Landslide stability in reservoir area is one of the key problems studied in water conservancy and hydropower engineering. The indicators affecting landslide stability are different and time-varying, and it is difficult to achieve dynamic evaluation of stability by traditional methods. By improving the extension theory, a multi-level and multi-variable dynamic assessment model of landslide stability was established. In the proposed method, the effective rainfall model was introduced to replace the daily rainfall model, the difference coefficient was used to replace the standard deviation of the traditional CRITIC method, and the dynamic weight calculation method based on the correlation function matrix was established. Through analyzing sheep song hydropower station reservoir area within a certain landslide surface macroscopic deformation, real-time monitoring data analysis, the applicability of effective rainfall model was proven, the effective rainfall period was defined, and the extension theory and effective rainfall model coupled dynamic evaluation on the stability of this landslide were eventually improved.The results and monitoring data reflect regularity has good consistency. The reliability and accuracy of the method proposed in this study are proved. In addition, the proposed model can quickly analyze the scene real-time data, implement change with time, environment of dynamic evaluation of landslide stability state for landslide prevention, and control work to provide a kind of objective and accurate analysis method.