加氢裂化过程多采样率数据建模方案研究  

Research on Multi-sampling Rate Data Modeling Scheme for Hydrocracking Process

作　　者：苏灿 Su Can(CNOOC Huizhou Petrochemical Co.,Ltd.,Huizhou Guangdong 516086,China)

机构地区：[1]中海油惠州石化有限公司,广东惠州516086

出　　处：《山西化工》2024年第11期142-144,共3页Shanxi Chemical Industry

摘　　要：为了通过控制过程参数提高石油炼化产品的质量,需要针对加氢裂化工艺构建多采样率软测量模型。研究过程构建了多采样率堆栈自编码器算法模型,介绍了该模型的网络结构、训练方法和软测量建模流程。在性能测试阶段,采集43个加氢裂化过程变量,针对柴油产品设置十六烷值和50%回收温度两个质量变量,通过实验确定了网络结构,与对照组算法进行性能对比,发现此次建立的算法模型达到了最低的均方根误差,质量预测效果最佳。In order to improve the quality of petroleum refining products by controlling process parameters,it is necessary to construct a multi sampling rate soft sensing model for hydrocracking process.The research process constructed a multi sampling rate stack autoencoder algorithm model,and introduced the network structure,training method,and soft sensing modeling process of the model.In the performance testing phase,43 hydrocracking process variables were collected,and two quality variables,cetane number and 50% recovery temperature,were set for diesel products.The network structure was determined through experiments,and the performance was compared with the control group algorithm.It was found that the algorithm model established this time achieved the lowest root mean square error and the best quality prediction effect.

关 键 词：加氢裂化 多采样率数据 软测量建模 

分 类 号：TE624.432[石油与天然气工程—油气加工工程]