Abstract:The diversified development of natural gas sources in the national natural gas pipeline network provides multiple guarantees for the stable supply and consumption of natural gas in cities, but there are great differences in different gas source components and temperament characteristics. In this paper, a variety of machine learning models were established for the dual gas source mixed pipeline system, and it was found that the prediction performance of linear regression and support vector machine models was better, and the calculation deviation of calorific value and methane components were both less than 0.04% and 0.5% respectively. Aiming at the natural gas pipeline system with multi-source and multi-sink dynamic changes, an intelligent learning model of multi-source and multi-sink gas mixture components and calorific value based on spatiotemporal deep learning technology is established. The deep convolutional neural network model has the best prediction performance, and its calculation deviation of calorific value and components is less than 0.5%.