Recovering lost performance ratio on an under-maintained rooftop plant

The starting point

The plant was a factory rooftop installation that had been running for several years under a basic cleaning contract, with no performance monitoring beyond the inverter's own local display. Nobody was tracking performance ratio against a baseline, and there was no record of when strings had last been inspected. The only signal anyone had was that the monthly generation figure "felt lower than it should be," without a number to back that up.

Running the audit

We started by establishing a design-baseline performance ratio from the original panel specification and site irradiance data, then instrumented the plant for a short monitoring window to get a clean read on actual output. The gap between the two was significant, and the next step was attributing it: how much was soiling, how much was faults, how much was downtime that had gone unlogged.

A site walk turned up two string-level faults that weren't causing full outages, so they hadn't triggered anything the existing setup would have flagged. Soiling accounted for a meaningful share of the loss on its own, unsurprising given the cleaning cadence had never been checked against actual soiling rate.

What got fixed, in order

The two string faults were prioritized first since they were the fastest, highest-confidence fix. A combiner box connection issue turned up during that work as well and was corrected at the same time. Cleaning came next, timed to the measured soiling curve rather than the old fixed schedule. Monitoring was the last piece: the plant was brought onto the Soligrid Uptime Platform so future drops would show up as an alarm instead of a vague feeling.

What the numbers looked like afterward

Performance ratio moved up meaningfully after the string faults were resolved, before the cleaning cycle even ran, which told us those faults had been the larger contributor. The subsequent clean added a further uplift. Combined, the recovered output accounted for the majority of the original gap between design baseline and actual generation.

What held, six months later

The real test of a recovery isn't the number right after the fix, it's whether it holds. Six months on, performance ratio has stayed within a narrow band of the post-fix level, with soiling-driven dips showing up and getting cleaned out on schedule instead of accumulating unnoticed. The plant now has an alarm history instead of a blank log, which is the more durable outcome of the two.