Soiling is a slow, quiet loss
Nothing breaks. There is no alarm. Production simply runs a few percent below model, and keeps running there until something washes it off. Because the loss is gradual and the baseline drifts with it, soiling is easy to normalize and hard to notice.
The IEA Photovoltaic Power Systems Programme puts typical annual energy loss from soiling at 3–5% of production, and notes that in the dustiest parts of the United States the figure reaches 7%.
Annual cleaning is not the same as clean
A once-a-year clean does less than it sounds. Soil accumulates continuously, so what matters is the average level of soiling across the year, not how clean the array is on the day after service.
An NREL model illustrates the gap: on an array that would accumulate soil blocking about 1.9% of sunlight over a year, a single annual cleaning holds the average loss near 1.5%. Real, but a long way from eliminating the problem. Frequency is the lever, not the existence of a cleaning program.
Location changes the arithmetic completely
Soiling rates vary by roughly an order of magnitude between environments. Arid, dusty sites accumulate several times faster than temperate urban ones, and rainfall patterns matter as much as dust load — light rain can be worse than none, lifting dust and depositing it unevenly.
This is why a national average is close to useless for a specific site. The number that matters is yours, and it is measurable.
Working out your own number
- Compare against model, not against last month. Persistent deviation from expected production is the signal.
- Clean a reference section and watch the delta. The cheapest soiling measurement available: clean part of the array, compare it against the rest.
- Track recovery after cleaning. How quickly production falls back tells you your accumulation rate, which sets your economic cleaning interval.
- Price the loss in revenue, not percent. Three percent of a large array is a real annual number and makes the comparison against cleaning cost obvious.
Where robotic cleaning changes the case
Manual cleaning does not scale well across utility-scale sites, and the labor cost per pass is what usually pushes operators to clean less often than the economics justify. Lowering the cost per pass changes the optimal frequency, which is where most of the recovered production comes from — not from any single clean being better.
That is the honest case for automation here. It is not that a robot cleans better than a person. It is that it makes cleaning often enough affordable.
See solar panel cleaning equipment, or the wider solar industry overview.