In recent years, computational intelligence and machine learning techniques have gained popularity to facilitate very fast dynamic security assessment for earlier detection of the risk of blackouts. However, many of the current state-of-the-art models usually suffer from excessive training time and complex parameters tuning problems, leading to inefficiency for real-time implementation and on-line model updating. In this study, a new transient stability assessment model using the increasingly prevalent extreme learning machine theory is developed. It has significantly improved the learning speed and can enable effective on-line updating. The proposed model is examined on the New England 39-bus test system, and compared with some state-of-the-art methods in terms of computation time and prediction accuracy. The simulation results show that the proposed model possesses significant superior computation speed and competitively high accuracy.