Latest / The Edge Computing Podcast with Fexingo: Local Compute, CDNs, and Distributed Infrastructure / How Edge Computing Is Reshaping Real-Time Smart Grid Resilience
Transcript
- Lucas: There's a substation outside Tampa that, until last year, would take about forty-five minutes to isolate a fault on a feeder line. Forty-five minutes of lights flickering, or worse, a full blackout spreading. Today, that same substation can sense a fault and isolate it in under two hundred milliseconds, and it does it without phoning home to the utility's central control room. Luna: Under two hundred milliseconds? That's faster than a human can blink. Lucas: Exactly. It's edge computing at work on the power grid. And it's not just Tampa — utilities around the country are starting to deploy low-latency compute nodes right at the substation, or even on pole-top transformers, to handle protection and control decisions locally. The central cloud still gets the data, but it's no longer in the critical path. Luna: And before we get deeper, just a quick thought: conversations like this, where we unpack real tech that's changing infrastructure — they take time to research and produce. If you've gotten value from these shows, a couple of dollars a month genuinely keeps them going. You can support at buy me a coffee dot com slash fexingo. It's small, but it makes a real difference. Lucas: Yeah, exactly. And we keep it ad-free because of listeners like that. So, back to the grid — why now? Why are utilities suddenly willing to put compute at the edge instead of keeping everything centralized? Luna: I think the big driver is reliability. The old model where every sensor and relay reports back to a central SCADA system works fine when everything is stable, but when a hurricane hits, or a wildfire threatens a line, those central connections can drop. Edge nodes keep running. Lucas: Right, and they have local power, too. Many of these nodes are paired with a small battery and a solar panel, so even if the grid goes down, the edge node stays up. That local intelligence can keep sectionalizing the grid, isolating faults, and reconfiguring feeders without waiting for a human operator to log in from a control center that might also be offline. Luna: There's a specific example I came across — Florida Power and Light has been piloting this. They installed edge compute units at over two hundred substations across the state. After Hurricane Ian in 2022, they credited those units with reducing outage durations by something like forty percent in the areas where they were deployed. Lucas: Forty percent is huge. And that's not just convenience — for hospitals, water treatment plants, cold storage warehouses, those hours of avoided downtime translate directly into lives and money saved. Luna: The economic angle is interesting too. A study from the Brattle Group estimated that if the entire US grid adopted edge-based fault isolation, utilities could save about one point two billion dollars annually in avoided outage costs. That's just the direct costs — not counting things like lost business revenue or spoiled food. Lucas: So we're talking about a technology that pays for itself, and fast. But it's not just about hurricanes. Think about the increasing frequency of 'load shedding' events during heat waves — edge nodes can help manage demand response locally, too. Instead of the utility sending a signal to a million smart thermostats over the cloud, the edge node at the substation can broadcast a local frequency signal that devices respond to in real time. Luna: That brings up an interesting tension. Utilities have historically been very conservative about letting distributed intelligence make decisions without central approval. There's a real cultural shift happening. Lucas: It's a shift forced by physics as much as economics. As we add more renewables — solar and wind — the grid becomes less predictable. A cloud passing over a solar farm can cause a sudden dip in generation. Central control loops that poll every four seconds can't react fast enough. Edge nodes operating on sub-cycle timescales can balance local supply and demand before the central system even notices the imbalance. Luna: Sub-cycle meaning in under one sixtieth of a second? That's the grid's frequency cycle. Lucas: Yes. Sixty hertz. A full cycle is about sixteen milliseconds. These edge nodes can detect a frequency deviation and adjust within a few milliseconds. That's the kind of speed that prevents cascading failures — the kind that caused the 2003 Northeast blackout. Luna: And all of this is happening on hardware that's not much bigger than a shoebox. The edge compute units I've seen are fanless, ruggedized, and designed to sit outside in a substation yard for twenty years. Lucas: They're essentially hardened servers running real-time operating systems. Some use FPGAs for deterministic latency. And the software stack is open source in many cases — Linux-based, using protocols like DNP3 and IEC 61850 that are