We size the arrayto the load,not to the roof.

Where

Packing sheds, offices, schools

Pattern

A moving shadow that crosses a few modules in one or two strings for part of the day

Typical loss

Typically 4 to 9 percent of array output, concentrated in the affected strings

What we specify

String inverter with DC optimisers on the affected strings only

The shadow is localised and moves. Optimisers on the shaded strings recover most of the loss without paying for module-level electronics across the whole array, and they keep the simple, serviceable string architecture everywhere else.

Where

Any roof with a raised perimeter

Pattern

A hard band of shade along the north and west edges in winter, largely gone by May

Typical loss

Predictable and seasonal, worst in December, near zero at the solstice in June

What we specify

Plain string inverter, with the first row set back out of the winter shadow

Parapet shade is geometric and entirely predictable, so the answer is layout rather than electronics. We calculate the winter shadow length and start the array beyond it. Electronics cannot recover energy from a module that should not have been placed there.

Where

Tilted rooftop rows and ground-mount fields

Pattern

Row shading row in the first and last hours, worst around the winter solstice

Typical loss

Set by row pitch, not by equipment: from under 1 percent to well over 5 percent annually

What we specify

String inverter with rows wired along the shade line

The fix is row pitch and string wiring, not hardware. We space rows against the winter-solstice sun altitude, then wire each string along the row so the whole string dims together rather than one module dragging a healthy string down.

Where

Processing plants and older industrial roofs

Pattern

Many small, scattered shadows falling on single modules throughout the day

Typical loss

Individually tiny, collectively 3 to 8 percent, and spread across most strings

What we specify

Microinverters, or module-level electronics across the affected zone

Scattered single-module shading is the one case that genuinely justifies module-level conversion. When the losses are spread across every string, per-string optimisation has nothing clean to protect, and per-module independence recovers the most energy.

Where

Ground-mount edges, school and civic sites

Pattern

A slow seasonal encroachment that gets worse every year as canopies grow

Typical loss

Starts near zero and grows, which is what makes it dangerous in a twenty-year model

What we specify

Keep the array out of it, and write the vegetation management into the O&M contract

The mistake is modelling today's shadow for a twenty-five year asset. We model the mature canopy, set the array boundary against it, and put the trimming schedule in the maintenance agreement so the loss does not creep back in.

Where

Everywhere in the San Joaquin Valley

Pattern

Uniform film plus a heavier band along the lower module edge where dust collects

Typical loss

Not shading in the electrical sense, but it behaves like it on the bottom cell string

What we specify

No topology fixes this. Scheduled washing, timed against measured loss, does

Dust builds through a dry season that runs May to October. The lower-edge band is the part that matters, because it can put the bottom cell string into bypass and cost far more than the uniform film alone. It is a maintenance problem, not an equipment problem.