What is the accuracy of the PVsyst simulation results ?

[André Mermoud]

This is a very difficult question, which strongly depends on the input meteo data, and simulation parameters.

Validation of the model

To evaluate the accuracy of the simulation, the simulation results should be compared to measured data.

The difficulty lies in obtaining high quality measured data for this assessment. The data recorded on existing plants are usually proprietary, and we have no access to them.

The accuracy evaluation is made of 2 components: the measurement accuracy and the modelling accuracy :

Measurement accuracy: irradiance measurements are not easy to perform, and require very well calibrated instruments, which is rarely the case.

The measurements of electrical data are usually more accurate. However misfunctions of the system are often not well documented, and may significantly affect the results which are then compared to the simulation output. This is especially the case when the comparisons are performed on accumulated data (not in hourly values).

Modelling accuracy: With given meteorological data, the main uncertainty is probably the PV module's performance, which is based on

- STC values provided by the manufacturer and temperature coefficients.

- The additional parameters Rshunt and Rserie, which may be either default values or established according to Low-light irradiance performance data (measured on 1-2 modules - representative ?)

- Deviation of the performance of the installed modules with respect to these specifications: is the tolerance respected ? What is the LID or PID or degradation effect ?

The shadings are evaluated by complex models, however their impact on the performance are usually less than 5 to 10%, so that inaccuracies should be less than 1-2%.

Other losses are specified by user defined parameters (wiring loss, inverter behavior, other components, soiling, unavailability), and may be set at any value, therefore not really significant to the accuracy of the simulation process.

Simulations for yield forecast

When using the simulation for predicting the yield of an installation, the main uncertainties are:

- The input meteorological data: nobody knows the wheather for the coming years, and there is some discrepancy between the available historical climatic data (see "Meteo Data Comparisons" on the PVsyst site)

- The real PV module's behaviour with respect to the specified parameters; and the PV module temperature (depending on the mounting mode and possibly the wind speed).

- The operating conditions (soiling, unavailability, etc).

NB: If you have to present yield warrantiess to a customer, you are advised to get rid of the meteo variability (and unprevisibility), by proposing a yield related to the effective irradiation during operation. This will require to foresee a reference Meteo data source (own measurements or satellite) in your contract, for the renormalization of the real yield with respect to the original simulation. The renormalization is rather easy, since for grid connected systems, the yield can be assumed to be proportional to the input irradiation.

According to our own experimental results, we estimate that the simulation process itself has an accuracy of the order of 1 to 2% for the yearly yield.

You can see for example our analyzis of an installation of amorphous modules, but sorry, only available in French.