Melbourne Water is leveraging cutting-edge technology to predict recycled water quality up to two days in advance with an impressive 75 percent accuracy rate. This innovation is a significant step towards providing timely warnings to farmers, businesses, and households relying on Class A recycled water for non-drinking purposes, ensuring water quality meets the required standards.
The predictive analytics initiative became essential as Melbourne Water needed to anticipate turbidity levels (water clarity), which could negatively impact water production. If water quality falls below the desired level, it becomes unsuitable for processing in their tertiary treatment plant, necessitating customers to seek alternative water sources.
Melbourne Water had already invested in IoT sensors and an in-house virtual model to gain real-time insights into treatment plant discharges. The addition of a new buoy with advanced technology at Boags Rocks further enhanced their capabilities. This buoy captures live data on water quality, contributing to the calibration of a sophisticated 3D hydrodynamic model of the wastewater discharge environment.
With around 700 sensors and meters in their water treatment plants, Melbourne Water transitioned from real-time data capture to predictive analytics. They partnered with AWS partner Arq Group to establish an organization-wide platform for a unified view of all data sources, using technologies like Snowflake and AWS services. This platform includes a data lake and analytics platform called the unified data store, enabling efficient data ingestion, transformation, and delivery to various reporting, analytics, and AI/ML tools.
The digital twin, a virtual representation of Melbourne Water’s recycled water production, was developed using AWS TwinMaker. This digital twin relies on AWS IoT SiteWise to collect and analyze real-time IoT sensor data, laboratory data, and weather data. Amazon SageMaker is employed to build machine learning models that predict the impact of various factors on water quality.
The outcome of this project is remarkable, as Melbourne Water can now predict water quality conditions two days ahead with 75 percent accuracy. This enables them to detect and address water quality issues promptly and plan for resuming water production when conditions support it.
This initiative aligns with Melbourne Water’s broader digital transformation roadmap, emphasizing the convergence of IT, OT, and IoT capabilities across their operations. The next phase involves developing a digital twin project using real-time video analytics for drones.
Overall, Melbourne Water’s commitment to leveraging technology to enhance water quality prediction demonstrates their dedication to delivering reliable and high-quality services to their customers while advancing the efficiency of their operations.
