The work aims to construct a prediction model for the corrosion depth of buried pipeline and predict the remaining service life of the corroded pipeline. According to the ASME B31G residual strength evaluation standard, the maximum allowable corrosion depth calculation method of pipeline was given. The generalized regression neural network (GRNN) was introduced to construct the buried pipeline corrosion depth prediction model, and the particle swarm optimization (PSO) algorithm was used to optimize the GRNN network parameters. Combined with the prediction method of pipeline corrosion development trend, the residual life of buried weak pipelines after corrosion was predicted. With a buried oil pipeline in Shaanxi Province as the example, eight major external corrosion factors were selected to construct an external corrosion index system. With the help of Pycharm programming simulation and buried chip test, the prediction results of the model were verified and analyzed, and the remaining service life of corroded sections was predicted. Compared with the BP model, the maximum relative error of the pipeline corrosion depth predicted by the PSO-GRNN model was controlled within 13.77%, and the average relative error was only 6.63%. From the prediction on service life, the remaining service life of some sections failed to reach the expected value. The prediction performance of the prospected model is obviously better than that of BP model. The prediction accuracy is higher, and the maximum corrosion depth and future corrosion development law of buried pipeline can be better predicted. The prediction result of remaining life is close to the actual value, which provides guiding basis for maintenance and replacement of pipeline and has certain application value in actual engineering.