不同输入方案对径流预测精度的影响研究
2021-08-11刘振男周靖楠陆之洋徐桂弘
刘振男 周靖楠 陆之洋 徐桂弘



摘 要:径流预测对合理利用有限的水资源至关重要。基于成因分析法、主成分分析法(PCA法)、核主成分分析法(KPCA法)分别构建3种不同的模型输入方案,并采用自适应模糊推论系统(ANFIS模型)对河南省北汝河汝州水文站月径流量进行预测,依据均方根误差与相关系数对预测精度进行评价,从而明晰不同变量选择方法在径流预测当中的应用效果。结果表明:ANFIS模型适用于研究区的径流预测。PCA法、KPCA法分别构建的模型输入方案与成因分析法得到的方案相比,不但变量数目大幅减少,而且径流预测精度亦有大幅度的提高。与此同时,PCA法较KPCA法更适合重建研究区的径流预测变量方案。另外发现,模型运行时间与输入方案中的变量个数关系紧密,即变量个数越少,运行时间越短。
关键词:径流预测;主成分分析法;核主成分分析法;自适应模糊推论系统;预测因子
中图分类号:TP391.9 文献标志码:A
doi:10.3969/j.issn.1000-1379.2021.07.008
引用格式:刘振男,周靖楠,陆之洋,等.不同输入方案对径流预测精度的影响研究[J].人民黄河,2021,43(7):41-44.
Abstract: Runoff prediction is very important for rational utilization of limited water resources. Based on the cause analysis method, PCA, KPCA and ANFIS model, the monthly runoff of RuzhouHydrology Station on the Beiru River in Henan Province was predicted. By means of root-mean-square error and correlation coefficient, the influence of different input schemes selected by different variable selection methods on runoff prediction accuracy was studied. The results show that the ANFIS model is suitable for runoff prediction in the study area. Compared with the schemes obtained by cause analysis, the input schemes constructed by PCA and KPCA respectively not only have a sharp decrease in the number of variables, but also greatly improve the accuracy of runoff prediction. Meanwhile, PCA is more suitable to reconstruct the runoff prediction variable scheme than that of KPCA. In addition, it is found that the running time of the model is closely related to the number of variables in the input scheme, that is, the smaller the number of variables, the shorter the running time.
Key words: runoff forecast; PCA; KPCA; ANFIS; forecasting factor
隨着我国经济快速发展,各行各业对水资源的需求量越来越大,因此合理有效地利用水资源至关重要,而径流的准确预测对于高效地分配有限的水资源具有重要的现实意义[1]。众所周知,大量不确定因素会对径流量的多少产生影响,给径流预测工作带来诸多挑战,如何有效地提高径流预测精度已成为水文预报研究领域的热点。传统的预测方法主要是根据河川径流自身存在的连续性、周期性等特点进行预测,如成因分析法、数理统计法以及时间序列法[2]。……
