Accurate and fast location of the disturbance source of forced power oscillation is the key to suppressing such oscillations. In the process of forced power oscillation, the disturbance source generators continuously inject energy into the power grid, while the non-disturbance source generators continuously absorb energy. There is a causal relationship between the power oscillation response of the disturbance source generators and the non-disturbance source generators. In this paper, a method for fast locating the forced oscillation source in power systems using the kernel density estimation (KDE) transfer entropy (TE) causal analysis is proposed considering casual relationships. This method first uses wide-area measurement information of the power system to construct a KDE transfer entropy model between the active power pairs of each generator in the system. Then, a sliding window method is used to calculate the transfer entropy between each generator pair. By quantitatively comparing the statistical results of the transfer entropy calculated for different generator pairs, the causal transfer direction between the system’s measurements is determined, thereby achieving rapid location of the disturbance source of forced power oscillation in the system. Finally, the feasibility and effectiveness of the proposed method are verified through simulation data from the IEEE 4-machine 2-area system and the IEEE 16-machine 5-area test system.