How a Plain Distribution Panel Becomes a Smart One — From a Single Sensor to the Cloud, Where Electrical, Electronics, and IT Meet

In the last post I argued that electronics, IT, and electrical power look a lot more alike at the system level than people assume.

Electronics moves signals. IT moves information. Electrical moves energy.

Different subject matter. Same job description: deliver a resource, split it up, control it, protect it, and watch it.

So here's a more concrete version of that question.

What would it actually take to turn the ordinary breaker panel bolted to a wall in a house, an office building, or a factory into a "smart panel"?

My first instinct was to think big.

Do I need to redesign the panel? Develop a proprietary breaker? Build everything from the sensors to the comms module myself?

Then I looked at it again with my IT eyes, and I started to suspect the question itself was wrong.

Why would you rebuild electrical gear that already exists and already works?

Maybe what's needed isn't a new panel at all. Maybe it's getting the panel already on the wall to tell you how it's doing.

The panel you already have is doing its job

One thing to get straight up front.

"Not smart" doesn't mean "not good enough."

A breaker panel has a clear purpose. Split incoming power across circuits. Protect those circuits against overcurrent and short circuits. Cut the faulty one loose when it has to.

Existing electrical technology already does this well.

Which is why "let's build a smart panel" turning into "let's replace the protection scheme with software" is a dangerous move.

AI replaces the breaker? Protection stops working when the cloud connection drops? The panel is unusable when the server dies?

That's strictly worse than what you started with.

Protection stays with the electrical hardware.

You add one layer on top.

A data layer.

Smart starts with measurement, not control

Say "smart panel" and most people picture flipping breakers from their phone.

I think there's something you have to do first.

Measure.

It's hard to control a system properly when you don't know what it's doing right now.

Same story in IT. Nobody bolts automation onto a service when they don't know its CPU usage, memory usage, response times, or error rate. Anyone who's tried it and gotten paged at 3am knows why.

Observe first.

Same order for the panel.

Think about what you'd collect: voltage, current, real power, energy, power factor, frequency. Add internal panel temperature, per-circuit consumption, breaker state. That's plenty to start with.

Once you're measuring, the panel on the wall starts to change character.

It stops being a thing electricity passes through and becomes a thing that produces data.

How do you measure current?

The classic answer is a CT, a current transformer.

Roughly: instead of shoving a huge current straight into your measuring gear, it converts that current into a signal you can actually measure.

Depending on the product and the layout, there are split-core CTs that clamp onto the conductor without tearing up existing wiring.

That matters more than it sounds.

There's no law saying the smart panel market has to open with "rip out the old panel, install a new smart one."

In plenty of cases you can retrofit — leave the equipment alone and add metering on top.

Translated to IT, that's closer to installing a monitoring agent on a server you're already running.

Temperature data matters more than you'd think

In electrical gear, heat is a signal.

Connection quality, load, ambient conditions. Lots of reasons a specific point starts running abnormally hot.

But nobody can stand in front of a panel taking temperature readings around the clock.

Put a temperature sensor on it and collect continuously, and the picture changes.

Say a point normally hovers around 35°C.

Then at some point it starts drifting: 35 → 38 → 42 → 47°C over a long stretch.

"Alert above 50°C" and "this equipment has started deviating from its own normal pattern" are two completely different problems.

The pattern-deviation side is where AI eventually earns its seat.

You don't need to get to AI yet, though.

Getting the data comes first.

Bolting on a sensor doesn't make it a smart panel

This is where electronics and IT walk in.

Sensors produce data. Data stuck inside a sensor is worthless. Something has to read it.

So you need an MCU, a meter, a controller.

The chain looks like this:

Power → CT / power metering / temperature sensor → MCU or controller

Up to here, we're mostly in electronics and embedded territory.

But you want to see that data on a phone or in a control room.

Which means it has to get out.

Enter the gateway

Walk into an industrial site and you'll run into equipment speaking Modbus all the time.

So you build a structure where the meter holds the data and a gateway reads it and hands it upstream.

Sketching it conceptually:

Panel → smart meter / sensor → RS-485 / Modbus → edge gateway → Ethernet / LTE / Wi-Fi → server / cloud

From this point on, it's an IT system I recognize.

The gateway's job is to read data and ship it to a server.

That could be an HTTP API. It could be messaging like MQTT.

Protocols and network topology shift with what the site demands. The job doesn't change.

The physical state of electrical equipment becomes digital data and lands inside an IT system.

That's the thing that actually changes in a smart panel.

