论文部分内容阅读
针对小波变换方法的不足,运用EMD方法对黄河兰州以上二级水资源分区45年(1956- 2000年)的年降雨量序列进行多时间尺度分析,发现该区域年降雨量存在准3年、准4-8年、准11年波动周期,并探讨了各IMF分量的物理背景及其趋势变化;然后以年降雨量的EMD分量为输入,以相应的年径流量为输出,建立了基于EMD的年降雨径流BP神经网络预测模型.研究结果表明:EMD作为一种全新的信号处理方法,可以对水文时序进行精确的多时间尺度分析,进而掌握其局部变化规律,为人工神经网络提供高质量、多层次的输入变量,显著提高模型质量.
Aiming at the shortcomings of the wavelet transform method, the multi-time scale analysis of the annual rainfall series over the 45 years (1956-2000) of the secondary water resources division of the Lanzhou Yellow River above the Yellow River by using the EMD method shows that the annual rainfall in the region exists for three years, 4-8 years and quasi-11-year fluctuation period, and explored the physical background and trend changes of each IMF component. Then, taking the EMD component of annual rainfall as input and the corresponding annual runoff as output, an EMD-based Annual rainfall runoff BP neural network prediction model.The results show that: EMD as a new signal processing method, the hydrological time series can be accurate multi-time scale analysis, and then grasp the local variation law, to provide artificial neural network with high quality, Multi-level input variables, significantly improve the quality of the model.