基于改进PSO-SVR的气温预测模型研究

Research on Temperature Prediction Model Based on Improved PSO-SVR

近些年来的诸多研究结果表明机器学习在气象预测领域有着广阔的应用前景。文章使用支持向量回归(SVR)构建模型,预测气温要素变化情况,首先利用粒子群优化(PSO)算法优化SVR,标准PSO算法在寻找最优参数过程中有陷入局部最优的缺点,考虑在改变粒子权重因子(ANDVW)和搜索邻域(ANS)的基础上综合改进,尽量使寻优过程逼近全局最优;最后建立ANDVW-ANS-PSO-SVR模型进行实验对比。利用研究区累计10年间的连续气象观测数据进行模型训练和测试,实验结果表明,改进的ANDVW-ANS-PSO-SVR模型对气温的预测精度有一定的提高。

Many studies in recent years have shown that machine learning has expansive application prospects in the field of meteorological prediction.This paper uses support vector regression(SVR)to construct a model to predict the change of temperature elements.First of all,the particle swarm optimization(PSO)algorithm is used to optimize the SVR.The standard PSO algorithm has the disadvantage of falling into local optimization in the process of finding the optimal parameters,considering the comprehensive improvement on the basis of changing the particle weight factor(ANDVW)and search neighborhood(ANS),and try to make the optimization process approximate the global optimality.Finally,the ANDVW-ANS-PSO-SVR model was established for experimental comparison.Using the continuous meteorological observation data accumulated in the study area for 10 years,the experimental results show that the improved ANDVW-ANS-PSO-SVR model has a certain improvement in the prediction accuracy of temperature.