Solar Panel Soiling and Robotic Cleaning: The Real Size of the Loss

Soiling loss is 3–7% globally; in arid sites the daily rate reaches 0.3–0.5%. Rain's cleaning threshold, the hot-spot risk from bird droppings, a water consumption comparison, and cleaning frequency optimization — with data from the literature.

Soiling is the largest single factor affecting PV performance after irradiance, and its magnitude varies from site to site not by a factor of two but by a factor of ten. Of two plants built with the same technology, one may lose under 1% a year and the other 39%. This application note lays out the real size of the loss, a comparison of cleaning methods, and the economic logic of robotic cleaning using data from the literature.

The two components of soiling

This distinction is economically decisive: in modeling studies, losses of the persistent type come out roughly 2.4 times higher than the washable type.

Measured annual losses

Region / studyAnnual loss
Global average (IEA-PVPS Task 13, 2022)3–5%
Global (IEA-PVPS fact sheet, 2025)4–7%
Typical US assumption (NREL)5% — literature range 2–25%
Atacama Desert, northern coast (12 months uncleaned)39%
Atacama, high altitude / south≤3%
Atacama, bifacial, 1 yearfixed system 5.8% · single-axis tracker 3.7%
Northern India10.2% monthly total
European modeling (washable type)average ~1.25–5% · Germany ~0.6–3% · southern Spain ~2–4%

Daily soiling rate

LocationSoiling rate
Doha, Qatar0.50%/day (28 days) · 0.52–0.55%/day (long-term monitoring)
Qatar site average0.4%/day
Western Senegal (29.5 MWp)0.33–0.49%/day (seasonal)
Morocco, semi-arid0.24%/day
India overall (fleet average)0.051%/day — 26% of systems above 0.1%/day
Extreme dust conditions (upper bound)up to 1%/day

A simple rule: in an arid, dusty region, 0.3–0.5%/day means 9–15% loss over 30 days without cleaning. In a temperate, rainy region, 0.05%/day can keep the annual loss below 1%.

Data for Turkey

No published, field-measured annual % loss or %/day soiling rate value for Turkey could be identified — so no figure is given here. The available findings are at the level of accumulated mass: in a field study on five rooftop PV systems in Bursa, winter particle accumulation was measured at 0.098 g/m² and summer at 0.051 g/m²; accumulation roughly doubled in winter, but the lower module temperature partly offset the loss. In laboratory work, a very high correlation (R² ≥ 0.965) between soiling ratio and performance has been reported for industrial dusts in Turkey.

How much does rain clean?

FindingValue
Threshold range in the literature0.3 – 20 mm/day
Observed minimum for a full clean3 mm (at that site's conditions)
Washable type calibration (example site)2.6 mm/day
Persistent type calibration (example site)18.5 mm/day
Adverse effectdrizzle below 0.5 mm/day does not clean, it increases dust accumulation

Practical takeaway: the summer periods when monthly rainfall stays below the threshold are the window in which soiling builds up and mechanical cleaning delivers the highest return.

Bird droppings and hot spots: an asset loss, not an energy loss

When a cell is shaded it becomes the current limiter in a series string, goes into reverse bias, and dissipates the power produced by the string as heat within itself. That local heating is the "hot spot."

Conclusion: the economic case for regular cleaning is not only energy but asset life.

Cleaning methods: water, labor, cost

MethodWater consumptionLabor
Manual water + brush3–5 L/panel (7–8 L at some sites)High, labor intensive
Tanker + pressure washingup to ~5,000 L in a single pass per 1 MWMedium-high + vehicle/fuel
Water with chemical additivedown to 1.53 L/panel (56% less than water alone)Medium
Robotic / dry cleaningNear zeroVery low — one operator, multiple robots

On the cost side, the operating cost of manual cleaning is reported in the 1,000–4,500 EUR/MW range depending on the region. At the 29.5 MWp site in Senegal, a single cleaning cycle cost approximately 6,100 EUR.

Water quality: less water does not mean no spotting

If the total dissolved solids of the mains water exceeds 50 ppm, a deionized water rinse or squeegeeing the water off before it evaporates is necessary. Calcium and magnesium in hard water leave a permanent white mineral haze, and that layer creates a more persistent optical loss than the dirt did. Detergent leaves a sticky residue that attracts more dust and can put the warranty at risk.

A robot's minimal-water dispenser gives a two-way advantage here: both the water volume and the total mineral load left behind (volume x TDS) drop. Even so, if hard water is used, deionization or immediate drying is required.

