输入多采样率系统随机梯度辨识算法研究  

Research of Stochastic Gradient Identification for Multi-Rate Input System

作　　者：徐鹏[1] 彭小峰[1] 李山[1] 

机构地区：[1]重庆理工大学,重庆400054

出　　处：《微特电机》2014年第5期1-3,16,共4页Small & Special Electrical Machines

基　　金：重庆市科委自然科学基金项目(cstc2012jjA40066);重庆市教委科学技术研究项目(KJ130807)

摘　　要：经典参数估计往往建立在系统输入输出相同采样频率下,而工业过程往往都是滞后系统,具有输入多采样率性特性。结合多项式变化技术,得到永磁同步电动机输入多采样率数学模型,并采用随机梯度算法对该模型参数进行辨识,同时在遗忘因子作用下能提高了收敛速度和辨识精度,仿真结果表明该算法的有效性。To the parameters identification the input and output data was sampling at the same rate in classic identification theory.Otherwise industry processes were usually the time-delay systems,and then had multi-rate sampling characteristic on the input.The input multi-rate mathematical model of PMSM was induced with polynomial transformation technique.Stochastic gradient（SG） algorithm was proposed and used to the transformed model parameters identification.With the forgetting factor,the proposed algorithm has more outstanding performance on convergence speed and precision.Simulation results show the effectiveness of the proposed algorithm.

关 键 词：输入多采样率 多项式变换技术 永磁同步电动机 随机梯度 遗忘因子 

分 类 号：TM351[电气工程—电机] TM341