基于神经网络的天然气双燃料发动机性能预测

Performance Prediction of Natural Gas Dual-fuel Engine Based on Neural Network

基于柴油/天然气双燃料发动机测试台架试验数据,以发动机扭矩、转速、喷油时刻、喷油压力和天然气替代率、过量空气系数为模型输入参数,以发动机燃油消耗率和CO、NO_(x)、总烃(THC)排放体积分数和碳烟排放量为模型输出,构建了基于逆向传播(BP)神经网络和遗传算法(GA)优化的GA-BP神经网络的预测模型,并将预测结果与试验值进行对比验证。研究结果表明:GA-BP神经网络模型相比BP神经网络模型具有更好的预测性能;GA-BP神经网络模型对5个输出参数预测的平均绝对百分比误差M_(APE)均小于6%,并且决定系数R^(2)均大于0.97,模型具有较高的预测精度和泛化能力。

Based on the test data of diesel/natural gas dual-fuel engine,a prediction model of GA-BP neural network was established based on BP neural network and genetic algorithm optimization with engine torque,speed,fuel injection timing,fuel injection pressure,natural gas alternative fuel and excess air coefficient as input parameters,and the brake specific fuel consumption,CO,NO_(x),THC emissions and soot emission as outputs,and the predication results were compared with test values for verification.The research results show that GA-BP neural network model has better predictive performance than BP neural network model.The Mean Absolute Percentage Error(M_(APE))predicted by the GA-BP neural network model for the five output parameters is less than 6%,and the coefficient of determination R^(2)is greater than 0.97,and the model has high prediction accuracy and generalization ability.