PROACTIVE DUAL-HORIZON ADAPTIVE CONTROL FOR 10 MW SURFACE-MOUNTED PMSG SYSTEMS UNDER SEVERE LVRT FAULTS AND WIND TURBULENCE

Ultra-large offshore wind energy conversion systems (WECSs) utilizing 10 MW direct-drive Surface-Mounted Permanent Magnet Synchronous Generators (SPMSGs) experience severe dynamic power imbalances during Low-Voltage Ride-Through (LVRT) grid faults under turbulent offshore wind profiles.Deep grid voltage sags severely constrain Grid-Side Converter (GSC) active power transfer, while turbine rotor inertia maintains full mechanical power input, leading to critical DC-link overvoltage surges, converter over-currents, and excessive dynamic braking chopper thermal stress.Standard fixed-gain Proportional-Integral (PI) controllers suffer from integrator windup and sluggish post-fault recovery.To address these challenges, this paper proposes a Dual-Horizon Proactive Reinforcement Learning (RL)-based adaptive control strategy using a continuous-action Actor-Critic architecture.The proposed framework integrates two predictive time scales: a fast electrical horizon (15 ms look-ahead) for grid fault detection (Egrid < 0.85 Enom) and a slow aerodynamic horizon (300 ms look-ahead) for incoming wind power gradient tracking (Δwind/dt).