The technology opportunity of water management 

When it comes to embracing digital tools, the water sector has traditionally lagged. However, digital technologies offer significant potential for water companies to improve environmental protection, water quality, customer service, and cost efficiency. Tania Flasck of Sand Technologies explores the opportunities and challenges of digital transformation in the water industry.

Many heavily regulated sectors have embedded digital technology across their business to drive value, particularly in the areas of customer service and engagement and are embracing sharing and opening up data. The water sector still has some way to go based on the results of an Ofwat commissioned Open Data Assessment Report by pwc. The good news is that digital technologies have a very real ability to substantially improve performance, affording significant opportunities for water companies to transform and achieve more.

The multiple benefits of AI

In the water sector, there hasn’t been a common approach to ways of working or adopting digital practices, and there’s little collaboration in an area that would greatly benefit from companies working together. Some progress is being made with the regulatory push for opening up data, and initiatives like Stream - funded by the Ofwat Innovation Fund in collaboration with Northumbrian Water and other partners - are helping to share best practice in how to make data available in a consistent format. To really maximise the benefits of AI and digital technologies, data collection, availability and use needs to go further.

There are many benefits of an AI-enabled company, including:

• Managed extraction to closely match consumption needs, effectively reducing factors of safety so more water stays in the environment for longer
• Predictive understanding of potable water needs and quality
• Reducing leakage to much lower levels using better sensor technology, predictive analytics and organisational work processes that match digital enablement
• Greatly reduce network discharges by maximising existing capacity and identifying best benefit capital improvements
• The effective elimination of wastewater treatment works failure though digital twin technology to optimise performance to prevent failures
• Reduced carbon emissions and embrace circular economy principles by operating assets with much greater understanding, maximising opportunities from waste streams
• Significantly lower capital cost of new assets by adopting generative design techniques, which will also alleviate capacity constraints

The first and often underrated step of adopting AI requires water companies to develop data platforms that can handle the vast amounts of data being received from multiple sources including from the past, present, and into the future. This explosion of data availability can and should enhance decision-making for operational improvements and strategic interventions. However, it necessitates updating both technology and company processes to fully benefit from these advancements.

The first step: AI platform development

The foundation of AI solutions is a robust, scalable Enterprise Data Platform (EDP). The efficacy of AI is reliant on large volumes of high-quality data.

At Sand Technologies, we use a Source, Ingest Enrich, Serve & Engage (SIESE) data ecosystem framework. This is a generally applicable abstraction of the activities associated with developing data products and helps to keep focus on building solid foundations:

Data acquisition and storage is centred around effectively managing diverse data sources, ensuring their coherent integration into a single platform. This methodology is particularly relevant for environments where data comes from multiple origins (e.g. internal systems, third-party sources, sensors and cloud environments). The most effective data-management programs include three key aspects:

• Integration Capability: It’s critical to create seamless connections between different data formats and sources using data pipelines and environments. Good examples of this have been seen with some large water companies that developed an asset health system, clean water network risk visualisation system and pumping station risk viewer via Microsoft Azure. In order to support the applications, they developed a platform to collect, clean, process and store asset-health data from 50 sources, built a network risk visualisation system from 84 data sources, and took additional data from 18,000 sewer level monitors, 5,000 pumping stations and hundreds of sewage treatment works.

An integrated platform enabled the seamless movement of ~20 million data records a day (more than 7 billion records per annum), allowing the water companies to meaningfully use insights from this huge volume of data in applications to drive better business and performance outcomes.

• Data Governance: Measures should be put in place throughout the data lifecycle, from acquisition to processing, to ensure data accuracy, quality and usability. This is a vital step to not only maintain data reliability, but also to ensure its usefulness for analysis and decision-making. Data security is addressed through encryption, access control, and regular monitoring, aiming to prevent unauthorised access and data breaches, which is an in-built feature of a data ecosystem framework.

• Scalability and Flexibility: The number and sources of data have grown exponentially, and that trend is poised to continue. Any data platform must therefore be built with an eye toward the future and growth through scaling. One strategy is to adopt a data mesh (hub and spoke model) to help ensure scalability over time. This enables a decentralised approach for data ownership allowing different groups to own and manage their data repurposing for different use cases. As data volume and variety grow, the mesh-enabled system can expand without overburdening a central point, ensuring sustained performance and adaptability.

Once data is accessible and useable, the next steps are to bring that data into an application allowing decisions to be made that were otherwise hidden or not obvious, or just too complex to process. The typical applications usually start with a visualisation platform that allows different data sources to be viewed together through layered applications and then progress onto reporting functionality through generative AI, other AI and Machine learning applications and digital twins.

System visualisation and understanding of risk

Building a suite of products necessitates a holistic view of the network. Each asset can be displayed on a geospatial map, and telemetry data from different assets can be studied simultaneously:

Building a suite of products necessitates a holistic view of the network. Each asset can be displayed on a geospatial map, and telemetry data from different assets can be studied simultaneously.

