Abstract: Landslide deformation prediction has significant practical value that can give the guide for preventing the disaster and protecting the people’s life and property safety. Form the point of system theory, the displacement time-series can be divided into trend item and deviate item. Combined with rheological theory of rock and soil and principle of time-series, these two items can be researched based on deep research of main environmental variables that have obvious influences on landslide deformation. The trend item of displacement time-series can be extracted by grey system model. Evolutionary neural network, constructed by combining genetic algorithm and artificial neural network, can be used to construct the nonlinear relationship between main environmental variables and deviate item. According to creep stage and deformation response to the environmental variables, the comprehensive model must be changed in real time. So the dynamic gray model-evolutionary neural network prediction model of landslide deformation can be constructed by this train of thought. Applying this thinking and method into research of some landslides in the Three Gorges reservoir area, the validity and practical value of this model can be demonstrated. And it also shows that the dynamic prediction of landslide deformation is very crucial.

Key words: landslide, deformation prediction, rheological theory, environmental variables, grey system, evolutionary neural network