How to Power Remote IoT Devices

The real question is energy, not connectivity

We talk about remote IoT as a coverage problem. Does the site have 4G? Would LTE-M reach it? Is this a job for LoRaWAN, or for satellite? Those questions matter, but there is one that usually comes first and quietly decides the rest: how much energy is available where the device has to live?

A conventional industrial 4G or 5G router designed to stay online continuously has nothing in common, electrically, with an NB-IoT sensor that wakes four times a day. A camera is different again. And a tag that only needs to answer when interrogated may need so little energy that a battery becomes unnecessary. Energy is not a footnote to the connectivity decision. It shapes it.

There are really four ways to solve it, and they form a ladder. You can bring energy to the site, store enough of it in a battery, harvest what is already there, or engineer the device to need almost none. The further down the ladder you go, the less power has to be delivered, and the stranger the engineering becomes.

Solar and the energy-balance model

Solar has become the default for a reason. Panels are cheap, proven, have no significant moving parts, and scale from tiny cells to arrays that run cameras and actuators. Royal Mail is a good, familiar example of the problem it solves: its rollout of 3,500 solar-powered smart postboxes fits computers, a scanner and a modem to street furniture that was never meant to hold any of them. Nobody was going to trench a mains supply to every pillar box. Sunlight was already there.

The architecture is familiar: panel, charge controller, rechargeable battery, then the equipment. Notice that the panel does not power the load directly. It refills a store. In daylight the system usually makes more than it uses; overnight the battery carries it. That distinction is the whole game. A remote system is not designed around its instantaneous power draw. It is designed around its energy balance over time.

Think in joules, not battery capacity

Imagine a device that receives the equivalent of ten units of energy a day and consumes eight. That is sustainable. If it receives ten and consumes twelve, it eventually stops, and a bigger battery only postpones the failure. The long-term equation has to balance: energy harvested must meet energy consumed plus system losses, with enough storage to survive the lean spells. It sounds obvious. A surprising number of badly sized solar installations quietly ignore it.

Solar in Britain is not solar in Arizona

This deserves care for UK sites. A system should not be sized around how the panel performs at lunchtime in June. The real question is whether it survives December: short days, low sun angle, cloud, shading, leaves, dirt, snow, panel orientation, battery temperature, conversion losses and ageing all bite at once. Research into solar-powered air-quality sensing in Scotland has shown how quickly the budget tightens when you bolt continuous communications onto a modest sensor. A 100 microwatt sensor and a 10 watt gateway are both “solar IoT”, and they live in completely different worlds.

Sometimes the right solar panel is no solar panel

Suppose a sensor sends a handful of bytes twice a day. You could add a panel, a charge controller, a rechargeable cell, brackets, cables and glands. Or you could fit one high-quality primary lithium cell that runs the device for years with fewer parts and fewer things to fail. For very low duty-cycle LPWAN devices on NB-IoT, LTE-M or LoRaWAN, the long-life battery is often the more honest engineering answer. Solar is not automatically better.

Do not give every sensor its own power

Here is where architecture starts to matter more than hardware. Picture monitoring a railway cutting with thirty tilt sensors. You could give each one a panel, a battery, a modem and a SIM. It would work, and it would be needlessly complex. The alternative is to let thirty ultra-low-power battery sensors talk locally over sub-GHz radio or LoRaWAN to a single solar-powered gateway, and let that one gateway carry the expensive wide-area link.

Now the power-hungry radio exists once rather than thirty times. Variants of this pattern are already standard in railway and structural monitoring. It is one of the most important lessons in the field, and it applies just as much on a farm, where a soil probe that sleeps almost all its life has no business carrying its own mast. The same logic runs the other way too: if there is only one postbox on a street corner, a direct LTE-M or NB-IoT link makes far more sense than building a local network around it. There is no single correct architecture. The available energy helps decide it.

When the thing being monitored can power the sensor

Solar harvests energy from the environment. But many industrial assets already contain usable energy of their own. A railway vibrates. A pipe carries moving liquid. A steam trap runs hot. A power conductor throws off a magnetic field. A bridge flexes as traffic crosses. Instead of asking where a solar panel could go, you can ask what energy already exists here. The answers are unexpectedly elegant, because in each case the thing you want to monitor is also the thing that powers the monitor.

Vibration: the railway powers its own monitoring

When a train passes, rails, sleepers, wagons and structures all move. Piezoelectric materials generate charge when strained; electromagnetic harvesters turn movement into electricity with magnets and coils. Both have been demonstrated for railway condition monitoring. The relationship is neatly circular: the train stresses the track, the vibration is harvested, and the harvested energy runs the sensor that measures the stress. The event that loads the infrastructure helps pay for measuring it.

