论文部分内容阅读
南京炼油厂计划把原有的柴油手工调和工艺改为自动化调和工艺 ,实现对调合组分配方的优化控制。建立能适应调合工况变化、并能正确测定调和柴油质量指标的数学模型是关键。根据实测的南京炼油厂柴油调合生产数据 ,建立了以常二线、常三线、催化柴油和精制柴油 4路组分柴油的流量和倾点为输入参数 ,调和柴油倾点为输出参数的柴油调合RBF神经元网络模型。与回归模型和BP神经网络模型比较 ,RBF模型无论对训练数据集还是对检验数据集 ,均能更精确地预测实际生产过程的调合柴油产品倾点值 ,显示了良好的工程应用前景。
Nanjing refinery plans to hand the original reconcile the diesel process to automatically reconcile the process to achieve the optimal control of blending components formula. Establishing a mathematical model that can adapt to changes in blending conditions and correctly measure the quality of reconciled diesel fuel is key. According to the measured diesel blending production data of Nanjing refinery, the flow and pour point of the diesel of the second, the third, the third, the catalytic diesel and the refined diesel are set as the input parameters, and the diesel pitch Combined RBF neural network model. Compared with the regression model and the BP neural network model, the RBF model can forecast the pour point of blended diesel products in actual production process more accurately both for the training dataset and the test dataset, and shows a good engineering application prospect.