基于RBF神经网络的SVPWM过调制  

Design of SVPWM Overmodulation Based on RBF Neural Networks

作　　者：易灵芝[1] 何素芬[1] 李举成[1] 彭寒梅[1] 李明[1] 陈彦如[1] 

机构地区：[1]湘潭大学信息工程学院,湖南湘潭411105

出　　处：《控制工程》2009年第6期696-700,共5页Control Engineering of China

基　　金：湖南省学位与研究生教育教学改革研究基金资助项目(06A09);湖南省自然科学基金资助项目(08JJ6029)

摘　　要：线性调制状态下的逆变器,存在不能充分利用直流母线电压的问题,为了获得尽可能大的输出电压,一般对逆变器进行过调制控制。由于过调制的非线性,计算复杂,提出新的基于神经网络的SVPWM逆变器,采用线性调制和过调制2种模式,通过限定轨迹双调制模式法,实现在整个调制范围内线性控制。采用资源分配法确定径向基函数的网络结构和参数,设计出较精简的网络实现SVPWM;并将这种逆变器应用于异步电动机控制系统中。最后在Matlab环境下建立基于神经网络的SVPWM逆变器供电的异步电机控制系统仿真模型,仿真结果表明,该方法简单、高效、控制效果良好,能提高直流母线电压利用率,降低输出电流谐波含量和电机转矩脉动。The problem of obtaining the higher output voltage value in linear modulation condition PWM inversion is discussed. A complete artificial-neural-network space vector pulse width modulation （ANN-SVM）controller for a voltage source inverter is presented. The operation is very well both in under-modulation and over-modulation regions. The linear control is realized in the entire modulation scope by limit trajectories. The resources method of distribution determination RBF network architecture and the parameter are used. The RBF neural network realize the voltage space vector modulate to control the inverter, which is applied to the control system of asynchronous machine. The simulation results show that the method features simplicity, high efficiency and excellent control effect. The use ratio of DC bus voltage is raised, and the output current harmonics content and the torque pulsation are reduced.

关 键 词：径向基神经网络 空间矢量调制 欠调制 过调制 逆变器 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]