无轴承异步电机BP神经网络PID控制

The PID Control for a Bearingless Induction Motor Based on BP Neural Network

针对无轴承异步电机(BIM)传统比例积分微分(PID)控制中存在最佳调节参数难以获得的问题,提出了一种基于反向传播神经网络(BPNN)PID控制新策略。在建立BIM数学模型的基础上,建立3层BPNN PID控制器,其控制算法综合了BPNN和增量式PID两部分,将转速、位移的给定值、测量值及两者的偏差输入BPNN,通过设定性能指标函数,按照梯度下降法修正网络的权值,实现对PID控制器参数的在线调整。仿真和实验结果均表明,该控制策略不仅能快速实现转子的稳定悬浮,而且启动超调小,突加负载后转子只发生微小抖动,相比于传统的PID控制系统具有更好的动、静态性能。

In order to solve the problem that the optimal adjustment parameters are difficult to obtain in traditional proportional integral differential（PID） control for a bearingless induction motor（BIM）,a new PID control strategy ba- sed on back-propagation neural network（BPNN） is proposed.Firstly, based on the establishing of the mathematical mo- del of the BIM, a three layer BPNN PID controller is established.The control algorithm of controller combines BPNN and incremental PID.Then, input the given value, measured value and the deviation between them of speed and dis- placement to BPNN.By setting the performance index function,the weights of the network are corrected according to the gradient descent method, and the online tuning of PID parameters are realized.Finally, simulation and experimental results are shown that the control strategy can quickly achieve the stability of the rotor suspension and small start overshoot, as well as sudden load after the rotor only a small jitter.The control system has a good dynamic and static performance when compared with the traditional PID.