改进小脑模型网络的干式变压器卷线机跑偏信号谐波分析

提出一种基于改进的小脑模型控制器（CMAC）神经网络的干式变压器卷线机跑偏信号谐波分析方法．该方法在检测到干式变压器卷线机跑偏信号的基础上，对不同频率的谐波进行了分析、推论，将常规CMAC网络的学习因子改成随学习误差的变化动态调整，然后采用基于改进的CMAC神经网络对跑偏各谐波分别辨识，再取主次非线性谐波叠加．辨识结果表明，这种方法不仅能方便地识别出最大跑偏信号谐波基频的最小频率范围，而且比在相同情况下采用常规反向传播（BP）的网络辨识的精度高，学习速度提高20％，同时得到了最大跑偏信号谐波的最简单模型．

Harmonic Wave Analysis Method of Improved Cerebellar Model Articulation Control for Coil Winder of Dry-Type Transformer

For coil winder of dry-type transformer,a harmonic wave analysis method of improved cerebellar model articulation control (CMAC) is presented,where the harmonic wave is analyzed and reasoned based on measuring offset signals;the learning parameter of the normal CMAC network is improved into an adjustment varying with learning error;and then the harmonic wave of offset signals are identified,the primary nonlinear components are superposed.This method enables to identify the scope of a minimum frequency of the harmonic wave for coil winder with higher accuracy and identifying rate than the normal back propagation network identification learning.Under the same condition the simplest harmonic wave model of coil winder of dry-type transformer is obtained simultaneously.