基于神经网络理论的开河期冰坝预报研究

在北方高寒地区的天然河道,开河期冰坝形成和导致凌汛的机理复杂,目前的冰水动力学模型难以模拟和预报其发生、发展和溃决的过程,可用的冰坝预报多采用传统的统计学方法和经验判别式法,为应对严重的防凌形势,迫切需要找到冰坝预报的新方法。本文在对开河期冰坝成因及机理研究的基础上,建立了基于神经网络理论的冰坝预报模型,并将其应用到黑龙江上游凌汛灾害频发的漠河江段冰坝预报中。通过神经网络聚类法预报冰坝是否发生,神经网络聚类法预报精度为85%,高于传统统计学的几率分析法62%的预报精度。通过预报开河日期实现了对冰坝发生时间的预报,开河日期预报平均预见期为10天,最大误差2天,预报合格率100%。该模型提前准确预报2017年黑龙江漠河江段开河冰坝发生情况。及时、准确的冰坝预报能为提前制订主动防凌方案和采取必要防凌措施提供重要的依据。

Prediction of breakup ice jam with Artificial Neural Networks

Breakup ice jams occur during periods of thaw in many northern river courses,which is hard to predict because the formation and progression of breakup ice jams result from a complex interaction between hydrologic, hydraulic, and meteorological processes. Ice jam prediction methods based on artificial neural networks (ANNs) are desirable to provide early warning and to allow rapid,effective ice jam mitigation. The soft computing through artificial neural networks and Clustering method is applied to predict break up ice jam in Mohe reach of the Heilongjiang River (Amur River). The ANNs predict with qualified rates of 85% to the occurrence of breakup ice jams proves to be more accurate than the statistical methods with qualified rates of 62%. The prediction of breakup dates is qualified according to the national standard published as Hydrographic Forecast Standard, with the errors of less than 2 days between the forecasted and measured results and the average 10-day forecast period. The forecast on the breakup ice jam in 2017 was released 24 days ahead on Apr.1, 2017, which provides the accurate results for the breakup data and the occurrence of breakup ice jams.