It is necessary to accurately identify the five key parameters in the Jiles-Atherton (J-A) hysteresis model when it is applied to analyze the hysteresis loop of current transformer. An improved particle swarm optimization algorithm (PSO) is proposed to identify the key parameters in the J-A hysteresis model to solve the problems of computation time-consuming and poor optimization ability existing in the current identification methods. The genetic selection strategy is introduced into the PSO algorithm to improve the global search ability of the algorithm by increasing the diversity of the PSO, so as to improve the accuracy of key parameter identification of the J-A hysteresis model. In this paper, the identification speed and accuracy of the proposed improved algorithm (GSS-PSO) are compared with other intelligent algorithms in identifying the key parameters of J-A hysteresis model. The results show that the error between the hysteresis loops obtained by the improved algorithm and the measured hysteresis loops is the minimum, and the identification efficiency is high, which proves the accuracy and effectiveness of the proposed improved algorithm in the parameter identification of J-A hysteresis model.