改进CS优化支持向量回归的汽车热舒适性预测

Prediction of Automobile Thermal Comfort Based on Support Vector Regression Optimized by Improved CS

热舒适性指标PMV参数间互相迭代、计算复杂、不易实时预测,而采用支持向量回归(SVR)进行数据拟合时,预测效果易受SVR参数的影响。针对以上问题,提出一种改进布谷鸟算法(CS)优化SVR参数的PMV预测模型。改进CS算法采用自适应步长对Lévy全局随机游动的步长进行调节,并用非洲野狗算法(DOA)的生存行为替换CS算法偏好局部随机游动行为,以提高CS算法寻优能力。实验结果表明,CS优化SVR模型预测值的RMSE为0.00742,比DOA优化SVR模型的RMSE低0.00481;使用自适应步长或DOA算法改进CS优化SVR模型比CS优化SVR模型的RMSE分别降低了32.34%和40.83%;融合两种策略改进CS优化SVR模型预测值的RMSE整体降低了60.52%。融合两种策略改进CS算法优化SVR预测模型具有更高的PMV预测精度。

The thermal comfort index PMV parameters are iterative,complex to calculate,and difficult to predict in real time.When using support vector regression(SVR)for data fitting,the prediction effect is easily affected by the SVR parameters.To solve the above problems,a PMV prediction model based on improved cuckoo search(CS)is proposed to optimize SVR parameters.The improved CS algorithm uses an adaptive step size to adjust the step size of Lévy is global random walk,and replaces the CS algorithm’s preference for local random walk behavior with the survival behavior of the African wild dog algorithm(DOA)to improve the optimization ability of the CS algorithm.The experimental results show that the predicted RMSE of the CS optimized SVR model is 0.00742,which is 0.00481 lower than the RMSE of the DOA optimized SVR model.Using adaptive step size or DOA algorithm to improve the CS optimized SVR model reduces the RMSE by 32.34%and 40.83%,respectively,compared to the CS optimized SVR model;Integrating the two strategies to improve the prediction value of the CS optimized SVR model,the overall RMSE decreased by 60.52%.Integrating two strategies to improve CS algorithm to optimize SVR prediction model has higher PMV prediction accuracy.