How Energy Harvesting Powers Self-Sustaining Industrial IoT

Maintenance-free operation is an ambition of industrial IoT. Energy harvesting is the technology most often held up as the path to that goal, drawing power from the light, vibration, or heat already present in the environment. For a long time, it sat closer to research than to product. A mix of more capable low-power components and better materials has started to close that gap.

Designing around harvested energy reaches back into the earliest architecture decisions in embedded engineering: how much of a power budget you actually have, what you store energy in, which wireless protocol you can afford to run, and how long the processor is allowed to stay awake. It reframes the work from building a device that runs to building one that sustains itself on whatever energy it can scavenge.

Dunstan Power, Director at ByteSnap Design, has spent time designing systems to that brief. We spoke with him about the mindset energy harvesting demands, how to match a harvesting method to its environment, the part storage and wireless choices play in a realistic power budget, and broader trends in embedded design.

ByteSnap design released an engineering blueprint, "Energy Harvesting: The Key to Maintenance-Free Industrial IoT. What problem is ByteSnap Design trying to solve with this blueprint, and why is battery maintenance such a critical issue for industrial IoT right now?

The reality is that battery replacement is a massive silent killer for Industrial IoT deployments. When you design a sensor and put a battery in it, it sounds fine on paper. But when a client deploys five thousand of those sensors across remote pipelines, high up on factory gantries, or deep inside hazardous zones, reality hits. Every single battery becomes a ticking clock. 

Sending a technician out with a ladder, safety gear, or a permit to work in a hazardous area just to swap a cheap battery is incredibly expensive. Sometimes it costs hundreds in labour and downtime just for one node.

Multiply that by thousands of devices, and the business model completely breaks down. That is the exact headache we are solving with this blueprint. We want to show engineers how to build self-sustaining, fit-and-forget systems that can run for a decade or more without anyone having to touch them.

What are the biggest mistakes engineers make when they first approach power budgeting for distributed sensor networks?

The biggest mistake is sheer optimism about how much power they can harvest.

A common customer initial requirement is to build a device that continuously draws tens of milliamps. Maybe they want to stream constant data or run heavy processing, and they assume a tiny solar cell or a vibration harvester can feed it. It just doesn’t work that way.

If your device needs continuous milliamps, energy harvesting is going to disappoint you. You have to design around a power budget of microamps during sleep mode, allowing for only tiny, fast milliamp pulses when transmitting data. That requires an absolute shift in mindset, as you aren't just designing for power generation; you are designing strictly for extreme energy conservation.

Your blueprint argues that energy harvesting is becoming practical for maintenance-free IIoT. What’s changed recently to make that possible?

It's a combination of silicon becoming incredibly efficient and materials technology making huge leaps. For years, for example, indoor solar performance was poor. Take, for example, photovoltaic cells under office or warehouse lighting, which provide around 15% to 25% efficiency. Recently, commercial perovskite PV cells have hit the market, pushing efficiency up to 32% to 38%. That's a game-changer for indoor industrial setups.

At the same time, we've seen ultra-low-power silicon and Power Management ICs (PMICs) mature to the point where they can wake up, process, and go back to sleep on absolute table crumbs of energy. What used to be academic laboratory experiments are now rugged, commercial-grade baselines we can actually design into real products

How should engineers think about choosing between solar, vibration, thermal, or hybrid energy harvesting approaches?

You have to look entirely at the physical environment of your specific deployment. The surrounding ecosystem often matters more than the technology itself. Solar (PV) is the most mature option. If it’s outdoors, it's a no-brainer. If it's indoors, we now look at those high-efficiency perovskite cells.

Vibration (Piezoelectric) is perfect if you’re mounting directly onto heavy, rumbling industrial machinery or conveyor belts. Vibration harvesting generally falls in the range of 10–100  µW/cm³, though the real figure for any given installation depends heavily on the vibration source, its frequency, amplitude and how directly the harvester is coupled to it. That's plenty for intermittent sensor updates, but it's not a number you should quote as fixed; it needs confirming against the actual machinery in each deployment. 

Thermal (Thermoelectric) is ideal if you have a reliable heat differential. If you have a hot pipe, a boiler, or an exhaust stack, even a tiny 5°C temperature difference can generate usable power via the Seebeck effect.

