A Kernel Entropy Method and its Application in Monitoring and Assessment of Wind Turbine Degradation Performance

Abstract

To overcome the problems of wind turbine (WT) degradation assessment, a new kernel entropy method based on supervisory control and data acquisition (SCADA) was proposed. This approach can be used to effectively monitor and assess WT performance degradation. First, a new condition monitoring method based on a kernel entropy component analysis (KECA) was developed for nonlinear data. Then, the squared prediction error (SPE) was used to monitor the WT health state. Due to the diversity and nonlinearity of SCADA data, fault features are easily overwhelmed by other vibration signals. To address this, a new kernel entropy partial least squares (KEPLS) algorithm was introduced. The proposed kernel entropy method improves the performance prediction by considering higher order information. Furthermore, changes in the prediction residual can be used to define certain limits to realize early warning of WT faults. Finally, the method was applied to actual SCADA data of a wind farm. The results show that the method can accurately evaluate the health state of WTs, thus verifying the effectiveness and feasibility of the proposed method.