PV module cleaning: when to automate the operation at solar plants

Key takeaways

• PV module cleaning is an O&M decision with a calculable payback, not a one-off field task.
• In STEMIS's technical model, automating starts to pay off at around ~6 MWp with 2 cleanings/year (payback under 5 years). Below that, manual cleaning is usually still cheaper.
• On a 10 MWp plant with 2 cleanings/year and average soiling, automation saves roughly 27 person-days and 320,000 liters of water per year, with a payback of ~3.1 years on a R$72,900 robot.
• Choosing between manual, outsourced, and robotic cleaning should compare time, water, safety, standardization, and recovered generation — not just the day-rate price.
• In an independent UNICAMP test (LESF report No. 25-043), cleaning recovered +3.1% power with no mechanical or electrical damage to the modules.

Introduction

Every solar plant accumulates dirt. Dust, soot, pollen, bird droppings, and suspended particulate form a film that blocks part of the irradiance before it reaches the cells. In some regions rain takes care of it; in others — agricultural areas, dirt roads, industrial zones, semi-arid climates — soiling comes back fast and becomes a fixed line item in the O&M budget.

The problem is that most content about "solar panel cleaning" stops at the obvious: *dirt reduces generation, so clean it*. That doesn't help whoever has to decide. The question an asset manager actually asks isn't "do I need to clean?" — it's: at what point does manual cleaning stop paying off and automation start paying for itself?

This article answers that with numbers. Instead of generic claims, we use the technical model behind the TCR-W1 ROI calculator to show exactly where that tipping point is — including the cases where automating isn't worth it.

What is PV module cleaning?

PV module cleaning is the process of removing soiling from the panel surface to recover the irradiance reaching the solar cells. It can be done three ways:

• Manual — operators with brushes, water, and support equipment.
• Outsourced — an external crew contracted per campaign.
• Robotic — dedicated equipment that repeats the process at scale, with a constant standard.

There's also the distinction between dry cleaning (enough when dust is loose and lightly adhered) and wet cleaning (needed when there's crusting, fine compacted dust, or droppings). The method you choose determines cost, water use, and how much of the lost generation you actually recover.

Why soiling becomes a financial decision

Soiling has a cost, but so does cleaning. A plant can lose energy from being dirty — and lose money from cleaning without criteria. A mature O&M routine needs to weigh, side by side:

• How much soiling is reducing generation (soiling loss).
• How much each cleaning campaign costs (labor, water, mobilization).
• How many times a year cleaning needs to happen.
• How much performance is actually recovered after each cleaning.

That last point is often ignored. In a test run by the Photovoltaic Energy and Systems Laboratory at UNICAMP (LESF Report No. 25-043), cleaning resulted in an average gain of +3.1% power, with 100% integrity confirmed by electroluminescence and no mechanical or electrical damage detected. That number is the honest baseline for any calculation: cleaning recovers measurable generation — and the only remaining question is *which method makes that cheapest*.

Manual, outsourced, or robotic: the comparison that matters

• Water per module — Manual / outsourced: ~10 L; Robotic (TCR-W1): ~2 L (up to 80% less)
• Time per module — Manual / outsourced: ~26 s; Robotic (TCR-W1): 5–8 s, depending on soiling
• Standardization — Manual / outsourced: Varies by crew and shift; Robotic (TCR-W1): Constant, row by row
• People exposure — Manual / outsourced: High (field, height, heat); Robotic (TCR-W1): Low (app-monitored operation)
• Dominant cost — Manual / outsourced: Crew day-rates; Robotic (TCR-W1): Robot CAPEX, amortized over time

The water and time figures above are the constants in the TCR-W1 technical model. They explain why the automation gain grows with scale: the more modules and the more campaigns per year, the more person-days and liters of water the manual operation consumes — and the faster the robot pays for itself.

When to automate: the math, not the guess

Here's the part that turns a decision into more than a hunch. Using STEMIS's technical model (assumptions at the end of the section), you can project the savings for any plant.

Example 1 — a 10 MWp plant (automation pays off)

Input: 10 MWp, 2 cleanings/year, average soiling.

• Modules: 20,000 — Robot: 6.5 s/module · Manual: 26 s/module
• Manual campaign time: ~39 days/year → robotic: ~12 days/year
• Time saved: ~27 person-days/year
• Water saved: ~320,000 L/year (8 L per module, two campaigns)
• Annual savings (labor + water): ≈ R$23,500
• Investment: 1 robot = R$72,900
• Payback ≈ 3.1 years · 5-year ROI ≈ 60%

Example 2 — a 2 MWp plant (automation does NOT pay off)

Input: 2 MWp, 1 cleaning/year, average soiling.

• Time saved: ~2.7 person-days/year · Water saved: ~32,000 L/year
• Annual savings: ≈ R$2,350
• Payback ≈ 31 years → in this scenario, stick with manual cleaning.

