The secure operation of a hybrid AC/DC grid requires not only a stable/unstable decision but also the voltage recovery margins of vulnerable buses. This paper presents a cascaded dual-criterion assessment and severity warning framework using a 0.75 p.u./1 s low-voltage-duration check and a 0.8 p.u./10 s recovery check. For trajectories labeled stable by the offline criterion, a multi-output ConvLSTM estimates the normalized margins of 22 monitored 500 kV buses from approximately 0.25 s of post-disturbance measurements comprising 18 electrical variables. These margins are mapped to four stable-state warning levels; the first-stage stability decision assigns unstable cases to the fifth emergency level. Evaluation uses 11,305 simulated CEPRI-VC scenarios and a source-reported random in-distribution split. On this split, ConvLSTM achieves an RMSE of 8.949 × 10−3, an MAE of 5.531 × 10−3, and an R2 of 99.684%, reducing RMSE and MAE by 18.7% and 22.2% relative to ConvGRU. These values quantify stable-state margin regression within the represented operating domain; a separate reference evaluation of the first-stage classifier gives 98.28% accuracy and a 3.05% missed-instability rate. Reference grouped hold-outs, repeated runs, and disturbance tests further characterize the dependence on data grouping and input quality; complete five-level end-to-end accuracy remains to be established. The framework provides bus-wise margin estimates and candidate weakest-bus information without demonstrating generalization to unseen topologies or converter control settings.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: