Module Cleaning Cost-Benefit Analysis for EU Utility-Scale Solar

Module cleaning at utility-scale solar parks is often scheduled by calendar, as a fixed number of cycles per year in the O&M contract. A module cleaning cost-benefit analysis asks a sharper question: will the energy this cleaning recovers, valued at the price the plant actually earns, pay for the cleaning before rain or the next cleaning would have recovered it anyway? In rainy climates the answer is often no for ordinary dust and often yes for the sticky, rain-resistant layer left by pollen, farming, traffic and birds. Every variable in that decision can be measured or priced, so it should not be left to the calendar.

How large soiling losses are in Europe

The IEA PVPS Task 13 review estimates that soiling cost at least 3-4% of global PV production in 2018, worth roughly 3 to 5 billion euros, with the share expected to rise to 4-5% ^[3]. Those figures are driven by arid, high-irradiance markets, so desert cleaning schedules do not transfer to Europe.

The first continental assessment, by Fernández Solas and co-authors, calibrated a soiling model against field sensors across Europe. It found an average annual loss of 0.9% with perfect rain cleaning, rising to 5.3% if rain removes only 10% of the deposit ^[1]. In southern regions the losses are strongly seasonal and peak around August, at the end of the dry summer. In France, Germany, Ireland and the Benelux countries they are spread more evenly through the year because rain is frequent ^[1].

Local sources often matter more than climate zone. Monthly soiling rates compiled by the IEA for two US region types reach 2% in natural deserts but up to 11.5% in dry agricultural areas ^[3]. European examples include an energy loss of 3-4% attributed to high pollen concentrations in Belgium, agricultural soiling at a Norwegian farm, and Saharan dust episodes reaching both southern and northern Europe ^[1]. Within one site, deposition rates can differ by a factor of 1.5 to 2 depending on prevailing winds and dust sources ^[3].

Rain removes most dust, but not all of it

Simple soiling models reset losses to zero after any day with rain above a threshold, for example 1 mm per day in one approach described by the IEA ^[3]. That works for loose dust, not for deposits that bond to the glass.

The clearest European example is a 60 kWp array in Burgdorf, Switzerland, near a railway line and farmland, documented in the IEA report. The site receives more than 1,000 mm of rain a year, so natural cleaning was considered sufficient. After a dry late summer in 1997, a compact strip of dirt formed along the lower frame edge of the landscape-mounted modules, where the cells sit only 1-2 mm from the frame, and the loss reached 10% by spring 1998. Periodic cleanings between 1998 and 2020 recovered between 2.3% and 10%, although from 2012 on they no longer produced a lasting gain, possibly because the glass surface had aged or been cleaned inadequately ^[3].

Ruiz-Donoso and co-authors modeled this by letting rain clean only partially and letting a fraction of the deposit adhere permanently. With a persistent soiling type, modeled European losses were 2.4 times higher than with a removable type, and the authors conclude that for at least some soiling types mechanical cleaning is necessary even in rainy Central Europe ^[9]. Their parameters come from two African sites, so the maps show plausible cases rather than the most likely European values ^[9]. SolarPower Europe adds that long gaps between cleanings let bird droppings etch the glass and lichen grow, and that parts of one site may need cleaning more often than others ^[6].

Plant geometry shifts the balance. Soiling losses generally fall as tilt increases, dirt tends to collect on the lower parts of modules, and on trackers a tilted night-time stow position is a simple way to reduce deposition ^[3]. Bifacial modules soil mainly on the front: a two-month study in Santiago, Chile, measured a rear-side soiling rate about 8.8 times lower than the front ^[10].

The practical consequence is to split soiling into two parts. Washable dust is reset by rain at no cost. Persistent soiling stays until something mechanical removes it. Only the persistent part creates a gap that a cleaning can monetize over months.

Measuring soiling before deciding to clean

IEC 61724-1:2021 defines the soiling ratio as the power of a soiled PV array divided by its power in clean conditions, and the 2021 edition updated the soiling measurement requirements ^{[3, 4]}. A soiling station compares a clean reference device, which has to be cleaned regularly, with one left to soil. The IEA review notes that IEC 61724-1 recommends more than one soiling sensor for sites above 5 MW with expected yearly losses above 2%, and that sensors measuring only short-circuit current can understate the power loss when soiling is uneven, as with edge bands ^[3].

