[关键词]
[摘要]
由于特殊的地理位置、水文气象条件及河道特性的影响, 黄河内蒙段几乎每年都会发生凌汛。对黄河内蒙段 主要控制站的气象水文等实测数据进行分析后, 发现近年来随着凌期气温升高, 流量增大, 流凌、首封日期推后, 开 河日期提前, 且最大冰厚明显变薄。为此, 以黄河内蒙段巴彦高勒站为例, 通过相关分析选取合适的预报因子, 采用 基于遗传算法的神经网络方法建立了凌情智能耦合预报模型( GA2BP 模型) , 对流凌、封河、开河日期进行预报。对 比不同模型的预报结果, 发现多元线性模型、BP 模型和GA2BP 模型合格率分别为80%、861 7%和931 3%, GA2BP 模型的预报精度较高。因此, GA2BP 模型可以为黄河内蒙段的凌汛灾害防治提供重要支持。
[Key word]
[Abstract]
Due t o the special geog ra phical po sition, hy dr o2meteo rolog ical co nditio ns, and r iver course characterist ics, ice floo d al2 mo st occur s ev ery year in the Inner Mong olia reach of the Yellow Riv er. The meteo rolog ical and hydro lo gical data at t he ma in co nt rolling statio ns in t he Inner M ongo lia reach wer e analy zed, w hich sug g ested that the temper ature and flow dischar ge in2 cr ease in recent year s, the ice run date and fr eeze2up date push back while the br eak2up dat e br ings fo rw ard, and the max imum ice thickness t hins obv iously . In this paper, the appr opriate predictio n facto rs w ere selected by the co rrelatio n analysis. Ice re2 g ime intellig ent coupling fo recast model w as built using the neur al netw or k metho d based o n the g enetic algo rithm. The model was applied to for ecast the ice run date, fr eeze2up date, and break2up date at t he Bay ang aole station in the Inner Mongo lia r each. The forecast r esults o btained from different models wer e compared, and it show ed that the mult iple linear model, BP mo del, and GA2BP model have hig h passing percentages, which ar e 80%, 86. 7%, and 93. 3%, respectively. GA2BP mo del has the hig hest forecast accur acy and can pro vide the cr itical suppo rt fo r the pr evention of ice disasters in the Inner Mongo lia reach o f the Yellow River.
[中图分类号]
[基金项目]