A diagnosis method based on bond graph model is proposed for fault detection and isolation (FDI) of wind turbine components. Bond graph model can be used to locate fault components for it is a multidisciplinary component-level model, and to derive analytical redundancy relations (ARRs) for it can represent clearly the relationships between components. ARRs are often used to fault detection and isolation. In order to obtain as much ARRs as possible to improve the isolability, first of all, this paper builds the bond graph model of wind turbine, then uses the model to obtain temporal causal graph (TCG), according to TCG to obtain relations graph of variables. Finally, it eliminates unknown variables of junction functions of bond graph by relations graph of variables. The simulation results show the ARRs of this method can be used to parameter fault diagnosis of the wind turbine system. This work is supported by National Natural Science Foundation of China (No. 60974048), Open Fund Project of Hunan Universities Innovation Platform (No.11K027), and Postgraduate Innovation Fund of Hunan University of Schence and Technology (No. S140022).