To achieve accurate online acquisition of time-varying cable forces in cable-stayed bridges, a new method for real-time identification of online time-varying cable forces is proposed based on the data-driven adaptive chirp mode decomposition (DD-ACMD) algorithm. This method adopts the sliding window technique to update the vibration signal of the inclined cable, and determines its prior information and target modal components through the power spectrum density (PSD) analysis of the vibration signal. Afterwards, the DD-ACMD algorithm was used to identify the instantaneous frequency of the cable vibration, and the time-varying cable force of the inclined cable was calculated using the axial loading beam theory. The accuracy of the method was tested by a numerical case of inclined cable, and the results showed that the average error in identifying the time-varying cable forces under high noise level is 0.45%, with a maximum error of 1.84%.