催化裂化MIP工艺原料油聚类研究

摘要

以催化裂化MIP装置工业数据为基础,选取原料油性质中的密度、饱和烃含量、芳烃含量、(沥青质+胶质)含量、镍含量、钒含量、残炭7个变量,分别采用K-means和FCM算法对原料油性质进行聚类.K-means聚类法将原料油性质的95组样本分为4类,FCM聚类法将原料油性质的95组样本分为5类.聚类结果中的每一类原料油特征都比较明显,表明K-means和FCM聚类法对于原料油性质的聚类分析均具有较好的适用性.以此为基础,可以针对每一类原料油建立相应的产品分布优化智能模型,为寻找使目的产品收率最大化的操作条件提供指导.%Based on the commercial data from a FCC MIP process unit,seven properties of feedstock including density,saturated hydrocarbons content,aromatics content,asphaltene plus resin content,nickel content,vanadium content and residue carbon were used to cluster for the feedstock oils for MIP process by K-means clustering method and fuzzy c-means clustering(FCM)method,respectively.The ninety-five data of feedstock properties were classified into four categories by K-means clustering algorithm and into five categories by FCM clustering method.The results showed that the characteristic of every category of the feedstock oils is obvious,indicating the good applicability to the feedstock properties by each method.On the basis of the works above,the product distribution intelligent model for every category of feedstocks can be established to find the optical operation conditions for MIP process.