AI-Assisted Virtual Synchronous Machine and Hydrogen Reserve Coordination for Bangladesh Grid Stability Under Rooppur Nuclear Power Plant Integration

This paper presents a Simulink-based dynamic stability framework for the Bangladesh national grid incorporating the 2400 MW Rooppur Nuclear Power Plant (RNPP) alongside a 2024-calibrated generation mix of 20648 MW total dispatch on a 25000 MVA base. A Virtual Synchronous Machine (VSM), a hydrogen fuel-cell reserve (H2/FC), a Mamdani Fuzzy-Logic Grid Stress Index (FGSI), and a Load Frequency Controller (LFC) are co-designed and evaluated under four case studies and seven disturbance scenarios (5002400 MW, including RNPP trip and Adani outage). Baseline comparisons conrm that the proposed VSMH2LFC architecture improves the frequency nadir by up to 0.076 Hz relative to LFC-only, limits peak RoCoF to 0.097 Hz/s (≪ 0.5 Hz/s Grid Code), and keeps the FGSI below the alert level of 0.8 for all disturbances up to 1000 MW. The Mamdani FGSI replaces the previously heuristic scalar with a rule-based fuzzy inference engine, providing a transparent and interpretable real-time stress measure. The proposed architecture oers a replicable blueprint for large-nuclear integration into weakly interconnected grids.