A wastewater network, for example, is comprised of interconnected assets. To understand how to improve the network in the long run and react to immediate issues, it is imperative to understand how the network’s assets are connected. In parallel, sewer level monitors installed at scale can allow water companies to combine sewer level monitor data, sewage pumping station data, and sewage treatment works data with data about rainfall, groundwater and soil moisture deficits to better understand the causal relationships across the network. This understanding improves the operational response to immediate issues while improving longer-term capital investment strategies by highlighting priorities.

Adopting a user-centric design approach enables the creation of applications that are not only effective but also intuitive and aesthetically pleasing. Focusing on the end-user experience ensures that visualisation tools are tailored to meet the specific needs of those managing and operating networks and treatment works.

The most important aspect in developing an end-end visualisation of the network is the interplay between the geospatial view and telemetry data. Operational teams have little time to decide on a course of action so providing key information to make the right decision first time is key. A prime example is raising proactive blockage investigation jobs using sewer level monitor data. In the past, users might need about 45 minutes to raise jobs. Following the development of a purpose built data product, the process has been streamlined to the point where users can confirm that high sewer levels are indicative of a blockage, assess the priority of the blockage investigation job required, understand the history of the site and raise a job, all from within the same tool, generally in under one minute.

The ability to study data in detail is essential in the development of effective AI/ML models. It is impossible to build highly accurate models without a deep understanding of the data, especially when dealing with complex problems such as the dynamics of a network. Comprehensive visualisation tools are therefore not only essential for end users, but also essential for data scientists building models.

Generative AI

We can all imagine the demands on a control room operator particularly when an incident occurs. The pressure to make the right decisions quickly using multiple sources of data and applications including spreadsheets is high. Visualisation tools that provide situational awareness can support multiple stakeholders to understand the risks and make timely decisions that are more likely to generate better outcomes. Along with this, generative AI can provide on demand and real time data capture to quickly generate reporting requirements to multiple stakeholders.

Regulatory reporting is another aspect that can be greatly enhanced with both open data and generative AI capabilities. A culture of radical transparency is attainable and would build trust with the public and regulators alike. The tremendous amount of resource requirements to manually review incidents and data as well as provide reports could be dramatically reduced with the right data foundations, platforms and use of AI.

The other exciting application of generative AI is in the ability to rapidly generate preliminary design estimates for capital investment scenario planning and cost estimates to accelerate engineering design stages significantly reducing time and resource requirements. Preliminary designs can be provided in hours and days rather than weeks and months.

AI & ML model development

AI and Machine Learning (ML) Models have a lot of hype associated with them but can support the predictive ability of a system providing a future view of the state of networks. By leveraging advanced machine learning and AI capabilities, these models can rapidly develop with proprietary ML frameworks that can ingest and create models an order of magnitude faster than traditional techniques. This can be a powerful tool to scenario plan and prepare for many different potential outcomes such as forecasting different supply demand scenarios or likely compliance failures at treatment works. Through continual feedback loops to help models learn, confidence factors can be built to the level that automated responses can be added. So a model can predict where a blockage is likely to have occurred in a sewer network and automatically raise a job directly to a sewer cleaning crew to quickly address the issue without the need for human intervention in the job raising process.

Consider healthcare in parts of Africa - the application of AI and ML models are revolutionising the provision of healthcare leapfrogging advancements made in other parts of the world. By reimagining healthcare, models can predict when and where disease outbreaks are likely to occur which allows for proactive intervention. Needed equipment is deployed by drones in near real time to health outposts also positioned using data insight on a population served by foot basis. This allows health care workers to proactively manage outbreaks resulting in positive outcomes to the economy and society as a whole.

Digital twins

Digital twins of clean and wastewater networks enable capabilities otherwise not possible and have contributed significantly to understanding network dynamics. Through the development of a clean water network digital twin, an application was built that increased the productivity of finding and fixing leaks by 30% which in turn led to a 2 ML/d reduction in leaks.

A digital twin is essential to understand the behaviour of connected assets in networks at an enterprise scale. For example, in order to concretely determine if sewage pumping stations influence sewer flooding, the connectivity of the network needs to be made visible through a digital twin (which makes drawing insights at scale feasible). A digital twin differs from geographic information system data in that a digital twin represents which assets are connected and the data that’s assigned to those assets. For example, data from sewer level monitors upstream in the catchment can be used to determine if parts of the catchment suffer from infiltration, while pinpointing the specific sewers affected by infiltration can facilitate remediation. Using this approach, all of the data from all of the assets and monitors in a catchment can be integrated for improved analysis and decision making.

Digital twins can also enable users to foresee site or asset behaviour using data modelling, and they can allow for the testing of various scenarios. This is crucial to understanding and predicting the operational dynamics of treatment facilities and providing a single source of truth.

The final word

We know how to solve many challenges when it comes to water management. It is not a single solution though and requires the human factor throughout to ensure that bias is not perpetuated and outcomes make sense.

The answer lies in working together to understand how the actions of one person, company, sector or government can impact many, then working collaboratively to integrate and use a range of solutions to solve our shared challenges. The amount of data involved can be huge, but we have the means to navigate this complexity and use AI to deliver real impact by bringing people together to support equitable water access and sanitation for all.

Tania Flasck is Director, Market Development (Utilities) at Sand Technologies. She is also Non-Executive Director for British Water and part-time chair of Waterwise.