Negative "POA horizontal ratio" when sun behind array

[adriesse]

Describe the bug
The function poa_horizontal_ratio gives negative values when the sun is behind the array.

To Reproduce
pvlib.irradiance.poa_horizontal_ratio(60, 90, 60, 270)
-0.99999999999999933

Expected behavior
pvlib.irradiance.poa_horizontal_ratio(60, 90, 60, 270)
0.0

Versions:

• pvlib.__version__: '0.5.2'
• pandas.__version__: '0.19.2'
• python: 3.5.1

Additional context
Similar problems were fixed in other irradiance functions already (#526).

[wholmgren]

I think we considered this behavior in #535 and decided against limits in this function. 9a74404

[adriesse]

Sorry, I realize I ran the example on an older version, but the new version looked the same here:

https://pvlib-python.readthedocs.io/en/latest/_modules/pvlib/irradiance.html

I guess that address currently is not showing the latest release?

[wholmgren]

I believe that link shows the latest GitHub master. #535 went into 0.6.0, so I think release vs. master does not matter in this case.

I might not have been clear before. My initial sky diffuse bug-fix proposal included limits in poa_horizontal_ratio. That's shown in commit 9a74404. Based on reviewer feedback (in particular this comment from @cwhanse), I ultimately left poa_horizontal_ratio unchanged and instead applied limits within each diffuse function. That's shown in final #535 diff.

There may be applications where the negative value is useful (filtering data, bifacial), so I'm not convinced that this is really a bug. If we keep the behavior we could add a sentence to the documentation explaining the output when the sun is behind the panel.

[adriesse]

Ok, I'll just close this then.

[wholmgren]

@adriesse would you suggest that we change the behavior or at least the documentation? If so, happy to keep the issue open.

[adriesse]

Slightly changing both would be best in my view. Allowing aoi_projection to go negative is ok, but the docstring could be expanded to explain the meaning of negative values. I think that poa_horizontal_ratio should only use non-negative values of aoi_projection like the other users of that function. But I wouldn't put a high priority on it--I just noticed it as I was browsing the file looking for something else.

[kandersolar]

poa_horizonal_ratio was removed in #2021. Shall we close this issue?