Volume 57

Prediction of CO2 storage with Enhanced Natural Gas Recovery Based on Backpropagation Neural Network Honglue Huang, Shuyang Liu, Junrong Liu, Zhiqiang Wang

Abstract

CO2 storage with enhanced gas recovery (CSEGR) offers the dual benefits of boosting natural gas production while achieving effective carbon sequestration, presenting broad prospects for industrial application. In this study, a numerical model of CSEGR was constructed using geological data from the Dongfang 1-1 gas field in the South China Sea to assess how different injection–production schemes and reservoir conditions affect both gas recovery and CO2 storage. To further enhance predictive capabilities, a backpropagation (BP) neural network–based surrogate model was developed. The AI-driven model accurately captured the nonlinear relationships between geological and engineering parameters and the performance outcomes. Results show that gas recovery increases with CO2 injection rate and permeability, while decreasing with bottomhole pressure and porosity. Conversely, CO2 storage performance improves with higher injection rates and bottomhole pressures but declines with greater permeability and porosity. The BP neural network achieved an average prediction accuracy of 95%, highlighting its effectiveness as a reliable tool for forecasting CSEGR performance under complex reservoir conditions.

Keywords CCUS, CO2 enhanced gas recovery, CO2 storage, BP neural network