Data-driven fault diagnosis of wind turbines using RMS-based residual filtering and Extreme Learning Machine

Abstract

This paper proposes a data-driven fault diagnosis framework for wind turbine benchmark models that combines Root Mean Square (RMS)–based residual filtering with the Extreme Learning Machine (ELM) algorithm. The main objective is to improve diagnostic accuracy while significantly reducing fault detection time, with particular emphasis on scaling sensor faults, which remain challenging for many conventional approaches. In the proposed methodology, residual signals derived from the system model are processed using a Moving RMS technique to extract robust and discriminative fault features. These features are subsequently classified using the ELM algorithm, enabling rapid and reliable identification of normal operation as well as sensor, actuator, and system fault conditions. The effectiveness of the proposed approach is validated through comparative experiments against conventional classification methods. The results demonstrate superior performance in terms of both detection speed and classification accuracy, confirming the robustness and efficiency of the proposed framework. Overall, this work contributes to the development of intelligent fault diagnosis strategies and offers practical insights for enhancing the reliability and maintainability of wind turbine systems.