基于差分进化算法和相关向量机的车辆油耗预测

Vehicle Fuel Consumption Prediction Based on Relevance Vector Machine and Differential Evolution Algorithm

为了对车辆实际油耗进行预测,总结了车辆实际运行特征和设计参数中影响油耗的特征参数,并利用互信息方法确定了油耗敏感特征。以200辆车的试验结果为训练样本,以敏感特征和型式认证油耗为输入,车辆实际油耗为输出,建立了基于相关向量机的车辆油耗预测模型,并采用差分进化算法对模型参数进行优化。对建立的预测模型进行验证并与BP神经网络法进行了对比,结果表明:本文所建立的模型具有较高的预测精度和可靠性,优于BP神经网络等典型方法。

In order to predict the actual fuel consumption of vehicles,this paper firstly summarized vehicle actual operating characteristics and design parameters,which had influence on vehicle fuel consumption,and used mutual information approach to determine sensitive feature of fuel consumption.Secondly,it used200vehicles test results as training samples,with sensitive features and fuel consumption of type approval as the input parameters,and the actual vehicle fuel consumption as output parameters,then established vehicle fuel consumption prediction model based on relevance vector machine that was then optimized by the differential evolution algorithm.The predication model was verified and compared with BP neutral network method.The results show that the fuel consumption prediction model has higher prediction accuracy and reliability than some traditional methods such as BP neural network.