Data-Driven Identification of Key Parameters for Grid-Forming Devices Affecting Grid Voltage and Frequency Stability Boundaries

Abstract

With the large-scale integration of renewable power plants, modern power systems are characterized by reduced system inertia and deteriorated voltage and frequency stability, which poses substantial challenges to conventional grid-following control strategies. Grid-forming devices (GFDs), with the capability to autonomously establish voltage and frequency, have emerged as a pivotal solution to enhance the resilience of low-inertia power systems. This paper proposes a data-driven framework to identify critical control parameters that govern voltage and frequency stability boundaries in power systems with high penetration of GFDs. By integrating multi-source operational data and electromagnetic transient simulation data, typical and extreme operating scenarios are extracted using the Local Outlier Factor (LOF) and K-Medoids clustering algorithms. A full-order small-signal model is then established to conduct sensitivity analysis, quantifying the impacts of droop coefficients, virtual inertia, and reactive power control gains on system transient performance. Furthermore, quantitative indicator-based simulation analysis method is adopted to mine dominant dynamic equations and identify key parameters under noisy measurement conditions, thereby improving model interpretability and robustness. Validation results based on a real-world model of Jilin Provincial Power Grid demonstrate that the proposed method can effectively capture transient overvoltage and frequency deviations caused by severe disturbances such as HVDC blockage, enabling accurate characterization of stability envelopes. The findings provide a theoretical basis and technical approach for optimal parameter tuning and real-time stability assessment of GFDs, supporting the intelligent operation of power-electronics-dominated power systems.