基于BP神经网络的黑龙江漠河段冰坝预测

黑龙江干流上游在开江流凌期经常出现冰坝，并产生凌汛灾害。为预测黑龙江漠河段的开江日期和冰坝发生情况，利用黑龙江漠河段1960—2010年的水文气象数据建立基于BP神经网络的冰坝预测模型，预测该河段2011—2015年的冰坝发生情况及开江日期，并和实际情况进行比较。研究结果表明:该模型的预测精度较高；通过对27年的气温转正日期和开江日期的分析发现，二者的日期均趋于稳定，且在气温转正后的15 d左右黑龙江漠河段会顺利进入开江阶段；预测2011—2015年开江日期的最大误差为3 d，根据水文情报预报规范，此次预测为甲等预测方案且预测结果均合格。

Forecast of ice dam in Mohe reach of Heilongjiang River based on BP neural network

Ice dams often occur in the upper reaches of the main stream of the Heilongjiang River during the ice flood period. In order to predict the date of ice breakup and illuminate the flood process caused by ice dams in the Mohe reach of the Heilongjiang River, an ice dam prediction model was established based on BP neural network through analyzing hydrometeorological data from 1960 to 2010 in the Mohe reach in order to predict the occurrence of ice dams and the date of ice breakup from 2011 to 2015 compared with the actual situation during the period. The research result indicates that the model has high predictive accuracy. Moreover, the dates of ice breakup and temperature normalization have been stabilized based on the analysis of the 27-years temperature data, and ice breakup would occur in the Mohe reach 15 days later after the temperature normalization. In addition, the dates of ice breakup from 2011 to 2015 were predicted, and the maximum error was 3 days. According to the hydrological information forecast specification, this prediction was a first-class prediction and the prediction results were qualified.