Taking the characteristic that GPU is particularly suitable for vector parallel computing into consideration, this paper presents a kind of fast calculation methods for transient stability which are based on the predictor-corrector strategy. The proposed method uses the Taylor series method to obtain an approximate value at first, then the approximate value is corrected by the implicit high-order finite difference scheme which is based on the diagonal Padé approximation. This method combines the advantages of the Taylor series method and implicit high order integral method, such as high calculation speed, high precision, good stability and so on. What’s more, it is suitable for vector parallel computing. The proposed algorithm has been tested on the IEEE145 nodes system. The simulation results show that the calculation accuracy of the new method is higher than the Taylor series method as it almost does not increase the amount of calculation. And compared with the CPU calculation, the new method can obtain more the 10 times speedup when it runs on the GPU. This work is supported by National Natural Science Foundation of China (No. 51309258).