Data-driven sewer management

Tracking and measuring the invisible

Our sewerage systems are hidden almost entirely from the outside world. But that doesn’t mean we can’t measure and monitor them. Today, a huge number of measuring points and sensors in the system help us understand and maintain unbroken visibility of what’s really going on underground.

They supply daily flows of data relating to everything from water levels and pump frequencies, to system energy consumption. But despite this modern approach to monitoring, most of this data still has to be interpreted manually by a trained professional – making it both a time-consuming and error-prone process.

That’s where data science can help. A machine learning model can process and interpret the raw data gathered from sensors, providing sewerage managers with immediate visibility of patterns and correlations. These models can then self-learn over time, using algorithms to understand past events and forecast more accurately for the future. Plus, they make it simple to establish relationships between complex variables and influencing characteristics such as the weather and pumping capacity.

Three sewerage data science applications

There are a huge number of ways you could apply data science to help simplify and improve sewer management. Here’s a look at three capabilities we have successfully developed and deployed so far:

1) Smart alarms and predictive maintenance

By using algorithms to process large data sets gathered from sewer system monitors and external sources, Royal HaskoningDHV Digital has created software that can recognise abnormal situations in the sewer system in real-time and alert managers immediately.

In practice, that means managers can take action before the system or any machinery within it fails. Pumps can be prevented from running without water in them, sensors can be replaced before they fail, and teams can immediately detect the sources of issues threatening system operations.

2) Detecting non-sewage water

In collaboration with the Drents Overijsselse Delta water authority, we recently developed a solution that can recognise abnormal volumes of water in the sewer system – often caused by well-draining, faulty connections, acute leaks or negative overflows.

The algorithm in the tool predicts the correct current flow rate based on historical data and compares it with measured values and precipitation data to detect anomalies. With this, the water authority can see spikes in flow rates immediately, and take action before non-sewage water impacts system efficiency.

3) Real-time station control

We’re currently working with multiple water authorities to develop software that will optimise how pumping stations are controlled based on machine learning. By predicting flow levels using an approach similar to the non-sewer water use case, the solution will help keep supply even, and maintain even flows at treatment plants.

By maintaining more even flow levels, pumps and other machinery will remain useable for longer, and unnecessary flood risk can be effectively reduced. This will also help purification operators optimise their process, because they can anticipate the flow rates they need to manage.

The future

Each one of those data science innovations is already having a significant impact on how sewer systems are managed. But this is just the beginning of what we want to achieve.

We believe that by bringing domain experts, data scientists, and leading technology together, we can completely transform how sewer systems are managed. And with this unique combination of capabilities and expertise, we can help water authorities not only meet today’s biggest challenges, but operate more efficiently than ever before.

If you’re experiencing any of the challenges discussed here, or if you’d like to connect and exchange ideas on the future of water treatment and sewage management, contact us.