Wind Turbine Condition Monitoring Based on Improved Active Learning Strategy and KNN Algorithm

Abstract

As the damage of the gearbox of wind turbines (WTs) will cause economic losses, it is necessary to conduct online condition monitoring (CM) on the gearbox. Most WTs are equipped with SCADA system, and CM method based on Supervisory Control and Data Acquisition (SCADA) data is one of the most economical methods. K Nearest Neighbor (KNN) algorithm has good robustness, and WTs are typical nonlinear objects. Based on this, KNN regression model is established for CM, and Distance Correlation (DC) coefficient is used to select modeling variables to improve the shortcomings of traditional feature selection algorithm. A large amount of redundant data will be generated during the operation of WTs, and the efficiency of KNN algorithm is affected by the size of training set. Therefore, an active learning (AL) algorithm combining multiple strategies is proposed to select high-quality training data. The validity of the proposed method is verified by the data of an actual WT. The experimental results show that the method presented in this paper performs well in the comparative experiments, and the online CM results are about 20 days earlier than the SCADA system.

Keywords

• Wind turbine gearbox
• condition monitoring
• K-nearest neighbor algorithm
• active learning
• distance correlation coefficient