摘要
以西江迁江站、柳州站和武宣站1952—2005年实测洪水数据为基础,通过分析上下游站点洪水流量的相关关系选取洪水预报特征因子,采用随机森林算法建立了武宣站洪水预报模型。结果表明:模型率定期预测的武宣站12~48 h洪水过程确定性系数大于0.98、合格率大于98%,验证期预测的12~24 h洪水过程确定性系数大于0.72、合格率大于82%,预报精度较高,预报结果不确定性较小,可为洪水预报提供方法参考。
Based on the measured flood data from 1952 to 2005 in Qianjiang Station,Liuzhou Station,and Wuxuan Station of Xijiang River,this paper selects the characteristic factors of flood forecasting by analyzing the correlation of flood flow at upstream and downstream stations.Meanwhile,the random forest algorithm is adopted to build a flood forecasting model for Wuxuan Station.The results are as follows.The certainty coefficients of the 12~48 h flood process in the studied station during the calibration period are more than 0.98 with the pass rates more than 98%.Additionally,the certainty coefficients of the 12~24 h flood process during the verification period are more than 0.72,with a pass rate of more than 82%.Thus,the proposed model has high forecast accuracy and little uncertainty and can provide a reference for flood forecasting methods.
作者
刘和昌
赵博华
孙波
LIU Hechang;ZHAO Bohua;SUN Bo(Technical Advisory of Pearl River Water Resources Commission(Guangzhou)Co.,Ltd.,Guangzhou 510611,China;Comprehensive Technology Center of Pearl River Water Resources Commission of the Ministry of Water Resources,Guangzhou 510611,China)
出处
《人民珠江》
2023年第10期132-139,共8页
Pearl River
关键词
洪水预报
随机森林算法
相关性分析
数据挖掘
预报特征因子
flood forecasting
random forest algorithm
correlation analysis
data mining
forecast characteristic factors