Machine learning and flow analytics for water network management

The purpose of the project HERAtech: Machine learning for water network improvement (final result at the Digital360 Awards 2020 in the category Machine Learning and Artificial Intelligence) was to limit the drain on resources and actions of countless irrelevant alerts or to avoid hours of work by operational teams called upon to verify hypothetical scenarios that do not require immediate intervention, diverting attention from events worth considering.

HERAtechits technological heart IRA Groupone of the main Italian multi-utility companies active in the environment, water and energy sectors, in partnership with Engineerssupplier of IT solutions, first created a Poc (proof of concept) and then put into full operation a tool applied to this type of installation, with the aim of controlling the entity parameter of the product injection with the aim of guaranteeing the disinfection of the water.

The main elements of the project

In plants, the percentage of disinfected is subject to fluctuations, even significant ones, after sharp fluctuations in the water flow or at the time of the injection itself. These conditions often lead to the creation of “spikes” that trigger alarms because they cause the limits set in the water quality monitoring system to be exceeded. The ordering element of the solution is the ability to identify reliable alarms. In this context, the application of machine learning has made it possible to improve traditional logic and algorithms by extending the observation of more measures. All this would not be possible with traditional platforms and technologies.

Only the use of machine learning algorithms for time series analysis allowed us to bring significant results.

Specifically, the ML model was trained through supervised learning by analyzing alarm return states without operator intervention compared to those for which intervention was necessary.

To make it happen, implemented a streaming analytics platform for real-time data processing. The platform is able to: acquire historical sensor data; determines chlorine values ​​in water automatically and in real time; analyze hundreds of values ​​from sensors. false alarm detection.

Finally it was created a new front end interface that makes it easier to track its levels chlorine and allows you to share them with operators, maintenance, remote control, laboratories analyses.

Thanks to the interface, management of the chlorination process and alarms is now available to individual operators in a transparent manner.

The platform was integrated with the existing SCADA system.

The project took place over a period of approximately 16 months. It started with a Poc of 6 plants in the provinces of Padua and Trieste.

The proven advantages

The project achieved a very high reduction of false positives, on average higher than 85%, yielding clear benefits.

The scalability resulting from the application of this experience to other installation situations and to the set of automation logics governing the various commands managed by SCADA, certainly represents the most attractive and insightful aspect.

The project has led to many benefits, starting with functional ones such as: fewer exits in the field for false alarms; greater operator focus on real alarms; processing hundreds of data in real timethe analysis of which can optimize facility management processes. greater control over alarms by operators thanks to the front.

On the other hand, business benefits are determined, among others, by a increasing the predictive ability of alarm systems to predict situations in facilitieswhich promotes the reduction of management time, cost and resources (less interventions in the field).

The The benefit to citizens consists in optimizing the drinking water service, thanks to an optimal disinfection that guarantees water quality.

Finally, is a solution characterized by maximum scalability that allows it to be extended to all areas covered by HRA, which, through its Remote Control Center, manages over 6,700 units.

The Digital360 Awards 2020 are here sponsors

Leave a Comment