Pvwattsv8 broken with major version upgrade from 3.0.2 to 4.x

[joules-michael]

Hello, I am finding that my Pvwattsv8 models are outputting all zeroes after upgrading the package from version 3.0.2 to any of the currently available versions 4.x.

Repro case shown in Google Colab. I can get results comparable to those from the online calculator with 3.0.1 and 3.0.2, but not 4.0.0 through 4.2.0.

None of the parameters referenced in the change log for Pvwattsv8 since 3.0.1 appear to be causing the issue.

[cpaulgilman]

@joules-michael I confirmed these results and am investigating. Will follow up as soon as I know more.

[brtietz]

This rhymes with NREL/ssc#973. That said I'm having trouble identifying which inputs would have changed between versions.

[cpaulgilman]

This is caused by NREL/ssc#830, which breaks pvwattsv8 for non-annual simulations.

A temporary workaround is to use the output ac_pre_adjust in place of gen.

[joules-michael]

Thank you for your suggestion.

I am finding that I get slightly different results across the two major versions when I compare ac_pre_adjust. Is this expected?

I modified the final line of the forecast function in the repro case, from

        return np.array(system_model.Outputs.gen).reshape(-1, 1)

to

        ac_pre_adjust = np.array(system_model.Outputs.ac_pre_adjust).reshape(-1, 1)
        gen = np.array(system_model.Outputs.gen).reshape(-1, 1) * 1000.0
        return np.concatenate((ac_pre_adjust, gen), axis=1)

So it looks like ac_pre_adjust is equivalent to gen (at least for my toy example), but the values of ac_pre_adjust differ across version 3.x vs 4.x

[cpaulgilman]

PySAM Version 4.2.0 (SSC 280) includes this fix to the solar position algorithm that causes small differences in the calculated output for the Pvwattsv8 and other solar performance modules compared to older versions.

[joules-michael]

Great, thank you for clarifying!

[dguittet]

@joules-michael We fixed another issue with non-annual simulations for version 5.0.0