论文部分内容阅读
首先对ECMWF不同物理量场预报因子群进行自然正交展开,选取能充分反映每个预报因子场主要信息的第一主分量作为模型输入.进一步利用粒子群算法对支持向量回归机的相关参数进行优化,以南宁市8个气象站单站逐日降水作为预报对象,建立粒子群-支持向量回归集合预报模型,进行单站逐日降水的数值预报产品释用预报方法研究.利用模型对2015年5-6月南宁市8站进行了逐日降水预报业务试验,结果表明,模型具有较好的预报效果.并提出了利用隶属函数建立可信度函数对不同的预报模型进行评价.
Firstly, the forecasting factor groups of ECMWF physical fields are naturally and orthonormally expanded, and the first principal component which can fully reflect the main information of each forecast factor field is selected as the model input.Furthermore, the PSO algorithm is used to optimize the related parameters of SVR , A single station daily precipitation of 8 meteorological stations in Nanning City was used as a forecast object to establish a set of forecast models for particle swarm-support vector regression ensemble forecasting and single-station daily precipitation for numerical forecasting products.Using the model, The experiment of day-by-day precipitation forecast service in 8 stations of Nanning city in a month shows that the model has a good forecasting effect and proposes the use of membership function to establish the credibility function to evaluate different forecasting models.