clearsky algorithm with POA data?

[toddkarin]

What is the expected behavior of the clear sky algorithm applied to plane-of-array data?

[cwhanse]

Are you asking about features.clearsky.reno?

[toddkarin]

Yes, either with fixed tilt or single-axis tracking.

[cwhanse]

If the clear-sky profile is transposed to the plane-of-array, then I think the algorithm would reasonably label clear-sky periods in plane-of-array data. To be explicit: the plane-of-array irradiance is passed to ghi, and the transposed clearsky GHI is passed to ghi_clearsky.

The thresholds aren't exposed in this interface. The thresholds are probably OK for plane-of-array data, ideally at 1 minute. If the irradiance data are at longer intervals, out to 15min, there's evidence that the same thresholds work, but no formal validation.

[wholmgren]

I did this a few years ago and found the thresholds were difficult to tweak. I had more success converting POA to GHI using pvlib.irradiance.gti_dirint and then applying pvlib.clearsky.detect_clearsky.