In order to improve the prediction accuracy of inner wall corrosion depth of natural gas gathering and transportation pipeline, a Gauss residual process corrosion prediction method was proposed. In this method, the importance of residual error was inferred by PF(particle filter), the trust of data-driven model was evaluated by combining physical model and historical data, and the weight of corresponding physical model and Gaussian process model was modified by physical model, so as to achieve accurate prediction of internal corrosion depth. The effectiveness of the method was verified by the corrosion monitoring data of a natural gas pipeline. The results show that the maximum relative error of this method is 1.276% and the minimum relative error is 0.045%. In addition, the root-mean-square error(E_RMSE) of the method is 0.92 μm and the mean absolute percentage error(E_MAPE) is 0.69%. It can be seen that the proposed method has high prediction accuracy and can accurately predict the corrosion depth of the inner wall of the pipeline effectively and for a long time.