基于联合互信息的水文预报因子集选取研究

发布时间：2018-05-16 13:40

  本文选题：水文预报 + 预报因子集　； 参考：《水力发电学报》2017年08期

【摘要】：预报因子集是预报因子的集合。作为预报信息的来源,因子集对预报结果有着重要影响,增加因子集包含的预报信息量能够有效地提高预报精度。针对现有方法侧重于对单个预报因子进行研究的不足,本文从整体的角度考虑,提出了基于联合互信息的预报因子集选取方法。首先介绍了互信息并将其扩展到高维情景,引出条件互信息与联合互信息,并采用Parzen窗估计法对其进行计算;其次以水文预报为背景,建立最大联合互信息模型,根据条件互信息进行求解,并耦合反向传播(BP)神经网络对计算结果进行检验;最后对雅砻江流域泸宁水文站进行实例计算,并将计算结果与现行方法进行比较。结果表明,新方法能够为预报模型提供更加科学的输入,提高模型的预报精度。
[Abstract]:The set of prediction factors is the set of prediction factors. As the source of prediction information, factor sets have an important impact on the prediction results, and increasing the amount of forecast information contained in the factor sets can effectively improve the prediction accuracy. In view of the shortcomings of the existing methods which focus on the study of single prediction factors, this paper proposes a method of selecting prediction factor sets based on joint mutual information from the overall point of view. First, the mutual information is introduced and extended to high-dimensional scenarios, then conditional mutual information and joint mutual information are derived, and the Parzen window estimation method is used to calculate the mutual information. Secondly, the maximum joint mutual information model is established based on hydrological forecast. The solution is based on conditional mutual information and coupled with backpropagation neural network to test the calculation results. Finally, a case study of Luning hydrologic station in Yalong River basin is carried out, and the calculation results are compared with the current method. The results show that the new method can provide more scientific input for the prediction model and improve the prediction accuracy of the model.
【作者单位】： 华北电力大学可再生能源学院;
【基金】：“十三五”国家重点研发计划课题(2016YFC0402208) 中央高校基本科研业务费专项(2016XS46;2016MS51)
【分类号】：P338
，

本文编号：1897022