摘要: |
【目的】为了提高降水量的预测精度,反映降水量的实际特征。【方法】基于经验模态分解对非线性时间序列的分析和处理的优势,对黄河三角洲气象站点1954—2018年连续65 a月均降水量数据进行经验模态分解(Empirical Mode Decomposition, EMD),得到了系列本征模态函数(Intrinsic Mode Function,IMF),然后对IMF进行Hilbert变换,在此基础上建立了两种黄河三角洲降水量多尺度预报模型。【结果】黄河三角洲降水量存在着9、13、23、76、135月左右的周期,并以9个月的波动为主;65年月均降水量数据预测结果显示:模型一的相对误差在0.9%~9.8%之间,模型二的相对误差在1.6%~11.8%之间,在建模时不考虑初相位的模型一平均预测误差为2.70%,整体预测精度要优于考虑初相位的模型二。【结论】两种模型的拟合精度及显著性均符合要求。 |
关键词: 降水量;时间序列;多尺度;EMD;预测 |
DOI: |
分类号:黑体,小五 |
基金项目:新乡黄河湿地生态服务价值评估与修复技术研究 |
|
In yellow River delta Precipitation Forecast Bsed on Nonlinear Multi-scale Model |
muyuzhu
|
Xinxiang Bureau of Hydrology and Water Resources Survey
|
Abstract: |
【Background】Under the influence of climate change and human activities, precipitation in various regions has also changed. Precipitation is closely related to human production, life and ecology. The change of precipitation is related to the sustainable utilization of regional water resources, the protection of ecological environment and the development of economy and society, The research on the variation characteristics and evolution trend of precipitation has become a hot topic in the field of climate and water resources. Scholars and researchers are very concerned about the accurate prediction of precipitation. 【Objective】In order to improve the prediction accuracy of precipitation, Reflect the actual characteristics of precipitation. 【Method】Based on the advantages of empirical mode decomposition in the analysis and processing of nonlinear time series and other fields, Empirical Mode Decomposition (EMD) was carried out for the monthly average precipitation data of the Yellow River Delta Meteorological Station from 1954 to 2018 ,and a series of eigenmode functions were obtained. Hilbert transform is performed on IMF, and on this basis, two multi-scale forecast models of precipitation in the Yellow River Delta are established. 【Result】The results of the study show that there are periods of September, 13, 23,76 and 135 months in precipitation in the Yellow River Delta, and 9-month fluctuations are the main ones; the 65-year monthly average precipitation data is predicted, Results show, The relative error of the model 1 is between 0.9% and 9.8%, and the relative error of the model 2 prediction is between 1.6% and 11.8%. When modeling, the average prediction error of Model 1 without considering the initial phase was 2.70%, and the overall prediction accuracy was better than that of Model 2 considering the initial phase.【Conclusion】The fitting accuracy and significance of the two models meet the requirements. |
Key words: precipitation; time series; multiscale; EMD; predict |