Suppression of Broadband Forced Torsional Vibration of Doubly-Fed Wind Turbine Shaft System Based on RBFNN-MRAC

The transmission chain of wind turbine shaft system exhibits wideband forced torsional vibration in the low-frequency range, which affects the stable operation of the turbine. In this paper, an RBF neural network is incorporated into the forward channel of the model reference adaptive control (MRAC) with input-output, forming the RBF neural network model reference adaptive control (RBFNN-MRAC). The three inputs of RBFNN-MRAC are identified, and the controlled object composite control signal, and adaptive control law are reshaped. This enables adaptive tracking of the gain coefficient, damping ratio, and center frequency of the band-pass filter in active damping control, ensuring that the controlled object tracks the reference model. Simulation results demonstrate that the proposed method effectively suppresses wideband forced torsional vibration, reduces torque fluctuations in the transmission shaft, and achieves smoother and more reliable generator output power.