Prediction of Storage Efficiency Index Based on Machine Learning: CO2 Storage Factor Analysis of Depleted Gas Fields

Recently, CO2 geological storage has been increasingly utilized as a technical solution for achieving carbon neutrality. Depleted gas fields are particularly beneficial because of their integrity and stability. However, no analyses have been conducted regarding storage stability and efficiency through residual and solubility trapping. In this study, injection simulations on a depleted gas reservoir model were performed to identify the factors affecting CO2 storage. Machine learning models were developed and evaluated using these data to reliably predict the storage efficiency index. The results confirmed the impact of the injection rate, maximum reservoir pressure, porosity, permeability, and temperature on storage efficiency. The applicability of random forests, support vector machines, extreme gradient boosting, and convolutional neural networks for predicting the storage efficiency index was reviewed. This study provides foundational data for the reliable prediction of storage efficiency using information from depleted gas fields.