Steve Hogg looks at how machine learning and artificial intelligence are best implemented into water services and discusses how our own innovative data analysis tool, FlowBot, can achieve quicker, accurate data analysis.
16 Apr 2020
AMP7 will deliver significant financial and operational challenges for Water Companies. With £51 billion to spend following Ofwat’s Final Determinations and a series of challenging performance commitments, the need to make the most of what we know and how we use it to drive performance improvement will be a critical success factor in the coming years.
Alongside these performance commitments, we are also at a point where the wastewater sector is embracing the ‘big data’ challenge, moving into the cloud and starting to adopt Machine Learning (ML) and Artificial Intelligence (AI) approaches to support activities and decision making. Perhaps as an industry we’re guilty of being ‘data rich, but information poor’ over recent years and therefore missing opportunities to maximise the benefit of the data being collected and the decisions that were made as a result – and at times, this resulted in significant financial penalties for Water Companies. This needs to change if we’re to be successful in AMP7, maximising the value we can get from existing systems, incorporating the new systems and adopting these new approaches as business as usual to drive efficiency and serviceability improvements.
As an industry however, we need to understand where these principles could and should be used, but also understand what these approaches mean and how we use them. Whilst we may still be some time away from AI thinking systems, we are in a position to build approaches with that aspiration in mind, whilst exploiting machine and deep learning opportunities.
Machine Learning algorithms provide significant benefit in interpreting large datasets, and we’ve expanded our sensor network in the last few years with the Event Duration Monitoring (EDM) programme and commitments to extend monitoring within AMP7.
With RPS having an integrated team covering both Operational and Consultancy services we’re in the unique position to think about challenges across the asset lifecycle, from ‘Strategy to Operations’, to help shape change and develop these approaches to benefit the whole asset lifecycle.
Initially RPS developed FlowBot (a bespoke data analysis tool which uses ML) to review flow monitor information as part of our integrated consultancy and survey business, delivering benefit to existing short and long-term monitoring projects. FlowBot’s Machine Learning algorithms are built on 30,000 monitor days of data and has automated the manual review of data and event quality driving an 86% efficiency in the data review process. End users now have faster access to the data and ultimately can make faster, more accurate decisions. This was an important step towards building confidence in automated data.
We are now taking the next step to build on these data quality algorithms and collaborating with Water Companies to develop predictive and pro-active alerting processes linked to live monitoring systems. This includes EDM locations, blockage monitoring and understanding and building algorithms to understand deviation and alert operatives.The WaterNet Pro system based in Azure, delivers this information to the user via a web app and will integrate its rapid geospatial visualisation with dashboard and alert prioritisation once live data is connected to the system. Our WaterNet Pro software is currently in development and is an extension of our existing bespoke Waternet software.
Building rules for EDM and blockage monitors is a logical step in the process, especially where there is enough historic training data for the algorithms. But, due to our network modelling experience we are also able to understand how these rules may evolve or change, and where these changes pose a risk to serviceability or are part of the annual risk profile. Again though, the challenge is maximising benefit. EDMs were in place primarily to enable the assessment of spill performance to meet Water Framework Directive and Storm Overflow Assessment Framework requirements. But overflows are the principal pollution points in the network, so integrating rainfall into the alerting system enables both functions to be managed by the same monitor in the same system. Within WaterNet Pro both functions will be undertaken, enabling real time spill assessment and operational interventions to eliminate spills due to operational conditions.
A key part of this picture for RPS is not just the ability for a system to provide an alert for immediate action, it’s the ability to understand why problems are occurring. The link to wider information, be it assets, performance or wider catchment information is a crucial part of the development of WaterNet Pro - being able to generate alerts on a live stream, and present the operative with the supporting information to understand why the incident is potentially occurring. With time, Machine/Deep Learning algorithms will evolve to consider upstream and downstream changes e.g. an increase in the density in fat/oil/grease generating properties upstream and the risk of blockage formation and serviceability risk further downstream.
During the AMP6 cycle, we spent time moving towards pro-active intervention, and to meet the serviceability challenge of AMP7 we’ll need to continue this work towards source control and solving the root cause of issues as a result of combining and expanding our knowledge and including it in our Machine Learning and Artificial Intelligence processes.
As a modeller myself, I believe hydraulic models are still part of our toolkit, but we need to understand how to drive more value from them, especially as Drainage and Wastewater Management Plans, which are likely to become regulated? In AMP7 have again evolved our thinking around catchment management - whether this is automating the generation of region wide flood / pollution risk through the correct structuring of a model library and Rubyscript tools to generate outputs, or global scenario testing of future pressures on a network impacted by climate change, growth and urban creep scenarios, for example.
While there is a perception of the need for a silver bullet solution that delivers everything, the efficiency challenge facing Water Companies in AMP7 means that all processes need to be as lean as possible and focused on delivering benefit. The use of new technology and innovative approaches will be key to achieving this and it’s crucial that we are in a position to look up, think differently and apply these emerging approaches to deliver the AMP7 serviceability challenge.
For more information please contact Steve Hogg, Water Consultancy Director, RPS, steve.hogg@rpsgroup.com
Steve Hogg
Water Consultancy Director
Get in touchYour contact information:
All fields are mandatory *
RPS is committed to protecting and respecting your privacy. We will only use your personal information to administer your account and to provide the products and services you have requested. We would also like to contact you about our products and services, as well as other content that may be of interest to you.