Particle Swarm Optimization and Hybrid Reinforced Learning–Based Current Compensation for Brushless Doubly-Fed Induction Generators Under Low Voltage Ride-Through and Transient Conditions
Abstract
Modern smart grids with large-scale wind energy systems need strong fault ride-through capability to maintain stability during voltage fluctuations. Brushless Doubly-Fed Induction Generators (BDFIGs) are increasingly adopted because they eliminate brush gear and require less maintenance. However, their dynamic response during transient and low-voltage ride-through (LVRT) events remains challenging. Existing PSO-only and purely analytical controllers cannot combine global parameter optimization with real-time adaptability, leaving a gap in fast, self-tuning LVRT control. To address this gap, this study proposes a hybrid Reinforcement Learning–Particle Swarm Optimization (RL–PSO) technique for adaptive current compensation in BDFIGs. The controller coordinates reactive current injection using an RL agent trained to minimize voltage variation and total harmonic distortion (THD), while PSO tunes the control-loop reference parameters offline to improve dynamic response. To capture transient behavior during LVRT, the study develops a reformulated mathematical model with re-indexed electrical variables. Compared with PSO-only and analytical controllers, the hybrid RL–PSO controller, implemented in MATLAB/Simulink 2024b, achieves 36% faster voltage recovery, a 14% reduction in THD, and a 62% reduction in peak current overshoot. These findings indicate that the proposed framework can improve renewable-energy-integrated grid reliability.