水文时间序列多尺度效应分析与改进预测方法  

Analysis of Multi-scale Effects in Hydrological Time Series and Its Improved Prediction Methods

作　　者：何中政 李祥吉 王永强[3] 路佳豪 HE Zhong-zheng;LI Xiang-ji;WANG Yong-qiang;LU Jia-hao(School of Engineering and Construction,Ministry of Education,Nanchang University,Nanchang 330031,China;Key Laboratory of Poyang Lake Environment and Resource Utilization,Ministry of Education,Nanchang University,Nanchang 330031,China;Changjiang River Scientific Research Institute,Changjiang Water Resources Commission,Wuhan 430010,China)

机构地区：[1]南昌大学工程建设学院,江西南昌330031 [2]南昌大学鄱阳湖环境与资源利用教育部重点实验室,江西南昌330031 [3]长江水利委员会长江科学院,湖北武汉430010

出　　处：《水电能源科学》2025年第4期35-39,共5页Water Resources and Power

基　　金：国家重点研发计划(2023YFC3209400);国家自然科学基金项目(52209024,42271044);江西省自然科学基金项目(20243BCE51170)。

摘　　要：提升水文预测技术水平一直是水文水资源研究领域的重点和难点,而在水文预测中,往往短临时段的水文预测效果高于中长期。对此,提出了大时间尺度的序列自回归特性弱于小时间尺度的猜想,以及基于自相关性系数的多尺度效应分析及其改进水文时间序列预测方法。并选定长江流域8个典型水文站点为研究对象,以3 h、6 h、日、周、月为预测时间步长开展实例分析。相关试验结果表明,不同站点的ACF(p)平均绝对值从大到小为3 h、6 h、日、周、月;各个站点不同时间尺度下改进预测方法结果相比传统预测方法,均方根误差平均降低20.76%,平均绝对误差降低20.70%,确定性系数平均提高6.65%,合格率平均提高9.77%,最大误差平均降低18.16%。验证了所提出的猜想及其改进预测方法,可为水文预测研究提供新的思路。How to improve the level of hydrological prediction technology has always been the focus and difficulty in the field of hydrology and water resources research.In hydrological forecasting,the effect of hydrological prediction in the short period is often higher than that in the medium and long term.In view of this phenomenon,this paper proposed the conjecture that the autoregressive characteristics of large time scales are weaker than those of small time scales.A multiscale effect analysis method based on autocorrelation coefficients and an improved hydrological time series prediction method was put forward.The eight typical hydrological stations in the Yangtze River basin were selected as the research objects.The case analysis was carried out with 3 h,6 h,day,week and month as prediction time steps.The results of correlation experiments show that the average absolute value of ACF(p)at different sites ranges from high to low at 3 h,6 h,day,week,and month;Compared with traditional prediction methods,the improved prediction methods at different time scales of each site showed an average reduction of 20.76%in E_(RMSE),20.70%in E_(MAE),6.65%in CDC,9.77%in RQR,and 18.16%in E_(ME).In summary,the conjecture and its improved prediction method proposed in this article have been validated,providing new research ideas for the field of hydrological prediction.

关 键 词：水文预测 自相关性 自回归模型 多尺度效应 

分 类 号：TV124[水利工程—水文学及水资源]