Computational Risk Analysis for Digitizing Sustainable Urban Water Supply Systems

Abstract

The movement of global urbanization is severely challenging many aspects of our modem lives, especially after the 1960s. To address these spawned problems, in 2015, the United Nations published a series of Sustainable Development Goals (SDGs ), including 17 topics as general visions for a higher level of human wellbeing. The implementation of these goals heavily relies on the urban infrastructures which have unfortunately become obsolete and deprecated due to long-time usage. Urban Water Supply Systems especially need the most urgent attention. The motivations of this thesis stem from practical demands in the digitization process of Norwegian urban water supply systems. These systems aim to bring sufficient, safe, and sustainable water resources for urban activities. However, risks come from many aspects. The potential dangers in the complex supply process become increasingly difficult to evaluate and predict. This thesis aims to study and develop information knowledge and technologies for systematic and sustainable risk detection and control in the context of digitizing the aging urban water supply systems. The first research question is to build a digitized urban water supply system architecture concerning sustainable development goals. This thesis proposes a novel 5-layer architecture based on Cyber-Physical Systems to integrate a data-driven framework across the water supply process and realize the transformation between data, information, and knowledge domains. Additionally, a loop working mechanism from sustainable objective settings to concrete water supply indicators’ control and feedback information flow is constructed to reflect the relationship among new architecture, goals, and system implementation. Furthermore, we construct a computational risk evaluation model considering influential indicators, system objectives, and constraints in the urban water supply. In this thesis, a systematic analysis was conducted to build different quantity, quality and sustainability indices with respect to domain and empirical knowledge and data collection feasibility. Moreover, correlation analysis methods are employed to formulate a tensor-based risk model. Besides, the thesis aims to support early warnings in the water supply system by introducing a data-driven risk analysis and prediction scheme. This scheme covers the whole process from raw data collection to final output. For risk prediction, this work investigates the current historical records and the data quality constraints. In addition, the spatial-temporal, frequency features of risks, and collaborative analysis methods are taken into account for a better prediction performance. Finally, to verify the proposed architecture and methods, this thesis looks into several industrial urban water supply systems in Norwegian municipalities, including Oslo, Bergen, Strjllmmen, and Alesund and experiment the proposed framework and methods on the real data and perform analysis thoroughly, in a close collaboration with a water engineering research group led by the Co-supervisor Razak Seidu. A practical prototype platform is implemented and deployed as the preliminary service in the local Alesund city management.