BMA集合预报在淮河流域应用及参数规律初探

发布时间：2018-07-10 09:59

本文选题：集合预报 + 洪水预报　；参考：《湖泊科学》2017年06期

【摘要】：以淮河流域吴家渡水文站作为试验站点,采用基于贝叶斯平均法(BMA)的集合预报模型处理来源于马斯京根法、一维水动力学方法、BPNN(Back Propagation Neural Network)的预报流量序列,通过分析BMA的参数以及其预报结果,对各方法在淮河典型站点流量预报中的适用性进行验证与分析.经2003—2016年19场洪水模拟检验可知,BMA模型能够有效避免模型选择带来的洪水预报误差放大效应,可以提供高精度、鲁棒性强的洪水预报结果.通过进一步比较各模型统计最优的频率与BMA权重值之间的相关性,发现权重值不适用于对单场洪水预报精度评定,而适用于描述多场洪水预报中,模型为最优的统计频率;基于大量先验信息,提前获取BMA的权重等参数,将是指导模型选择、降低洪水预报不确定性、改进洪水预报技术的有效手段.
[Abstract]:The Wujiadu hydrologic station in Huaihe River Basin is used as the test station. The ensemble forecasting model based on Bayesian average method (BMA) is used to process the forecasting flow series derived from Muskinggen method and one-dimensional hydrodynamic method BPNN (back Propagation Neural Network). By analyzing the parameters of BMA and its prediction results, the applicability of each method in the flow prediction of typical stations in Huaihe River is verified and analyzed. After 19 flood simulation tests in 2003-2016, it is shown that BMA model can effectively avoid the amplification effect of flood forecast error caused by model selection, and can provide high precision and robust flood forecast results. By comparing the correlation between the statistical optimal frequency of each model and the weight value of BMA, it is found that the weight value is not suitable for evaluating the accuracy of a single flood forecast, but is suitable for describing the optimal statistical frequency of the multi-field flood forecast. Based on a large number of priori information, obtaining the weight of BMA in advance will be an effective means to guide the selection of the model, reduce the uncertainty of flood forecast and improve the flood forecasting technology.
【作者单位】：淮河水利委员会水文局(信息中心);沂沭泗水利管理局水文局(信息中心);
【基金】：国家重点基础研究发展计划项目(2016YFC0402703,2016YFC0400909)资助
【分类号】：P338

【相似文献】