改进PSO算法优化LSSVM的柴油机排气预测

Diesel Engine Exhaust Prediction Based on Improved PSO Algorithm Optimized LSSVM Model

针对柴油机SCR系统对排气流量、排气成分和排气温度控制精度要求高的特点,提出一种基于粒子群优化支持向量机的排气预测模型。该方法采用改进粒子群算法寻找支持向量机最优的惩罚参数和核函数参数,提高模型的泛化能力,根据柴油机运行时的转速和负载,实时精准测算出排气流量、温度以及氮氧化合物浓度。结合柴油机实际排放实验仿真表明,与未经优化的PSO-SVM模型相比,该方法对柴油机排气预测有很高的精确度,可以将误差控制在1.6%以内,平均误差仅为0.665%。

In view of the high demand of SCR system of diesel engine on exhaust flow,exhaust composition and exhaust temperature accuracy,a kind of support vector machine exhaust prediction model based on particle swarm optimization is presented.In this method,the particle swarm algorithm is used to find the optimal penalty parameters and kernel function parameters of the support vector machine,which makes the model show better generalization ability,so as to accurately measure the flow rate,temperature and concentration of nitrogen and oxygen compounds in real time according to the rotating speed and load of the diesel engine.Combined with actual emissions from diesel engines,experimental simulation show that compared with the LSSVM model without optimization,this method has high accuracy in the prediction of diesel exhaust and can control the error within 1.6%and the average error is only 0.665%.