Transient stability analysis of a VSG grid-connected system based on precise current-limiting switching and AI model identification

Constrained by the overcurrent capacity of semiconductor components, grid-forming (GFM) converters need to switch from constant-voltage control (CVC) mode to current-limiting control (CLC) mode during faults. This gives rise to issues such as current-limiting lockout and mode oscillation, while also reducing the transient stability of virtual synchronous generator (VSG) grid-connected systems. Therefore, to ensure the successful fault recovery of VSG grid-connected systems, it is necessary to simultaneously address two problems: the intelligent identification of current-limiting switching and of transient instability. To this end, this study illustrates the precise mode-switching conditions of VSG grid-connected systems based on phasor diagrams and fully discusses several segmented cases of the maximum deceleration area of VSG grid-connected systems under the influence of mode switching. An intelligent identification method for mode-switching timing based on the XGBoost machine learning algorithm is proposed. Taking easily measurable quantities such as power angle, voltage, and current as inputs, a classification surrogate model capable of quickly judging switching conditions is constructed. Furthermore, the critical conditions for current-limiting lockout and mode oscillation in VSG grid-connected systems are analyzed. Finally, the optimal current angle value for system transient stability is given under the premise of ensuring that no mode-switching problems occur in VSG grid-connected systems. This serves as a practical tuning criterion for engineers to achieve an optimal trade-off among transient stability enhancement, successful fault ride-through, and overcurrent protection of power electronic devices.