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
- Removable soiling: loose dust that rain largely clears.
- Persistent soiling: a cementing, adhering mineral and biological layer. Rain does not clean it.
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 / study | Annual 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 year | fixed system 5.8% · single-axis tracker 3.7% |
| Northern India | 10.2% monthly total |
| European modeling (washable type) | average ~1.25–5% · Germany ~0.6–3% · southern Spain ~2–4% |
Daily soiling rate
| Location | Soiling rate |
|---|---|
| Doha, Qatar | 0.50%/day (28 days) · 0.52–0.55%/day (long-term monitoring) |
| Qatar site average | 0.4%/day |
| Western Senegal (29.5 MWp) | 0.33–0.49%/day (seasonal) |
| Morocco, semi-arid | 0.24%/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?
| Finding | Value |
|---|---|
| Threshold range in the literature | 0.3 – 20 mm/day |
| Observed minimum for a full clean | 3 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 effect | drizzle 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."
- A reverse-bias hot spot can permanently destroy the cell, the encapsulant, and the cell interconnections, and creates a fire risk.
- In simulations, local shading from bird droppings can produce temperatures up to 200 °C in a single cell.
- Bird droppings are opaque — they block completely, not partially, which makes them more dangerous than dust.
- Once the bypass diode conducts, the shaded cell group drops out of circuit and the dissipation stops — that is protection, but the output of that string (typically one third of the module) is lost.
- Prolonged hot spots can lead to solder joint failure, delamination, browning, and burned bypass diodes.
Conclusion: the economic case for regular cleaning is not only energy but asset life.
Cleaning methods: water, labor, cost
| Method | Water consumption | Labor |
|---|---|---|
| Manual water + brush | 3–5 L/panel (7–8 L at some sites) | High, labor intensive |
| Tanker + pressure washing | up to ~5,000 L in a single pass per 1 MW | Medium-high + vehicle/fuel |
| Water with chemical additive | down to 1.53 L/panel (56% less than water alone) | Medium |
| Robotic / dry cleaning | Near zero | Very 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:
| Study | Result |
|---|---|
| Senegal, 29.5 MWp | Optimal cycle 14 days (20 cleanings per year; current practice 8). 31% reduction in total dust-related cost |
| Arid climate, techno-economic model of 10 techniques | For fixed tilt, the optimal interval is 6.48 days with truck washing and 99.01 days with mobile dry cleaning |
| Portfolio-scale optimization | 0.8% 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
- The anti-reflective coating on PV glass scratches easily; scrubbing or scraping the surface destroys that coating irreversibly.
- Scrubbing with water alone requires excessive mechanical force because there is not enough lubricity, which raises the scratch risk.
- Brush bristles should be chosen from soft types designed for PV, such as nylon or hog hair; abrasive pads and stiff bristles must not be used.
- Pressure washing damages the coating and drives moisture into electrical components — most manufacturers void the 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?
| Test | Value | What it means |
|---|---|---|
| IEC 61215 mechanical load | 2400 Pa uniform, 1 hour, 3 cycles on front and back faces | Severe wind capable of lifting the module; on the order of ~130–150 km/h depending on mounting |
| Optional snow load | 5400 Pa | A 1–1.5 m snow layer; equivalent to ~916 kg on a 60-cell module |
| Hot-spot endurance | IEC 61215-2 MQT 09 | Endurance against hot-spot heating caused by partial shading and soiling |
| IEC 61730 | Safety qualification | Fire, 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
| Feature | Basis in the literature |
|---|---|
| 800 mm industrial brush | Contact 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 labor | Manual 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 grid | Some robotic + electrostatic systems can consume 0.5–2% of PV output; a battery removes this parasitic load |
| Dual high-torque DC motors + non-slip wheels | Stable 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:
| Scenario | Loss recoverable by cleaning | Consistency with 15–30% |
|---|---|---|
| Temperate/rainy site, regular rainfall | 1–5% per year | Inconsistent |
| Global average | 3–7% | Inconsistent |
| Arid/dusty site, long period without cleaning | 0.4–0.5%/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.