摘要
径流序列可以看成是各种不同成分线性叠加构成的时间序列。利用小波变换良好的局部化时频分析能力，将年最大径流序列进行分解，使其趋势项、周期项和随机项得以分离。各子序列分别代表不同的时间尺度，反映了各种物理因素对径流过程的影响。然后根据各子序列的特性分别建立幂函数、周期函数或ARMA模型并进行预测。最后将各子序列的预测值合成，得到年最大径流序列的预测值。对宜昌站1991-2002年最大径流量的预测结果表明，该方法是切实可行的。并指出小波包变换在分析中、高频信息方面优于小波变换，有助于进一步提高预测的精度。
Runoff series can be regarded as a linear overlapping of different components. In this paper, the perfect localized time and frequency analysis capacity of wavelet transform is made use to separate the items of tendency, period, and random through decomposition of annual maximum runoff series. Sub-series represents different scale of time to reflect the influences of different physical factors on runoff process. Power function, periodical function, and ARMA model are established according to the characteristics of sub-series. The calculated results of sub-series are added to forecast the annual maximum runoff series. Through forecast of annual maximum runoff 1991-2002 in Yichang Station, the results show that the method is practical and feasible; wavelet package transform is superior to wavelet transform for analysis of information of high and middle frequency.
出处
《中国农村水利水电》
北大核心
2006年第7期10-11,14,共3页
China Rural Water and Hydropower
