Abstract: The horizontal displacement of the pump station buildings is an important monitoring item in the long-term operation of the pump station, and it is also an important index reflecting the operation status of pump stations. The fixed regression parameters of the traditional statistical model cannot describe the dynamic characteristics of the deformation behavior for the pump station structures. In view of that, this paper constructs a dynamic monitoring model of horizontal displacement for pump station based on forgetting factor recursive least square（FFRLS） method. By introducing forgetting factor, the correction ability of the new monitoring data to the model is enhanced and the parameters of statistical model are got dynamically to improve prediction accuracy. Then the validity of the proposed model is verified through a case study of a pump station project of South-to-North Water Diversion Project. The results show that the model can update the parameters of the statistical model adaptively with the addition of new data, and the regression and prediction accuracies of the statistical model are effectively promoted, which provides a new method to master the operation status of pump station buildings.