摘要
提出一种基于DBSCAN特征聚类的改进随机森林特征选择算法对SOFC系统的故障进行定位。该方法通过DBSCAN聚类算法挖掘变量间的相关关系,将变量聚类,再挑选出特征选择结果。实验表明:该方法不仅可以筛选出与故障强相关的特征,还能尽可能地减少特征间的冗余,可以高效、快速、准确地对故障进行定位。
A DBSCAN feature clustering-based improved random forest feature selection algorithm was proposed to locate faults of SOFC system.Through making use of DBSCAN clustering algorithm to explore the correlation among variables,clustering the variables and then sorting out the results of feature selection were implemented.The experimental results show that,this method can select features which are strongly related to faults and can reduce redundancy between features as much as possible and it can locate faults efficiently,quickly and accurately.
作者
秦超
杨煜普
QIN Chao;YANG Yu-pu(School of Electronic Information and Electrical Engineering,Shanghai Jiaotong University)
出处
《化工自动化及仪表》
CAS
2019年第5期371-376,共6页
Control and Instruments in Chemical Industry
基金
国家自然科学基金项目(51777122,61273161)