Frequency Adaptive Torque Ripple Suppression for Electrical Drives Using Radial Basis Function Neural Network

Torque ripple suppression in electric drives employing a resonant controller or observer involves harmonic frequency knowledge. The frequency-locked loop technique can be used to achieve frequency adaptive control, however performance suffers when the system contains numerous harmonics. In this paper, a radial basis function neural network (RBFNN) is used to achieve frequency adaptive torque ripple suppression for electric drives. The RBFNN is combined with active disturbance rejection control (ADRC) to provide good rejection properties for both constant and ripple disturbances. Unlike the ADRC system based on RBFNN with offline learning, the suggested method can update the weights vector online, considerably improving system robustness and flexibility. Various experiments are carried out to validate the effectiveness of the proposed method.