<![CDATA[Objective: Inverter-dominated microgrids exhibit low rotational inertia and fast electromagnetic dynamics, making transient stability assessment significantly more challenging than in synchronous machine dominated systems. This paper proposes a physics informed deep learning (PIDL) framework to estimate post disturbance stability status and critical clearing time (CCT) directly from short window dy namic trajectories while embedding nonlinear inverter dynamics into model train ing. A total of 1200 disturbance scenarios were generated from reduced order time domain simulations of droop controlled inverter microgrids with randomized virtual inertia, droop damping, fault severity, clearing time, and stochastic renewable fluctuations. The proposed architecture combines a bidirectional temporal encoder with dual output heads and physics residual regularization, followed by two stage optimization (Adam and L-BFGS). On the held out test set, the model achieved 95.56% accuracy, 97.40% precision, 97.40% recall, and 97.40% F1-score for transient stability classification, with CCT error of 29.67 ms MAE and 45.55 ms RMSE. Inference speed reached 1.57 ms per sample, outperforming direct numeri cal simulation (8.37 ms per sample), and robustness testing under ±20% parameter scaling maintained 95.56% accuracy. These results indicate that integrating phys ical constraints with deep learning yields a practical and computationally efficient tool for real-time transient stability monitoring in inverter dominated microgrids.]]>
Copyrights © 2026