Bridges take the same idea further. Continuously wiring sensors across an existing structure is expensive, so researchers have built self-powered wireless bridge-monitoring systems where piezoelectric harvesters draw energy from structural vibration, store it, and use it for measurement and transmission. It is not perpetual motion. The passing vehicle supplies the mechanical energy; the sensor captures a minute fraction of it.

Heat: the hot machine powers its own sensor

Industrial kit sheds another kind of energy: heat. A thermoelectric generator exploits a temperature difference, so if one side is hotter than the other, you get electricity. Everactive has commercialised batteryless industrial sensing on exactly this principle, including steam-system monitoring, where the sensor draws its power from the temperature gradient it is already sitting in. Instead of replacing thousands of batteries across a plant, the estate harvests tiny amounts of energy from the process it monitors.

Magnetic field: the cable powers its own sensor

High-voltage infrastructure offers the same trick. Current in a conductor creates a magnetic field, and inductive harvesting captures a sliver of it. Fraunhofer IMS has built overhead-line condition-monitoring modules powered this way, drawing the few milliwatts they need by induction from the field around the conductor and dispensing with batteries entirely. The harvester even survives short-circuit currents, and the readings go back over a low-power wide-area link. The electricity cable powers the sensor watching the electricity cable.

Water flow: the pipe powers its own meter

Water networks may offer the cleverest version of all. Rather than fitting a battery or panel to a flow meter, you put a tiny generator in the flow. Researchers have demonstrated turbine-based flowmeters where the moving water both provides the measurement and generates the power for the electronics, with more ambitious designs adding pressure sensing, LoRaWAN and even valve control. For distributed water networks, irrigation and remote pipelines, that is a compelling proposition: water flows through the pipe, a micro-turbine makes electricity, the sensors measure the system, and a radio reports back. Similar work on oil, water and gas pipelines uses fluid movement, pressure and vibration, and has looked at powering systems such as cathodic protection from the flow itself.

Harvesting ambient energy

Beyond the asset itself, the wider environment is full of energy that a sufficiently frugal device can live on.

The sea never stops moving

Consider a buoy 100 kilometres offshore. No mains, often no cellular, and an expensive boat trip to change a battery. Solar buoys are common, but researchers are also harvesting the thing a buoy has in abundance: motion. Wave-energy harvesters, sometimes paired with wind, turn the sea into electricity, and recent marine systems combine both with environmental sensing and wireless links. In some designs the electrical output of the harvester itself reveals the wave and wind conditions, so the generator doubles as a sensor. High-latitude work makes the point sharply: in an Arctic winter there is barely any sun, but the ocean still moves, and drifting buoys using triboelectric wave harvesting have stored enough energy to run sensors and transmit by satellite. Pair harvesting with a low-power sensor and a non-terrestrial network link and a device can exist almost anywhere on Earth with neither local power nor local terrestrial coverage.

Indoor light is energy too

Solar does not have to mean outdoors. Photovoltaic materials can harvest indoor lighting, and while the available power is far lower than direct sun, so is the appetite of a modern building sensor. Companies such as EnOcean have long commercialised energy-harvesting building sensors that run on light, small temperature differences and the mechanical energy of a switch being pressed. That last one is energy harvesting in its purest form: the person pressing the switch provides the energy that sends the command, and there is no battery to change and no cable to run.

The strangest power station is the soil

It gets stranger. Microorganisms generate current as they metabolise organic material, and a microbial fuel cell captures a little of it. Researchers have built these for environmental monitoring, and recent work has looked specifically at soil microbial fuel cells for autonomous agricultural sensors: microbes in the soil break down organic matter, an electrochemical process produces electrons, the trickle is harvested and stored, and eventually a sensor measures and reports. Plant variants extend the idea, using the compounds plants release around their roots. None of this is ready to replace a battery tomorrow, but it belongs in an agricultural IoT story because it shows how far the question travels once a device needs only microwatts. We normally think of a field as the place that needs power. In these systems, the field is part of the power supply.

Store it in bursts, then send

Harvesting rarely delivers power neatly. Sunlight shifts, machines vibrate intermittently, trains pass now and then, RF comes and goes. Yet radios often need a relatively large current for a fraction of a second. The bridge between the two is storage.

A vibration harvester might make a fraction of a milliwatt continuously. A capacitor gathers it, and when the voltage reaches a threshold the device wakes, reads the sensor, fires a transmission and shuts down. This is why supercapacitors and power-management ICs matter so much: they let a device whose radio briefly draws tens of milliwatts run on a source that produces almost nothing on average. The trick is that peak draw and average draw are different numbers.

The cheapest watt is the one you never need

Before increasing generation, cut consumption. Do not leave a processor running if it can sleep. Do not keep a modem attached if the job is one report an hour. Do not power a sensor continuously if it needs five seconds to take a reading. And do not transmit raw data if useful information can be extracted locally. Remote power design should start with what can be turned off, not with how big a panel will fit.

