Multi-Physics Graphical Model-Based Fault Detection and Isolation in Wind Turbines

Abstract

Early detection and isolation of faults in wind power generators increases the availability of wind turbines and reduces their down times and maintenance costs. Wind turbines constitute a complex system that includes sub-systems from different physical domains such as electrical, mechanical, and hydraulic systems. They constitute hybrid dynamical systems comprising discrete parts due to presence of power electronic switches and continuous parts such as induction machines, pitch actuator and mechanical drive-train. In this paper, a multi-physics graphical model-based fault detection and isolation (FDI) method is developed for doubly fed induction generator-based wind turbines. The model of the wind turbine is developed using hybrid bond-graph theory that captures causal, temporal, and structural properties of the system. Causality inversion method is then employed to derive analytical redundancy relations (ARRs) based on the developed model. The FDI is performed based on the changes in the values of ARRs. A systematic approach based on Chi-square criterion is developed to determine the values of thresholds, based on which the change in ARRs due to occurrence of faults are detected. The capability of the proposed method in accurate and fast detection and identification of mechanical, electrical, and hydraulic faults is demonstrated by the simulation results.