Large-scale, multi-regional, highly coupled interconnected power systems are increasingly facing challenges such as strong perturbation responses, nonlinear time-varying behavior, and complex oscillation modes. To extract oscillation information under fault conditions from the massive data accumulated by wide-area measurement systems (WAMS) and to achieve fault source localization, a negative damping oscillation fault source location method based on oscillation mode identification is proposed. First, a data-driven mode identification (DDMI) method is proposed to identify modal parameters of low-frequency oscillations with negative damping, thereby filtering out the dominant modal signals. Next, based on damping torque analysis, the calculated formula of generator oscillation energy in different oscillation modes is defined. Furthermore, the oscillation signals for different modes are reproduced using the identified modal parameters to compute the dominant mode oscillation energy, which enables the localization of the oscillation-causing generator. Finally, simulations are conducted on a New England power system model with wind power integration to verify the effectiveness of the DDMI method. The results show that the proposed localization method can accurately locate the fault source under multi-machine fault scenarios.