基于BP神经网络的船舶主机能效状态评估  

Evaluation of main engine energy efficiency based on BP neural network

作　　者：陈伟南[1] 黄连忠[1] 张勇[1] 路通 CHEN Weinan;HUANG Lianzhong;ZHANG Yong;LU Tong(College of Marine Engineering,Dalian Maritime University,Dalian 116026,China)

机构地区：[1]大连海事大学轮机工程学院,辽宁大连116026

出　　处：《中国舰船研究》2018年第4期127-133,160,共8页Chinese Journal of Ship Research

基　　金：中央高校基本科研业务费专项资金资助项目(3132016336);工信部高技术船舶科研计划项目(工信部联装(2014)508号)

摘　　要：[目的]在船舶航行期间,需要通过分析船舶和主机的运行参数来客观判断主机当前的工作情况,从而准确评估主机的能效状态。[方法]以状态良好的船舶运行记录为样本,结合主成分分析法和BP神经网络算法,构建船舶的航行状态识别模型和主机油耗模型,并在船舶航行期间对船舶实时运行参数进行分析,得出船舶主机在当前工况下的油耗量正常值。以某30万吨级远洋散货船为例开展模型计算验证,将正常油耗值与实际油耗值进行对比,以二者的残差值为依据,进而评估当前的主机能效状态。[结果]计算结果显示,航行状态识别模型的正确率为98.05%,油耗模型的平均误差为3.47%,2种模型的可靠性均较高。[结论]研究成果可为智能船舶的能效管理提供一定的参考。[Objectives]During the voyage of a ship,the operating parameters of the ship and its main engine need to be analyzed so as to objectively judge the current working state of the main engine and accurately evaluate its energy efficiency.[Methods]Taking good ship operation records as samples,and combined with principal component analysis and the BP neural network algorithm,a ship navigation state identification model and main engine fuel consumption model are built.During the voyage of a ship,these two models are used to analyze the real-time operating parameters of the ship in order to obtain the normal value of the main engine fuel consumption under current working conditions.The model of a 300 000-ton ocean bulk carrier is used to calculate and verify the models,the normal value is compared with the actual fuel consumption value of the main engine,which is based on the residual value of the two,and the current energy efficiency state of the diesel engine is then evaluated.[Results]The validation results show that the correct rate of the navigation state identification model is 98.05%and the average error margin of the fuel consumption model is 3.47%,thus proving the two models to be more reliable.[Conclusions]The results of this research can provide references for the energy efficiency management of intelligent ships.

关 键 词：BP神经网络 主成分分析法 油耗模型 船舶能效管理 

分 类 号：U676.41[交通运输工程—船舶及航道工程]