高速公路雾天能见度预测方法

Forecasting Method of Highway Visibility in Foggy Weather

以多要素气象检测器采集的样本数据为基础,将温度、风速及湿度作为输入变量以及雾天能见度作为输出变量,分别采用三层结构BP神经网络和支持向量机非线性回归预测方法,建立雾天能见度的预测模型;将预测结果与实际数据进行对比分析的结果表明:BP神经网络和支持向量机均能较好地预测雾天能见度,其中BP神经网络和支持向量机模型预测值与实际值的相关性分别为0.895和0.978.支持向量机预测结果的误差更稳定,因而更适于处理非线性小样数据.

Taking temperature,wind speed and humidity as input variables and fog visibility as output variables,the forecasting model of highway visibility in foggy weather was established based on the sample data collected by multi-element meteorological detector with the three-layer BP neural network and nonlinear regression of support vector machine.The results showed that the visibility of the fog can be forecast by the BP neural network and the support vector machine,whose correlation values between the predictive and the actual were 0.895 and 0.978.The error of support vector machine prediction is more stable,so it is more suitable for dealing with small sample,nonlinear,high dimension and local minima.