多模式集成的RBF神经网络天气预报

摘要

针对复杂庞大的多模式数值预报数据，提出一种径向基函数(RBF)神经网络集成天气预报模型。根据天津市预报站点采用的WRF模式、RUC模式等数值预报数据的特点，将多种单模式数据作为RBF神经网络输入，网络输出为集成预报结果。实验表明：RBF神经网络集成预报模型降低了单模式预报误差，更加贴近了真实数据，并且在稳定性和实效性方面均有良好表现。%An integrated forecast model based on radial basis function(RBF)neural network was proposed for large com-plex multi-model numerical forecasting data. According to the characteristics of the numerical model forecast data of WRF model and RUC model used in Tianjin,numerical data of several models were chosen as the input of the RBF neural network,and the output is the integrated result. Experiments of temperature integration show that the RBF neural network integration method can reduce the error of the single model. The integrated result does good work in simulating real data. The method also has stability and effectiveness.