The doubly fed induction generator (DFIG) and its relative technique, especially in the case where crowbar protection is provided for enhanced low-voltage-ride-through (LVRT) performance, have been hot issues for a long time. In this paper, the wind farm LVRT standard and the dilemma that the conventional crowbar protection encounters are introduced in the first place. Then, the variation of the DFIG rotor faulted current is analyzed, with part of its duration being considered as a black box. Therefore, the radial basis function neural network (RBFNN) is utilized for the black box modeling. Based on above studies, an adaptive crowbar protection scheme is finally put forward and assessed. Emphasis is mainly placed on the black box modeling, error analysis, and the investigation of its impact on the wind farm LVRT capability. DIgSILENT-based simulation results indicate that, with the novel scheme, it becomes possible for the crowbar to be adaptively switched out at a certain reasonable moment, so as to generate reactive power within the faulted duration. As a result, the steady-state voltage of the point of common coupling (PCC) is enhanced and the safety of the rotor-side convertor (RSC) can be ensured at the same time. © 2015 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.