甲醇-柴油反应活性控制压燃发动机排放特性与经济性预测及优化

Prediction and Optimization of Emission Characteristics and Fuel Economy for Methanol-Diesel RCCI Engine

基于甲醇-柴油双燃料反应活性控制压燃(reactivity controlled compression ignition,RCCI)发动机台架试验数据,建立了基于粒子群算法(particle swarm optimization,PSO)优化BP神经网络的甲醇-柴油RCCI发动机排放特性、经济性预测智能模型,以发动机负荷、甲醇替代率、EGR率为输入参数,NO_(x)、烟度、CO、THC排放和当量有效燃油消耗率为输出,预测模型的决定系数(R^(2))分别为0.99,0.97,0.99,0.98和0.96,平均绝对百分比误差(mean absolute percentage error,MAPE)为6.46%,0.56%,3.12%,1.21%和0.3%,表明构建的PSO-BPNN模型能够有效预测甲醇-柴油RCCI发动机的NO_(x)、烟度、CO、THC排放和经济性。基于偏最小二乘回归(partial least squares regression,PLSR)分析了不同控制参数对发动机污染物排放和经济性的相关性,将PSO-BPNN预测模型与NSGA-Ⅱ算法相结合,以NO_(x)、烟度和当量有效燃油消耗率为目标对负荷、甲醇替代率和EGR率进行协同优化,将最优控制参数组合标定至双燃料发动机的控制系统进行试验验证。结果表明:优化后烟度变化不明显,NO_(x)排放平均降低39.6%,当量有效燃油消耗率平均降低2.6%。

Based on the test data of methanol-diesel dual-fuel reactivity controlled compression ignition(RCCI)engine bench,the prediction intelligent model of emission and fuel economy was established based on BP neural network optimized by the particle swarm optimization(PSO)algorithm.The inputs of model were engine load,methanol replacement rate and EGR rate,and the outputs of model were NO_(x),smoke,CO,THC emission and equivalent effective fuel consumption.The determination coefficients of model were 0.99,0.97,0.99,0.98 and 0.96 respectively,and the mean absolute percentage errors were 6.46%,0.56%,3.12%,1.21%and 0.3%,which indicated that the constructed PSO-BPNN model could effectively predict the NO_(x),smoke,CO,THC emissions and economy of methanol-diesel RCCI engine.Based on the partial least square regression(PLSR),the correlation of different control parameters to engine pollutant emissions and economy was analyzed.The PSO-BPNN prediction model was combined with the NSGA-Ⅱalgorithm,the collaborative optimization of load,methanol replacement rate and EGR rate was conducted based on the objective of NO_(x),smoke and equivalent effective fuel consumption,and the optimal control parameter combination was finally calibrated to the control system of methanol-diesel dual-fuel engine for the experimental verification.The results showed that the optimized smoke had little change,NO_(x)emissions reduced by 39.6%,and the equivalent effective fuel consumption reduced by 2.6%.