standard in the utility world. Luna: So it's not bleeding edge in the sense of unproven technology. It's mature, but it's been slow to deploy because of the regulatory and institutional barriers. Lucas: Exactly. The technology has been ready for years. What changed is that after several high-profile blackouts — Texas in 2021, California's public safety power shutoffs — regulators started asking harder questions about resilience. And utilities realized that waiting for the cloud was no longer acceptable. Luna: There's also the cost aspect I mentioned. With the price of edge compute hardware dropping, the business case becomes clear. A single edge node might cost a few thousand dollars, but if it prevents a single major outage, it pays for itself a hundred times over. Lucas: And the software-defined nature means utilities can add new capabilities without swapping hardware. Want to run a machine learning model that predicts transformer failures based on harmonics? You can deploy that as a container on the existing edge node. Luna: That's a huge shift from the old model where protective relays were hard-coded and any change required a truck roll. Lucas: Right. So the edge is making the grid not just faster, but more adaptable. And I think we're going to see this accelerate. The Department of Energy's Grid Modernization Initiative has been funding pilot projects that specifically require edge computing. Luna: One of those pilots I know about is in Chattanooga, Tennessee. The local utility, EPB, already has a fiber network that reaches every home, and they've been using that backbone to deploy edge nodes that enable real-time voltage optimization. They've reduced energy consumption by something like three percent systemwide just by fine-tuning voltage at the edge. Lucas: Three percent might not sound huge, but for a utility serving a hundred and fifty thousand customers, that's millions of dollars in savings, and less generation needed, so lower carbon emissions. Luna: And it's all done with local compute and local data. The edge node knows the voltage at the end of the feeder, can adjust the tap changer on the transformer, and never sends that data to the cloud unless there's a fault. Lucas: That's important for cybersecurity too. The less data that travels over wide area networks, the smaller the attack surface. Edge nodes can encrypt and authenticate locally, and they can even operate in a 'dark' mode where they don't communicate externally at all unless there's an event. Luna: I think that's a point that doesn't get enough attention. When people talk about grid security, they often focus on the control center. But the edge nodes are like a distributed immune system — they can contain an infection locally. Lucas: Exactly. And they can also support microgrids. If a hurricane takes down the main transmission line, an edge node can island a section of the grid that has local solar and battery storage, keeping power on for critical facilities even if the rest of the city is dark. Luna: That's exactly what happened in a small community in Puerto Rico after Hurricane Maria. They had a microgrid controller with edge compute that islanded automatically and kept a hospital running for three days until the main grid was restored. Lucas: So the pattern is clear: edge computing isn't just a nice to have for the grid. It's becoming a fundamental enabler of resilience in an era of more extreme weather and more distributed generation. Luna: And the interesting thing is, the same technology stacks — ARM processors, container orchestration, real-time Linux — are being used in industrial automation, autonomous vehicles, and telecom. It's a convergent platform. Lucas: That convergence means the cost keeps coming down and the capabilities keep going up. Five years ago, you couldn't get a sub-millisecond control loop on a passive-cooled edge node. Now it's off the shelf. Luna: So where do you see the biggest adoption in the next two years? Is it hurricane-prone coastal utilities, or are we going to see it in the Midwest for winter storms? Lucas: Both, but I think the wildfire risk in the West is going to drive a lot of deployments. Utilities in California are already using edge nodes to trip lines when they detect a fault that could spark a fire — and they have to do it faster than a human can react. That's a killer app. Luna: And those edge nodes are also being used to monitor vegetation growth with AI, analyzing camera feeds locally to decide if a tree branch is too close to a power line. Lucas: Right. So the same hardware does protection, control, and asset management. That's the vision — a unified edge computing platform that makes the grid smarter, safer, and more resilient, all without a dependency on the cloud. Luna: It's a good reminder that edge computing isn't just about faster Netflix streaming or lower latency for gaming. It's about foundational infrastructure that keeps the lights on. Lucas: Literally. And it's happening now, substation by substation.