A Novel Torque Ripple Suppression Scheme of Switched Reluctance Motors Based on Unipolar Sinusoidal Current Excitation

This paper presents a novel control strategy for switched reluctance motors under unipolar sinusoidal excitation to minimize torque ripple at medium and low speeds. This strategy focuses on enhancing both the accuracy of the conversion process and the performance of current tracking. For the conversion scheme, a modified back propagation neural network that considers the magnetic saturation characteristic is proposed. The lightweight network structure of this conversion allows for seamless integration into the unipolar sinusoidal excitation strategy, enabling fast and accurate online calculation of the q-axis current. Additionally, to improve the current tracking capability, an improved predictive current controller is designed, with a Luenberger disturbance observer incorporated to address predictive model mismatch. The stability of the system is proven in the discrete-time domain using the Jury criterion. Experimental results demonstrate that the proposed method can effectively suppress both torque ripple and disturbances, achieving good overall performance.