基于WOA-RF算法的船舶柴发配电系统故障诊断  

Fault diagnosis of ship diesel power distribution system based on WOA-RF algorithm

作　　者：李维波 高峰 肖朋[1] 黄康政 阮道杰 高俊卓 LI Weibo;GAO Feng;XIAO Peng;HUANG Kangzheng;RUAN Daojie;GAO Junzhuo(School of Automation,Wuhan University of Technology,Wuhan 430070,China;College of Electrical Engineering,Northwest Minzu University,Lanzhou 730124,China)

机构地区：[1]武汉理工大学自动化学院,湖北武汉430070 [2]西北民族大学电气工程学院,甘肃兰州730124

出　　处：《中国舰船研究》2025年第2期77-88,共12页Chinese Journal of Ship Research

基　　金：国家重点研发计划项目(2020YFB1506802);国家自然科学基金面上项目(51977164);湖北省科技计划项目(2024BAB067)。

摘　　要：[目的]船舶柴发配电系统对航行稳定性至关重要,海洋工作环境的严苛性致使其故障频发,为此提出一种基于鲸鱼优化算法的优化随机森林(WOA-RF)算法,用以开展船舶柴发配电系统故障诊断。[方法]首先,基于Matlab/Simulink仿真软件搭建船舶柴发配电系统模型,采集其故障工况和正常工况的数据;然后,对收集的数据进行预处理以提取时域特征,并使用随机森林算法提取重要特征,从而减少数据维度;最后,使用WOA优化后的随机森林模型对船舶柴发配电系统运行数据进行故障识别、诊断和分类。[结果]仿真模拟试验表明:采用WOA-RF算法识别故障状态和正常状态的准确率为100%,区分12种故障类型的诊断准确率为98.26%;在原始数据集中,与9种不同算法对比,WOA-RF算法的准确率最低提升了4.86%,最高提升了34.37%;在添加10dB噪声数据后,与6种不同算法对比,WOA-RF算法的准确率最低提升了2.43%,最高提升了18.40%。[结论]基于WOA-RF算法的故障诊断方法在复杂海洋环境下展示了优异的准确性和鲁棒性,结果可为船舶电力系统故障的可靠识别提供参考。[Objective]The marine diesel generator(DG)power distribution system is crucial for ship navigation.However,due to the harsh marine environment,frequent failures occur.Therefore,a fault diagnosis method based on whale optimization algorithm-optimized random forest(WOA-RF)is proposed for the marine DG power distribution system.[Methods]The marine DG power distribution system model is built using Matlab/Simulink simulation software.First,fault and normal condition data are collected.Then,the collected data is normalized,time-domain features are extracted,and important features are selected using random forest to reduce data dimensionality.Finally,the WOA-optimized random forest model is used for fault identification,diagnosis and classification.[Results]Simulation results show that the WOA-RF method can identify fault and normal states with 100% accuracy.It can classify 12 different fault types with an accuracy of 98.26%.In the original dataset,the accuracy of WOA-RF improved by at least 4.86% and by up to 34.37% compared to nine different algorithms.In the dataset with 10 dB noise,the accuracy of WOA-RF improved by at least 2.43% and by up to 18.40% compared to six different algorithms.[Conclusion]The WOA-RF-based fault diagnosis method demonstrates superior accuracy and robustness in complex marine environments,providing a reliable solution for fault identification in marine power systems.

关 键 词：船舶柴发配电系统 故障分析 故障诊断 鲸鱼优化算法 随机森林算法 SIMULINK模型 特征提取 

分 类 号：U665.14[交通运输工程—船舶及航道工程]