Abstract: The petroleum pipelines in use urgent need to be ascertained the degree of safety. And they need a kind of method to optimize and direct the maintenance. According to the actual risk factors of the Qinhuangdao-Beijing pipeline, the paper offers a neural network model to the risk analysis and the forecast of every segment of long-distance petroleum pipeline, based on BP algorithm. It proves the practicability and reliability of the method by calculating the Qinhuangdao-Beijing oil pipeline.