Energy
E.ON Energy wants to use AI to help fix your energy smart meter, and it’s hoping to save £10 million in the process, to be passed on to customers of course.
The UK energy company, a subsidiary of the German E.ON, is trialling a new video analysis tool able to diagnose smart meter problems in as little as five seconds.
At the AWS Summit in London, Sam Charlton, product manager at E.ON, said: “When a customer calls us and says something’s not working, if we can fix it instantly, without having to ask them to take a days holiday in two weeks' time when we visit their house … that’s a massive win for [them] and for us.”
See also: The energy sector’s digital transformation and data blockers
While it is yet to be rolled out, the tool, developed by AWS in just a few weeks, could see users required to upload just a seven second video of their fault meter to an AWS/E.ON platform.
Then, using AWS’ Textract, AI takes over the job of an engineer and analyses the video, pulling specific frames and using a "signal intensity score" to read the Meter’s five flashing error lights and compare them to the error manual to diagnose the problem.
With more than 1,000 possible light combinations, AWS Data Scientist Tan Takher said "clearly there’s a need for some kind of automation here".
In the best case scenario (a seven second video showing one low intensity error light), the tool can pull just three frames to analyse and reduce processing cost by 70x he claimed.
£10 million in savings?
Faulty smart meters are a big issue, energy companies have faced a lot of criticism over their rollout of the devices, which the government wants in 75% of homes by the end of 2025, and many have received fines for their poor efforts.
E.ON itself has been slapped with a few over the last decade, including a £1.72 million payment ordered by energy regulator Ofgem in 2023 for falling short on installation targets.
So, though the AI tool may seem fairly simple, Charlton and E.ON calculate it could save the company, and by proxy customers, around £10m per year through fewer home engineer visits, longer lifecycles for smart meter devices, and warranty payments on fault devices.
As he explained to The Stack, the savings calculation is based around the six million smart meter devices overseen by E.ON, costing £100+ each, the 249,000 that are faulty in an average year, and the 2,600 ‘meter health’ visits E.ON engineers carry out each week, costing £200 on average.
The exact calculation he couldn’t pass on, but it's success likely depends on how accurate the prediction is that E.ON could remotely fix 500 of those faulty meters per week through its headend network centre, the “golden fleece” of solutions according to Charlton.
However, people would also be surprised how often a connectivity issue can be fixed by simply moving something like an ironing board away from your device, he said.
A new approach
While the tool does sound useful, could E.ON not simply just digest the error messages sent by the smart meter to its network and diagnose the issue from there?
It’s a problem many energy companies have attempted to work on but, thanks to the lingering question of how to process the tons of error and action data sent by the devices everyday, has yet to be cracked, as far as Charlton knows, he told The Stack.
So, for now at least, he’ll stick with a user-submitted video solution and hopes to get backing for a national rollout of the diagnostics tool, potentially next year if it can pass exec approvals and boost its current 84% accuracy in field trials.
After that, E.ON and AWS said they could see many more use cases for the tool, telling the London audience about its potential to diagnose issues with energy distribution systems, EV chargers and telco networks.