The wind power plant (WPP) operates in a wide range of operating points, resulting in diverse small-signal stability performances in a high-dimensional parameter space. The mapping between the stability performance and the operating point is implicit, which makes an obstacle for the stability region and margin assessment. Furthermore, the implicit mapping cannot guide the system operation in real-time applications. In this paper, a data-driven method is developed for the stability assessment and enhancement. Regarding the stability assessment, a regression-based analytical model is proposed to assess the stability region and margin as functions of multiple parameters, such as the active power, the reactive power, and the wind speed. The stability assessment makes use of the impedance data of the WPP. Based on the impedance data at diverse operating points, the generalized Nyquist criterion is analytically reformulated with the supportive vector regression method. Minimum characteristic locus is defined as a single-parameter index of the relative stability, and the stability region boundary is developed. Besides, the dominated oscillation frequency in stable and unstable regions is analytically formulated. The analytical solutions yield an explicit mapping between the stability performance and the operating points. Moving forwards to the stability enhancement, with the help of the explicit mapping achieved in the digital twin (DT) system, the real-time stability margin of the physical WPP could be enhanced by optimizing the operating point of the WPP. The proposed data-driven method for stability assessment and enhancement is implemented in a DT system of WPP. The DT system provides the impedance data of WPP and the effective interaction between the physical and virtual systems, offering opportunities for online applications. The validation is performed on a single wind turbine generator (WTG) and a WPP, respectively. With numerical simulations and experiments, the stability region is proven to be accurately and efficiently observed, and the stability margin is enhanced by adjusting the operating point.