基于孤立森林方法的催化裂化装置排污数据异常识别  

Detection of Pollution Emission Data Anomaly of FCCU Based on Isolated Forest Algorithm

作　　者：陈冲 何为 钟田福 王晶[1] CHEN Chong;HE Wei;ZHONG Tianfu;WANG Jing(School of Information Science and Engineering,China University of Petroleum (Beijing),Beijing 102249,China;HSE Testing Center,CNPC Safety and Environmental Protection Technology Research Institute Co.,Ltd.,Beijing 102206,China)

机构地区：[1]中国石油大学(北京)信息科学与工程学院,北京102249 [2]中国石油集团安全环保技术研究院有限公司HSE检测中心,北京102206

出　　处：《西安石油大学学报（自然科学版）》2021年第4期119-126,共8页Journal of Xi’an Shiyou University（Natural Science Edition）

基　　金：中国石油天然气集团有限公司直属院所基础科学研究和战略储备技术研究基金(2017D-5008)。

摘　　要：基于分裂准则与相对质量算法对孤立森林算法进行了改进,优化了孤立森林算法的分支步骤与局部度量方式。利用标准数据集(Shuttle、Satellite、Annthyroid)验证了算法的有效性,并分析了算法复杂度;应用改进的孤立森林算法对某炼化企业催化裂化数据进行异常识别,与经典的孤立森林算法、SCiForest及ReMa-iForest的异常识别效果进行对比分析。结果表明:改进的孤立森林算法能够在提高模型精确度的同时提升算法的执行效率。研究提出的异常识别方法将为催化裂化监测数据的分析提供可靠数据支撑。The iForest algorithm is improved by combining split criterion and relative mass which provide novel branching scheme and local ranking measure,respectively.The effectiveness of the improved iForest algorithm is validated using standard datasets(Shuttle,Satellite,Annthyroid)and the complexity of it is analyzed.The data anomaly of FCC in a refinery is identified using the improved iForest algorithm,and the identified result are compared with those using several basic algorithms(iForest,SCiForest and ReMa-iForest).It is shown that the improved iForest algorithm can improve the accuracy of the model and the efficiency of the algorithm.The proposed data anomaly identification method will provide reliable data support for the analysis of FCC monitoring data.

关 键 词：异常识别 孤立森林 催化裂化 污染外排 

分 类 号：TE624.4[石油与天然气工程—油气加工工程]