神经网络技术在柴油加氢精制装置生产中的应用

针对柴油加氢精制过程的产品质量难以优化和预测的问题，提出了人工神经网络模型。根据国内某石化企业1.0 Mt/a柴油加氢精制装置生产操作数据，分别应用动量BP神经网络、LMBP神经网络和RBF神经网络建立了用于预测柴油加氢产品硫含量的模型。并对建立的RBF神经网络模型的泛化能力进行了考察。结果表明，动量BP神经网络、LMBP神经网络和RBF神经网络预测的平均相对误差分别为3.50%，2.30%，2.18%，RBF神经网络模型的预测性能最佳，且具有良好的泛化能力，能够在工艺操作参数变化时准确地预测柴油产品的硫含量，为柴油加氢精制装置的良好运行和优化操作提供了指导。

APPLICATION OF NEURAL NETWORK TECHNOLOGY ON DIESEL HYDROFINING UNIT

In view of the product quality was difficult to be predicted in diesel hydrofining process, the artificial neural network model was proposed. Based on the production operation data of 1.0 Mt/a diesel hydrofining unit in a petrochemical enterprise, the model for predicting sulfur content of diesel hydrogenation products were established by using momentum BP neural network, LMBP neural network and RBF neural network. The generalization ability of the RBF neural network model was also investigated. The results showed that, the average relative errors of the prediction of momentum BP neural network, LMBP neural network and RBF neural network are 3.50%, 2.30% and 2.18%, respectively. The RBF neural network model has the best prediction performance and good generalization ability. The RBF neural network could predict the sulfur content of the diesel product accurately when the process parameters changes. The work provides guidance for the better operation of the diesel hydrofining unit.