基于知识发现与动力统计集成的暴雨短期预报建模研究

The rainstorm is affected by many factors, whose relationship is very complex. And how to improve the heavy rainfall forecast ability is the worldwide difficult problem with considerable theoretical and applied significance for a long time. Especially, the rainstorm short-term forecast is very difficult. Currently, short-term forecast ability of rainstorm is weak and it is so far from the social demand. This project will do some innovative researches on rainstorm short-range forecasting method from a data driven and knowledge discovery perspective, in order to build a rainstorm short-range forecasting methodological framework named as Knowledge-power-statistical optimization integrating forecast, so as to improve the short-range forecasting ability of rainstorm. It mainly includes using some technologies with meteorological expert knowledge guidance, such as Gene-expression Programming, Wavelet Analysis, Manifold Learning and methods of Multivariate Time-Series Data Mining, to do following researches: (1) knowledge mining and modeling from the main relative factor and Physical Features of rainstorm evolution data,(2)a new intelligent nonlinear Interpretation model of numerical prediction product,(3)building a comprehensive forecast modeling integrated qualitative and quantitative based on (1)and(2),(4) studying and analyzing the uncertainty of model theory. The scientific significance of this project is that it would not only supply new effective methods both for the rainstorm short-range forecasting and other weather forecasting, but also promote the development of the research on rainstorm and disaster weather forecasting, data mining & knowledge discovery and some related weather forecast modeling theory and technology.

如何提高暴雨预报能力长期以来是一个理论性和应用性都很强的世界性难题。尤其是暴雨短期预报难度较大，当前预报能力较弱，离社会需求的差距较大。本课题拟从数据驱动与知识发现的角度进行暴雨短期预报方法的创新研究，以便构建一套知识-动力统计集成的新预报方法体系，提高暴雨预报能力。主要包括在气象专家知识的指导下利用基因表达式编程、小波分析、流形学习等理论和方法，及多元序列数据挖掘技术开展以下研究：（1）暴雨发生发展过程的关键要素场数据的演变规律和知识挖掘，及其预报建模研究；（2）数值预报产品的智能释用预报建模研究；（3）在（1）和（2）基础上以复杂性科学理论为指导，探究基于知识发现的知识－动力统计综合集成的新的暴雨短期预报建模理论和方法；（4）模型的不确定性理论分析研究。本研究对暴雨等灾害性天气预报、气象数据挖掘和预报建模等相关理论和技术研究的发展，有着重要的科学理论意义和应用参考价值。

如何提高暴雨预报能力长期以来是一个理论性和应用性都很强的世界性难题。尤其是暴雨短期预报难度较大，当前预报能力较弱，离社会需求的差距较大。本课题拟从数据驱动与知识发现的角度进行暴雨短期预报方法的创新研究，以便构建一套知识-动力统计集成的新预报方法体系，提高暴雨预报能力。主要包括在气象专家知识的指导下利用基因表达式编程、小波分析、流形学习等理论和方法，及多元序列数据挖掘技术开展以下研究：（1）降雨发生发展过程的关键要素场数据的演变规律和知识挖掘，及其预报建模研究；（2）数值预报产品的智能释用预报建模研究；（3）在（1）和（2）基础上以复杂性科学理论为指导，探究基于知识发现的知识－动力统计综合集成的新的暴雨短期预报建模理论和方法；（4）模型的不确定性理论分析研究。经过本课题组4年的共同努力，现已基本完成本项目的预期研究工作，在降雨关键要素场数据分析、挖掘与预报建模、数值预报产品的智能释用预报建模、气象信息智能分析处理、模型优化等方面获得了一系列优秀的理论成果，已发表高质量学术论文19篇，另有两篇在投稿中，申请相关发明专利3项，并已获授权1项。所获得的研究成果对暴雨等灾害性天气预报、气象数据挖掘和预报建模等相关理论和技术研究的发展，有着重要的科学理论意义和应用参考价值。

内容获取失败，请点击重试