Latest / Internet of Things with Fexingo: Connected Devices, Sensors, and Industrial IoT / How IoT Sensors Prevent Wildfire Ignition on Power Lines
Transcript
- Lucas: California's 2021 wildfire season included a fire called the Dixie Fire. It burned nearly a million acres, destroyed over a thousand structures, and the cause was traced to a power line — a conductor making contact with a tree. Luna: That's the classic failure mode. A line sags in the heat, brushes against vegetation, and the arc ignites dry grass below. It's terrifyingly simple. Lucas: Exactly. And utilities have known about this for decades. But the traditional response was reactive — you patrol lines after a fire starts, or you preemptively shut off power to millions of people during high-risk conditions. That's the Public Safety Power Shutoff, or PSPS. Luna: Which is incredibly disruptive. Hospitals, water pumps, people on life-support equipment — all lose power. Lucas: Right. So a growing number of utilities are now deploying IoT sensor networks that detect the conditions before a fire starts. I want to focus on one specific system: line-mounted sensors that measure vibration, conductor temperature, sag angle, and electrical current in real time. A company called GridSens — they've installed sensors on tens of thousands of miles of distribution lines across California and Australia. Luna: These are the little hockey puck looking devices clamped directly on the conductor, right? Lucas: Exactly. Each sensor is about the size of a soda can, self-powered from the line's electromagnetic field, and it communicates wirelessly back to a central platform. The key is that they measure things a human inspector can't see from the ground — like the exact tension on the conductor. If a line sags too much due to thermal expansion, the sensor detects the angle change and flags it. Luna: So you get an alert that says 'this span is approaching critical sag' — hours or days before it actually contacts a tree. Lucas: That's the idea. And there's another mode: high-impedance faults. When a conductor breaks and falls to the ground, the current doesn't spike — it actually drops, because the ground contact is high resistance. Traditional circuit breakers don't trip. The sensor can detect that subtle drop and report it within seconds. Luna: Because it's measuring not just current but also harmonics — the electrical noise signature of an arc. Lucas: Right. The sensor's onboard processor runs a machine learning model that's been trained on thousands of arc events versus normal switching or lightning strikes. It can distinguish a tree branch brushing against the line from a squirrel touching two phases. And it sends an alert to the utility's operations center. Luna: How fast does that alert go out? Lucas: They claim under 30 seconds from event to notification. That's fast enough for a crew to be dispatched while the arc is still happening — or at least before it ignites dry fuel. In some cases, the system can even trigger a remote-controlled switch to de-energize that section of line automatically. Luna: That's a big shift from the standard approach, where you wait for a fire report or a customer call. Proactive vs. reactive. Lucas: It is. And it's not just California. Australia — specifically Victoria and New South Wales — has mandated similar systems after the 2009 Black Saturday fires that killed 173 people. Their utilities now deploy millions of sensors across high-risk bushfire zones. Luna: What about the data volume? Tens of thousands of sensors, each reporting every few seconds — that's a lot of data. Lucas: It is. Most of the processing happens at the edge — on the sensor itself. Only alerts and summary statistics are sent to the cloud. The raw waveform data is stored locally and only uploaded if an event is flagged. That keeps bandwidth and power consumption manageable. Luna: Makes sense. So you get a real-time heartbeat of the grid without drowning in telemetry. Lucas: Exactly. And this is the kind of technology conversation that — if it gave you something useful today — I'll mention that we keep this show ad-free intentionally. No commercial breaks, no sponsor reads. If you want to support that choice, the link is buy me a coffee dot com slash fexingo. Luna: Yeah, it's a model we believe in. Keeps the focus on the engineering and the impact. Lucas: So back to the grid. One challenge I find really interesting is how these sensors survive extreme conditions. They're clamped on lines that can reach 80 degrees Celsius in summer, and they have to endure ice, salt spray, and vibration. Luna: And they're expected to run for years without maintenance, I assume. Lucas: Right. The manufacturer claims a ten-year battery-free life, because they harvest energy from the line's magnetic field. But if the line goes dead — say, during a PSPS event — they have a small backup battery that keeps the sensor reporting for about 48 hours. Luna: So even when the grid is off, you still get situational awareness. That's smart. Lucas: It is. And the data from those periods is actually valuable — you can see which lines are still under tension from wind or thermal contraction, and whether any vegetation is still in contact. Luna: Let's talk about cost. These sensors aren't free. What's the business case for a utility? Lucas: A single sensor is roughly two to three hundred dollars, and you might need one every mile on high-risk lines. For a utility with 100,000 miles of line, that's tens of millions. But compare that to the cost of a single major wildfire — liability settlements can run into the billions. PG&E alone paid over 13 billion dollars from the 2017 and 2018 fire seasons. Luna: So the ROI is almost immediate if you prevent even one catastrophic fire. Lucas: Exactly. And regulators are starting to push. The California Public Utilities Commission now requires utilities to submit wildfire mitigation plans that include early detection technologies. So it's moving from voluntary to mandatory. Luna: What about false positives? If every squirrel triggers a crew dispatch, you'd burn out your field staff. Lucas: That's the machine learning challenge. The model needs to be trained on local conditions — a line in the Sierra Nevada has different vegetation and wind patterns than one in the Central Valley. Utilities have to tune the detection thresholds over months, using historical data from past fires and near-misses. Luna: And you can't just set the threshold too high, or you'll miss real events. Lucas: Right. It's a balancing act. But utilities report that after tuning, the false positive rate drops below one percent of all alerts. And crews know that when an alert comes in, it's almost certainly real. Luna: That's impressive. So the next frontier — can these sensors predict vegetation growth rates? If you know a tree is growing toward the line, you can schedule trimming before it becomes a threat. Lucas: Some are doing that. The same sag-angle data can be correlated with temperature and line current to infer clearance distance. If you see the clearance decreasing over weeks, even if absolute sag is still safe, you know a tree is encroaching. That gives you weeks of lead time. Luna: Predictive vegetation management. That's a whole new capability from what started as a simple contact sensor. Lucas: It is. And it shows how IoT networks evolve. What begins as a fire prevention tool becomes a grid asset management platform — you can also detect things like insulator degradation, corrosion, even bird strikes. Luna: So we're moving from 'detect the fire' to 'prevent the conditions that cause the fire.' That's a profound shift. Lucas: It really is. And I think the next five years will see these sensors become standard on every overhead line in fire-prone regions. The technology is proven; now it's about deployment scale and regulatory will. Luna: And the data those sensors generate — anonymized and aggregated — could also help climate scientists model fire risk more accurately. Lucas: That's a great point. Some utilities are already sharing aggregated sag and temperature data with research institutions. It's a positive feedback loop: better data leads to better models, which leads to better prevention. Luna: Alright, I think we've covered a lot. From a hockey-puck sensor to a wildfire prevention ecosystem. Lucas: Exactly. And if you're curious, a lot of the technical details are in the filings utilities make with the California Public Utilities Commission — they're public records. The GridSens case studies are also worth a read. Luna: Good resources. Thanks, Lucas. Lucas: Thanks, Luna. Talk next time.