calcparams_desoto gets lost in the darkness

[markcampanelli]

#340 added support for solving the single-diode equation for an "ideal" PV device that has infinite shunt resistance (equivalent to zero shunt conductance). I see several pvlib issues and discussions concerning very low and zero irradiance causing problems for pvlib.pvsystem.calcparams_desoto, and I just hit this issue myself. Concerning this edge case, which appears to lead to a lot of hacky workarounds, has the community discussed/tried making calcparams_desoto return float("inf") for R_sh when the irradiance is zero?

[markcampanelli]

In a similar fashion, pvlib.pvsystem.max_power_point does not seem to do well when photocurrent is zero.

[cwhanse]

Behavior of calcparams_desoto depends on effective_irradiance: inf when pd.Series, warning when np.array, error when float.

Gracefully handling the float case seems like verging on a hacky workaround.

Edit: comment is misplaced, sorry.

[adriesse]

#340 added support for solving the single-diode equation for an "ideal" PV device that has infinite shunt resistance (equivalent to zero shunt conductance). I see several pvlib issues and discussions concerning very low and zero irradiance causing problems for pvlib.pvsystem.calcparams_desoto, and I just hit this issue myself. Concerning this edge case, which appears to lead to a lot of hacky workarounds, has the community discussed/tried making calcparams_desoto return float("inf") for R_sh when the irradiance is zero?

I think it returns inf as it should unless all the inputs are floats, because in which case numpy is not used. Shouldn't be too hacky to fix this.

[claythe3ed]

This inconsistency is a classic edge case. The cleanest fix in calcparams_desoto would be to cast scalar floats to a 0-d numpy array internally, or explicitly handle the float case by returning np.inf for R_sh when effective_irradiance == 0. This avoids the error without changing the physics. I'd be happy to submit a small PR for this if the maintainers think it's worth it.