Enhanced fault detection and diagnosis in wind turbine systems using canonical variate analysis and reconstruction-based contribution method
摘要
Fault detection and diagnosis (FDD) is critical for ensuring the performance, safety, and reliability of industrial systems, especially in the expanding wind energy sector. As wind turbine installations continue to increase globally, maintaining operational reliability has become increasingly important due to their complex and nonlinear nature. Traditional FDD methods often underperform in such systems due to challenges in handling high-dimensional, nonlinear data. This paper proposes a robust methodology for fault detection and diagnosis in wind turbines. The proposed approach employs canonical variate analysis (CVA) for fault detection by analyzing multivariate data, and reconstruction-based contribution (RBC) for fault isolation by quantifying the contribution of individual variables. The methodology was validated using benchmark data from a real wind turbine system. The results demonstrate high effectiveness in detecting and diagnosing faults, highlighting the approach’s capability to manage complex, nonlinear systems. Simulation results show that the proposed methodology outperforms traditional techniques such as PCA, PLS, EMPRM, and TPCR achieving higher fault detection rates and improved sensitivity, with detection accuracy exceeding