Abstract: The integration of large-scale wind farms into the grid leads to subsynchronous oscillation events, which poses a consistent threat to the safety and stability of the power system. The accurate location of the oscillation source plays an important role in mitigating oscillations. Therefore, in this paper we introduce a synchrophasor data-driven method for the localization of forced subsynchronous oscillation. Firstly, we derive the analytical expression for the oscillation within thermal power units caused by subsynchronous harmonic disturbance in the wind power grid-connected system. Subsequently, the low-rank characteristic matrices are constructed for forced subsynchronous oscillation components. On this basis, the sparse weights of self-augmentation and robust principal component analysis are combined to extract the main features of forced subsynchronous oscillation and achieve accurate localization in the wind-thermal bundled system. Finally, a three-machine system and an IEEE 39-node system are constructed to verify the improved robust principal component analysis on PSCAD/EMTDC. The simulation results demonstrate that the improved robust principal component analysis can intuitively locate the source of forced subsynchronous oscillation in the absence of grid topology information.