CAJ | 학술논문

监测序列经小波分解后可以得到各层分量。对低频分量采用灰色GM （1，1）模型进行建模预测，对高频分量采用BP神经网络进行建模预测，最后将各分量进行小波重构，得到监测序列的预测值。将预测值分别与没有进行小波分解直接用GM （1，1）模型预测的值和经小波分解的低、高频系数都采用GM （1，1）模型预测的值进行对比，发现经小波分解的灰色‐神经网络组合模型预测精度更高。
감측서렬경소파분해후가이득도각층분량。대저빈분량채용회색GM （1，1）모형진행건모예측，대고빈분량채용BP신경망락진행건모예측，최후장각분량진행소파중구，득도감측서렬적예측치。장예측치분별여몰유진행소파분해직접용GM （1，1）모형예측적치화경소파분해적저、고빈계수도채용GM （1，1）모형예측적치진행대비，발현경소파분해적회색‐신경망락조합모형예측정도경고。
Monitoring sequences obtain each component through wavelet decomposition .The low frequency components adopt GM (1 ,1) model to make predictions ,and the high frequency components with BP neural network .Then the predicted data are obtained after reconstructing the comporents .The result shows the gray‐neural network combination model based on wavelet has high prediction accuracy comparing with the GM (1 ,1) model without wavelet decomposition and the model that uses GM (1 ,1) model predicting the low frequency components and the high frequency components after wavelet decomposition .