Predicting the effectiveness of geological CO₂ storage and evaluating the field application of successful CO₂ sequestration require a large number of case studies. These case studies that incorporate geologic, petrophysical, and reservoir characteristics can be achieved with an artificial neural network. We created an artificial neural network model for geological CO₂ sequestration in saline aquifers (ANN-GCS). To train and test the ANN-GCS model, data of residual and solubility trapping indices were generated from a synthetic aquifer. Training and testing were conducted using Python with Keras, where the best iteration and regression were considered based on the calculated coefficient of determination (R²) and root mean square error (RMSE) values. The architecture of the model consists of eight hidden layers with each layer of 64 nodes showing an R² of 0.9847 and an RMSE of 0.0082. For practical application, model validation was performed using a field model of saline aquifers located in Pohang Basin, Korea. The model predicted the values, resulting in an R² of 0.9933 and an RMSE of 0.0197 for RTI and an R² of 0.9442 and an RMSE of 0.0113 for STI. The model was applied successfully to solve a large number of case studies, predict trapping mechanisms, and optimize relationships between physical parameters of formation characteristics and storage efficiency. We propose that the ANN-GCS model is a useful tool to predict the storage effectiveness and to evaluate the successful CO₂ sequestration. Our model may be a solution to works, where conventional simulations may not provide successful solutions.

预测地质CO ₂储存的有效性和评估成功隔离CO ₂的现场应用需要大量案例研究。这些结合了地质，岩石物理和储层特征的案例研究可以通过人工神经网络来实现。我们创建了一个人工神经网络模型，用于咸水含水层中的地质CO ₂隔离（ANN-GCS）。为了训练和测试ANN-GCS模型，从合成含水层生成了残留和溶解度捕获指数的数据。使用Python与Keras进行培训和测试，其中根据计算出的确定系数（R ²考虑最佳迭代和回归）和均方根误差（RMSE）值。该模型的体系结构由八个隐藏层组成，每层包含64个节点，其R ²为0.9847，RMSE为0.0082。对于实际应用，使用位于韩国浦项盆地的盐水层的现场模型进行了模型验证。该模型预测了这些值，从而得出RTI的R ²为0.9933，RMSE为0.0197，R ²STI为0.9442，RMSE为0.0113。该模型已成功应用于解决大量案例研究，预测圈闭机理以及优化地层特征物理参数与储层效率之间的关系。我们建议，ANN-GCS模型是预测存储效率和评估成功的CO ₂隔离的有用工具。我们的模型可能是工作的解决方案，而传统的模拟可能无法提供成功的解决方案。