Online Health Monitoring and Incipient Fault Detection for Large Wind Turbine Based on a Data-Driven Method

Faults can cause a reduction or complete interruption in the wind power generation of wind turbines, making it crucial to detect faults at its early stage to allow for timely maintenance and replacement decisions, and to prevent significant economic losses. However, many fault detection methods that rely on analyzing Supervisory Control and Data Acquisition (SCADA) data are not sensitive enough to detect incipient faults. In this paper, a data-driven fault detection method is proposed to address the detection of incipient faults in wind turbines. The collected multivariable data from the wind turbine SCADA system is effectively reduced in dimensionality using the Principal Component Analysis (PCA) method. In addition, secondary dimensionality reduction and feature enhancement can be performed on the feature variables in the principal subspace and residual subspace using Hotelling T2 and Squared Prediction Error (SPE) algorithms, respectively. Based on the feature-enhanced feature vector, small deviations caused by incipient faults are accumulated and amplified using the cumulative sum (CUSUM) algorithm so that the incipient faults can be detected. Experimental verification is conducted using multitemperature data can successfully detect the incipient fault, providing a warning 43 days earlier than the SCADA system. Our proposed method can effectively extract and enhance incipient fault features, with great prospects in online health monitoring of wind turbine systems and incipient fault detection in wind turbines.

I. Introduction

Wind energy, as a renewable and clean source of energy, is a growing contributor to the world's energy markets [1]. Wind turbine is a device converting wind energy into electricity efficiently. However, due to the complexity of its structure and the diversity of control methods, it has a high failure rate leading to the increase of capital investment in wind farm equipment operation and maintenance year by year. Monitoring the operating status and incipient fault detection, and providing appropriate repair and replacement strategies in time can effectively reduce the operation and maintenance costs of wind turbines and prolong their service life [2].