With high proportion of the wind farms connected to the power system, subsynchronous oscillation resulting from the interaction within the wind farms or between the wind farm and the power grid becomes increasingly prominent. It is difficult to online monitor and control subsynchronous oscillations due to broad application of the power electronics, complex and diverse control links and AC/DC hybrid connection in modern power grids. The data-driven method mines the characteristics of subsynchronous oscillation according to the correlation between the operation data of wind power grid-connected system, and realizes the online analysis of subsynchronous oscillation from the perspective of data. As a result, a new research idea for ensuring the secure and stable of power grid is provided. This paper reviews the research results both from China and foreign countries in recent years. From four aspects of modal parameter identification, wide-area phasor calculation, system operating state classification, and system parameter estimation, the research status of data-driven online analysis methods for subsynchronous oscillation in wind power grid-connected system are summarized. The application of various methods in multiple scenarios such as subsynchronous oscillation monitoring, state assessment, early warning, source tracing and propagation of wind power grid-connected system is summarized and analyzed. Based on the results, the advantages and limitations of online application of various methods are summarized. The research trends and application prospects of the existing methods on the subsynchronous oscillation of wind power grid-connected system are prospected.