Virtual sensing is one of the cutting-edge applications of artificial intelligence (AI) with the greatest potential for impacting on the field of water cycle management. Virtual sensors provide information on physical parameters that are difficult and expensive to monitor by replacing hardware sensors with AI model-based software sensors trained with historical data. In technical terms, a virtual sensor determines correlations between the measured parameter and other parameters of interest, providing an indirect value for monitoring the target parameter.
As a water technology centre, Cetaqua works to apply AI in the different processes in the complete water cycle, thus providing viable, practical and innovative solutions that continue ensuring optimal water quality and the monitoring of relevant parameters.
Trihalomethanes (THMs) are a mandatory control parameter regulated by Royal Decree 3/2023. Thus, an important measure to control their presence and continue ensuring these components do not appear in the short term is to analyse the potential for THM formation at the plant output.
In the digital department, working together with the critical infrastructure and resilience management department, we developed an AI model based on supervised learning that acts as a virtual sensor and is capable predicting this formation potential 24 hours in advance. It was rolled out in the cloud in a scalable environment that permitted real-time data viewing.
The developed model, based on virtual sensing techniques, uses operational data from the plant’s information systems and the sensors installed by Cetaqua. All this provides an automatic prediction of the potential for THM formation with a periodicity that can be highly customised and adapted to the plant’s needs. In doing so, it replaces a manual measurement process that was expensive and impractical for plant personnel. The results of the predictions are logged in a database and can be viewed in a data viewing environment for plant operators.
Work is currently being done on a new version to improve the results of the current model and to develop prediction models of the potential for THM formation at different points and processes in the water cycle, expanding the availability of sensor information.
Authors:
Laura Bonastre, Sergi Baena and Miquel Tomé.
Virtual sensing is one of the AI applications with the greatest potential for impact on the management of the water cycle. In this article we talk about how this tool can help optimise water use and improve its management.