Laser ranging, a high-precision measurement technique, is widely utilized in satellite orbiting, space debris tracking, lunar exploration, and deep space missions. Due to their high mobility and adaptability, mobile stations have become critical components of space target monitoring networks. However, the instability of the ground and the frequent changes in operational conditions exacerbate the issue of laser pointing deviations in mobile stations, compromising both measurement efficiency and accuracy. Current correction methods primarily rely on spot image processing to dynamically adjust the laser pointing by extracting the position of the laser spot. However, the accuracy of spot extraction is often impaired under conditions of weak spot visibility or high noise levels. Moreover, these approaches fail to address the underlying system errors contributing to the deviation. In this study, we propose an automatic correction method based on a laser pointing deviation model that accounts for these factors. The model enables real-time prediction of pointing deviations during observations and provides feedback to the control system for correction.By analyzing the system errors between the laser emission axis and the mechanical axis of the telescope, a theoretical model of laser pointing deviation is established. In continuous operation mode of the laser, dense sampling is conducted within the ranges of azimuth (0° to 360°) and elevation (20° to 80°), with an azimuth interval of 3° and an elevation interval of 2°. After sampling, the images are processed in batches to extract the laser spot, as shown in the flowchart (Fig.3). Once batch processing is complete, the data is cleaned using a moving window smoothing method to remove outliers, and data points with deviations exceeding 3 are excluded through pre-fitting of a function. After data cleaning, the theoretical model is fitted using the measured data to obtain a high-precision pointing deviation model.The laser pointing deviation model is nonlinear and sensitive to both initial values and boundary conditions. To derive a high-precision model, the laser spot distribution at different elevation angles was fitted to determine the center and radius of the circular distribution. The center was then refined to obtain the initial model parameters, as shown in Fig.5. Appropriate boundary conditions were applied, and the initial values were input into the model, which was subsequently optimized using the least squares method. The resulting accuracy of the model is better than 2.5".This paper proposes a method for predicting the laser pointing deviation in a ranging system. The theoretical model derived from the angular deviation between the optical axis and the mechanical axis matches well with the measured data. The prediction accuracy of the deviation model reaches the arc second level, comparable to the precision of real-time processing of the light spot image. The ranging station can deploy the model to the computer and dynamically correct the laser pointing based on the predicted deviation. This method exhibits strong robustness, even under conditions of faint light spots or strong background noise, and can effectively improve the success rate and efficiency of the ranging process.