A novel methodology is proposed to evaluate the available transfer capability (ATC) of prescribed interfaces in connected power networks. With respect to the major uncertain factors affecting the ATC value, a realistic stochastic model is formulated in which availability of generators and circuits are considered as random variables following binomial distributions. With reasonable assumptions, fluctuations of loads are taken into consideration as normally distributed variables. To deal with the complex problem formulated with both discrete and continuous variables involved efficiently, a hybrid stochastic approach which takes advantage of both two-stage stochastic programming with recourse and chance constrained programming is proposed. The necessity of considering the impact of uncertainties on ATC assessment is demonstrated clearly by using the IEEE 118-node standard. Numerical results also illustrate the capability and efficiency of the proposed methodology.