A High-Accuracy Cost Prediction Model for Shale Gas Drilling in Southern Sichuan Using PCA and BP Neural Network

Shale gas, as a typical low-quality marginal hydrocarbon resource, faces persistently high drilling costs, which have become one of the main bottlenecks restricting its large-scale development. The Southern Sichuan region of China holds enormous shale gas reserves and is a strategically important area for achieving cost-effective large-scale development. However, as production capacity construction intensifies and the volume of investment and cost data increases, traditional data processing methods can no longer meet the timeliness and accuracy requirements for handling massive data. Accurate prediction of oil and gas drilling costs will help in making scientific decisions and evaluations. In this study, based on the costs and engineering parameters of settled wells in the Southern Sichuan Block N shale gas field, we established a Back-Propagation (BP) neural network model incorporating principal component analysis (PCA) to achieve accurate prediction of single-well drilling costs. Results show that: (1) PCA can effectively extract useful information from the shale gas drilling cost influence factors. Specifically, the number of fracturing stages, drilling duration, well depth, total proppant volume, horizontal section length, etc., are identified as key parameters affecting single-well drilling cost. (2) Using Matlab programming and a graphical user interface (GUI), we developed an integrated shale gas single-well cost prediction software system that combines data import, model training, cost prediction, and results export. The BP neural network model’s predictions achieved an average relative error of only -0.57\%, demonstrating convenience, practicality, and high accuracy. This system can provide a basis for investment decision-making in the Southern Sichuan shale gas block and has value for commercial application.