基于数据分析技术的水文组合预报应用研究

Research on applications of data analysis technology to hydrological combined forecasting

水文组合预报方法是对多种预报模型的预报结果进行组合分析的一种预报方法.针对历史洪水数据丰富和相对贫乏两种情况,分别提出基于多目标模糊优选和基于贝叶斯分析的组合预报模型.前者是选用多目标模糊优选模型根据各预报方案在不同流量级别下的预报精度确定方案优属度,而后加权平均的组合预报模型;后者则是以贝叶斯分析为基础,同时结合专家经验、马尔可夫蒙特卡罗模拟、Gibbs抽样法,并引入实时校正的组合预报模型.以嫩江流域为实例,分别对两种组合预报模型的精度进行了验证.验证结果表明:两种模型可行而且实用,预报精度均明显高于单个模型的预报精度.

Hydrological combined forecasting is a method giving summary and analysis to different forecasting results, produced by different predication models. Aiming at the two situations-abundance or lack of history flood data, the combined forecasting models separately based on multi-objective fuzzy optimization theory and Bayesian analysis theory are proposed correspondingly. The former model introduces multi-objective fuzzy optimization theory to find out optimal relative membership degree of each projection on some precision at different discharges, and then by means of weighted average to confirm the optimal forecasting result; the latter model is based on Bayesian theory, combined with experts＇ experiences, MCMC simulation, Gibbs sampling and real-time auto-tuning technology. Taking the drainage area of Nenjiang for instance, the precision of the two integrated models was tested, and the result indicates that the established models are available and practical, with higher precision than that of any single model.