Plant data offers an independent check. The stochastic rate and recovery (SRR) method, developed at NREL, estimates soiling directly from system yield without precipitation data. It detects soiling intervals between cleaning events, generates a distribution of possible soiling profiles, and was validated against soiling stations ^[5]. The open-source RdTools package implements SRR and returns an insolation-weighted soiling ratio with a confidence interval, although the developers still label the soiling module experimental ^[11].

Averaging matters. A plain average of daily soiling ratios underestimates the energy loss, because the highest losses often coincide with the sunniest seasons ^[1]. For merchant plants I go one step further and weight by revenue, because production multiplied by price is what a cleaning actually recovers.

What a cleaning cycle really costs

Cleaning cost combines labor and materials, is usually quoted per square meter, and depends on water availability, labor rates and site access ^[3]. Costs compiled by Ilse and co-authors range from 0.07-0.09 EUR/m² in Türkiye to 0.46-0.92 EUR/m² in the Netherlands, about 0.3 to 4.3 EUR/kW per cleaning for 21.4% efficient modules ^[1]. Higher-efficiency modules have less glass per kW, so they cost less to clean per kW and recover more energy per cleaning ^{[3, 12]}.

Equipment ranges from manual brooms and water-fed brushes through vehicle-mounted brushes, which may carry a water tank, to semi-automatic and fully automatic robots. Most mechanized systems use rotating brushes with varying amounts of demineralized water, and the water should be close to module temperature to avoid thermal shock ^[3]. SolarPower Europe recommends ultra-pure water rather than tap, mains or borehole water, biodegradable detergents, planned runoff, and minimal water use where water is scarce ^[6].

Warranty conditions belong in the analysis. Manufacturers typically specify a maximum cleaning pressure, and warranties may be voided if it is exceeded ^[3]. SolarPower Europe advises confirming that the method is in line with the manufacturer’s warranty and with IEC 61215 specifications such as maximum pressure load ^[6]. For example, a 2017 First Solar guideline for its thin-film modules caps nozzle pressure at 35 bar, requires motorized cleaning tools to be approved to keep warranty coverage, and for anti-reflective coated modules prohibits wet contact cleaning and limits dry cleaning with soft cloths to six times a year ^[7]. Limits differ between manufacturers, so the current manual for the installed module is the reference.

Abrasion is the hidden cost of frequency. There is no international performance standard with pass/fail criteria for PV cleaning ^[3]. In an 18-month field test of dry-brush robots in Doha, five of nine module types lost more performance with more frequent cleaning, and the authors recommend field-testing each robot and module pairing ^[8]. That was a desert test, but the lesson carries over: each pass removes soiling and may also remove a little coating. Targeted spot cleaning of individual defects is a separate decision, covered in the article on AI thermal inspection and spot cleaning.

The price side: tariffs, merchant exposure and negative prices

The simplest rule in the IEA report is to clean when the cost of production losses, the energy lost to soiling times the value of each kWh, exceeds the cost of one cleaning. Optimizers that account for free cleaning by rain do better; in one Californian example the optimal day cut the cost of soiling by 17% compared with the day the simple rule was first met. Lower electricity prices make cleaning less attractive, because each recovered kWh earns less ^[3].

Under a fixed tariff or a contract for difference, every recovered kWh earns a known price, subject to contract terms. Under merchant or spot-indexed PPA exposure, recovered energy follows the production profile, so its value is the plant’s capture price, and in curtailed or negative-price hours it is worth nothing or less. ACER reports that negative-price occurrences across the EU and Norway rose twelvefold in 2023 and by half again in 2024, driven by low demand, high supply and subsidized renewables with guaranteed revenue ^[2]. In Hungary, the Ministry of Energy counted 306 hours with zero or negative prices in 2024 and 309 in January to August 2025 alone ^[13].

The choice of metric also changes the answer. For a 1 MW site in southern Spain, Micheli and co-authors found that LCOE stopped recommending any yearly cleaning above about 0.11 EUR/m², while NPV still favored cleaning up to about 0.25 EUR/m² ^[12]. The optimal number of cleanings rose with electricity price and fell with cleaning cost. A single yearly cleaning beat no cleaning only within about plus or minus 31 days of the optimal summer date, and the actual cleaning, done seven days late, cut the profit gain by 0.2% ^[12]. An asset manager judged on cash flow should use the NPV view.

