In this paper, parameters of the Jiles-Atherton (J-A) hysteresis model are identified using a stochastic search algorithm called differential evolution (DE). The J-A hysteresis model's parameters are identified by DE in such a way, that best possible agreement is obtained between the measured and model calculated hysteresis loops. This agreement is furthermore increased by improving the J-A hysteresis model. The improvement is achieved by replacing a constant pinning parameter in the J-A hysteresis model with a variable one. Here, the variable pinning parameter is written as a function of a magnetic field. By DE identified parameters are used in the J-A hysteresis model, which is included in the dynamic model of a single-phase transformer. The effectiveness of the improved J-A hysteresis model and parameters identification approach is verified with experiments and simulations.