Synthesized data doesn't change

[S.Faulkner]

I would like to see how much effect the synthesis of hourly data has on the results (for a large grid-connected tracking system, in particular the inverter clipping loss), so I tried producing multiple hourly datasets from the same monthly data. I used "Synthetic hourly data generation" under Databases, changed the file name and reran the synthesis (Execute Generation) multiple times. However the resulting .met files all were the same. Confirmed by exporting to csv files.

How can I get the synthesis to run again to produce different .met file results?

I'm using V6.39.

Help says "Please note that this generation is a fully random process: two successive generations performed from the same monthly data will result in completely different years."

[André Mermoud]

The values were indeed random in the version 5, with the algorithm of PVsyst. This resulted in a variation of about 0.5% to 1% annually with grid-connected systems (may be more with stand-alone systems, where non-linearities are higher).

Now PVsyst uses the algorithm from the Meteonorm DLL. In this tool they have chosen to always take the same initial number for the random generator.

We have asked them for the possibility of changing the random generator initialization. However we did not yet implement this in PVsyst.