基于改进的PSO优化SVR的机场道面积冰预测  

Prediction of Airport Road Surface Icing Based on Improved PSO Optimized SVR

作　　者：王立文[1,2] 刘建华 陈斌[1,2] 李庆真 WANG Li-wen;LIU Jian-hua;CHEN Bin;LI Qing-zhen(Institute of Aviation Ground Special Equipment,Civil Aviation University of China,Tianjin 300300,China;College of Electronic Information and Automation,Civil Aviation University of China,Tianjin 300300,China)

机构地区：[1]中国民航大学航空地面特种设备研究基地,天津300300 [2]中国民航大学电子信息与自动化学院,天津300300

出　　处：《计算机仿真》2021年第9期45-50,89,共7页Computer Simulation

基　　金：国家自然科学基金委员会与中国民用航空局联合资助项目(U1933107);中央高校基本科研业务费资助项目(201938)。

摘　　要：针对冬季机场道面积冰厚度的问题,建立了粒子群优化支持向量回归参数的预测模型。并且针对粒子群算法在参数寻优的过程中易陷入局部最优和过早收敛的问题,对经典学习因子寻优的范围进行了改进并引入惯性权值。在人工气象室中搭建模拟机场道面积冰环境实验系统,通过正交实验法进行实验方案设计,根据极差的大小判断影响积冰厚度的主次顺序为:降雨量-温度-风速-湿度。经过模型仿真结果的均方误差对比,基于改进的PSO优化SVR的模型预测精度高于常规的PSO-SVR模型和BP神经网络。Aiming at the ice thickness problem of the airport runway area in winter,a prediction model of support vector regression parameters based on particle swarm optimization was established.In order to solve the problem of local optimum and premature convergence in the process of parameter optimization of particle swarm optimization,the range of classical learning factor optimization was improved and inertia weight was introduced.An experimental system for simulating the ice environment of the airport runway area was built in the artificial weather room.The experimental scheme was designed by the orthogonal experiment method.According to the magnitude of extreme difference,the order affecting ice thickness is rainfall-temperature-wind speed-humidity.Comparison of mean square error of model simulation results that the prediction accuracy of the model based on improved PSO optimization SVR is higher than that of the conventional PSO-SVR model and BP neural network.

关 键 词：机场道面 积冰预测 支持向量回归 粒子群优化 正交实验 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]