基于改进神经网络的两电机同步系统张力辨识  

Tension Identification Based on Improved Neural Network of Two-Motor Synchronous System

作　　者：袁骏[1] 刘国海[1] 

机构地区：[1]江苏大学,镇江212013

出　　处：《微特电机》2016年第12期77-80,共4页Small & Special Electrical Machines

基　　金：国家自然科学基金项目(61273154;51577084);江苏省高校自然科学研究重大项目(15KJA470002);江苏省"333工程"科研资助项目(BRA2015302)

摘　　要：张力辨识是实现两电机同步系统无传感器运行的重要步骤。为了辨识出张力值,在系统数学模型左逆存在性已证明的前提下,针对BP算法训练慢且精度不高,两电机系统存在负载扰动和系统噪声的特点,提出一种基于跟踪微分器——粒子群优化BP网络左逆软测量辨识方法。仿真结果表明,该方法辨识出的张力可以在存在负载扰动情况下精确跟踪张力实际值,抑制外在扰动对于张力辨识的影响,为实现两电机无张力传感器系统的控制提供了现实可行性。Tension detection is a key to achieve sensorless operation of two-motor synchronous system. After proving left-invertibility of mathematical model, to identify the tension of belt, a novel identification based on tracking differentiator particle swarm optimization BP network left-inversion is proposed, considering that the convergence speed of BP network is slow and accuracy is low, two-motor system exit load disturbance and system noise. The simulated result shows that, the tension identified by the method can track the actual value exactly under the load disturbance, and also can restrain the in- fluence of the system noise. The resuh offers practical feasibility for achieving the control of two-motor synchronous system under sensodess operation.

关 键 词：两电机同步系统 张力辨识 神经网络左逆 跟踪微分器 粒子群算法 

分 类 号：TM346[电气工程—电机]