黄河内蒙古河段凌汛期首封影响参数辨识和预报模型研究

Ice prediction and identification of influence parameters affecting the initial freeze-up of the Inner Mongolia reach of the Yellow River

黄河上游河道首封时间和地点准确预报是制定科学防凌预案的关键。本研究在总结黄河内蒙古河段首次封冻地点基础上,总结了首封河段同河道弯曲、束窄、桥梁及河心滩等河道特征的关系,定量总结了易发生首封河道的地形特征为:河段处弯曲系数大于1.5、束窄系数在0.25~0.60范围内,为首封地点分析预测提供依据。在总结首封日期影响因子特征基础上,分别基于随机森林和BP-DEMATEL模型构建首封日期分析模型,得到影响首封日期的主要因子权重,并应用到首封日期预报中。通过因子选择和模型比较两方面对比,随机森林模型选取出的关键因子得到的预报结果与实测值吻合率100%,其在首封分析中的应用效果最优。

Accurately predicting the time and location of the initial freeze-up of the Yellow River is the key to formulate scientific ice prevention plans.On the basis of summarizing the initial freeze-up location of the Inner Mongolia reach,this study summarizes the relationship between the initial freeze-up section and the river channel characteristics such as river bends,bank narrows,bridges,and midstream beaches.It quantitatively reveals bend coefficients and bank narrows at the initial freeze-up section,providing a basis for analyzing and predicting the initial freeze-up location.On the basis of summarizing influencing factors of the initial freeze-up date,the initial freeze-up date analysis model was constructed based on Random Forest and BP-DEMATEL models respectively,and the weights of the main factors were obtained to apply to the forecast of the initial freeze-up date.The results showed that the forecast results of the key factors selected by the Random Forest model were completely consistent with the measured values,and its application effect in the initial freeze-up analysis was better than that of the BP-DEMATEL model.