Abstract: In this paper, the fractal geometry theory is used to predict dam deformation, which can help to monitor the dam safety accurately. It is difficult to use a constant dimension fractal model to analyse and forecast the dam deformation data. Therefore, a variable dimension fractal model is applied to analyse dam deformation. Firstly, the monitoring data are taken into a transform of cumulative sum of N order. Then, we calculate fractal dimension sequence of each order and choose the best sequence. Finally, the dam deformation forecasting data are calculated in turn. Considering the problem that the traditional prediction method of the fractal dimension is uncertain and that the amount of required monitoring data is large, the prediction of the fractal dimension is made to establish an improved dam fractal geometry monitoring model by use of a grey model. Analysis results show that, based on practical engineering, by comparing the prediction precision of the traditional fractal geometry monitoring model and the improved model, the prediction accuracy of the improved model is better, and the model has the stability of prediction and the resistance to volatility.