The practical rule

In STEMIS's model, with average soiling and 2 cleanings/year, payback drops below 5 years starting at ~6 MWp. Below that — or with only one cleaning a year — manual operation usually wins. More campaigns per year (semi-arid climate, agriculture, proximity to dirt roads) bring that tipping point forward; plants that rely on rain push it back.

Model assumptions: 2,000 modules/MWp · crew day-rate R$750 · water R$0.01/L · 5.5 sun-hours/day · robot R$72,900/unit. Adjust for your own numbers in the ROI calculator before deciding.

Checklist before adopting a cleaning robot

Before evaluating equipment, gather these answers — they feed the calculation above:

1. How many modules (or MWp) does the plant have?
2. How many cleanings per year are expected?
3. What's the current cost per campaign (manual or outsourced)?
4. How much water does each campaign use?
5. Does the layout allow continuous row-by-row operation? (Tilt, spacing, obstacles.)
6. Does the plant follow the 2-P standard (two vertical rows)?
7. Does the team already measure generation before and after cleaning?

If you can't answer item 7, start there: without a soiling-loss baseline, no cleaning decision — manual or robotic — is defensible.

Where the TCR-W1 fits into the operation

The TCR-W1 is an autonomous robot for large-scale PV module cleaning, designed to turn cleaning into a recurring, standardized process. It comes in after the O&M math, not instead of it. The attributes that matter for the decision:

• Water: ~2 L/module for wet cleaning — up to 80% less than the traditional method; a 1,000 L tank feeds extended autonomy.
• Standardization: constant row-by-row cleaning, with automatic end-of-row detection; the 2-P version cleans two vertical rows in a single pass.
• Safety and integrity: soft nylon bristles (0.3 mm) and pressure sensors; a 40 kg aluminum chassis with IP65 protection. Validated by UNICAMP with no module damage.
• Operational control: mobile app to schedule cleanings, configure routes, track in real time, and receive alerts.
• Independent validation: the LESF/UNICAMP Report No. 25-043, and STEMIS's track record of 600+ projects / 2.1+ GWp / 3.5M+ modules monitored.

These aren't loose sales pitches — each one addresses a variable in the payback math (water, time, standardization, generation recovery). See the full technical specs.

How to measure the result after cleaning

Generation varies with irradiance, temperature, availability, and inverters. Looking only at total daily energy leads to the wrong conclusions. A reliable evaluation compares:

• Generation curves before and after, under comparable weather conditions.
• Cleaned areas versus still-dirty areas (an internal control within the same plant).
• Rainfall history for the period.
• Total execution time, water used, and rework.

To close the loop, connect cleaning to the rest of your monitoring: thermal inspection via ZenVision helps separate soiling loss from real-fault loss, and SCADA supports the before/after comparison with continuous operational data. Cleaning without measurement is a cost; cleaning with measurement is a decision.

Next steps

If your plant still cleans manually or via an outsourced crew, the first step isn't buying a robot — it's mapping the real cost of your current routine: annual recurrence, campaign days, water use, cost per module, and the estimated gain after cleaning.

With that baseline, run the numbers through the TCR-W1 ROI calculator. If the plant has scale (≳6 MWp), recurrence (≥2 cleanings/year), and a compatible layout, automation tends to reduce variability, control resources, and professionalize the operation. If it doesn't, the math will tell you clearly to stick with manual — and that's also a correct decision.

The strongest choice isn't the most modern method. It's the method that improves plant performance with predictability, safety, and a measurable result.

CTA: Map out the current cost of your cleaning and simulate the scenario in the ROI calculator. If you'd like, our team will build the manual-vs-robotic comparison for your plant.

Frequently asked questions

From what plant size does a cleaning robot make sense? In STEMIS's technical model, with average soiling and 2 cleanings/year, payback falls below 5 years starting at around 6 MWp. Smaller plants, or those with only one cleaning a year, usually get better cost-benefit from manual cleaning.

How much water does the robot save? The TCR-W1 uses about 2 L per module for wet cleaning (closer to 1.5 L depending on soiling and speed), against ~10 L for the manual method — up to 80% savings. On a 10 MWp plant with 2 campaigns/year, that's roughly 320,000 liters saved per year.

Can robotic cleaning damage the modules? No. The TCR-W1 uses soft nylon bristles (0.3 mm) and pressure sensors, with IP65 protection. In UNICAMP's test (LESF No. 25-043), there was no mechanical or electrical damage, with 100% integrity confirmed by electroluminescence.

How much generation does cleaning recover? It depends on the soiling level, but UNICAMP's independent test recorded an average gain of +3.1% power after cleaning. The ideal is to measure before/after on your own plant, under comparable conditions.

Does the robot replace the O&M team? No. It replaces the repetitive, standardizable part of cleaning. Measuring soiling loss, prioritizing, and tracking results remain O&M work.