A Predictive Fault Diagnose Method of Wind Turbine Based on K-means Clustering and Neural Networks

• DOI: 10.6138/jit.2016.17.7.20151027i
• Publication Date: 2016-12-01
• Authors: Xu Liu; Li Mei; Shiyao Qin; Xiaojing Ma; Wenzhuo Wang
• Keywords: SCADA; Data pre-processing

Maintenance costs can be greatly reduced by improving the prediction accuracy of wind turbine faults. In this paper, a new method combining K-means clustering and Neural Net classification methods are adopted to predict fault condition turbine. First, data preprocessing has been done with from SCADA (Supervisory Control And Data Acquisition) system. According internal association data, adopting unsupervised learning K-means, similar features transformed one cluster, after which BP is made based on converted clusters. Compared traditional methods, proposed mentioned above improve 3.5%, abnormal state mechanical faults determined more accurate timely degree.