基于组合聚类RBF网络的开关磁阻电动机建模仿真  

Modeling and Simulating of Switched Reluctance Motor Based on RBF Neural Network with Combined Clustering Algorithm

作　　者：周素莹[1] 林辉[1] 

机构地区：[1]西北工业大学,陕西西安710072

出　　处：《微特电机》2009年第10期24-26,59,共4页Small & Special Electrical Machines

摘　　要：在已知SRM静态磁特性曲线的基础上,将一种自适应RBF网络应用于SRM的建模中。该网络采用组合聚类的方法动态确定RBF神经网络的隐层节点数及网络中心,克服了传统RBF网络把每个数据点都作为隐层节点,当数据量较大时致使网络结构冗余较大、学习速度较慢的缺点,同时又保持了传统RBF网络的优点。与传统的RBF网络相比,所设计的自适应网络节点少,精度能够满足要求。仿真结果表明,该模型能够较好地反映电机磁特性,所建立的SRM驱动系统仿真模型具有较好的通用性,为SRM的设计分析与新型控制策略的验证提供了基础。RBF neural network with self-adaptive structure was used to model for SRM based on the static magnetiza- tion curves. Combined clustering algorithm was presented here to self-adaptively determine the node number of bidden layer and center of RBF neural network. Thus, the difficulty of determining node number of hidden layers and center, slow learning speed and weaken generalization ability of RBF neural network when the input data was generous and complex were solved. Simulation results prove that the network designed has less node numbers than general RBF network when approximation precision is satisfied. The simulation results also show that the proposed model can be used to accurately predict the rela- tionship between flux-linkage and current. The general model for switched reluctance drive developed here can be widely used and is easy to modify, which offers a good platform for the design of SRM and validating of new control strategies.

关 键 词：组合聚类自适应RBF网络 开关磁阻电动机 建模与仿真 

分 类 号：TM352[电气工程—电机]