simulation results depend on many parameters, which may be set at any value in order to get the expected result... (for example: PV module quality loss, or mismatch, or soiling),

- The real meteo values when running are rarely known (or sometimes not recorded with sufficient care) and operating parameters measurements are also subject to errors.

- The real performances of the components used (especially the PV modules) is rarely checked in detail at the installation time.

For getting reliable conclusions, the measurement conditions and the validation process should be clearly defined. Namely comparisons between measurements and simulation should be performed in hourly values.

We present here validations performed with old versions of PVsyst (1996) on 7 Swiss installations. The yearly power was predicted with an annual accuracy of the order of +/- 5%, except with an installation involving amorphous modules (which were not well modeled in this early version).

We analyzed very roughly several plants in Geneva, over 2-3 running years, usually designed with PVsyst using the straightforward simulation (i.e. with all "default" loss values), and normalizing the results to the real monthly irradiation the yearly results were within +/- 5%.

More recently (2008, V 4), we also analyzed the data of a 10 kWc system of amorphous modules, and closely compared them to the simulation. The conclusion is that the calculation procedures are relatively reliable (within 1-2% MBE accuracy over one year). PVsyst tries to use the best models (or the most suited) for simulating each part of the system, and taking each behavior into account.
But the parameters you put in the simulation are the main sources of uncertainty (mainly the meteo data, usually known at 5-10%, and also the real behavior of the PV modules with respect to the specifications). For the modules, PVsyst uses a PV model (experimentally studied by myself) with a very good accuracy, provided you put the good parameters...

The meteorological annual variability is around +/-5%, but the last few years are sometimes considered exceptional. Therefore any production warranty should always be done under condition of re-normalizing the real results according to actual meteo data (which become more and more available from satellites data, but often not free).