Identifying early defects of wind turbine based on SCADA data and dynamical network marker

Abstract

Defects Identification is of great significance to prevent wind turbines (WTs) accidents and improve its operation reliability, and it keeps challenging due to the complex relationship between internal faults and external observed data. A new method to identify early defects of WTs is presented based on Dynamical Network Marker (DNM) by adopting only the data of supervisory control and data acquisition (SCADA). In the presented method, WT is mapped into a multi-node complex network according to the corresponding relationship between the internal structure topology of WT and the monitoring variables of its SCADA system. Then the dominant nodes in the network under different states are screened out through the correlation and cross-correlation analysis of de-nosed SCADA monitoring data series and prediction data series to form a key subnetwork, and the dynamical network marker (DNM) is constructed as warning signal of defects of WT. The proposed method is tested with the SCADA data of a WT with known faults. The results illustrate that the proposed method can not only give an early warning signal when WT under defect state but also further determine the defect location. Moreever, the proposed method only needs the SCADA monitoring data of WT itself, which is convenient and easy to be popularized.