船舶溢油事故等级评估方法比较  

Comparative Research on Evaluation Methods of the Ship Oil spill Accident Threaten Degree

作　　者：张欣[1] 施欣[1] 

机构地区：[1]上海海事大学交通运输学院,上海200135

出　　处：《环境科学研究》2007年第6期164-168,共5页Research of Environmental Sciences

基　　金：上海市基础研究重点项目(05JC14073);国家自然科学基金资助项目(70273019)

摘　　要：船舶溢油事故等级评估是船舶溢油应急计划研究的重要内容,它受到溢油量、事故发生位置等多种因素的影响,是复杂的多属性分类问题.利用熵值法、关联度法和BP神经网络法3种目前常用的事故等级评估方法,对上海港1984─2003年发生的12起船舶溢油事故进行了危害等级判定,以期探索溢油事故等级合理量化评估技术方法.对比了3种方法在结果一致性、误差精度上的优劣性及实际适用性.结果表明:BP神经网络和关联度法适用性较好,熵值法必须结合专家打分才能得到较好的判定结果;同时BP神经网络法的评估结果一致性比率较高,误差率较低,但存在建模成本高和对噪声较敏感的缺点.As one of the key part of ship oil spill emergency plan, the evaluation of ship oil spill accident threaten degree （TD） which is influenced by various factors such as oil spill amount and accident location, is an complicated multi-attribute classification issue. Basing on three current popular evaluation methods including entropy method, related degree analysis method and BP neural network, a TD classification was made for 12 ship oil spill accidents which happened in Shanghai Port from 1984 to 2003. Through comparative analysis, the advantages and disadvantages of each method were discussed in result accordance and error. It is found out that the applicability of BP neural network and related degree analysis method are much better than entropy method. Additionally, the result of BP neural network is closer to the fact situation and its error is also low. But the modeling cost of BP neural network is very high and the model is easily influenced by system noise.

关 键 词：安全评估 熵值法 关联度法 BP神经网络 船舶溢油事故 

分 类 号：X736.3[环境科学与工程—环境工程]