The question the industry could not answer
When electric vehicles arrive in numbers on a residential street, what actually happens to the low voltage network? And can smart charging manage the evening peak instead of digging up roads to reinforce it?
Modelling could not settle it. So Western Power Distribution built the trial that would. The collaboration partners were EA Technology, DriveElectric, Lucy Electric GridKey and TRL. Two smart charging platforms ran side by side, CrowdCharge and GreenFlux. Between February 2017 and July 2018, 673 smart chargers went into homes across the Midlands, the South West and South Wales.
The trial went on to produce more than 130,000 charging events and close to two million hours of charging data. None of it existed unless every charger could reach its back office, every day, from wherever the homeowner happened to have put it.
Why the obvious answer was the wrong one
The project ruled out running ethernet through people's homes early. Drilling through walls for a benefit the householder does not directly feel is a poor trade, and participants might want the kit removed at the end of the trial. Mobile data was considered and rejected: coverage on the private low power networks was patchy, and per-device data cost would have undermined the economics of any eventual national rollout.
That left consumer Wi-Fi, which sounds fine until you try it across 673 houses. At 2.4GHz and 5GHz a signal has to survive cavity walls, foil-backed insulation and metal garage doors, and then compete with whatever the neighbours are running on the same channels. It works in a demonstration. It fails across 673 houses.
What we specified
The final report credits The Tech Factory with identifying the solution, citing our event Wi-Fi work and our experience with awkward wireless data challenges.
We specified a secure Wi-Fi bridge: a paired transmitter and receiver running on a lower frequency channel than domestic broadband equipment uses.
The engineering trade becomes obvious once you look at what the application actually needs. A lower frequency buys range and considerably better penetration through building fabric. It costs you data rate. But a charger sends plug-in events, meter values and heartbeats, and receives current allocation instructions. The packets are tiny. Throughput was never the constraint. Getting a usable signal from a garage to a hallway router was the constraint, and the bridge solved it.
The bridge went into 549 of the 673 installations. Hardwired ethernet covered 89 where the router happened to be close enough, and power line carrier units 31.
From September 2017 we supplied a relay unit so installers could push the range further on the difficult properties, instead of falling back to a less reliable method.
Configuration, logistics and the hard yards
We configured the CrowdCharge controllers before dispatch and handled hardware distribution to the installation companies. Five separate installers covered four licence areas, so equipment had to arrive configured, matched and ready to commission, with procedures the installers could follow.
When a charger went offline and could not be recovered remotely or by talking the participant through it, somebody had to attend. The report is explicit that this work came to us rather than to the electrical installers, because diagnosing it needed detailed communications knowledge.
We completed 346 site visits. 286 of the 673 chargers were visited at least once, by us or by Siemens, who were separately commissioned by Alfen to update firmware on 59 units. The split tells its own story: 199 of the 328 CrowdCharge installations needed a visit, against 87 of the 345 GreenFlux ones, because GreenFlux faults were more often fixable remotely.
Visit volume was 160 in 2017 and 109 across the whole of 2018. That decline is the number we are proudest of. It tracks a steady rise in fleet communications reliability, because each visit fed procedure changes, configuration fixes and installer guidance, so the same fault stopped recurring.
One recurring finding: qualified electricians are not communications engineers. We found poorly terminated RJ45 plugs, commissioning steps skipped, and a general reluctance to use diagnostic software on site. That is not a criticism of the installers. It is a supply chain that had never needed those skills, and a pricing model that did not pay for the extra time a smart install genuinely takes.
Deploying connected devices into premises you do not control?
Talk to our engineering team
The outcome
By the second half of 2018 communications reliability across both platforms had stabilised at around 80 per cent. That was enough to move 80 per cent of CrowdCharge participants and 86 per cent of GreenFlux participants into the demand management trials, which is where the project's findings came from.
Those findings are now part of the evidence base for UK smart charging policy. Every EV charging at home is roughly equivalent, in energy terms, to adding a new house to the network. There is real flexibility in when people charge. Time of use incentives combined with smart charging move demand off the evening peak effectively, and acceptably to customers.
What we would do differently
The report's lessons learned section names us alongside the charger manufacturer for configuration problems: controllers despatched with the wrong configuration, and mismatched controller and charger sets. It is a fair criticism, and we would rather state it here than have you find it yourself on page 87.
Three things we changed as a result, and still apply:
- Configuration is a build process, not a task. Every unit that leaves us goes out against a serial-matched build record with a verification step, so a device and its paired hardware cannot be separated in transit.
- Design for two paths from the start. The GreenFlux chargers carried a multi-network SIM as backup and needed a third of the site visits. Dual-path communications cost more per unit and far less per estate. This is the reasoning behind the SpeedFusion architectures we build today.
- Commissioning has to be proven, not declared. The procedure said confirm the connection before leaving site. Sites where that was skipped generated visits for two years. Proof of commissioning now means a back office record, not a tick box.
Why this matters now
When Electric Nation started, smart charging was a research question. It is regulated now, and the volumes are not 673 units. Charge point operators, distribution network operators, metering providers and vehicle to grid platforms are all pushing tens of thousands of connected devices into premises they do not control, over connections they do not own.
The failure modes have not changed. Radio paths through domestic building fabric. Single-path backhaul. Unverified commissioning. Configuration drift across a supply chain.
We have run that problem at national scale, and the results are in a 591 page public report rather than a testimonial.
Sources
Every figure on this page comes from the project's published final report. Nothing is estimated.
- Electric Nation Customer Trial Final Report, WPD_NIA_013, 24 October 2019. The Tech Factory is named in Figure 4-1 (page 20), Section 4.3 (page 23), Section 5.3.1 (page 74), Section 5.5 (page 81) and Section 5.7 (page 87).
- Published with the permission of the project team. The Tech Factory was a supplier to EA Technology on this project, not a named collaboration partner.
Related reading
- Network design for distributed estates. How we architect resilience across sites we do not control.
- SpeedFusion and multi-path resilience. The dual-path principle Electric Nation taught us, built properly.
- Enterprise multi-site connectivity. The same problem shape, in retail, utilities and industry.
- All case studies.