基于不同因子筛选指标的丹江口入库月径流预报研究  

Monthly Runoff Forecasting of Danjiangkou Reservoir Inflow Based on Different Factor Screening Indicators

作　　者：张宁玥 陈元芳[1] 刘勇[2] ZHANG Ning-yue;CHEN Yuan-fang;LIU Yong(College of Hydrology and Water Resources,Hohai University,Nanjing 210098,China;The National Key Laboratory of Water Disaster Prevention,Nanjing Hydraulic Research Institute,Nanjing 210029,China)

机构地区：[1]河海大学水文水资源学院,江苏南京210098 [2]南京水利科学研究院水灾害防御全国重点实验室,江苏南京210029

出　　处：《水电能源科学》2024年第5期39-42,共4页Water Resources and Power

基　　金：国家重点研发计划(2022YFC3202802,2021YFC3000102)。

摘　　要：鉴于筛选识别适宜的预报因子对于提升中长期径流预报精度的重要性,以丹江口入库月径流预报为例,选择Pearson、Kendall、Spearman相关系数及随机森林因子重要性作为因子筛选指标,利用多元回归和随机森林模型,开展基于不同因子筛选指标的丹江口入库月径流预报研究。结果表明,大气环流仍是研究流域降水及产汇流的重要影响因素,部分月份径流与前期海温关系较密切;Spearman相关系数筛选下的随机森林模型全年平均预报效果最优,全年平均合格率为72.02%,因子重要性筛选下的随机森林模型在主汛期效果更优,主汛期平均合格率为69.64%;综合预报因子下的随机森林模型精度有一定的提升,全年平均合格率为75.00%,主汛期平均合格率为71.43%,在全年内不同月份的预报效果更稳定,测试期内12个月合格率的标准差下降较显著。Screening and identifying suitable forecasting factors are crucial for improving the accuracy of medium and long-term runoff forecast.Taking the monthly runoff forecasting of Danjiangkou as an example,Pearson,Kendall,Spearman correlation coefficients and Random Forest Factor Importance were selected as the factor screening indexes,and multiple regression and Random Forest model were used to carry out a study on the monthly runoff forecasting of Danjiangkou Reservoir based on the screening indexes of different factors.The results show that the atmospheric circulation is still an important influencing factor for the study of basin precipitation and production and sink flow,and part of the monthly runoff is more closely related to the SST in the previous period.The average annual forecast effect of the Random Forest model screened by Spearman correlation coefficient is the best,and the average annual pass rate is 72.02%.The average pass rate of the Random Forest model screened by factor importance is better in the main flood season,and the average pass rate is 69.64%.The accuracy of the Random Forest model under the comprehensive forecasting factors improved to some extent,with the average pass rate of 75.00%in the whole year and 71.43%in the main flood season.The forecast effect is more stable in different months throughout the year,and the standard deviation of the 12-month pass rate during the test period decreases significantly.

关 键 词：月径流预报 因子筛选指标 随机森林 多元回归 丹江口水库 

分 类 号：TV121[水利工程—水文学及水资源]