A novel stochastic multi-area unit commitment (MAUC) framework is proposed to coordinate scheduling of generators and tie lines. In consideration of the randomness and volatility characteristics of wind energy, a worst-case based scenario selection method (SSM) based on the peak and valley shaving of the system, the ramping-up/down rates of net load, and the dispersion of uncertainty factors is presented to reduce the number of scenarios and improve the robustness of unit commitment (UC) schemes. Regarding the balance of efficiency and flexibility of tie-line scheduling, tie-line operation modes are established on the basis of the tie-line load rate (TLLR). Besides, the number of reversals of tie-line power flow directions during the entire scheduling period is also modeled. By linearizing the tie-line operation modes’ equations, the MAUC can be converted into a mixed-integer linear programming (MILP) formulation. Case studies show that more wind energy can be consumed by connecting power systems through tie lines and coordinating scheduling of generators and tie lines. Expected energy not supplied (EENS) of scenario reduction techniques based on the clustering method, i.e. k-means, Gaussian mixture model (GMM) and Fuzzy C-means (FCM), are 147%, 38%, and 130% higher than that of the SSM.