Dynamic Optimization of Rotor-Side PI Controller Parameters for Doubly-Fed Wind Turbines Based on Improved Recurrent Neural Networks Under Wind Speed Fluctuations

This paper investigates a doubly-fed wind turbine generation system (DFIG) where the rotor-side control parameters have a significant impact on the effectiveness of the DFIG due to the adoption of its inner-loop current and outer-loop power control strategies. Under rated operation, the original DFIG parameter adjustment relies mainly on manual adjustment. In this paper, mathematical models are established through literature research and data search, and neural networks are found to have unique advantages in dynamic automatic parameter tuning. First, a mathematical model of DFIG based on PI controller is established in this paper, and then the improved recurrent neural network is applied to the parameter tuning control of rotor-side PI controller, and an experimental model of DFIG simulation based on the improved recurrent neural network is established in MATLAB/Simulink. By comparing the DFIG models before and after the improvement, the simulation experiments verify that the DFIG system based on the improved recurrent neural network (CLR-DRNN) has significant control advantages under the wind speed fluctuation. The simulation experimental results show that the DFIG system based on the improved recurrent neural network achieves significant improvement in wind energy utilization coefficient, active power, reactive power, response time of rotor speed, overshoot and static error compared with the conventional PI-regulated DFIG system.