With the rapid development of wind power generation, the wideband oscillation issues have become increasingly serious. Impedance-based stability analysis method is an efficient tool for analyzing wideband oscillation issues of wind power systems. In order to address the difficulties of accurate wideband impedance/admittance acquisition of wind farms, a combined data-driven and knowledge-driven (physics-informed) online identification method is proposed in this paper. Based on the data-driven method, the neural network-based wideband impedance/admittance identification models of wind turbine generators (WTGs) are first established. Then, based on the combined data-driven and knowledge-driven method, the real-time wideband impedance/admittance characteristics of the wind farm can be identified with a versatile node admittance matrix-based impedance network aggregation algorithm. On this basis, the online stability assessment of wind farm integration system (WFIS) can be carried out. The effectiveness of the proposed method is verified by a case study of a permanent magnet synchronous generator (PMSG)-based wind farm built in MATLAB/Simulink.