In view of brushless DC motor speed control for highly nonlinear characteristics, a self-adaptive DRNN control system was proposed. The controller consisted of a PI feedback controller and a self-adaptive DRNN feed-forward controller. The PI feedback controller used target speed and actual speed error for input to improve the stability of control system. The feed-forward controller was trained by using feedback controller outputs as learning error to improve transient performance of control system. The simulation and experimental results illustrate that the control system can track the target speed, reduce the torque ripple and improve the current waveform under load disturbances, which has strong robustness. 

brushless DC motor, self-adaptive, diagonal recurrent neural network(DRNN), robustness

针对无刷直流电机速度控制存在高度非线性特性,提出了基于自适应DRNN（diagonal recurrent neural network）的“前馈+反馈”控制方法。反馈控制器以目标转速与实际转速的误差为输入量,采用PI控制来提高控制系统的稳定性。前馈控制器采用DRNN,以反馈控制器的输出作为性能误差进行自适应控制,以提高控制系统的瞬态响应性能。仿真和实验结果表明：该控制系统能较好地跟踪目标转速,在突变负载扰动下,能有效地改善相电流波形,降低电机电磁转矩脉动,而且该控制系统具有较强的鲁棒性。

无刷直流电机, 自适应, DRNN, 鲁棒性