基于遗传算法与神经网络的费托合成负荷优化  

Loading optimization of Fischer-Tropsch synthesis using artificial neural networks and genetic algorithm

作　　者：王瑞航 丁文瑶 温润娟 廖祖维[1] 李虎 郭中山 WANG Ruihang;DING Wenyao;WEN Runjuan;LIAO Zuwei;LI Hu;GUO Zhongshan(College of Chemical Engineering and Biological Engineering,Zhejiang University,Hangzhou 310058,China;CHN ENERGY Ningxia Coal Industry Co.Ltd.,Yinchuan 750411,China)

机构地区：[1]浙江大学化学工程与生物工程学院,浙江杭州310058 [2]国家能源集团宁夏煤业有限责任公司,宁夏银川750411

出　　处：《高校化学工程学报》2023年第4期608-614,共7页Journal of Chemical Engineering of Chinese Universities

基　　金：宁夏回族自治区重点研发计划(2019BFH02016)。

摘　　要：生产负荷是大型费托合成流程的重要指标。高负荷运行不仅可以提高产能,还可以降低综合能耗。针对费托合成流程工艺复杂、最佳生产操作条件难以确定的问题,提出了以数据驱动方式破除大型费托合成流程负荷难以提高的瓶颈。以400万t×a^(-1)费托合成流程为研究对象,建立了负荷预测的神经网络模型,利用遗传算法优化了神经网络的权值和阈值初值,提高了模型预测准确度。再次利用遗传算法对神经网络输入变量进行全局寻优,确定了负荷最大时各项输入变量取值,优化后最大负荷提高了11.5%。针对6个输入变量的灵敏度分析表明,氢气温度与循环气流量对负荷影响较大,这2个变量在模型取值区间内变化时负荷分别改变了13.9%和13.7%,对负荷的提高提出了指导性建议。Production load is an important index of large-scale Fischer-Tropsch synthesis processes.High load operation can improve productivity and reduce comprehensive energy consumption.A data-driven method was used to improve load of large-scale Fischer-Tropsch synthesis processes which have problems of complicated processes and challenging optimization.A neural network model for load prediction was established based on a 4 million tons per year Fischer-Tropsch synthesis process.Genetic algorithm was used to optimize the initial weights and thresholds of the neural network to improve the accuracy of model prediction.It was also used to optimize the input variables of the neural network,and the values were determined when the load was at its maximum.The load was increased by 11.5%after optimization.The sensitivity analysis of the six variables shows that hydrogen temperature and circulating gas flow rate have a great effect on load,and the changes of these two variables within the interval make the load change of 13.9%and 13.7%respectively,which provides suggestions for load increase.

关 键 词：费托合成 神经网络 遗传算法 模型 优化 

分 类 号：TQ[化学工程]