基于遗传算法优化小波神经网络预测GPS可降水量  

Prediction of GPS Precipitation Based on Modified Genetic Wavelet Neural Network

作　　者：熊红伟 郑进 Xiong Hongwei;Zheng Jin(WSGRI Smart City(Wuhan) Engineering Technology Co.,Ltd,Wuhan 430080,China;Hubei Institute of Land Surveying and Mapping,WuHan 430010,China)

机构地区：[1]中冶智诚(武汉)工程技术有限公司,湖北武汉430080 [2]湖北省国土测绘院,湖北武汉430010

出　　处：《城市勘测》2019年第4期136-140,共5页Urban Geotechnical Investigation & Surveying

摘　　要：大气水汽变化,有着高度的随机性与时空剧变性;通过GPS遥感的方式获取得到的GPS可降水量是一个剧烈变化的非线性系统。针对GPS可降水量这一特点以及现有预测模型的局限性,采用遗传算法的全局搜索能力和小波神经网络良好的逼近与容错能力相结合的遗传小波神经网络进行建模。用所建模型对不同特点的GPS可降水量的时间序列进行分析。结果表明遗传小波神经网络在预测的精度、稳定性和预测的步长上均优于现有的PWV预测模型。The variation of water vapor has a high degree of randomness and time-space dramatic. The GPS precipitable water vapor (PWV) obtained by remote sensing is a severely changed non-linear system. Aiming at the characteristics of GPS precipitation and the limitations of existing prediction models,the global search ability of genetic algorithm and the combination of good approximation and fault tolerance of wavelet neural network were used to model.The model is used to analyze the time series of GPS precipitation with different characteristics. The results show that the genetic wavelet neural network is better than the existing PWV prediction model in the accuracy,stability and step length of prediction.

关 键 词：GPS可降水量 预测 遗传算法 小波神经网络 

分 类 号：P228.TP183[天文地球—大地测量学与测量工程]