摘要
风能如今在电网系统总容量中占有很大比例,但由于风力涡轮机频繁故障,导致非计划停机时间,从而带来高运行维护成本,风机故障检测是解决方法之一。为了解决其中因故障日志缺失和不适宜的特征选择问题,使用DBSCAN聚类算法对原始数据进行异常聚类,将数据标记成正常和故障两类。其次,提出基于LSTM模型进行故障检测,无需手动提取特征,学习多变量时间序列数据中隐藏的依赖关系,使得整个模型有较好的故障检测能力。
Nowadays,wind turbines account for a large proportion of the total power grid outage time,and therefore one of the solutions to the problem is that the wind turbine takes up a large proportion of the total power grid outage time,which leads to the maintenance cost.In order to solve the problem of missing fault log and unsuitable feature selection,DBSCAN clustering algorithm is used to cluster the original data ab⁃normally,and the data are marked as normal and fault.Secondly,the fault detection based on LSTM model is proposed,which does not need to extract features manually,and learn the hidden dependency relationship in multivariate time series data,so that the whole model has better fault detection ability.
作者
胡翔
殷锋
袁平
HU Xiang;YIN Feng;YUAN Ping(College of Computer Science,Sichuan University,Chengdu 610065;College of Computer Science and Technology,Southwest University for Nationalities,Chengdu 610041;College of Mathematics and Information Engineering,Chongqing University of Education,Chongqing 400067)
出处
《现代计算机》
2021年第8期36-40,共5页
Modern Computer
基金
四川省科学技术厅省重点科技专项(No.2020YFG0254)。
