人工神经网络在汽柴油混合加氢脱硫中的应用

Application of Artifical Neural Networks to Hydrodesulfurization of Gasoline and Diesel Mixtures

在温度320～360℃、空速1.2～2.0 h-1、氢油体积比350～550、压力6.0～8.5 MPa的条件下,采用Ni-Mo-P/Al2O3加氢精制催化剂在100 mL加氢评价装置上,对5种劣质汽柴油进行混合加氢脱硫评价。应用LMBP神经网络建立了用于预测汽柴油混合加氢脱硫率的模型,并应用LMBP神经网络考察了原料油性质和工艺条件对加氢脱硫反应的影响。实验结果表明,LMBP神经网络对脱硫率和脱硫反应温度的预测精度较高,平均相对偏差分别为0.55%和0.28%;原料油性质对加氢脱硫影响大小的顺序为:密度>溴值>90%馏出点>氮含量>硫含量>运动黏度,工艺条件对加氢脱硫影响大小的顺序为:温度>空速>氢油比>压力,为汽柴油混合加氢脱硫工艺条件的优化提供了指导。

The hydrodesulfurization（HDS） of the five mixtures of low-quality gasoline and diesel over Ni-Mo-P/Al2O3 catalyst were carried out in a 100 mL high-pressure continuous flow fixed- bed device under the conditions of 320-360 ℃, WHSV 1.2-2.0 h^-1, volume ratio of hydrogen to oil 350-550 and 6.0-8.5 MPa. A predicting model for the HDS was established by means of LMBP neural network. The predicted results show that the average relative errors of the desulfurization rate and reaction temperature were 0.55% and 0.28% respectively, which can meet the industrial prediction requirement. The influences of the feedstock properties on HDS are in order of density 〉 bromine value 〉 90% distillation point 〉 nitrogen content 〉 sulfur content 〉 kinematics viscosity, and the influences of reaction conditions are in order of temperature 〉 space velocity 〉 ratio of hydrogen to oil 〉 pressure, which can provide guidance for optimizing the HDS process conditions.