A dynamic knowledge graph approach for studying the decarbonisation of power systems

Abstract

This thesis introduces a dynamic knowledge graph framework, providing a reusable, interoperable, and extensible approach for cross-domain analyses in power system studies, facilitating research on decarbonisation. Domain ontologies were designed to conceptualise the power system and used to create a knowledge graph featuring linked data representations spanning from concrete entities to abstract concepts. The knowledge graph framework incorporates relationships between power system assets and their respective administrative regions, structured by the official administrative hierarchy. By employing geospatial queries, this framework enables tasks like estimating the population surrounding power plants, evaluating the quantity and capacity of these plants, and pinpointing demand in different regions. Computational agents were designed to interact with the knowledge graph, performing dynamic operations such as data uploading, updating, retrieval, processing, model construction, simulation, and optimisation. Modularised like Lego bricks, these agents allow customizable assembly for specific tasks. Agents encapsulate mathematical algorithms and models based on their functionalities. For example, some agents wrapped algorithms for power flow (PF) and optimal power flow (OPF) analyses. Another agent incorporates a new optimization algorithm for selecting nuclear Small Modular Reactor (SMR) sites, balancing objectives of minimising transmission losses and population risk. The framework was implemented in the UK power system, populating its data into the knowledge graph and instantiating two transmission grid models. A case study explored clean energy transition in the UK using SMRs, assessing the impact of a CO2 tax on promoting SMR adoption over fossil fuels. Multiple scenarios were analysed, examining different SMR placement policies and renewable power availability assumptions. The study found that deploying SMRs at a balanced distance between high-demand and densely populated areas is the most cost-effective strategy, thereby identifying the optimal SMR sites/areas in the UK and determining the retirement order for fossil-fired generators. Results suggest a carbon tax threshold for incentivising SMR adoption, linked to SMR levelized cost of electricity (LCOE) and minimally affected by renewable power availability. SMR introduction affects energy system performance, costs, and regional energy balances with diverse renewable energy levels, offering policymakers insights for carbon-neutral electricity generation integration.