Smart Water Magazine
Connecting Waterpeople
Smart Water Magazine
Subscribe to the newsletter with the latest and most relevant news of the global water sector.

You are here

Machine learning tool predicts sewage overflows

  • Machine learning tool predicts sewage overflows
  • It’s not uncommon for water utilities to experience sewage overflows during extremely wet weather, but a new piece of software will help predict them before they occur, potentially avoiding them altogether.

About the entity

Unitywater provides water supply and sewage treatment services to the Moreton Bay, Sunshine Coast and Noosa regions of South East Queensland.
Schneider Electric
· 50


Unitywater and GHD have created a sewer overflow prediction tool using machine learning technology.

Unitywater Manager Asset Knowledge and Performance Ivan Beirne said the tool was a win for the environment and the utility.

“While this tool is still in the testing phase, it has demonstrated the ability to accurately predict wet weather impacts faster than previous hydraulic models used to predict sewage overflows,” he said.

“Sewage overflows are priority 1 jobs which must be responded to within one hour and can pull our field crews away from other planned work. If we can predict sewage overflows and manage our network accordingly, we can avoid diverting crews to these jobs and this is really beneficial for us.”

Mr Beirne said the tool used radar precipitation echo images and harnessed the relationship between rainfall patterns over time with sewer overflow events.

He said it consisted of two machine learning components, a radar echo ‘nowcasting’ model and a sewer overflow prediction model.

The ‘nowcasting’ model produces short term (0-6 hour) rainfall radar image forecasts using the previous two hours of rainfall obtained from the Bureau of Meteorology (BOM) and the overflow model then predicts the likelihood of a sewage pump station overflow.

“Once used operationally, the model will feed live rainfall radar data from BOM to produce overflow risk scores,” Mr Beirne said.

“We can then determine those pump stations most at risk and take preventative action.”

GHD Digital Solutions Lead (Water and Innovation) and Director Digital Labs Saskya Hunter said: “Working on this project with Unitywater has been extremely valuable, both in terms of the digital nature of the solution that has been developed and the opportunity to explore new ways of working through AquaLAB, our co-creation program for the water industry.

“We are excited by the promise that the prediction tool shows and the broader benefits that it may provide to the environment and the water industry. 

“GHD Digital is working hard to make co-creation a reality through our AquaLAB initiative and we are pleased that Unitywater has been prepared to come on the journey with us for our pilot project.”