How to measure PV soiling: Soiling stations, reference cells, optical sensors and IEC 61724-1

Three numbers: the soiling ratio, the soiling loss and the soiling rate. With those you can put a rand value on dirt, time the next clean and prove whether a product works. Without them, a cleaning programme runs on the calendar and on how the glass looks. The formulas and worked examples are in soiling ratio, rate and loss explained.

SCADA measures production, not soiling. A drop in performance ratio can come from soiling, but also from heat, clipping, outages, degradation or a faulty string. To separate soiling from the rest, you need a reference that isolates it.

Performance ratio can still help, if the irradiance sensor is clean. IEC 61724 recommends cleaning irradiance sensors weekly, and IEA PVPS (2022) notes that temperature-corrected performance ratio can then give a good estimate of soiling in dry seasons. Even with a clean sensor, though, soiling shows up only as a lower performance ratio, mixed in with every other loss. Chemitek makes the same point in a post titled “Your SCADA measures production. It doesn’t measure soiling.”

What does IEC 61724-1 require? IEC 61724-1:2021 is the international standard for PV performance monitoring. Whether it requires soiling measurement on your plant is in soiling ratio, rate and loss explained.

For other plants, the economics make the case. On an illustrative 10 MWp plant near Upington, each 1% of soiling loss is worth about R0.10–0.87 million a year, depending on the REIPPPP tariff. The workings are in soiling losses in South Africa.

Uncertainties from Dunn et al. (2018) as reported by IEA PVPS (2022), where the best setups reach about ±1%; DustIQ figure from its 2025 manual.

Which one to choose? Where soiling drives a large cleaning budget, the lower uncertainty of a power-based pair is worth its upkeep. Optical sensors earn their place where cleaning a reference device on schedule is impractical, or as extra measurement points across a large site. Performance analysis is still useful: it’s how you check that the sensors agree with what the inverters see.

The rule is simple: a sensor must soil like the modules it stands for. In practice:

• Same geometry. Mount reference devices at the tilt and orientation of the array they represent, and on trackers, on a tracker. Geometry changes the answer.
• Same exposure. Put sensors where the dust is: beside the access road, downwind of a quarry, cement works or farmland. Soiling is not uniform across a site, so a single sensor in the cleanest corner will under-report. See soiling ratio, rate and loss explained.
• No false signals. Avoid spots that soil for reasons the array doesn’t share, such as below a bird perch or beside a vehicle turning area.
• Safe from the crew. The soiled reference must never be washed by accident, and the clean one must be washed on schedule. Label both and write them into the O&M procedure.

How many sensors a plant needs, and why, is in soiling ratio, rate and loss explained. If your plant is large and soiling is material, plan for more than one.

Beyond that, count the distinct soiling zones. This is how we size a measurement layout:

How long a dry spell you need to fit a soiling rate is in soiling ratio, rate and loss explained. Over shorter windows, day-to-day swings in weather and temperature can hide the trend.

Then think in seasons. In the South African PV studies, the worst losses came in the dry winter months. The measured rates and rain findings are in soiling losses in South Africa. A full year of data shows you both seasons.

There is also a data gap. We found no peer-reviewed PV soiling rates for Upington, Kathu, Prieska, Kimberley, Namibia or Botswana. For plants there, your own station is the only data you will get.

Two approaches are in use in South Africa:

• A trigger. A rule of thumb reported in South Africa is to clean when a constant gap of more than 50 W/m² opens between the cleaned and uncleaned reference cells (Naicker 2018). It’s simple and better than the calendar, but it ignores the tariff and the cost of a clean.
• A cost optimum. Compare the rand value of the soiling loss with the cost of cleaning, and clean when the balance tips. The right interval depends on your tariff, soiling rate and what a clean costs, and it changes with the season. The calculator works it out for your plant.

Both depend on a daily soiling ratio. The cleaning frequency guide sets out how to use the result.

A trial asks a different question. It isn’t how dirty the plant is, but how much better one method or product is than another. The measurement has to rule out the sun, the season and the crew.

1. Baseline first. Choose test strings and matching control strings, and record their output under the same sun before any treatment.
2. Control for irradiance. Compare treated and control strings in the same irradiance window, so a passing cloud can’t pose as a gain.
3. Match the measurement to the question. String current or inverter data shows the immediate gain after a clean. Soiling references beside treated and untreated rows show how fast each soils again, which is where a coating earns its keep.
4. Run it long enough. A re-soiling rate needs a dry spell long enough to fit; see soiling ratio, rate and loss explained. Short tests flatter coatings: they can show more than 80% less soiling, while long-term field results are typically 20–50% (Ilse et al. 2019).
5. Record water and time. Log the water used and the cleaning time per string, not only the output. Savings there can matter as much as yield.

This is the protocol behind our free pilot for utility-scale and large C&I plants.

Chemitek’s case studies show each of these rules at work:

• Controls make the result. At Green Energy Park in Morocco, five identical strings each ran on their own inverter with DC monitoring and a meteorological station on site, and I–V curves were corrected to standard test conditions under IEC 60891. The uncleaned and water-cleaned control strings are what separate the numbers. See the Morocco trial.
• An unwashed string is a cheap irradiance reference. The Malaysia case used an unwashed string that way.
• Match exposure, and record a baseline. At a Scatec-operated plant in Honduras, the treated string sat nearer a road and was subject to much more dust than the two controls, and the two water-only controls also differed from each other.
• Without soiling meters, soiling is judged by eye. On a 330 MW plant in Vietnam, energy was measured at the inverter, but the plant had no soiling meters, so the soiling itself was assessed only visually. A soiling station would have shown how fast each set of strings soiled again.

A sensor on the array shows how fast the glass soils again after each clean, and that rate is what a cleaning programme should be built on.

RenewChem installs reference panels and soiling stations, supplies soiling measurement equipment and reads the data with your O&M team. See soiling measurement and monitoring, or start with a site soiling assessment.