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.