Torque Observer of SRM Based on BP Neural Network Optimized by Bat Algorithm

The traditional torque observation of switched reluctance motor (SRM) has disadvantages such as low accuracy and complex calculations, which may easily lead to poor control influence on minimizing torque ripple. To solve this problem, a new method based on BP neural network optimized by the bat algorithm (BA-BP) to establish the torque observer of SRM is proposed. The bat algorithm (BA) is used to optimize the initial weights and thresholds of BP neural network to achieve a much more accurate non-linear fitting of switched reluctance motor current, rotor position angle to torque, and construct a torque observer to achieve accurate SRM torque calculation. Finally, the direct torque control (DTC) system of switched reluctance motor is built, the designed neural network torque observer is applied to estimate the torque and minimize the torque ripple of the motor. Simulation results show that the proposed torque observer method can calculate the torque with much higher accuracy than the traditional BP neural network torque observer. Compared with DTC based on torque look-up table method or traditional BP neural network torque observer, DTC based on BA-BP neural network torque observer has better effect on minimizing the torque ripple of SRM.