A k-Nearest Neighbor Wind Turbine Fault Detection Method Based on Deep Metric Learning and Domain Feature Discrimination

Abstract

The reliability of wind turbines (WTs) is directly related to the safe and stable operation of wind farms. However, existing data-driven fault detection methods are challenging in coping with the complex operating conditions of WTs, which affects the ability to distinguish fault samples. To this end, a k-nearest-neighbor (kNN) fault detection method based on domain feature discrimination is proposed. A clustering-based data screening method is adopted, deep metric learning (DML) is introduced to extract discriminative features, and a potential generalized feature data mining method based on generative adversarial network (GAN) is proposed and introduced into the kNN-based fault detection framework, which enhances the model’s ability to describe the complex working conditions. Through experimental verification of 10 common faults in megawatt-level WTs, the results show that the proposed method reduces the average false alarm rate and missing alarm rate to 0.48% and 1.28%, respectively, which is an overall decrease of 6.17% and 4.73% compared to traditional methods.

Keywords

• Wind turbines
• fault detection
• k-nearest neighbor
• generative adversarial networks
• deep metric learning