Wastewater monitoring has proved a difficult nut to crack for water companies in the face of increased pressure from regulators, government and customers to reduce pollution incidents across the UK. Yorkshire Water, in collaboration with Siemens and The University of Sheffield have developed and trialled an innovative new solution combining artificial intelligence (AI) and internet of things (IoT) to reduce wastewater network blockages and reduce pollution with a data driven approach.
Monitoring waste water is not a new problem and water companies all over the UK have invested millions in a variety of methods for gathering real-time data in a bid to better understand the way the sewer network operates under times of pressure such as heavy rainfall and storms.
Thousands of sensors operate across the wastewater network, providing mountains of data, but the challenge for water companies has been how to interpret that information when what is a ‘normal’ reading differs for each sensor location. Is the sensor in an urban or rural location? Is the topography flat or hilly? Is the sensor providing data from the start or the end of a network?
Each of these factors can create situations where sensors are providing data that appears to indicate a problem to the untrained eye, but it is just the unique conditions for that particular section. For operators the daily impact is an overwhelming number of false alarms in high rain fall that mask real issues and waste time.
In a bid to make better use of its data, Yorkshire Water teamed up with Siemens and the University of Sheffield to develop an innovative new blockage prediction tool – SIWA Blockage Predictor.
A trial of 70 sites across Yorkshire has been positive, with the tool finding nine in 10 potential issues – which is three times more successful than the existing Yorkshire Water pollution prediction processes that relied on statistical methods. The AI also reduced the number of false positive alerts by 50%.
How it works
No one data set of measurement can effectively predict potential problems within the sewer network, but by combining AI with real-time rainfall and sensor data the SIWA Blockage Predictor evaluates the characteristics and performance of the sewer network in real time and predicts problems like a network blockage before they happen, enabling Yorkshire Water to fast-track engineers to inspect and resolve issues.
The tool relies on an AI system to monitor every asset and learn the ‘normal’ behaviour of the network before the artificial neural network learns the way levels change based on the rainfall and the time of day. This can provide a prediction a short time ahead of any problem.
At this point a second part of the AI, a tool called fuzzy logic, compares the prediction with the actual level and considers if this difference is significant for that asset given the amount of rainfall. Up to two weeks of early warning can be given for building issues as well as identifying sites that require urgent action. Notifications alert Yorkshire Water staff and people can be sent to the location for additional checks to be carried out and any maintenance or repairs can be conducted quickly.
The analytics are embedded within a secure web application, enabling remote access on mobile devices or PCs and notifying users in advance of any issues.
Ultimately, the aim of the project is to reduce the number of pollution incidents from the sewer network and it is just one part of Yorkshire Water’s Pollution Incident Reduction Plan 2020-2025, which aims to reduce pollution incidents by 50% by focusing on early intervention.
The pilot has proven that the AI and fuzzy logic can find the needle in the haystack of a building or actual blockage amongst thousands of threshold-based false alarms. During a trial period the University of Sheffield evaluated 38 measuring points over 21,300 days and the AI found 88% of the confirmed issues. As well as identifying potential problems, the AI was able to reduce the number of false positives to 3%, meaning Yorkshire Water was able to attend and fix issues efficiently, without technicians visiting false alarms.