基于IFOA-FSVM的WRF模式台风期阵风精细化预报

Precision forecasting of typhoon wind speed in WRF model based on IFOA-FSVM

为提高天气预报模式(Weather Research and Forecasting Model, WRF)输出中台风期阵风的预测精度,将WRF模式输出与某观测站实况数据相结合,提出一种台风期阵风精细化预报方法.针对影响台风风速的因素众多,而传统依据人工经验预判的风速存在较大误差的现状,该方法构建了台风期阵风预测的模糊支持向量回归模型,同时为解决模糊支持向量回归模型中惩罚因子C和核参数g难于确定的问题,将果蝇优化算法(Fruit Fly Optimization Algorithm, FOA)引入到模糊支持向量机(FuzzySupportVectorMachine,FSVM)的参数寻优中,并根据风速回归的特点,把果蝇优化算法引入到三维空间,结合增强因子γ以提高传统果蝇优化算法的全局寻优能力.实验结果表明,本文构建的模型预测风速与实际风速基本一致,相关性达到99%,不仅提高了WRF模式风速的适用性,而且风速预测精度明显优于传统FOA-FSVM和FOA-SVM方法,具有更强的泛化能力.

For the sake of improving the accuracy of forecasting wind speed during typhoon strike in WRF model forecast, a new method for precision forecasting of typhoon wind speed is proposed by combining the data collected from the WRF model forecast and an automatic observation station. The method incorporates many factors influencing the typhoon wind speed. The wind speed which is obtained using the traditional human prediction produces large error when compared with actual wind speed. To address this issue, a fuzzy support vector regression model for wind forecasting is built. Considering the fact that the fuzzy support vector regression model is not adequately efficient in determining the punishment factor C and kernel parameter g, the fly optimization algorithm is introduced into optimizing the parameters of the fuzzy support vector machine. According to the characteristics of the wind speed regression, the fruit fly optimization algorithm is developed in three dimensional space, combining with the enhancement factor - for improving the global optimization ability of traditional fruit fly optimization algorithm. The results show that the forecasting wind speed and the actual one is in good agreement with each other, and the correlation is as high as 99 %. The presented method of wind speed prediction provides higher accuracy than that of traditional FOA-FSVM model and FOA-SVM model.