摘要
针对柴油加氢脱硫生产过程中出现的工艺参数和产品质量难以精准控制的问题,提出粒子群优化(POS-BP)神经网络。基于中国石油大庆石化公司1 300 kt/a柴油加氢脱硫装置生产工艺操作台账数据,选取生产过程中的易波动工艺参数构建训练样本集合和测试样本集合,采用PSO-BP神经网络预测生产操作参数变化时精制柴油产品中硫含量的变化,将POS-BP神经网络与神经网络(BP)和遗传算法优化(GA-BP)神经网络进行横向预测效果比较。实验结果表明,BP神经网络预测的均方误差为2.66×10^(-3),GA-BP神经网络预测的均方误差为2.94×10^(-5),PSO-BP神经网络预测的均方误差为2.41×10^(-5);PSO-BP神经网络预测值与实际值最为接近,且预测结果较佳,具有较好的稳定性和泛化能力,能够精确预测生产操作参数变化时精制柴油产品中硫含量的变化。
Aimed at that the process parameters and product quality in the hydrodesulfurization of diesel oil were difficult to control accurately,the particle swarm optimization（PSO-BP） neural network was proposed. Based on the operation ledger of 1 300 kt/a diesel oil hydrodesulfurization unit of Daqing Petrochemical Co.,training sample set and the test sample set were constructed by selecting the fluctuating process parameters. The PSO-BP neural network was used to predict the change of sulfur content in the refined diesel oil products with the change of the operating parameters. The back propagation（BP） neural network and genetic algorithm optimization（GABP） neural network were also used to predicted the change of sulfur content to compare the performances of the three networks. The results showed that,the mean square errors of the of the BP,GA-BP and PSO-BP predictions were 2.66×10-3,2.94×10-5 and 2.41×10-5,respectively. So the values predicted by the PSO-BP neural network was the closest to the actual values and it had good stability and generalization ability.
作者
田景芝
杜晓昕
郑永杰
李郁
荆涛
Tian Jingzhi Du Xiaoxin Zheng Yongjie Li YU Jing Tao(College of Chemistry and Chemical Engineering, Qiqihar University, Qiqihar Heilongjiang 161006, China College of Computer and Control Engineering, Qiqihar University, Qiqihar Heilongjiang 161006, China)
出处
《石油化工》
CAS
CSCD
北大核心
2017年第1期62-67,共6页
Petrochemical Technology
基金
黑龙江省自然基金项目(B201422)
关键词
人工神经网络
硫含量
柴油
加氢脱硫
artificial neural networks
sulfur content
diesel oil
hydrodesulfurization
