遗传小波神经网络在变形预报中的应用  

Application of genetic wavelet neural network in deformation forecast

作　　者：邓勇[1] 张冠宇[1] 李宗春[1] 杨振[1] 薛志宏[1] 

机构地区：[1]信息工程大学测绘学院,郑州450052

出　　处：《测绘科学》2012年第5期183-186,共4页Science of Surveying and Mapping

基　　金：矿山空间信息技术国家测绘局重点实验室开放基金(KLM201009)

摘　　要：目前常用的变形预报方法有BP人工神经网络和小波神经网络,但是都存在收敛速度慢且易受局部极值的影响。针对这两种算法的不足,本文利用遗传算法的全局寻优特性,将遗传算法与小波神经网络结合,形成遗传小波神经网,将其应用于变形预报,取得了良好的效果;并将算法的预报精度、稳定性、有效区间及运算时间作为评价算法优劣的4个标准,对BP神经网络、小波神经网络及遗传小波神经网络进行对比,结果表明遗传小波神经网络具有明显的优势。Based on the fact that BP Neural Network and Wavelet Neural Network have local minimum and slow convergence speed, this paper used the Genetic Algorithms to seek all the data in order to calculate the best result and combined the Genetic Algo- rithms and Wavelet Neural Network to form the Genetic Wavelet Neural Network for deformation forecast, and good effects were a- chieved. With the forecasting precision, stability, effective extent and calculation time as four standards to evaluate the arithmetic, it compared the BP Neural Network, the Wavelet Neural Network and the Genetic Wavelet Neural Network, the result showed that the Genetic Wavelet Neural Network had superiority.

关 键 词：变形预报 遗传算法 小波神经网络 遗传小波神经网络 稳定性 有效区间 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]