Abstract: In order to obtain the creep characteristics of the surrounding rock of the Dianzhong water diversion tunnel and to study the long-term stability of the surrounding rock of the tunnel, a fully automated triaxial developed by Wuhan Institute of Geotechnics, Chinese Academy of Sciences, is used to carry out the uniaxial creep test of green mud shale. The axial creep curve and isochronous stress-strain curve of green mud shale are obtained. And the long-term strength value of green mud shale is determined based on the characteristics of isochronous stress-strain curve. Based on the classical creep curve characteristics and a large number of experimental creep curves, it is found that the equations describing the attenuation creep curve can be treated by symmetry. Therefore, it is assumed that the acceleration curve and the decay creep curve are symmetric about the midpoint of the stable creep curve. An accelerated creep model based on the symmetry of the creep curve is obtained. And a set of methods to determine the parameters of the creep model is proposed based on the characteristics of the creep test curve. Finally, the parameters introduced into the accelerated creep model for sensitivity analysis. The parameters introduced into the model have a clear physical meaning. The results show that with the increasing axial stress, the instantaneous strain value and creep deformation value of the rock are also increasing, and the instantaneous strain of the rock under the first stage load accounts for the largest ratio of the total creep deformation. The established creep model can not only well describe the attenuation creep and stable creep deformation law of green mud shale, it also better make up for the defects of the Nishihara model that cannot describe the accelerated creep. The agreement between the model curve and the test curve is much higher than that between the model and the test curve, and the correlation coefficients between the model curve and the test curve under different stresses are all above 0.90. Meanwhile, the validation of different types of test curves and model curves also shows that the model can be applied to the prediction of creep curves of different types of rocks. Finally, the value of parameter j is introduced to control the deformation rate and the time to enter accelerated creep. The value of parameter k controls the creep time and the creep rate in the accelerated stage.