Severn Trent trials machine learning to detect leaks

Severn Trent has announced it has started to use machine learning to help transform the way it approaches leaks.

The water company says it has created an “advanced leakage detection model” that uses data from its network to help identify, locate and manage leaks.

Severn Trent teamed up with consulting and technology firm Capgemini to create the machine learning model which, the company claims, is already seeing notable early results.

In areas where the model has been used, Severn Trent says time to find leaks has “reduced by over half”, with engineers being able to pinpoint leaks more accurately.

It added that overall the network in those pilot areas saw a reduction in leakage of over 16 per cent.

Rob Ryder, technology and lab data manager at Severn Trent, said: “We’re really excited by this project, and it’s fantastic that we’re already seeing brilliant results.

“We’ve created the model by applying advanced analytics to data from our network of pipes collected through sensors, which gives us loads of valuable data which we can use to anticipate and manage leakage across our region.

“When using the model we’re getting the most out of our network data, giving us the upper-hand when it comes to locating and fixing leaks.”

Amit Ghosh, group account executive energy and utilities at Capgemini’s UK business unit, said: “The success of the leakage detection programme demonstrates strong partnership ethos between Severn Trent and Capgemini.

“We brought our deep expertise in data science and analytics, and our pre-built analytical models, which combined with Severn Trent’s subject matter expertise in leakage operations enabled them to fast track the rollout of the solution.”

Water companies are increasingly beginning to use technology to help detect and manage leakages in their systems.

Severn Trent is one of several companies to use satellites, with the company announcing it was trialling the devices in November last year in a bid to reduce leakage by 15 per cent.

The company, which already used drones to spot leaks from the air, said it is going a step further with its “leakage detection from space”.

In September Yorkshire Water announced the first stage of its £300,000 satellite leak detection trial had been “hugely successful” and had saved 0.5 megalitres of water per day.

Meanwhile in August Thames Water announced it had launched a three-pronged aerial attack in its hunt to find and fix leaky pipes. The company is using a fleet of drones, an aeroplane and a satellite to support its mission to get back on track with its leakage targets by 2020.

The industry has been challenged by the government and Ofwat to tackle leakage.