Forced oscillation (FO) has posed a significant threat to the power system stability. Accurate and fast FO source location (FOSL) plays an essential role in mitigating FOs. However

most existing FOSL methods fail in dealing with non-stationary FOs and exciter-induced FOs. To overcome the above issues

a multi-synchrosqueezing transform (MSST)-based data-driven approach is developed in this paper to locate the FO source in bulk power systems by using measurement data. Firstly

MSST matrices of multi-channel measurements are formed by MSST. Then

the bus time-frequency representation (BTFR) defined and calculated from the MSST matrices is employed to extract the multi-channel FO components via the ridge detection technology. Furthermore

the MSST-based dissipating energy flow (DEF) model in the time-frequency domain and the dissipating energy spectrum (DES) model in the frequency domain are derived from the conventional time-domain DEF model

and their corresponding FOSL criteria are established. Moreover

an auxiliary criterion of the proposed approach has been developed to overcome the malfunction of the conventional DEF in interactions between generators. The performance of the proposed FOSL method is evaluated with simulation data from the WECC 240-bus and WECC 179-bus test systems

as well as field measurements from ISO New England. The results demonstrate the efficiency and accuracy of the proposed method in locating FO sources.