基于粗糙RBF神经网络的气象数据插补方法研究

Meteorological data interpolation method based on rough RBF neural network

针对实时气象数据因通信和设备故障等引起要素值缺失的问题,建立了基于粗糙集-径向基神经网络的插补模型。该模型以单个站点的相对湿度缺失为例,利用粗糙集理论约简气象影响因素,以关键因素作为径向基神经网络的输入,进行缺失数据的插补。实验结果表明,该模型插补效果较线性插值法精度有显著提高,为实时气象数据缺失提供了有效的处理方法,同时为建立连续气象数据集奠定了基础。

Realtime meteorological data always have elements-value missing problem due to the communication and equipment failure, etc. For this problem, the interpolation model is established based on rough set-radial basis function neural network. In this model, the single site＇s relative humidity missing is taken as an example, first the rough set theory is used to reduce meteo rological influence factors, then the key factor is taken as the input of the radial basis function neural network, finally the inter polation of missing data is carried on. The experimental results show that, the model＇s interpolation effect is improved signifi cantly compared to the accuracy of the linear interpolation method. An effective processing method is provided to solve the real time meteorological data missing problem and to laythe foundation for the establishment of the continuous meteorological data set.