Improving Power System Resilience Against High-Impact-Low-Frequency Events

Abstract

The exponential growth of renewable energy has been observed worldwide, supporting the emergence of the new global energy economy. Strong policies and raised climate goals are driving renewables to new records, pushing the power systems to net zero, and raising the standards higher than ever for secure and efficient operation of the grids. Traditional models and methods start to falter, especially under the situation where the uncertainty of renewables and the possibility of High-Impact-Low-Frequency (HILF) events need to be seriously considered. This dissertation describes a holistic and systematic solution, covering models, methods, algorithms, computation, and tools, with the goal to help the industry prepare for the new operational challenges, while improving the power system resilience under high renewable penetration and HILF events. Considering the current grid infrastructure is heavily inter-coupled with the underlying communication and control system, this work starts with generating the realistic cyber networks using the proposed method to match the graph properties and characteristics of industrial communication systems. To facilitate the research on protecting the grid from a cyber-physical point of view, a high-fidelity cyber physical testbed is developed, and it is used to study the propagation process of adversarial cyber threats. Verified in the proposed testbed, an active routing scheme is developed to prevent the lateral movement of cyber vulnerabilities based on the feedback from physical state estimation system. To better evaluate the weather impact on power systems, especially under the extreme weathers, a fully-decoupled non-linear weather-dependent power flow model is developed to efficiently estimate the conductor impedance for large-scale power systems with geographically-varying weather data. And to handle the computation intensity required to analyze the grid with large renewables and distributed energy resources (DER), a sparse matrix pre-ordering method is proposed and proved to improve the computation efficiency up to 100% for power flow, and up to 40% for sensitivity analysis.