This is a very seductive subject to me and I’ve been contemplating this some more, as we’ve done elsewhere (perhaps too much):
Access to geographically proximate PV arrays is key in terms of doing predictive analysis for cloud motion and determining the consequent actions: “Charge EV now!”; “Pause the laundry”; “Switch to battery charging”; “Spin up the flywheel!” …
With that in mind, if I had a large enough property I’d be wiring up a widely spaced array (minimum of 4 panels I guess) … imagine a property centered on a lot and the “cloud detector PVs” at the four corners of the lot.
The other option is to scout out solar neighbors at some optimal distance (based on “normal” cloud speed = prevailing winds) and give them a Sense and share data!
[Utility-scale PV arrays, btw, are somewhat immune to this “cloud reflection spikes” phenomena for obvious reasons. As @MaheshAtSense has highlighted, many PV households are pushing the majority of their watts to the grid (less than optimal) … while individual households may well be able to time shift usage to exploit more of their PV output, it would seem that at neighbor(hood) scale the efficiencies must improve dramatically. Car sharing + charge sharing.]