Small-Signal Stability and Black-Box Modeling of Large-Scale Power-Electronic Converter Rich Power Systems

Modern electric power systems are undergoing a transformation driven by the integration of power electronics converters (PECs), such as inverter-based resources (IBRs) and power electronics interfaced loads, aimed at promoting sustainability and addressing global energy challenges. However, this rapid proliferation of PECs, leading to large-scale PEC-rich power systems, introduces significant challenges, including small-signal stability analysis and the development of accurate, computationally efficient transient simulation models.

The first major challenge is the development of a practical and computationally efficient small-signal stability approach for large-scale PEC-rich power systems. To address this, a partition-based method utilizing the nodal admittance matrix (NAM)-based stability criterion has been proposed. This technique divides the large-scale PEC-rich systems into manageable sub-areas and their interconnected networks, facilitating the assessment of small-signal stability at both sub-system and interconnection levels. Additionally, an issue related to the applicability of the NAM-based stability criterion in power systems with pure inductance models is identified. The dissertation not only highlights this problem through practical examples and mathematical demonstrations but also provides solutions to ensure the accuracy of stability analysis.

The second major challenge is the development of precise and computationally efficient models for transient simulations in large-scale PEC-rich systems. To this end, the study explores the utilization of black-box modeling as a promising approach. A comparative study method is introduced to evaluate the performance of three popular nonlinear black-box modeling approaches for modeling the transient behaviors of PECs. Based on the comparative study's findings, three deep learning (DL)-based black-box models (BBMs) are developed for three-phase inverters. The dissertation delves into the model architecture, data acquisition, and data preprocessing necessary to create accurate and reliable models. A switching model and another hardware-in-loop model of three-phase inverters are constructed to generate training and testing data. The performance of the developed DL-based BBMs is evaluated under a variety of transient conditions. Furthermore, the proposed DL-based BBM modeling approach has been extended to three-phase inverters under unbalanced grid conditions.