基于BP神经网络的开关磁阻电动机转矩脉动控制

Torque-Ripple Minimization in Switched Reluctance Motor Based on BP Neural Network

针对平均转矩控制的开关磁阻电动机(SRM)转矩脉动大的问题,提出了一种瞬时转矩控制的方案。首先在实测SRM静态转矩特性的基础上,基于Levenberg-Marquardt算法的BP神经网络建立了SRM转矩逆模型。该模型网络规模小,便于实时控制。然后基于转矩分配函数的瞬时转矩控制,通过优化电流波形,实现了减小转矩脉动。仿真结果验证了该控制方法的有效性。

To overcome the disadvantage of higher torque ripple in switched reluctance motor （SRM） based on average torque control technique, one method of instantaneous torque control is presented. After static torque characteristic of SRM having been measured, the torque reverse model was developed based on BP neural network of Levenberg-Marquardt arithmetic. The network scale of this torque reverse model is small and is convenient to real-time control. The torque ripple minimization can be achieved by optimum profiliing of the phase currents based on instantaneous torque control of torque sharing function. Simulation results verify the feasibility of this torque ripple minimization technique.