基于小波分解的害虫发生非平稳时间序列分析和预测  

Analysis and forecasting for non-stationary time series of pest occurrence degree based on wavelet decomposition

作　　者：朱军生[1,2] 翟保平[1] 刘英智 

机构地区：[1]南京农业大学昆虫学系,江苏南京210095 [2]山东省植物保护总站,山东济南250100 [3]烟台市植物保护站,山东烟台264500

出　　处：《南京农业大学学报》2011年第3期61-66,共6页Journal of Nanjing Agricultural University

基　　金：国家973计划项目(2006CB102007);国家科技支撑计划项目(2006BAD08A01)

摘　　要：将小波分解应用于害虫发生程度非平稳时间序列的分析和预测。通过小波分解,将非平稳时间序列分离为多个平稳分量,然后采用自回归滑动平均方法对各平稳分量分别进行分析和建模,最后将所有分量的模型进行组合,从而可以得到原非平稳时间序列的预测模型。在实例分析中,利用1959年至2004年烟台市一代玉米螟发生程度数据序列建立了预测模型,利用2005年至2009年的数据对模型进行了检验。检验结果表明:5年预测准确率达到了80%,预测效果令人满意。Wavelet decomposition was applied to analyze and forecast for non-stationary time series of pest occurrence degree.The non-stationary time series was decomposed into several stationary components with wavelet decomposition.Then,every stationary component was analyzed by auto-regressive moving average method and a model was established.Finally,the models of all stationary components were combined to obtain the model of the original non-stationary time series.The occurrence degree data series of Ostrinia furnacalis in Yantai from 1959 to 2004 was used to establish forecasting model,and the data from 2005 to 2009 was used to test the model.The test result showed that the forecasting accuracy of five years reached satisfying 80%.

关 键 词：小波分解 多分辨率分析 非平稳时间序列 玉米螟 预测 

分 类 号：S431.9[农业科学—农业昆虫与害虫防治]