永磁同步电机的改进对角递归神经网络PI控制策略  

Improved diagonal recursion neural network and PI control of permanent magnet synchronous motor

作　　者：彭熙伟[1] 高瀚林 PENG Xi-wei;GAO Han-lin(School of Automation, Beijing Institute of Technology, Beijing 100081,China;China Ship Development and Design Center,Wuhan 430060,China)

机构地区：[1]北京理工大学自动化学院,北京100081 [2]中国舰船研究设计中心,武汉430060

出　　处：《电机与控制学报》2019年第4期126-132,共7页Electric Machines and Control

基　　金：国家自然科学基金(61304026)

摘　　要：针对采用传统PI控制器的永磁同步电机交流伺服系统无法兼顾良好的速度响应性能和抗干扰能力的问题,提出一种将对角递归神经网络(DRNN)与PI控制相结合的控制算法,并引入学习率动态调整的思想对算法进行改进,解决固定学习率DRNN算法无法兼顾系统稳定性和较快学习速率的问题。建立永磁同步电机的仿真实验模型,并对传统PI控制器、固定学习率以及学习率可动态调整的DRNN-PI控制器的实验效果进行综合对比与分析,验证了采用改进后控制器的永磁同步电机交流伺服系统能够实现速度曲线无超调且不受负载转矩突变影响的良好控制效果。For the problem that the permanent magnet synchronous motor AC servo system with traditional PI controller cannot strike a balance between good response performance and strong robustness, a control algorithm combining diagonal recurrent neural network ( DRNN) and PI control was proposed. In addi-tion ,the idea of dynamic adjustment of the learning rate was introduced to improve the algorithm, which solves the problem that the DRNN algorithm cannot strike a balance between system stability and fast learning rate. The simulation model was constructed, and the comparison experiment between PI control-ler, DRNN-PI controller with fixed learning rate and DRNN-PI controller with variable learning rate was carried out. The experiment shows that the AC servo system using DRNN-PI controller with variable learning rate has a good speed performance. There is no overshoot in the speed curve, and the speed is not affected by load fluctuation.

关 键 词：伺服系统 永磁同步电机 神经网络 比例积分控制 计算机仿真 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]