The site and the clean
Eco Cover Energy looks after a rooftop PV plant in İzmir, Türkiye, for an owner Chemitek doesn’t name. Metal dust had settled on the modules and cured onto the glass as metal oxides. According to the report, no washing method worked, from conventional washing with water to cleaning robots with powerful brushes. The report gives neither the plant’s capacity nor the source of the dust.
- Modules: 1,230 on the contaminated roof.
- Product: Chemitek Metal Oxides Removal Agent (MRA). The deposit was severe, so Chemitek advised using it far stronger than the datasheet’s starting dilution of 1:10: 680 modules got the concentrate undiluted, and 550 got it mixed 1:1 with water.
- Dose: 500–800 mL of solution per module, which the report says solved the problem.
- Method: spray the product on, scrub with a cylindrical electric brush, then rinse with de-ionised water from a cleaning robot. The crew kept the glass wet throughout, because if the product dries the whole process has to start again.
- Measure: each month’s energy against the energy expected for that month. With every module contaminated there was no clean reference string, so the expected figures came from EU photovoltaic data. The report calls the trial 60 days long; its tables run from March 2024 to February 2025 and put the clean between August and September 2024.
Results
| Month | Expected | Produced | Gap |
|---|---|---|---|
| Mar 2024 | 63,752 | 58,078 | −8.9% |
| Apr 2024 | 75,258 | 66,904 | −11.1% |
| May 2024 | 86,360 | 75,427 | −12.66% |
| Jun 2024 | 89,964 | 77,882 | −13.43% |
| Jul 2024 | 96,722 | 82,869 | −14.32% |
| Aug 2024 | 89,087 | 75,319 | −15.46% |
| Average before | 83,524 | 72,747 | −12.64% |
| Sep 2024 | 74,568 | 72,182 | −3.2% |
| Oct 2024 | 60,058 | 57,716 | −3.9% |
| Nov 2024 | 43,439 | 41,397 | −4.7% |
| Dec 2024 | 35,585 | 34,482 | −3.1% |
| Jan 2025 | 38,124 | 36,675 | −3.8% |
| Feb 2025 | 43,585 | 41,798 | −4.1% |
| Average after | 49,227 | 47,375 | −3.76% |
Monthly energy as Chemitek reports it. The report doesn’t state the units, or whether the figures cover only the treated modules or the whole plant.
The O&M team reported that MRA broke the bond between the oxides and the glass and emulsified the debris, and that the crew got faster with practice. It decided to use MRA across the whole contaminated plant, then protect the modules with Solar Wash Protect or Antistatic Solar Armor, so that routine cleans would keep the oxides off without the remover.
How to read it
- Before cleaning, the loss grew every month, from 8.9% short in March to 15.46% short in August. Chemitek’s guidance adds that the longer panels go uncleaned, the harder they are to clean.
- After cleaning, the gap held between 3.1% and 4.7% for six months.
- Chemitek’s “almost 12%” matches the step from August to September. The six-month averages moved by 8.88 percentage points, and that is the figure we use.
- Two cautions. Without a clean reference string, the result depends on how well the benchmark matched the real weather. And the two periods cover different seasons, so part of the change could be seasonal.
What this means for a South African plant
- Metal dust comes from the neighbours. Plants near mines, smelters, steelworks and metal workshops can collect iron-rich dust that oxidises and bonds to the glass. Our metal oxides page covers how we trace the source.
- Don’t scrape it off. Water and brushes failed here, and scrapers or abrasive pads damage the glass coating and put the module warranty at risk. MRA dissolves the oxide instead.
- Find the weakest dilution that works. This roof needed MRA undiluted or at 1:1; most stains won’t. We start at 1:10 on a small area and strengthen only if the stain holds.
- Protect after you restore. Chemitek’s plan was to remove the oxides once and then coat, which is our Restore → Protect → Maintain programme.
- Measure against a reference. On a pilot we keep treated and untreated strings side by side under the same sun, which this roof couldn’t. Pilots are free for utility-scale and large C&I plants.
