Correct calculation of temperature factor

[cwhanse]

The use case descriptions I provided assumed that the temperature coefficient gamma was a positive number. That assumption conflicts with pvlib-python's conventions. This calculation should change to

return _zero_div(1 + gamma * (t_actual - t0), 1 + gamma * (t_ref - t0))

Is there a numerical test of compare_predicted_and_actual that checks correctness of the adjusted energy? Or should I come up with one?

[wholmgren]

Good catch.

I'm not very familiar with testing compute.py because of the heavy dependencies. I thought this test would fail if those signs are changed but it does not.

https://github.com/SolarPerformanceInsight/solarperformanceinsight/blob/f75f54e4ca45bed31925f7cd977bd809a76a2245/api/solarperformanceinsight_api/tests/test_compute.py#L785-L816

[wholmgren]

Ah, I forgot the monthly temperature correction is different. I think the - in this line should also be changed to a +:

https://github.com/SolarPerformanceInsight/solarperformanceinsight/blob/f75f54e4ca45bed31925f7cd977bd809a76a2245/api/solarperformanceinsight_api/compute.py#L816

The test in the comment above does fail when I apply that change.

>       assert ser.loc["weather_adjusted_energy"] == 56249.625
E       assert 56250.375 == 56249.625

source data for that test is

https://github.com/SolarPerformanceInsight/solarperformanceinsight/blob/f75f54e4ca45bed31925f7cd977bd809a76a2245/api/solarperformanceinsight_api/tests/test_compute.py#L150-L166

[wholmgren]

This test coarsely checks the correctness of the corrected energy

https://github.com/SolarPerformanceInsight/solarperformanceinsight/blob/f75f54e4ca45bed31925f7cd977bd809a76a2245/api/solarperformanceinsight_api/tests/test_compute.py#L1031-L1073

There are 270 permutations run within that test. Perhaps we need a more precise test for a single permutation.