用神经网络方法对雷达资料进行降水类型的分类  

Precipitation Echo Classification of Radar Reflectivity with Artificial Neural Network

作　　者：王静[1] 程明虎[1] 

机构地区：[1]中国气象科学研究院,北京100081

出　　处：《气象》2007年第7期55-59,I0001,共6页Meteorological Monthly

摘　　要：利用不依赖先验统计模型的多层前馈神经网络模型对合肥的新一代S波段A系列雷达2001—2003年的降水资料进行了三种降水类型的分类,并将训练完成后的网络应用于一次降水过程。利用单隐层的多层前馈神经网络模型,在取适当参数时,已经可以较好地对雷达资料进行对流云降水、层状云降水和混合云降水三种降水类型的分类。同时验证了:训练集样本的数量和顺序、隐层神经元的数目以及学习率的选择等都将影响分类的成功率。A Back-Propagation （BP） Model of Artificial Neural Network （ANN） is used for the partitioning of radar reflectivity into convective and stratiform-cloud precipitation classifications with the CINRAD-SA data from 2001 to 2003 in Hefei. The trained ANN is applied in a precipitation process. It is proved that the single hide-layer BP model of ANN can be used to classify the different precipitation echoes with a high success-rate. It is also validated that： the success-rate is influenced by following factors： the amount and the in-put-order of the training-database, the nerve cell number of the hided-layer and the choice of the learning rate.

关 键 词：神经网络 雷达资料 降水类型 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]