基于诊断比值的Logistic回归分析对中东原油的鉴别  

The Logistic Regression Based on Diagnostic Ratio for Identification of the Middle East Crude Oil

作　　者：黄义 王思童 张婷婷 公维民[1] 齐超越 李佳 刘晓星[1] 

机构地区：[1]大连海事大学环境科学与工程学院,辽宁大连116026

出　　处：《环境科学与技术》2017年第10期66-70,共5页Environmental Science & Technology

基　　金：中央高校基本科研业务费专项资金资助项目(01760516)

摘　　要：原油的诊断比值C_(17)/Pr、C_(18)/Ph和Pr/Ph具有良好的抗风化性,可作为海上溢油鉴定的重要参数。以诊断比值为参数考察了18种原油之间的修正余弦相似度,中东原油之间的相似度均值为0.76,表现出较高的相似性,而中东与非中东原油之间的相似度均值为-0.42,两者差异显著。以诊断比值为变量建立了判别中东原油的二元Logistic回归模型,C统计量为0.99(>0.90),表明诊断比值和模型预测概率之间的相对一致性很好。该模型对风化30 d油样的鉴别准确率为100%,对文献报道的9种原油的鉴别准确率也达到100%,它不仅适用于未风化原油的鉴别,也适用于海上短期风化原油的鉴别。Diagnostic ratio of crude C（17）/Pr,C（18）/Ph and Pr/Ph has good resistance to weathering and is an important parameter for identification of oil spills into the sea. Adjusted cosine similarity among eighteen types of crude oil was investigated with diagnostic ratio as a parameter. The Middle East crude oils show high similarity because their similarity average value is 0.76,while the Middle East crude oil and non-Middle East crude oil have significant differences since their similarity average value is-0.42. A binary Logistic regression model was established to identify the Middle East crude oil using diagnostic ratio as variables,C statistic is 0.99（0.90）,indicating there is good relative consistency between diagnostic ratio and model forecast probability. By this model,the identification accuracy is 100% for oil samples weathered for 30 days,and for 9 types of crude oil reported in the literature. This model is applicable not only for identification of crude oil,but also for crude oil weathered on the sea during short-term.

关 键 词：LOGISTIC回归分析 诊断比值 修正余弦相似度 中东原油 风化 

分 类 号：X74[环境科学与工程—环境工程]