Machine learning boosts leak control

More than 18 billion litres of water is supplied by water companies every day in the UK. But figures from Water UK estimate that three billion litres of that is lost through leaky pipes. 

Government and regulators have introduced ambitious plans to improve this statistic and reduce water consumption, with Ofwat challenging providers to reduce leaks by 16% by 2025.

But updating old pipe networks isn’t an overnight fix and water companies up and down the UK are looking at different innovative methods to help them meet Ofwat’s target. 

United Utilities, for instance, has installed more than 100,000 sensors across its network to detect leaky or breached pipes, by picking up the vibrations caused by the flowing water. 

Meanwhile, in February Yorkshire Water announced it was carrying out the UK’s most advanced smart water network pilot in Sheffield, using technologies such as acoustic loggers, pressure loggers and flow meters to monitor for leaks. 

Early results show these investments are working. Another report from Water UK found that utilities companies have reduced the amount of water leaking from pipes by 7% to the lowest levels since records began in the mid-1990’s. That’s the equivalent of 216 million litres being saved every day.

Tech investment is key

The common thread between United Utilities and Yorkshire Water’s response to leaking pipes is technology, and we’re seeing more and more water organisations investing in innovative and modern technologies to assist with pipe network maintenance. 

That was also the case for South East Water (SEW), who approached our organisation last year to support with their leak prevention strategy.

SEW wanted to provide the public with a platform where they can report any external leaks they come across, by taking a picture of it and sending it to their water company. The project’s aim is to build a machine learning system that can detect and analyse customer pictures and alert SEW’s maintenance team of the seriousness of the leak.

We’re now in the process of building that machine learning technology. To do it, we first needed to make sure the machine learning can identify water. It may seem simple, but if there is no water, there is no leak. We have inputted huge volumes of images of water into the system to allow the machine learning to teach itself to identify the liquid. Once the technology is able to do that, we can then teach the ML to assign a severity scoring against the leak itself, using a similar method. This will then enable the machine learning to alert South East Water’s maintenance team to fix the problem.

… And IT underpins tech

The successes so far couldn’t have been achieved without a modernised IT system that underpins the machine learning technology.

SEW’s five-year strategic plan included migrating its IT systems to the cloud to improve operations and help drive innovations, which ultimately allowed the business to utilise machine learning. 

Water companies are becoming increasingly creative and innovative with their approach to technology and it’s only a matter of time until we see technologies like artificial intelligence and machine learning be universally commonplace in the maintenance of our pipe networks. 

Alison King is account manager for public sector and government at CTS