Latest / The Operations Podcast with Fexingo: Process, People, and Profit in Modern Business / How a Solar Farm Cut Panel Cleaning Costs by 70 Percent
Transcript
- Lucas: If you own a solar farm, your biggest recurring operational headache is probably dust. Luna: Dust? I mean, I get that dirty panels produce less power, but is it really that big a deal? Lucas: It is. A panel caked with dust can lose 25 to 30 percent of its output. And on a utility-scale farm — say, ten thousand acres — you're talking about millions of dollars in lost revenue each year if you don't clean them. Luna: Right, so the cost-benefit of cleaning is real. But cleaning a thousand acres of panels sounds like a massive operation. Lucas: It was. Until recently, most large solar operators used either high-pressure water trucks or contract crews with mops and deionized water. Both are slow, expensive, and water-intensive. A single cleaning cycle on a medium-sized farm could cost half a million dollars and take weeks. Luna: So what changed? I'm guessing there's a robotics angle here. Lucas: There is. A company called Ecoppia — they're Israeli — developed a lightweight autonomous scrubber that rides along the edge of each panel row. It uses rotating microfiber brushes and a dry-cleaning method — no water at all. Luna: No water? That's huge, especially in desert environments where most solar farms are built. Lucas: Exactly. But the real operations innovation wasn't the robot itself — it was the scheduling algorithm that decides when to deploy it. The farm we looked at is in California's Central Valley, run by a large independent power producer. They installed Ecoppia units on about 200 megawatts of capacity — roughly 600,000 panels. Luna: And they had to figure out how often to clean each section? Lucas: Right. Dust doesn't accumulate evenly. It depends on wind direction, rainfall, soil composition, even the tilt angle of the panels. A fixed schedule — say, every two weeks — is wasteful. You're cleaning panels that aren't that dirty, and you might miss a section that got hammered by a dust storm. Luna: So they built a predictive model. Lucas: They did. The algorithm ingests data from local weather stations, particulate-matter sensors, and historical soiling rates. It predicts the optimal cleaning window for each block of panels — clean only when the expected energy loss exceeds the cost of cleaning. Luna: That's basically a dynamic threshold. How much did it save them? Lucas: Over the first year, they cut cleaning costs by 70 percent. The water savings alone were significant — they used zero water for cleaning, compared to 15 million gallons a year before. Labor hours dropped 80 percent because they no longer needed crews driving around at night. Luna: And the robots run overnight, I assume, when the panels aren't producing. Lucas: Exactly. They operate from 10 p.m. to 6 a.m., when the farm's output is negligible. Each unit covers about 2,000 panels per night. The farm now cleans more frequently — but only where it matters — and their annual energy output rose by an estimated 4 percent. Luna: Four percent on 200 megawatts — that's eight megawatts of additional generation, basically for free. Lucas: And they didn't have to build new panels or acquire more land. It's pure operations leverage. The algorithm itself isn't that complicated — it's essentially a cost-benefit optimizer — but the integration with the robotics and the farm's existing monitoring system was the hard part. Luna: If these kinds of conversations have helped you think differently about your own operations, it's worth mentioning that listener support is what keeps this show ad-free. A couple of dollars a month genuinely makes a difference — you can find us at buy me a coffee dot com slash fexingo. Lucas: Yeah, and we really appreciate it. It's what allows us to dig into these specific cases rather than just scratching the surface. Luna: So back to the solar farm — one thing I found interesting is that they also had to redesign the robot's pathing to avoid damaging the panels. Lucas: That was a huge issue. Early versions of the robot sometimes left micro-scratches on the glass, which actually reduced efficiency over time. The microfiber brushes had to be precisely calibrated — too soft and they don't remove dust, too hard and they scratch. Luna: So they added a feedback loop. The algorithm also monitors the brush pressure and the condition of the panels. Lucas: Exactly. They installed a camera on each robot that takes a picture of the panel before and after cleaning, then uses machine vision to verify that no scratches occurred. If it detects a scratch, that robot is flagged for maintenance. Luna: That seems like a lot of data per robot. How do they handle the bandwidth? Lucas: The robots process the images locally on a small edge computer — they only send alerts and summary metrics back to the central system. Each robot generates about 50 kilobytes of data per night, which is manageable even on a cellular network. Luna: Smart. So the whole system is really a coordinated set of operational decisions: when to clean, how to clean, and how to verify quality. Lucas: And that's the lesson for any business. The visible innovation is the robot, but the real leverage is in the algorithm and the process design. Solar farms are capital-intensive assets — a 2 percent improvement in yield can be worth millions. Luna: Do you think this approach scales to residential rooftop solar? That's a much more fragmented market. Lucas: Probably not economically — at least not yet. The robot costs about $15,000 per unit, and a residential rooftop might only have 20 panels. The cost per panel would be prohibitive. But for commercial rooftops or carport solar, it could work. Luna: Interesting. So the operations principle here — dynamic scheduling based on actual conditions rather than a fixed calendar — applies broadly, but the specific technology has to match the scale. Lucas: Exactly. And that's what I love about operations stories. The best solutions aren't always the most technologically advanced; they're the ones that fit the specific constraints of the system. Luna: Any idea what's next for this farm? Are they expanding the system? Lucas: They're rolling it out to another 300 megawatts this year. And they're experimenting with a version of the robot that can also do thermal imaging to detect hot spots — early signs of panel failure. Luna: So the robot becomes a mobile sensor platform. Lucas: Right. Once you have a robot that travels the rows every few nights, you can add sensors cheaply. That's the kind of compounding operational improvement that makes a real dent in the cost of solar energy. Luna: It's a good reminder that the green energy transition isn't just about building more panels — it's about running them efficiently. Lucas: Exactly. And that's where operations thinking comes in.