基于CEEMDAN-VMD-BP模型的月径流量预测研究  

Research on Monthly Runoff Prediction Based on CEEMDAN-VMD-BP Model

作　　者：王文川[1] 杜玉瑾 和吉[1] 邱林[1] WANG Wenchuan;DU Yujin;HE Ji;QIU Lin(College of Water Resources,North China University of Water Resources and Electric Power,Zhengzhou 450046,China;Xifeng Hydrology and Water Resources Survey Bureau,Yellow River Conservancy Commission,Qingyang 745099,China)

机构地区：[1]华北水利水电大学水资源学院,河南郑州450046 [2]黄河水利委员会西峰水文水资源勘测局,甘肃庆阳745099

出　　处：《华北水利水电大学学报（自然科学版）》2023年第1期32-40,48,共10页Journal of North China University of Water Resources and Electric Power：Natural Science Edition

基　　金：国家自然科学基金项目(51509088,51709108);河南省重点研发与推广专项项目(202102310259,202102310588);河南省高校科技创新团队项目(18IRTSTHN009)。

摘　　要：为进一步提高月径流量的预测精度,采用自适应噪声的完全集合经验模态分解方法(CEEMDAN)分解实测月径流数据,并计算各模态分量的样本熵(SE);然后利用变分模态分解(VMD)对样本熵最大的分量进行二次分解,以削弱序列的非平稳性;将各分量作为BP神经网络模型的输入进行训练、预测,再将各分量预测结果线性叠加得最终预测结果,提出了二次分解组合预测模型(CEEMDAN-VMD-BP)。以洪家渡水电站和长水水文站实测月径流数据作为研究对象,将提出的模型与BP模型、CEEMDAN-BP模型预测结果进行对比。结果表明:组合模型预测精度高于单一模型;融和VMD方法后的二次分解模型的拟合优度高于一次分解组合模型,验证了CEEMDAN-VMD-BP模型的高效性。In order to further improve the prediction accuracy of monthly runoff,the adaptive noise based on CEEMDAN was used to decompose the measured monthly runoff data,and the sample entropy(SE)of each modal component was calculated.The variational mode decomposition(VMD)is used to decompose the component with the largest sample entropy twice to weaken the non-stationarity of the sequence.Each component is used as input to the BP neural network model for training and prediction,and then the final prediction result is obtained by linear superposition of the prediction results of each component,and the model of CEEMDAN-VMD-BP is proposed.Taking the measured monthly runoff data of Hongjiadu Hydropower Station and Changshui Hydrological Station as the research object,the prediction results of the proposed model are compared with those of BP model and CEEMDAN-BP model.The results are as follows.The prediction accuracy of the combined model is higher than that of the single model.The goodness of fit of the secondary decomposition model combined with the VMD method is higher than that of the primary decomposition combination model,verifying the efficiency of the CEEMDAN-VMD-BP model.

关 键 词：月径流预测 二次分解 CEEMDAN VMD 样本熵 

分 类 号：P338[天文地球—水文科学]