In order to reduce the partial derivative errors in Preisach hysteresis model caused by inaccurate experimental data, the concept and correlative method of discretization of Preisach hysteresis model are proposed, the essential of which is to centralize the distribution density of Preisach hysteresis model in local region as an integral, which is defined as the weight of a certain point in that region. For the input composed of an ascending segment and a descending segment, a method to determine the initial weights together with an additional method to determine present weights is given according to the number of input ascending segments. If the number of input ascending segments increases, the weights of the corresponding points in updating rectangle are updated by adding the initial weights of corresponding points. A prominent advantage of discrete Preisach hysteresis model is its memory efficiency. Another advantage of discrete Preisach hysteresis model is that there is no function in the model, and thus, it can be expediently operated using a computer. By generalizing the above updating rectangle method to the continuous Preisach hysteresis model, identification method of distribution density can be given as well.