Solar Panel Cleaning Robots: The Real Blocker Isn't the Cleaning Itself

As someone who has never touched PV O&M at all, last week I tagged along with a friend who does power station O&M and squatted at a station in the northwest for half a day. They've deployed Airtouch's AT 4.0, a PV cleaning robot from Airtouchsolar, specifically for washing solar panels.

First, what this thing does. Panels get dusty, power output drops—that's common knowledge. The problem is that large stations have panels spread across the entire Gobi, and having people spray them with hoses just isn't realistic. Hence the direction of "robotized cleaning." AT 4.0 takes the waterless route—no hose hookup, it relies on brushing and suction to carry the dust away. Why emphasize waterless—I figure it's two things: first, the northwest is short on water anyway; second, water washing leaves water spots, and once dry they actually block light even more. My friend agrees—he says what he dreads most is washing and then a sandstorm drops another layer right after.

What really surprised me was the accompanying AT IoT cloud platform. O&M staff aren't on site—they can see each device's status, location, and task progress right from an office computer, and can remotely dispatch cleaning tasks. I watched for a while—the interface isn't flashy, but the information density is high; which unit is running, which one is erroring out, you can scan it at a glance. That approach is right—hundreds of devices can't realistically be inspected by people.

There was a small hiccup in the afternoon. One machine went offline—the interface showed it as offline, my friend clicked in to reconnect, and it took a few tries to get it back. He said this isn't common but not unheard of either—it happens when the on-site network signal is unstable.

I asked him what annoys him most. He said it's actually relocation and scheduling. After finishing one area you have to move to the next—how to order the sequence, how to avoid strings under maintenance—all of that still has to be decided by a human. The cloud can monitor, but the scheduling intelligence hasn't reached the point where you can just throw it in and let it run itself.

The advantages are clear. Waterless, suited to water-scarce regions; centralized cloud management compresses the O&M radius for hundreds of devices; scenario-specialized—it doesn't chase generality, it focuses on doing this one thing—panel washing—solidly. I usually work in embodied AI and I'm used to solutions that want to generalize everything, so I actually think this kind of convergence is the right call.

The downsides are obvious too. It has prerequisites for the station's upfront conditions—network coverage, device placement, regularity of the panel arrays—if any link falls short, using it gets awkward. Also, it only solves the "periodic cleaning" link; scheduling still depends on people.

So the conclusion is: it depends. For large ground-mounted stations with regular panel arrays and basic network coverage, it's worth a serious evaluation. For distributed rooftops, scattered arrays, or O&M teams that are just one or two people to begin with, the learning and deployment costs of this thing might outweigh the benefits.

For me, the value of this thing is making a task that must be repeated manageable—whether the robot is smart is secondary.