BP神经网络在三电平逆变器消谐中的应用  

Application of BP Neural Network in Three-level Inverter Harmonics Elimination

作　　者：秦昌伟[1] 王群京[1] 陈权[1] 胡存刚[1] 

机构地区：[1]安徽大学,安徽合肥230601

出　　处：《电力电子技术》2015年第4期7-9,14,共4页Power Electronics

基　　金：国家自然科学基金(51307002);安徽省自然科学基金(1308085ME81)~~

摘　　要：提出了应用反向传播(BP)神经网络消除中点箝位型三电平逆变器中的谐波技术。以消除5,7,11,13次谐波为例,在分析特定谐波消除脉宽调制(SHEPWM)技术原理的基础上,建立了用于求解开关角的BP神经网络模型,通过遗传算法以离线方式获取了测试样本,完成神经网络的训练。训练完成的BP神经网络根据给定调制度的大小,以非线性拟合的方式在线求解开关角。仿真和实验结果表明,该方法克服了传统数值法求解消谐方程时初值选择困难的情况,将离线计算和在线求解的方法相结合,可实现实时的开关角求解。An approach to eliminate harmonics in three-level neutral point clamped inverter using back propagation （BP） neural network is proposed.Taking the 5th,7th, l lth and 13th harmonics eliminated for instance, a BP neural ne- twork model to solve switching angles is set up based on the principle of selected harmonic elimination pulse width modulation（SHEPWM） technique.Test samples are acquired by means of genetic algorithm offline and the training of neural network can be completed.After training, the BP neural network can solve switching angles online through non- linear fitting, according to the value of given modulation depth.Simulation and experimental results demonstrate that this approach overcomes the diflicuhy of traditional numerical method in choosing initial value to solve harmonics e- limination equations, combines ogline calculation and online solving, and realizes real-time solving switching angles.

关 键 词：神经网络 反向传播 特定谐波消除脉宽调制 

分 类 号：TN721[电子电信—电路与系统]