Cleaning frequency optimization: the real economic logic of the robot

The optimum frequency is the point that minimizes the sum of revenue lost to soiling and the cost of cleaning. Concrete results from the literature:

StudyResult
Senegal, 29.5 MWpOptimal cycle 14 days (20 cleanings per year; current practice 8). 31% reduction in total dust-related cost
Arid climate, techno-economic model of 10 techniquesFor fixed tilt, the optimal interval is 6.48 days with truck washing and 99.01 days with mobile dry cleaning
Portfolio-scale optimization0.8% average revenue increase with a sector-based cleaning plan
NREL model (low-soiling site)On a system with 1.9% loss, 1 cleaning per year → 1.5%; 2 cleanings → 1.3%; 3 cleanings → 1.2%

The principle that follows is clear: as the marginal cost per cleaning drops, the optimum frequency rises and the average soiling loss falls. The economic value of robotic cleaning is not that it "cleans better in one pass" but that it can clean more often and more cheaply. The optimum frequency is also not a fixed number; it should change over the years with degradation, electricity price, and maintenance cost.

Surface damage and warranty

Manufacturer warranty documents generally use qualitative wording such as "soft-bristle brush, no abrasives, no high pressure"; no binding numerical hardness limit is published. For that reason, bristle material and contact pressure must be documented separately by the supplier for consistency with the warranty terms of the modules in use.

Mechanical load: what does robot weight do to the panel?

TestValueWhat it means
IEC 61215 mechanical load2400 Pa uniform, 1 hour, 3 cycles on front and back facesSevere wind capable of lifting the module; on the order of ~130–150 km/h depending on mounting
Optional snow load5400 PaA 1–1.5 m snow layer; equivalent to ~916 kg on a 60-cell module
Hot-spot enduranceIEC 61215-2 MQT 09Endurance against hot-spot heating caused by partial shading and soiling
IEC 61730Safety qualificationFire, electric shock, injury; breaks, cracks, and surface tearing are not acceptable on visual inspection

The critical distinction: 2400 and 5400 Pa are uniform load values, not point loads. Because a robot's weight is concentrated at the wheel contact areas, one cannot say "safe up to 2400 Pa"; the governing parameters are contact pressure per wheel and the bending moment on the glass. Static loading studies have also shown that non-uniform loading accelerates microcrack formation and electrical performance loss — and while the robot is working, wind load and robot load superimpose. Automatic stopping or parking above a given wind speed is therefore a mandatory safety feature.

Traction and fall safety on a tilted array fall outside the scope of PV standards; they must be addressed on the machinery safety side (safety line, motor braking, self-locking on power loss).

How ROBEG ELS-03 features map to the literature

FeatureBasis in the literature
800 mm industrial brushContact mechanical cleaning; bristle material and contact pressure must be documented for warranty compliance
Minimal water consumption (optimized dispenser)Manual methods use 3–8 L/panel; at water-constrained sites this removes the physical barrier to cleaning more often
Robotic operation, low laborManual cleaning runs 1,000–4,500 EUR/MW; as marginal cost drops, the optimum frequency rises and average loss falls
DC battery — independent of the gridSome robotic + electrostatic systems can consume 0.5–2% of PV output; a battery removes this parasitic load
Dual high-torque DC motors + non-slip wheelsStable travel on a tilted array; a machinery safety area not covered by the standards
Modular construction (brush / wheels / battery)The maintenance cost advantage is plausible, but no numerical basis for it could be found in the literature

The correct framing of the "15–30% efficiency gain" claim

This is a manufacturer claim and has not been verified by independent third-party measurement. Compared against the literature:

ScenarioLoss recoverable by cleaningConsistency with 15–30%
Temperate/rainy site, regular rainfall1–5% per yearInconsistent
Global average3–7%Inconsistent
Arid/dusty site, long period without cleaning0.4–0.5%/day x 30–60 days = 12–30%Consistent — but as the instantaneous recovery from a single cleaning, not an annual average

The correct statement: the 15–30% figure cannot be presented as an average annual energy gain. Presented as "the instantaneous production increase after a single cleaning on panels left uncleaned for a long period at an arid, dusty site," it is consistent with the literature. For an annual average, the realistic and defensible range is on the order of 2–7%, depending on the site's own soiling rate and cleaning frequency. When preparing a quotation, the right approach is to measure the site's own soiling rate and calculate the optimum frequency from that data.

Knowledge Center