A dynamic knowledge graph approach to creating interoperability in smart cities: Selected case studies towards holistic flood impact assessments and district heating operations

Abstract

Modern cities leverage information and communication technology to enhance residents' quality of life, thereby evolving into increasingly ‘smart’ entities. Advancing digitalisation generates extensive data with the potential to address pressing environmental and social challenges, boost innovation, and provide decision support, from strategic planning to day-to-day operations. Yet, the seamless integration of ever-increasing amounts of information in heterogeneous formats and types remains challenging, leading to a landscape of fragmented solutions and data silos. This thesis proposes a solution to these problems based on a dynamic knowledge graph and demonstrates its effectiveness using two interdisciplinary case studies - the assessment of potential flood impacts with regard to the population and built environment at risk and the resource-optimised operation of a district heating network. Specifically, a dynamic knowledge graph approach is investigated to align the representation of data and models using Semantic Web technologies, enabling the connection of siloed data sources and overcoming limited automation opportunities in non-interoperable tools and smart city applications. New domain ontologies are developed to capture relevant concepts and their dependencies, and to link related information. Unlike conventional knowledge graphs, the proposed approach includes semantic descriptions of software capabilities and embeds corresponding agents as an integral part of the graph to carry out computations, rendering it inherently dynamic. The semantically connected ecosystem of knowledge, data, and computational capabilities supports graph-native provenance and dependency tracking, which ensures that instantiated updates are automatically cascaded to provide dynamic up-to-date insights at all times. A set of linked software agents is developed to continually instantiate publicly available near real-time data for a more holistic perspective on flood risk in the UK. The implications of newly raised flood warnings are directly assessed in terms of the number of people and buildings as well as total property value at risk, enabling continuously updated impact assessments as flood hazards evolve. In a second example, the potential for cross-domain automation is highlighted by integrating airborne emission dispersion modelling with the dynamic control of a district heating network. An agent-based implementation is developed to forecast the anticipated heat demand, minimise associated total generation cost, and couple it with dispersion modelling of corresponding emissions to provide insights into air quality implications of various heat sourcing strategies. The effectiveness of the approach is demonstrated based on actual historical operations data from an existing heating network of a midsize town in Germany, identifying reduction potentials of around 20% in operating costs and 40% in CO2 emissions compared to baseline data.