5. Conclusion

In this article, the best-known wind turbine diagnostic methods have been presented, classified and compared in terms of their use for predictive maintenance of critical wind turbine components such as generators, gearboxes and blade steering systems. These methods were divided into two broad categories: model-free and model-based. Model-based methods are in turn divided into two categories: analytical model-based methods and artificial intelligence-based methods. Both categories of methods perform diagnostics by means of: 1) performance monitoring/analysis (power generated), 2) thermal monitoring/analysis and 3) vibration monitoring/analysis. The choice of one of these methods depends on the context of the application (SCADA data or CMS data), and its objectives (early diagnosis, reliable diagnosis), as well as the available knowledge of the system's behavior or dynamics (physical knowledge,...