The architecture, roughly

Stripped down to its bones:

[Power]        Mains → Breaker → Load

[Electronics]  CT / power metering / temp sensor → Controller

[Edge]         Gateway → preprocessing / comms

[IT]           Network → API / MQTT → Data Platform → Database → Dashboard / Alert

[AI]           anomaly detection → pattern analysis → predictive maintenance

Laid out like that, a smart panel isn't one electrical product.

It's closer to a small cyber-physical system that starts in power, runs through electronics and edge, crosses network and cloud, and ends up in data and AI.

Do you have to build all of it yourself?

One business question is left.

If you're going into smart panels, is it a good idea to develop your own CT, your own power meter, your own comms module, your own gateway, and your own cloud?

I don't think so, necessarily.

The market is full of sensors, power meters, comms devices, gateways. If there's a proven part, use it.

The IT industry is comfortable with this.

You don't write your own database because you're building a web service. You don't implement a container runtime because you're running Kubernetes. You pick the components you need, wire them together, and create new value on top.

The same path exists for smart panels.

Electrical manufacturing capability + off-the-shelf metering and sensors + off-the-shelf gateway + your own data platform + dashboard / app

That's a fine place to start.

The competitive edge doesn't have to live in the sensor.

What data you collect, how you connect it, what meaning you surface to the user, and what problem you solve with it. That's the edge I'd bet on lasting longer.

Should the panel stop when the cloud does?

If you've done IT, one thing nags at you immediately.

Failure.

Say the smart panel is connected to the cloud. The internet drops. The cloud server dies. The gateway fails.

Protection cannot go down with it.

So if I were designing this system, I'd nail one principle down before anything else.

Protection runs locally and independently. The IT system sits above it and does observation and analysis only.

Separate the safety layer from the intelligence layer.

Breakers and protective devices work with no network at all. If the sensors and gateway die, power delivery and basic protection keep working.

Without the IT system, you still have electricity. With the IT system, you understand the equipment better. Never invert that order.

You don't need remote control on day one

Build products long enough and you start wanting to cram features in.

Kill a breaker from your phone. Let AI decide on its own. Remote control from the cloud. Turn on the power with your voice.

I'm honestly on the greedy end of that spectrum. I have a habit of drawing up the feature list before anything else. Looking back, that order was a bad idea.

Electricity sits right next to safety.

For an early product, stripping features out is the better instinct.

The first stage can be just this:

Measure → store → visualize → alert

The user opens an app and checks.

How much power am I drawing right now? Which circuit is eating the most? Is it higher than usual? Is the panel's internal temperature climbing? Is there a strange pattern at certain hours?

Do only that much, and do it well, and you've already created something the old panel couldn't give you.

Control and AI can wait until you've accumulated data and operational experience.

Maybe the real product isn't the steel box

Get this far and you land somewhere interesting.

In traditional panel manufacturing, the product is something you can see.

Form the sheet metal into an enclosure. Assemble breakers and components. Wire it. Ship it.

Sell it once and the transaction is often over.

With a smart panel, the relationship continues after delivery.

Sensors keep generating data. The server stores it. A dashboard shows state, and alerts fire when something goes wrong. You analyze long stretches of history.

So what you're selling isn't one panel anymore.

Panel + metering + connectivity + data + software becomes a single product.

Monthly monitoring or maintenance service models come out of that.

Hardware sales widen into software and service.

Where electrical, electronics, and IT actually meet

Back to the original question.

How does an ordinary panel become a smart panel?

You don't have to rebuild the panel.

The existing panel keeps doing what it does. Split the power, protect the circuits.

You add sensors on top. Electronics converts physical state into data. An edge gateway carries it. An IT platform stores it and shows it.

Once enough data has piled up, AI looks for patterns.

One line for what changed:

The old panel was equipment that divided electricity. A smart panel divides electricity and reports its own state as data.

That's the part I find fun.

Electrical alone doesn't finish it. Electronics alone doesn't either. You can't build it with IT in hand and nothing else.

Electrical handles the energy. Electronics measures the physical world. IT connects the data, and AI finds the patterns inside it.

Technologies I used to think of as separate fields start meeting inside one small steel box on a wall.

Maybe the competitive edge in smart panels isn't a better steel box, but what you turn the power data coming out of it into. I'm not sure yet either.

Next up

Which leads straight to the next question.

On a server, I watch CPU, memory, and network in real time. On Kubernetes, I collect piles of metrics, put them on a dashboard, and get an alert when something breaks.

Why shouldn't electrical equipment be watched the same way?

The next post is a slightly absurd thought experiment.

"What if you put Prometheus on a breaker panel? — electrical equipment needs observability too"