基于ICPSO优化的PMa-BSynRM神经网络逆系统解耦控制

Decoupling Control of PMa-BSynRM Based on Neural Network Inverse System Optimized by ICPSO

针对永磁辅助式无轴承同步磁阻电动机(PMa-BSynRM)存在的强耦合和非线性问题,以及神经网络逆系统解耦控制收敛速度慢和易陷入局部极值的问题,提出了一种将BP神经网络与改进混沌粒子群优化(ICPSO)算法相结合的PMa-BSynRM解耦控制方法。基于PMa-BSynRM的结构和工作原理建立转矩和悬浮力的数学模型并进行可逆性分析,再建立BP神经网络作为逆系统,利用ICPSO算法优化其初始连接权值,通过仿真验证优化效果,经过充分训练后,将BP神经网络与原系统串联形成伪线性系统,从而实现PMa-BSynRM的解耦控制。仿真及试验结果表明,采用ICPSO算法优化的BP神经网络逆系统收敛速度和解耦性能均优于传统神经网络逆系统,改善了PMa-BSynRM的动态与静态性能。

To solve the problems of strong coupling and nonlinearity of permanent magnet assisted bearingless synchronous reluctance motor(PMa-BSynRM)and improve the deficiencies of slow convergence speed and propensity to fall into local extremum of neural network inverse system decoupling control,a decoupling control method combining back propagation neural network(BPNN)with improved chaos particle swarm optimization(ICPSO)algorithm is proposed for PMa-BSynRM.The mathematical models of torque and suspension force are established based on structure and working principle of PMa-BsynRM,and the reversibility analysis is carried out.Then,the BPNN is established as inverse system,and the ICPSO algorithm is used to optimize the initial connection weight of BPNN.The optimization effect is verified by simulation.After sufficient training,the BPNN is connected in series with original system to form a pseudo-linear system,so as to realize the decoupling control of PMa-BSynRM.The simulation and experimental results show that the convergence speed and decoupling performance of BPNN inverse system optimized by ICPSO algorithm are better than those of traditional neural network inverse system,and the dynamic and static performances of PMa-BSynRM are improved.