The stator of doubly-fed induction generator-based wind turbine (DFIG-WT) is directly connected to the power grid, and the rotor of DFIG-WT system exchanges power with main grid via a back-to-back converter. The power electronic switches of back-to-back converter are prone to open-switch fault, which affects the stable operation of the system. A deep belief network (DBN) based fault diagnosis method for the open-switch faults of converters of DFIG-WT system is present in this paper. Firstly, the output response of DFIG-WT system with single and double switch faults of rotor-side converter (RSC) and grid-side converter (GSC) is analyzed. Based on the open-switch fault data of DFIG-WT, multilayer restricted Boltzmann machines (RBMs) are constructed to extract the deep information of rotor currents and grid currents under various fault and operation conditions, which fully takes advantage of excellent pattern recognition ability of DBN to improve the fault diagnosis accuracy. The simulation results indicate that the proposed DBN based fault diagnosis method is able to precisely detect the single and double open-switch faults of DFIG-WT system.