{"id":81762,"date":"2025-09-16T14:51:28","date_gmt":"2025-09-16T12:51:28","guid":{"rendered":"https:\/\/www.kwrwater.nl\/?p=81762"},"modified":"2025-09-26T12:04:50","modified_gmt":"2025-09-26T10:04:50","slug":"highlights-from-the-21st-international-ccwi-conference","status":"publish","type":"post","link":"https:\/\/www.kwrwater.nl\/en\/actueel\/highlights-from-the-21st-international-ccwi-conference\/","title":{"rendered":"Highlights from the 21st International CCWI Conference"},"content":{"rendered":"
\n

The CCWI conference is a must for our researchers at KWR, and even more for our Water-Futures PhD students. This year we got to the 21st edition, which was held at the University of Sheffield, UK. Like every edition, this conference provides attendees with a thorough overview of the latest developments in the water distribution industry, covering diverse sessions ranging from design and modelling of pipe networks to digital transformation and water quality.<\/p>\n<\/div>\n

In this post, we share key takeaways from our Hydroinformatics team at CCWI25 and how they link to our ongoing research. The nine works featuring KWR researchers from the Hydroinformatics and Water Infrastructure Teams are also listed at the end, and if you are curious about previous editions, check out our coverage from CCWI23<\/a> and WDSA\/CCWI24<\/a>.<\/p>\n

Building the Foundation for reproducible Water Research<\/h2>\n

One presentation that particularly resonated was our Water-Futures colleague Andr\u00e9 (Bielefeld University), titled: “The WaterBenchmarkHub: A Platform for Benchmarks in Water Distribution Networks.” Andr\u00e9’s observation about the fragmented state of benchmark resources (network models, data, and metadata at large) in our field is extremely correct.<\/p>\n

The WaterBenchmarkHub is an open-source project that addresses this critical gap. It is a space where researchers and practitioners in urban water distribution systems can look up open-source data, download pipe network models, and inspect the documentation to use them. This isn’t just a technical convenience (you can easily install everything with a simple pip install command or rely on the online UI); it’s a fundamental step toward strengthening the scientific foundation of our discipline. Reproducibility and accountability are at the core of science, and Andr\u00e9’s effort represents a noble and invaluable contribution to our community in a time when scientists and researchers are being increasingly scrutinised.<\/p>\n

Be sure to check out the WaterBenchmarkHub.<\/a> Soon, Dennis and the Water-Futures team will publish a new initiative: the Battle of the Water Futures. This new competition in the \u201cBattle series\u201d from the WDSA\/CCWI conferences follows the previous, highly successful Battle of the Water Demand Forecasting (which will also be published on the Hub) and focuses on the challenge of staged design under deep uncertainty, exactly the mission of the Water-Futures ERC Synergy grant!<\/p>\n

Sensor Placement Science vs Practice<\/h2>\n

The CCWI conference also highlighted the importance of balancing theoretical, optimized solutions with real-world practicalities in water distribution network design. A presentation by Fabio Belotti<\/strong>, from Zurich’s water supply company (WVZ), demonstrated this by comparing an optimal water quality sensor placement plan with a practical one. The case study, focused on Zurich’s Glattzone pressure zone, showed that while a theoretical solution for sensor placement is a great starting point, real-world constraints often necessitate adjustments.<\/p>\n

The researchers used sophisticated methods to determine the optimal placement of 21 sensors to minimize the impact of potential contamination events. However, several factors prevented the Swiss water company from implementing the theoretical plan directly. Factors like limited space between pipes, difficult maintenance access, and the need for an electrical cabinet at street level led to a practical solution that deviated from the theoretical one.<\/p>\n

Despite the adjustments, the Zurich case study found the practical solution’s performance was not significantly worse than the theoretical one. This case study underscores that while advanced tools are a great starting point, the final decision on sensor placement must integrate academic theory with the on-the-ground realities faced by water professionals.<\/p>\n

Trustworthy Large Language Models for Urban Water Systems<\/h2>\n

Another presentation that sparked our interest was the one given by Grigorios Kyritsakas (TU Delft), who introduced the work by Taormina et al. on DRACO: a Drinking water and wAstewater COgnitive assistant. In this work, the researchers presented an LLM benchmark dataset for the water sector\u2014covering domain-specific texts, reasoning, and coding tasks for water networks and treatment systems\u2014and evaluated both open-weight and proprietary models on these tasks. Results showed that open models can already match, and sometimes surpass, commercial systems in water-related retrieval tasks, which is an encouraging sign for utilities interested in secure and customizable AI deployments. The presentation concluded with an open call to the community to contribute new questions and case studies to the benchmark, offering a valuable opportunity to build more reliable and trustworthy models for the sector. You can find more information about the DRACO project here<\/a>. At KWR, we are also experimenting with LLMs in practice: from chatbot development and future scenarios generation to code and literature review support. These experiences, discussed in our blog post Approach AI with an open mind<\/a>, underline both the potential and the limitations of generative AI (in general) in the water sector.<\/p>\n

Further Reading<\/h3>\n