KWR research on adaptive water infrastructure wins Silver at the 2026 IWA Project Innovation Awards

The IWA Project Innovation Awards are presented every two years at the IWA World Water Congress & Exhibition. They recognise and promote excellence and innovation in water management, research and technology, with a focus on creative solutions to shared water challenges. 

The research addresses a fundamental challenge for water utilities: infrastructure needs to be planned decades in advance, while many of the factors that determine what will ultimately be needed — including demand, climate, population, technology and regulation — are inherently uncertain. At the same time, infrastructure investments are costly and are planned over time horizons of 50 years or more. Decisions made today can therefore constrain the system for decades. 

Learning how investment decisions can adapt over time 

The research explores how Deep Reinforcement Learning can be used for long-term water infrastructure planning. Rather than asking an optimisation algorithm to identify a single ‘best’ infrastructure plan for an assumed future or a limited set of scenarios, infrastructure planning is formulated as a sequential decision-making problem. 

A reinforcement learning agent interacts with models of the water system, experiences many possible future developments and learns how investment decisions should evolve over time. The result is therefore not a single fixed infrastructure plan, but a strategy for adapting the network as the future unfolds. 

A key innovation is that the agent learns the adaptation strategy itself. The approach is designed to learn which early investments are useful across several possible futures, which decisions can safely be postponed, and when changing future conditions should lead to different infrastructure choices. 

The method was first demonstrated using established water distribution network benchmarks. In these tests, the agent was able to autonomously learn flexible strategies, recognise futures that required similar decisions, and remain effective as the number of scenarios and planning stages increased. 

Application to a real-world planning challenge with Vitens 

A major next step was the application of the methodology to a real utility planning problem in collaboration with Vitens. Vitens was already moving towards adaptive planning: a strategy-driven approach rather than a single fixed ‘2050 blueprint’. This aligned closely with the research approach, which focuses on resilience and adaptation pathways. 

The aim of the research is to provide utilities with something more useful than a single optimal answer based on assumptions about the future: a way of learning what to do now, what to keep open, and how to adapt as the future reveals itself. 

Part of Water Futures 

The research forms part of Lydia Tsiami’s PhD, supervised by Prof. Dragan Savic and Prof. Christos Makropoulos, and is part of Water Futures. This 10 million ERC Synergy Grant explores how water systems can be designed and managed under deep uncertainty. The real-world application was developed in collaboration with Vitens. 

The International Water Association selected AI Enabled Strategies for Future Proof Water Systems as one of four winning projects in the Breakthroughs in Research and Development category. Each category has one Gold, one Silver and two Bronze winners. According to IWA, the project was selected following a highly competitive judging process in which an international expert panel ranked it among the strongest submissions received this year.