improve ground reflection when sun is behind panels

[mikofski]

Is your feature request related to a problem? Please describe.
I think that the ground reflection is over-estimated when the sun is behind the panels.

The following image shows a panel facing the ground, tilted 160° down and facing northeast at 70° on Jan-1 from 8AM to noon in Oakland, CA.

The ground reflection on this panel seems quite high when you consider the diffuse sky for a similar panel, tilted 20° and facing southeast at 250°

Another way to think of these panels are as the backside and frontside of the same panel.

I think the reason the backside ground reflection is over-estimated is because the it is not considering that a portion of the ground it sees is shaded, so the GHI should not include the direct component. If I remove the beam and re-calculate I get what seems like a more reasonable value.

But the actual backside ground POA would be in between these because some of the ground visible to the backside is in shade and some is not.

Describe the solution you'd like
This may just be a continuation of #421 or some other bifacial issue or PR. But I still think it would be relevant for the get_ground_diffuse function anyway.

I think we should calculate where the shadow would be on a flat surface, divide the ground into two regions, and add the view factors algebraically.

Describe alternatives you've considered
Well for the front side this delta is obviously nearly zero and probably not measurable, so maybe just wait for #421 or some other bifacial solution to be completed.

Additional context
Probably related to #421

[cwhanse]

I think that the ground reflection is over-estimated when the sun is behind the panels.

I agree, if we're comparing model to reality. The ground-reflected irradiance model in use assumes:

1. the ground visible to the panel is the half-plane extending to the horizon (no obstructions), and
2. the irradiance on this half-plane is GHI (no shadows).
   With these assumptions, the model is correctly implemented in pvl_grounddiffuse.

A better model requires accounting for shadows on the ground that reduce reflected irradiance to the panel. That's the essence of the rear-side irradiance model anticipated by #421. Our analysis and measurements have shown that the rear-side irradiance is ~80% from ground reflections.

[kandersolar]

With infinite_sheds and pvfactors available now, shall we close this? I suppose we could note the simplifying assumptions (Lambertian surface reflectance, also) in get_ground_diffuse's docstring.

[mikofski]

yes, I agree! Thanks Kevin for keeping tabs on stale issues 👍