The full text of this item is not available at this time because the student has placed this item under an embargo for a period of time. The Libraries are not authorized to provide a copy of this work during the embargo period, even for Texas A&M users with NetID.
Data Creation and Decision Making for Improved Power System Security: A Machine Learning Approach
Abstract
In an era of rapid decarbonization and increasing adoption of renewable energy resources, the maintenance of power grid stability and security presents an evolving set of challenges and opportunities. Modern power grids, serving as a critical infrastructure for our society, are in the midst of significant transformations, driven by advancements in digitization and computational technology. Machine learning (ML) has emerged as a powerful tool that, when well-applied, has the potential to revolutionize power grid operations and planning, enhancing reliability, efficiency, and resilience. This dissertation is a comprehensive exploration of this burgeoning intersection of ML and power systems, with a particular focus on how ML can be leveraged to tackle challenges posed by the large-scale integration of renewable energy.
Chapter 1 of the dissertation provides a thorough overview of the functionalities of modern power grids and the transformations occurring due to decarbonization. It emphasizes the increasing importance of power grid digitization and the rise of ML-based solutions to address the emerging challenges. The opportunities and hurdles associated with the application of use-inspired ML in power grid security are also discussed.
Chapter 2 introduces an open-access, multi-scale time-series dataset for ML benchmarking, named PSML. This dataset is synthetically generated from a joint transmission and distribution electric grid model, designed to capture the increasing uncertainties and interactions of grid dynamics. This chapter emphasizes the importance of such a dataset in promoting the development of ML-based solutions and provides benchmark results for critical tasks in power grid operations using this dataset.
Chapter 3 tackles a significant hurdle in power grid research: the limited availability of eventful Phasor Measurement Unit (PMU) data. To mitigate this, Generative Adversarial Networks (GANs) and Neural Ordinary Differential Equations (ODEs) are leveraged to generate synthetic, eventful PMU data, demonstrating the potential of these methods on both a small-scale test system and a large-scale real-world dataset.
Chapter 4 addresses the necessity of identifying critical N-2 contingencies (two simultaneous contingencies) in real-time, ensuring reliable operation even with increasing penetration of intermittent renewable generation. It introduces and validates a physics-informed Graph Neural Network (GNN)-based screening method for these contingencies, applicable in the AC power flow framework.
In sum, this dissertation offers a holistic view of the ways in which ML applications can be deployed to enhance power grid security in the context of the energy transition. It underscores the importance of continual, use-inspired ML research in enabling the safe and effective decarbonization of our energy infrastructure.
Citation
Zheng, Xiangtian (2023). Data Creation and Decision Making for Improved Power System Security: A Machine Learning Approach. Doctoral dissertation, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /200079.