Wavelet neural network based reduced-ripple DITC of switched reluctance motors in electric vehicles.

Switched Reluctance Motors (SRMs) are potential candidates for high-performance and cost-effective electric drives owing to their simple structure, robustness, fault-tolerant control, high-reliability, and high-efficiency. However, their highly nonlinear magnetic characteristics and doubly salient structure inherently generate considerable torque ripples, which limit their widespread adoption across various industrial applications. This paper introduces an enhanced Direct Instantaneous Torque Control (DITC) scheme augmented with a Wavelet Neural Network (WNN) to minimize torque ripples. The proposed WNN is employed to dynamically compensate for the torque error, hence adjusting the reference torque signal and providing an optimal input for the hysteresis torque controller. This nonlinear mechanism can significantly mitigate the torque ripples, enhance the quality of torque profiles, and maintain the required average torque. A fixed-gain, 3 layers, 2 Neurons, 7 parameters, single WNN is implemented for compensating the torque error, considering the different operating conditions; the Equilibrium Optimizer (EO) algorithm is utilized to train and optimize the network parameters (weights, translation, and dilation of wavelet functions). The simulation and experimental results confirm the superior performance of the proposed DITC-WNN compared to conventional schemes. The proposed DITC-WNN shows experimental reductions in torque ripples of about 28.8% for heavy loads and 16% for light loads over a wide speed range, confirming its suitability for high-performance and reduced ripple SRM drives.