The observability and optimal monitoring of Urban Drainage Systems

Urban drainage system (UDS) overflows are a major driver of urban waterway pollution, largely caused by suboptimal use of network conveyance capacity during wet weather. Real-time control (RTC) mitigates overflows by optimizing the use of UDS hydraulic capacity, making it a critical tool for UDS management—but it relies on comprehensive network monitoring.High costs of procuring, installing, and maintaining monitoring equipment for large-scale UDS restrict real-time monitoring to only a limited number of strategic locations, posing key challenges: selecting appropriate sensor types, determining optimal placement, and extracting maximum actionable information from limited data to support RTC.This project aims to develop a theoretical framework and advanced algorithms to address these challenges: (1) optimally design UDS sensor networks (selecting types, quantities, and locations); (2) maximize reconstruction of UDS state variables (e.g., pipe flows, manhole water levels) from data collected via the optimized network.To achieve these goals, we will for the first time develop a UDS-specific observability concept and analysis framework—extending the well-established control domain observability theory to UDS, which are nonlinear networks with both free-surface and pressurized flows. This framework transforms sensor network design into a constrained optimization problem: we select sensor numbers, types, and locations to maximize UDS observability while adhering to resource limits (e.g., cost, device availability). We then construct a hybrid observer by integrating this framework with deep learning: it estimates full UDS states for fully observable systems or critical state subsets for partially observable ones, ensuring robust state reconstruction even with limited data. Finally, we validate these theories/methods via simulations on two benchmark UDS(Astlingen and Chaohu) and a real-life case study in Hong Kong (Shau Kei Wan), in collaboration with local drainage authorities.This project delivers significant value: it addresses a critical gap in the urban water sector, fills a void in UDS observability theory, and resolves longstanding theoretical bottlenecks by accounting for UDS nonlinearity and network topology—pioneering a new research direction. Its approaches also hold transferable value for other engineering systems facing similar nonlinear and topological challenges.The PI and Co-I, with strong expertise in UDS modelling and control, are well-equipped to lead the project to success. The PI’s collaborative ties with Hong Kong’s Drainage Services Department facilitate the real-world implementation and transfer of the project’s technologies.