A radically data-driven method for fault detection and diagnosis in wind turbines

Abstract

In order to improve the reliability of wind turbines, avoid serious accidents and reduce operation and maintenance (O&M) costs, it is important to effectively detect faults of wind turbines operating in harsh environment. This paper proposes a radically data-driven fault detection and diagnosis (FDD) method for wind turbines, which implements deep belief network (DBN). The DBN requires no knowledge of physical model, instead, it employs historical data without any pre-selection. The method has been evaluated in a wind turbine benchmark simulink model, in comparison with four model-based algorithms and four data-driven methods, and the results have shown that the proposed method achieves the highest accuracy. Moreover, extensive evaluation has been taken to analyse the robustness of proposed method, and the simulation results indicate the stable performance of proposed method in faults diagnosis of wind turbine.