>An independent internet of things on an island
Ynys Enlli (Bardsey Island) is a place steeped in history, 3km from the mainland and at the northern edge of Cardigan Bay. Off the northwest coast of Wales, it is situated across the surging waters of Bardsey Sound, which separate the island from the headland, called ‘Pen draw’r byd’ in Welsh, meaning literally ‘the end of the world’.
It is remote, it is idyllic, but it can also be brutal. In stormy weather, the rough waters of the sound can prevent boats from setting sail for months on end. Its topology excludes it yet further from the mainland, as the large hill at its north-eastern flank cuts off any sight or sound of the world beyond its shores for the handful of farmsteads which nestle there.
But Enlli is dear to our hearts, and this is the kind of challenge we relish. So, when we were asked if we could provide this remote island with Superfast Broadband and an ‘Internet of Things’ (IoT) setup, we couldn’t say no.
The long-range wide-area network (LoRaWaN) gateway already being provided, as part of a larger programme that we are running across the county of Gwynedd, would only be as ‘useful’ as the intelligence itself that we sought to provide is ‘useful’ to people, but it wasn’t immediately clear to us what might be needed. Until we met the islanders.
We quickly learned that many of the things that we take for granted on the mainland, such as running water, electricity, gas, and oil, are precious and finite resources here. A mundane but vital part of the islanders’ everyday lives is checking the levels of water, gas and oil tanks, as well as wells and batteries. A failing solar panel or battery, a leak or a rapidly emptying water tank - can all be catastrophic.
We also grasped immediately that the islanders have an impressive array of skills. Like farmers in other remote places, they have had to be adept at many things. They are able to repair, install, design, build and set up a wide range of machines and tools. It didn’t take long to see that IoT devices could be of real and immediate benefit once we had delivered a high-speed broadband link to this magical place.
But we had a dilemma! The island is famous for being disconnected and makes a virtue of it. Visitors know that when they arrive they will be safe from calls, text messages, emails and online meetings - there is no signal and there is no broadband. The Island Trust was explicit that only residents of the island should be able to access the Internet, and yet here we were delivering a wireless bridge to the mainland in order to deliver up to 150Mbps superfast broadband connections to the islander’s homes.
The challenge was our IoT solution needed to be independent of our broadband link to the mainland. While this may seem counterintuitive, if the link to the mainland were to go down, a traditional LoRaWaN system that uses The Things Network or other cloud solutions would fail, as there would be nowhere for the sensors to send their data to. So, a requirement for our solution for Enlli was that it needed to be able to operate continuously, gathering data from the devices, regardless of the status of the link to the mainland.
A further consideration was that there was a minimal udget available for this element of the project. Despite this constraint, we felt that it was important to get something of real value out to the islanders, rather than a standard ‘dashboard’ downloaded from the web. We had to be careful about how much time and resource we could devote to the system’s design, sensor testing and deployment
The majority of our experience of LoRaWaN and IoT up to this point had been gleaned using The Things Network and various cloud providers of IoT solutions, but these would not be resilient enough. So, to run something on the island independent of the link to the mainland we needed to develop a custom-made alternative.
Well, it had to be a Raspberry Pi based solution, didn’t it? A bespoke IoT system, crafted in Wales, for use on a Welsh Island, needed a Welsh Computer to run it all!
TThe network server/controller that was chosen was ChirpStack; it is open source, and it can run on a Pi. This compact unit sits at the heart of the island’s network. It will also soon be able to send data across the link to an AWS (Amazon Web Services) data store for backup and security.
In addition to this, the Pi runs a web application server that ingests sensor data, then processes and visualises it via an Enlli-specific bilingual information panel. This information panel can be accessed by the islanders via any browser on their desktop computers and mobile phones
A key design feature is that the islanders themselves can manage the addition of new and replacement sensors themselves, and the unit can be remotely updated and supported by our team.
The system already supports water level sensors for water tanks and wells, and we plan to supply a couple of prototype solar panel output sensors on our next visit. Future releases will include support for nesting box sensors, water flow, and humidity as well as other weather sensors.
In six months and with a team of three people (putting in hours outside our normal working day), Kodergarten have:
- Designed, built, and deployed an operational Pi-hosted IoT system for use on Ynys Enlli.
- Trialled four different ‘tank sensors’, and finally found one that worked!
- Set up ChirpStack on the Pi.
- Designed an IoT system that didn’t require a link to the mainland and a standard IoT network. li>
- Created a Data Model that supports different properties/places and sensor types.
- Delivered a system that can be controlled and augmented by the islanders themselves.
- Designed and built a prototype IoT DC clamp sensor for monitoring solar panel activity within a week of the requirement being suggested.
The system is now up and running and the sensore are being calibrated, and will, we hope, come to be an essential and reliable part of the island’s fabric. It has the capability to transform some small but essential parts of island life as well as the study of wildlife. It is a robust, low power, low cost, multilingual and multi-sensor solution that can be deployed in a range of remote places which do not have access to standard IoT network services.