This paper proposes a stochastic subspace identification (SSI) based modelling scheme to analyse small-signal stability (SSS) of wind power integrated power system. A reduced-order model which describes dominant dynamics of a power system is obtained by SSI. With this model, impact of wind farm controls on evolutions of critical modes of the power system is identified on-line. Superior to linearised model based analysis, tedious modelling procedures are significantly simplified. The proposed scheme can be implemented on-line without interfering normal operation of the system. Therefore, it can be used for power system dynamic security assessment. Furthermore, the relations between oscillation patterns and network variables can be analysed and quantized with network modal index provided by the proposed scheme. With the index, optimal network variable can be selected for online damping control design, which provides a potential way to improve power system SSS. Effectiveness of the proposed scheme is validated by simulations.