A break-even method for module cleaning cost-benefit analysis

From an asset management perspective (I currently have 170 MWp under management), the useful question is not “how dirty is the plant” but “what will this cleaning earn before the gain disappears”. Working per kWp makes plant size drop out. A cleaning pays when:

ΔL × Y_w × p_eff > c

• c: all-in cost of one whole-plant cleaning per kWp: contractor, water logistics, supervision, production lost during daytime cleaning, and an allowance for coating wear
• ΔL: average avoided soiling loss over the benefit window, as a fraction (expected soiled minus expected cleaned loss, irradiance-weighted)
• Y_w: expected clean-plant specific yield over the benefit window, in kWh/kWp
• p_eff: effective value of a recovered kWh in EUR/kWh: the tariff, or the capture price reduced for curtailed and negative-price hours

The break-even avoided loss is ΔL* = c / (Y_w × p_eff).

The benefit window ends when the gap would have closed anyway: at the next effective rain for washable dust, or at the next planned cleaning for persistent soiling. For persistent soiling building up at a linear rate r (fraction per day) on a plant with average daily specific yield y_d, the cost-minimizing interval is T* = √(2c / (y_d × p_eff × r)). Equivalently, clean when the revenue lost since the last cleaning, y_d × p_eff × r × T² / 2, reaches the cost of one cleaning. This textbook case ignores seasonality and rain, which is why the IEA reviews more detailed optimization models ^[3].

Worked example (illustrative inputs only)

The inputs below are illustrative assumptions chosen to show the method. They are not measurements from any plant.

Washable dust in a rainy climate almost never justifies a whole-plant cleaning, because the window is too short. Persistent soiling can, but the margin is thin and can flip on the revenue model alone. A cheaper contractor, or a cleaning timed just after the main deposition event, moves the break-even as much as better soiling data does.

With illustrative persistent soiling of 0.02% per day and y_d of 3.4 kWh/kWp per day, T* is about 220 days at 90 EUR/MWh and about 285 days at 54 EUR/MWh, roughly one cleaning a year. At 0.1% per day, about 3% a month and inside the range the IEA compiles for dry agricultural areas ^[3], T* drops to about 100 to 130 days.

The IEA report cites findings that cleaning is often not cost-effective in mid and northern Europe because of lower soiling and frequent rain, while it makes economic sense in southern Europe ^[3]. The same report recommends annual or at least periodic cleaning even where rain is frequent, to remove hard-to-remove soiling before it bonds to the glass ^[3]. The European assessment found cleaning cost-effective in most locations it priced only under partial rain cleaning; with perfect rain cleaning it paid mainly in Türkiye, Italy, Spain and the United Kingdom ^[1]. My position is that both views fit together: budget for a periodic cleaning aimed at the persistent layer, and treat every additional cycle as an option that must pass the break-even test.

Decision checklist for asset managers

1. Soiling measured, not assumed: at least two power-based sensors on a large park, clean reference maintained, plus an SRR cross-check from inverter or string data.
2. Correct weighting: loss irradiance-weighted, and revenue-weighted for merchant plants.
3. Persistent share known: the loss left after the last heavy rain is what a cleaning can monetize over months.
4. Local sources mapped: harvest calendar, pollen season, roads and rail, birds, construction, Saharan dust episodes.
5. All-in cost per kWp: contractor, water supply and quality, supervision, daytime production loss, coating-wear allowance.
6. Benefit window defined: weather forecast for washable dust, next planned cleaning for persistent soiling.
7. Realized price checked: negative-price treatment in the tariff or PPA, capture price, curtailment history.
8. Break-even computed: ΔL* compared with the measured avoided loss; no cleaning just before forecast heavy rain.
9. Warranty-safe method: water, pressure and tools confirmed against the current module manual, with extra care for anti-reflective coatings.
10. Contract structure: cleaning is generally included in O&M contracts or offered as a priced option ^[6]. A per-event price makes this decision possible; a fixed annual count does not.
11. Result verified: soiling ratio recorded before and after each cleaning, to calibrate ΔL and spot diminishing returns.

Conclusion

In Central Europe, rain already does most of the cleaning for free. A whole-plant cleaning earns its keep only against soiling that rain does not remove, over a window long enough and at a price high enough to cover a cost that varies widely between markets. Measure the persistent loss, value the recovered energy at what the plant actually earns, and let the break-even test, not the calendar, decide how many cycles to buy.