Only speak when something happens

A great deal of IoT traffic amounts to saying nothing has changed. A tilt sensor does not need to tell the cloud every minute that it has not tilted. Event-driven designs invert this: an ultra-low-power stage watches continuously, and only when a threshold is crossed does it wake the processor, confirm the event and fire the radio. For flood alerts, intrusion, valve state, landslip and asset movement, the radio becomes something used when needed rather than something always on.

Edge computing can cut the radio bill

Edge computing is usually sold on latency, but it is also a power technology, because sending data costs energy. A vibration sensor taking thousands of readings a second could stream them all over cellular, or a local processor could analyse the waveform and report only that bearing vibration has risen and now matches a developing fault. A little local computation can save far more communications energy than it costs, which is one reason the trade-offs between a smart module, an SBC and a router are worth thinking through before you specify the link.

Removing the battery: Ambient IoT and 3GPP

Push the logic to its conclusion and the battery disappears. Radio transmissions carry energy, and sufficiently low-power electronics can harvest some of it. Passive RFID has done a version of this for decades: the reader energises the tag. Ambient IoT takes the idea much further, with tiny devices that harvest RF, sense temperature, movement or environmental conditions, and report a reading, all without a conventional cell.

Part of what makes this possible is backscatter. A normal transmitter generates its own radio signal, which costs energy. A backscatter device instead reflects and alters an existing signal: the gateway shouts, and the device changes the echo. Because it is not generating a transmission of its own, its energy need can be extraordinarily low. Combine backscatter with RF harvesting and you approach a model where the network infrastructure supplies both the communications and the energy.

Walmart and the 90 million question

The economics are best seen at Walmart, which is deploying millions of battery-free Ambient IoT tags across its US supply chain, with a target of around 90 million pallets covered by the end of 2026 across 4,600 stores and more than 40 distribution centres. The point is not the technology on any one pallet; it is what removing the battery does to the sums. Maintaining 90 million battery-powered trackers would eventually be an operational nightmare, however long the cells lasted. Take the battery out and the sensible number of connected objects changes. You move from connecting valuable machines, to connecting pallets, to asking why almost any object could not carry some digital intelligence.

Now part of the cellular standards

This is no longer only a proprietary idea. 3GPP defined the first Ambient IoT device class in Release 19, describing devices that are batteryless or carry only limited storage and harvest energy from radio waves, light, motion or heat, at complexity and power levels orders of magnitude below NB-IoT and LTE-M. Release 19 reached fully implementable specifications at the end of 2025, and the work has since moved into Release 20, which is extending Ambient IoT through 2026 with a slightly higher-power device type and further architecture support. The distinction matters: this is not NB-IoT with a smaller battery, it is a different class of device.

A practical power hierarchy for remote IoT

Before specifying anything, it helps to work the problem in order rather than reaching for a panel by reflex.

1. Is usable power already there? If a reliable supply sits beside the equipment, use it. Do not engineer harvesting next to a socket.
2. How much power can you eliminate? Reduce duty cycle, sleep processors, switch sensors off, cut reporting frequency, process locally.
3. Could a primary battery last the service life? If the device sips energy, a long-life cell may be the simplest, most reliable answer.
4. What energy exists naturally at the site? Sunlight, indoor light, heat, vibration, movement, wind, water flow, pressure, magnetic fields, radio energy, even biological activity.
5. How should it be stored? Battery, supercapacitor, hybrid, or nothing at all for passive designs.
6. Which link fits the remaining budget? Local radio, LPWAN, cellular or satellite, chosen around the energy you have, not the other way round.
7. Can sensors share infrastructure? One solar gateway serving a hundred battery sensors usually beats a hundred independent solar systems.
8. What happens in the worst week of the year? Design for that week, not the average one.

Energy-harvesting technologies at a glance

What should store the energy?

Generation is only half the problem. The store often decides whether a design works at all, because it has to smooth an uneven supply and cover the short, heavy demands of a radio.

The bigger picture

Traditional computing starts by assuming you bring electricity and communications to the machine. Remote IoT increasingly starts somewhere else: find the energy and the connectivity already available at the asset, then design the device around them. Sometimes that means a substantial solar array and battery bank running a router and a camera. Sometimes it is a lithium cell quietly powering an NB-IoT sensor for years. Sometimes a hundred tiny sensors share one solar gateway. Sometimes a passing train supplies vibration, or flowing water turns a generator, or a conductor’s magnetic field powers the monitor watching that conductor. And increasingly, a device needs so little that ambient light or stray radio waves are enough.

That is why the most important development in remote IoT may not be a better battery or a more efficient panel. It may be our growing ability to make the endpoint need less and less energy in the first place. The question is shifting from how do we get power to this thing, to how little power does this thing actually need, and eventually to whether there is already enough energy around it. The Internet of Things has not made it practical to run electricity to everything. It is slowly making that unnecessary.

Frequently asked questions