摘要
针对传统PID控制方法不能对电机在工况状态变化时做出快速反应的问题,结合模糊控制和神经网络的特点提出一种智能控制方法。依据模糊神经网络算法组成新的速度控制器代替传统的PID速度控制器。通过RBF神经网络辨识器给出永磁同步电机的Jacobian信息,传递给模糊神经网络控制器,以此解决算法中转速对网络输出的偏导项无法计算的问题。通过自构式反馈来修正网络的拓扑结构,确定模糊神经网络中隐含层的神经元个数,避免因隐含层神经元个数设定不当引起欠拟合或过拟合。仿真和实验的结果表明,电机在启动时能够快速平稳地达到给定转速,超调量和稳态误差小,转矩脉动小、响应迅速。突加负载时速度变化量小且能快速回归平稳运行,突变转速时能快速稳定在变化后的给定转速。
In view of the problem that the traditional PID control method cannot respond quickly to the the change of the motor’s working condition,an intelligent control method was proposed combining the characteristics of fuzzy control and neural network.According to the fuzzy neural network algorithm,a new speed controller was formed to replace the traditional PID speed controller.The Jacobian information of the permanent magnet synchronous motor was given by the RBF neural network identifier and transferred to the fuzzy neural network controller,so as to solve the problem that the partial derivative of the speed to the network output in the algorithm cannot be calculated.The topology and the number of neurons in the hidden layer of the network was corrected through self-constructed feedback,to avoid under-fitting or over-fitting caused by improper setting of the number of neurons in the hidden layer.The simulation and experimental results show that the motor can reach the given speed quickly and smoothly when starting,with small overshoot and steady-state error,small torque ripple and fast response.The speed change is small and it can quickly return to smooth operation when a sudden load is applied,and will quickly stabilize at the given speed after the change when the speed is suddenly changed.
作者
康尔良
蔡松昌
KANG Er-liang;CAI Song-chang(Engineering Technology Innovation Center of High Efficiency Direct-Drive System in Universities in Heilongjiang,Harbin University of Science and Technology,Harbin 150080,China)
出处
《电机与控制学报》
EI
CSCD
北大核心
2023年第3期92-101,共10页
Electric Machines and Control
基金
国家科技主力经济2020(Q2020YFF0402198)
黑龙江省科技攻关资助项目(GC04A517)。
关键词
永磁同步电机
矢量控制
速度控制
动态控制
模糊神经网络
比例积分微分控制
permanent magnet synchronous motor
vector control
speed control
dynamic control
fuzzy neural networks
proportional integral differential control
