Abstract
With the development of big data informatics, sensors, embedded monitors, and other cyber devices have been installed as power systems' supervisory control and data collection components. Thus, the cooperation between cyber communication and physical control components has received significant attention as a critical technology to support the efficient operation and development of smart grids. Advanced technologies, such as the Internet of Things (IoT), have produced an exponential volume and variety of data from different parts of smart grids. The data is generally presented in different modalities, including size, format, and sampling frequency, while it may also include multiple modalities, such as graphs, textual records, numerical parameters, etc. This data presents both opportunities and challenges for power system researchers and operators.From the perspective of reducing the data abundance, conventional power system analysis methods such as load forecasting, state estimation, and optical power flow (OPF) study rely on centralised control and limited information with the same modality. These approaches are adequate for the traditional supervisory control and data acquisition (SCADA) system of power networks. However, the increasing penetration of cyber components in modern power systems now generates massive amounts of data from numerous sources, such as smart meters, embedded measurement units, weather stations, social media and other related IOT components. Therefore, this thesis aims to adopt Heterogeneous Data Fusion (HDF) to combine data from various sources with different characteristics and modalities to gain comprehensive insights and make informed decisions for power systems.
Data fusion, known as multi-source information fusion or data integration, is an advantageous technique used to combine data from multiple sources and leverage the diverse set of acquired information to improve the overall accuracy, efficiency, and reliability of cross-discipline systems, such as computer vision and meteorology. Compared to traditional approaches, this technique is crucial in harnessing the potential of multi-source information for power system development.
This thesis has four main objectives and contributions:
(1) To establish the importance of data fusion on power system, it is necessary discriminate and conclude the data fusion theory, algorithms, and frameworks. Referring to the previous work of data fusion on other disciplines, this thesis demonstrates the potential of data fusion and reviews its novel applications in the field of power system research.
(2) Identifying potential features and data correlations among multiple power system components in power systems and rationale the feasibility of data fusion on power system analysis. This thesis shows the unique advantage of data fusion through solving the blind-zone state estimation problem of distribution system which is inapplicable in traditional model-based power system analysis.
(3) Identifying the paucity of previous research on data fusion application on the field of power system cyber security. This thesis deploys the data fusion methodology to enhance the cyber security of power systems and showcase its significant improvement on the accuracy and reliability of the cyber-attack detection and recovery process in power system operations.
(4) Establishing the importance of data fusion on the cross-domain study in the field of power system research. This thesis deploys the data fusion methodology to integrate new data dimensions in different power system analyses to improve the accuracy of the demand forecast in distribution system operations.
In the HDF application, this thesis further deploys the data fusion along with other advanced techniques in extensive data analysis, such as deep neural network, to consist of an improved informatics framework that can handle multiple heterogeneous datasets and significantly improve the reliability of different power systems analysis, such as load forecasting and cyber-attack detection. This thesis proposes heterogeneous data fusion for system operators to support reliable decision-making and energy strategies by integrating massive amounts of information from different domains. For researchers, this thesis fully demonstrates the performance and reliability of heterogeneous data fusion on early detection of potential threats and precise forecasts of complex scenarios by incorporating data from diverse sources. Overall, the proposed HDF in power systems significantly contributes by improving decision-making, optimising operations, enhancing grid resilience, fostering innovation, and promoting energy efficiency. Its impact extends across the cyber and physical networks in power systems and achieves the data incorporation of power systems, society, and environments.
| Date of Award | 27 Mar 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Chenghong Gu (Supervisor) & Furong Li (Supervisor) |