基于GANN的SRM无位置传感器位置预估方法研究  

Study on Position Estimation Method of Sensorless SRM Based on GANN

作　　者：宋受俊[1] 葛乐飞[1] 王路生[1] 王琛[1] 

机构地区：[1]西北工业大学,西安710072

出　　处：《微特电机》2015年第2期1-3,7,共4页Small & Special Electrical Machines

基　　金：国家自然科学基金项目(51107100);陕西省自然科学基金项目(2011GQ7001);教育部博士点基金新教师类项目(20116102120033);西北工业大学本科生毕业设计(论文)重点扶持项目

摘　　要：神经网络具有强大的非线性映射能力,非常适用于对磁链特性高度非线性的开关磁阻电机(SRM)转子位置的预估。然而,由于其网络结构、初始连接权值和阈值的不确定性,很难一次获得理想的训练结果。提出了一种基于遗传优化神经网络(GANN)的SRM无位置传感器位置预估方法,利用遗传算法对神经网络的初始权值和阈值进行优化,在此基础之上,以磁链和电流为输入、转子位置为输出,建立了预估模型。仿真结果表明,该方法可以在不同转速下对转子位置进行准确的预估,其预估误差不大于2°。Neural network（ NN） has strong non-linear mapping ability,and is ideal for rotor position estimation of switched reluctance machine whose flux characteristic is highly non-linear. However,it ＇s difficult to obtain satisfying training results,due to the uncertainty of its network structure,initial connection weights and thresholds. Aposition estimation method for SRM based on genetic optimized neural network（ GANN） was presented. In the method,genetic algorithm was used to optimize the initial weights and thresholds of the neural network. Then,estimation model was built by using flux and current as input and rotor position as output. Simulation results verify that the presented method can estimate rotor position accurately under different operation speeds,and the estimated error is within 2°.

关 键 词：开关磁阻电机 神经网络 遗传算法 无位置传感器 位置预估 

分 类 号：TM352[电气工程—电机]