Hybrid is where the industry is moving to, though. To get true resilience, you often need to combine the other inputs, like putting solar and vibration together, so that if one ambient source drops out, the other keeps your system alive.

How do supercapacitors fit into these designs, and what should engineers pay closest attention to when selecting one?

Because ambient energy is unpredictable, you almost always need storage. If you're relying on light, what happens when the factory lights go off at night?

Supercapacitors are ideal for low-power devices because they handle rapid charge and discharge cycles beautifully and can provide those quick, high-current pulses needed when a wireless transmitter wakes up to send data.

But here’s the catch: you have to look closely at their operating temperature limits and long-term degradation. Supercapacitors degrade over time, especially in harsh, hot, or freezing industrial environments. If you want a product to actually last 10+ years, you cannot design right to the limit. You have to overrate your storage components from day one to build in a safety margin for the end of the product's lifespan.

Wireless protocol choice has a major impact on energy demand. How do LoRaWAN, Zigbee, BLE and NB-IoT compare when power is the priority?

Wireless choice matters, but it is not the whole story. What really determines viability is whether your energy storage can supply the burst current a protocol needs at the moment it wakes up.

LoRaWAN tends to be the easiest fit because it lets the transceiver sleep at microamp levels, wake in a millisecond, send a packet over long range, then drop straight back to sleep. 

BLE is excellent for short-range, low-power sensor-to-gateway links but will not cover a large industrial site. 

Zigbee's mesh routing keeps a toll on power. NB-IoT is appealing because it rides on existing cellular networks, but the peak current during network registration is high enough that your storage design has to be built to absorb that surge. 

Can you talk us through the rail case study and what it showed about the real-world viability of energy harvesting in harsh environments?

This was an exciting project for us. We were asked to design a trackside backup monitoring system to check the integrity of railway signalling control lines. From previous experience, we already knew that working near railway lines is incredibly restricted. You can't just run power cables out there, and you absolutely cannot send technicians out to change batteries every few months because of massive safety and logistical headaches. It had to be 100% self-sustaining.

We ended up designing a triple-hybrid harvesting system. We combined piezoelectric harvesters to capture the intense vibration of passing trains, solar panels for daylight, and even small wind turbines. We fed all of this into an Analog Devices ADP5091 PMIC, backed up by a robust supercapacitor network.

The supercapacitors stored enough energy during peak harvesting windows to reliably supply the high-pulse currents needed for the LoRa transceivers, even in the middle of the night when there was no wind and no trains were passing. It proved beyond a doubt that even in the harshest, most highly regulated environments, you can achieve reliable, zero-maintenance IoT.

ByteSnap also showcased Edge AI running on low-power hardware recently. How does that fit into the broader move toward ultra-efficient embedded systems?

It fits perfectly because harvesting energy is only half the battle. The other half is extreme conservation. Traditionally, IoT sensors gathered raw data and transmitted it all to the cloud for processing. But wireless transmission is the most power-hungry thing an embedded system does. If you are constantly radioing data back and forth, you're going to drain your system fast.

This is where Edge AI comes in. At the Design Engineering Expo, we demonstrated a real-time object detection vision system running entirely on-device using an ultra-low-power Infineon PSOC Edge processor. 

By running machine learning and computer vision locally at the edge, the device only needs to turn its radio on and transmit when something actually changes or when an anomaly is detected. This keeps the system in a deep sleep state for the vast majority of its life, saving massive amounts of power and making energy harvesting highly viable even for complex tasks.

Looking beyond this blueprint, what broader trends in embedded design are you watching most closely?

Two big trends are keeping us busy right now. First is component obsolescence. In our industry, keeping products alive for 10 to 15 years is a major challenge. Between supply chain shifts and high-performance components being discontinued, technical directors are constantly being pulled away from innovation just to deal with component status changes. It’s why we actually launched our own Obsolescence Management Service (OMaaS) recently. We wanted to take that administrative headache off engineers' plates so they can stay focused on next-generation designs.

Second is the structural shift we are seeing in the semiconductor market. Because of the insatiable demand for AI data centres, we are seeing serious constraints on memory availability in particular. It is threatening to disrupt manufacturing lines just like the pandemic-era shortages did. 

We are working closely with clients on 'Design Rescue' projects, proactively redesigning boards and rewriting firmware to support whatever components are actually available in the supply chain, rather than just waiting and hoping for things to recover.

‍