Abstract: To overcome the disadvantages of the existing pattern recognition-based transient stability assessment (PRTSA) methods, a new PRTSA method based on Gaussian process with composite kernel function (CKF-GP) is presented by considering the possible real-time information provided by WAMS from which a group of system-level classification features are selected in this paper. Compared to support vector machine, the presented method has such advantages as automatic parameter selection and prediction with probability interpretation. Furthermore, the classification ability of the proposed method is improved by combining different single covariance functions to construct a composite kernel function. The effectiveness of the proposed method is validated by the simulation results on the New England 39-bus system.