Supporting Better Public Response by Enhancing Wastewater Monitoring with Machine Learning

Nasrine Tomasi (Mott MacDonald), Chris Park (Mott MacDonald), Nathan Donald (Watercare)

Wastewater overflows can have a significant impact on the environment. During these overflows, sewage is released into nearby water bodies which can lead to several environmental problems and can affect human health. For years, water utilities have invested in improving the management of wastewater overflows by establishing long term monitoring plans of their constructed sewer overflow structures. These monitoring programmes alert operators of overflows occurring with the goal of better prioritising resources on the ground.

This paper outlines a recent trial undertaken to enhance wastewater monitoring datasets with the latest technologies in artificial intelligence (AI) and support operators with the early identification of potential blockages, forecasted overflows and trends in inflow/infiltration at a catchment level. Here are some ways AI and machine learning (ML) have been used for these purposes:
 Wastewater blockage detection: AI can be used to analyse data from sensors that measure sewer depth and velocity in wastewater pipes. By analysing changes in patterns in this data, AI can identify potential signs of blockages and alert operators when an inconsistency in the data is identified. This can help prevent minor or partial blockages from escalatingto overflows.
 Wet weather overflow forecast: AI can be used to analyse weather data inconjunction with sewer depth to create a model that predicts when and where wet weather events are likely to occur based on available rain forecasts. By combining this with the historical performance of the system during wet weather events, AI can forecast when and where overflows are likely to occur. This can help operators prepare for and respond to overflowsmore effectively.
 Inflow infiltration characterisation: Statistical models can be used to compare dry weather data with wet weather profiles to identify sources of inflow and infiltration (such as leaky pipes or illegal connections) and estimate their magnitude. Undertaking this analysis in near-real-time can help operators identify degrading performance and condition of the asset to support the prioritisation of repairs, renewals and maintenance to reduce the impact of inflow and infiltration on the system.

Overall, AI, ML and statistical models can be powerful tools for optimising theoperation and maintenance of wastewater systems and reducing the impact of wastewater on the environment. It can enhance traditional wastewater monitoring to proactively manage network issues before seeing their impact on the receiving environment and communities.