基于改进小波神经网络的新型PMSM速度控制  

Novel PMSM Speed Control Scheme Based on Improved Wavelet Neural Network

作　　者：周雅夫 赵洋 ZHOU Yafu;ZHAO Yang(School of Mechanical Engineering,Dalian University of Technology,Dalian 116024,China)

机构地区：[1]大连理工大学机械工程学院,大连116024

出　　处：《组合机床与自动化加工技术》2024年第10期105-108,114,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家自然科学基金项目(52172382)。

摘　　要：采用传统PI控制策略的永磁同步电机调速系统无法兼顾良好的动态响应性能与抗扰动能力,且在实际应用中参数整定繁琐。为改善以上问题,对速度环设计一种改进型小波神经网络PI控制器,该控制器基于自适应学习速率的梯度下降法并引入惯性项在线更新网络参数,通过权重因子扩大PI参数输出范围,以增强控制器性能;为消除电流环比例参数对速度环的影响,设计无差拍电流预测控制器,进一步提升系统动态响应性能。仿真结果表明,上述控制方案能够实现速度环PI参数在线自整定,明显提升系统动态响应性能,且具有良好的抗扰动能力。Using PI control strategy in the PMSM speed control system cannot guarantee good dynamic response performance and anti-disturbance ability at the same time,and the parameters setting is complicated in practical application.To improve the above problems,a modified wavelet neural network PI controller was designed for the speed loop.It used adaptive learning rate gradient descent with an inertial term for updating network parameters online and expanded the PI parameters output range by applying weight factors.These improvements enhanced the controller′s performance.In order to eliminate the influence of current loop proportional parameter on speed loop,a deadbeat predictive current controller was designed to further improve the dynamic response performance of the system.The simulation results show that the above control scheme can realize online self-tuning of the PI parameters of speed loop,significantly improve the dynamic response performance of the system,and guarantee good anti-disturbance ability.

关 键 词：PMSM 小波神经网络 自适应学习速率 PI控制 模型预测控制 

分 类 号：TH165[机械工程—机械制造及自动化] TG659[金属学及工艺—金属切削加工及机床]