Based on the model that combined by grey model and markov model,the prediction of corro-sion rate of nuclear power pipeline was studied.Works were done to improve the grey model,and the optimi-zation unbiased grey model was obtained.This new model was used to predict the tendency of corrosion rate,and the markov model was used to predict the residual errors.In order to improve the prediction precision,rolling operation method was used in these prediction processes.The results indicate that the improvement to the grey model is effective and the prediction precision of the new model combined by the optimization unbi-ased grey model and markov model is better,and the use of rolling operation method may improve the predic-tion precision further.