Median P50 AC Energy vs. AC Energy Injected into the Grid

[kjs55]

It is reported in literature [1] that the p50 (median) derived from a Monte Carlo- or convolution-based uncertainty analysis [2] is more appropriate to use (and not always the same) as the primary AC energy into grid value (E_Grid simul) that comes out of PVsyst. e.g., Not all the distributions that feed into the Monte Carlo are normal Gaussian distributions (e.g., the Gamma-shaped degradation rate distribution from Dirk Jordan [2]-[3]). Does the p50 in the PVsyst p50 estimation directly use E_Grid simul? Or is the p50 derived from the uncertainty analysis? Is it really a median p50 (or is it simply E_Grid simul)?

[1] https://www.researchgate.net/publication/318259323_A_Framework_to_Calculate_Uncertainties_for_Lifetime_Energy_Yield_Predictions_of_PV_Systems/

[2] https://www.eupvsec-proceedings.com/proceedings?paper=19917/

[3] https://onlinelibrary.wiley.com/doi/10.1002/pip.3566/

Edited September 8, 2022 by kjs55
Add convolution option (requires significantly less computational work)

[Michele Oliosi]

Hi @kjs55,

In PVsyst we assume the normal Gaussian distribution, and the assumed P50 value (since the distribution is Gaussian, it is the median and the average) is E_Grid simul. In order to perform a Monte-Carlo evaluation, multiple simulations should be run, which is not how the tool works in PVsyst at the moment.

Intuitively, I would assume that since the uncertainties are large and partially ad-hoc (and usually the main factor is the climate variability, for which I don't know the customary distribution -- I should check) that the details of the distribution don't matter much. However I can see how when you take into account aging, or when the weather statistics are well-defined, this could change.

But yeah, indeed, adding a Monte-Carlo tool, maybe on the basis of the batch mode, as well as a distribution function selection would be interesting.