AI shading analysis: what it can and cannot do today

Shading is where solar designs die quietly: a chimney, a parapet or the neighbour's ficus can knock double digits off production. It is also the part of PV modelling with the steepest expertise requirements, which makes it the natural target for AI.

What AI does well right now

Computer vision models segment roofs, detect obstacles and estimate geometry from satellite, aerial and LiDAR data with no manual tracing. From that 3D understanding, slope and orientation heatmaps show instantly which roof faces deserve modules and which do not.

For feasibility and sales, is this roof worth pursuing, roughly how much will it produce, this is not just faster than manual modelling; it removes the queue behind your one trained designer.

What still needs classical simulation

Bankable reports remain the territory of hour-by-hour (8,760 h) simulation engines validated over decades. If a bank or auditor must sign off on the production estimate of a large plant, they will ask for that lineage, and they should.

Our own roadmap at Heliocentrium includes hourly shading precisely because of this: AI gets you to a confident commercial answer today; certified simulation depth is where the category is converging.

The practical takeaway

Match the tool to the decision. Qualifying leads and closing rooftop deals needs speed and visual clarity, AI territory. Financing a solar farm needs certified hourly modelling. Teams that use each where it belongs quote faster than their competitors without ever presenting a number they cannot defend.