Clustering algorithm for calculating cross-axis tilt zones

[kurt-rhee]

A common method for modeling projects on terrain which has multiple cross-axis tilts is to create zones where each zone has it's own mono-slope backtracking algorithm. I think that an open question is how many zones might be needed to closely approximate a terrain adaptive backtracking strategy which takes into account all of the cross-axis tilts natively. This number of zones will increase or decrease depending on the complexity of the terrain.

I am working on using clustering algorithms to define which trackers should be included in what zones. If this is something that others might find useful I am happy to contribute the results to pvlib.

[cwhanse]

I've had a few conversations about interest in modeling PV systems on uneven terrain, and what that might take. The clustering algorithm sounds like one of several pieces that could make that possible.

[kandersolar]

Sounds interesting! Also sounds nontrivial enough that it would need a published reference for a pvlib implementation to point to.