Abstract:This paper presents a machine learning (ML) model designed to speed up the appraisal of geologic CO 2 storage sites by predicting the effectiveness in trapping and accommodating CO 2 in saline aquifers. Considering the urgency of de-risking as much geologic CO 2 storage resources as possible to help with CO 2 emission reduction Paris' goal, ML-based reservoir modelling has been documented as proper tool when a faster, good approximate, and less expensive approach is needed to surrogate multiple assessments of storage sites traditionally performed by long-timeframe and multi-stage geologic CO 2 storage numerical modelling approach. In this paper, a case study is presented. It consisted of a dataset comprised of six geologic aquifer parameters (CO 2 residual saturation, horizontal permeability, vertical to horizontal permeability ratio, porosity, brine salinity, and CO 2 flow rate) and elapsed time as input data, and as output data the CO 2 trapping mechanism indices (Solubility Trapping Index, Residual Trapping Index, and Structural Trapping Index) along with the dynamic storage capacity (CO 2 injected volume). Such dataset was used to train and test the artificial neural network (ANN) model. The dataset was generated from thousands of post-processed numerical realizations at several injection periods by applying design of experiment using a synthetic aquifer model derived from the Bunter Sandstone Closure 36 aquifer numerical model, from the Southern North Sea. The ANN architecture designed in Python consisted of 3 hidden layers and 40 nodes and its performance was assessed using the coefficient of determination (R 2 ) and root mean squared error (RMSE). The ANN performance showed accuracies (R 2 ) for training and testing with 96% and 95% of precision respectively. Practical application of the ANN model was successfully implemented to CO 2 storage aquifer sites selected from CO2Stored® database which lacking numerical models (Bunter Closure 3, 9, 35, and 40), obtaining at the end of 100-years injection case a Structural, Residual, and Solubility Trapping Index averaging 83%, 11%, and 6% respectively, with low variation coefficient indicating that trapping indices were predicted properly because aquifers selected for ANN model application have similar structures (dome-like shape) and reservoir properties. In addition, CO 2 injected volume predictions for 100-years injection case were ranging from 397 to 456 million ton (Mt) totalling 2.1 giga ton (Gt) of potential storage capacity which represents 70% of total theoretical volumetric capacity. These results show the significant impact to accelerate geologic CO 2 storage sites assessment by implementing ML-based modelling to preliminary de-risking groups of saline aquifers and reasonably consider them technically feasible CO 2 storage sites in UK.

Application of Heterogeneous Ensemble Learning for CO 2 –Brine Interfacial Tension Prediction: Implications for CO 2 Storage

Toward Predicting Interfacial Tension of Impure and Pure CO2-Brine Systems Using Robust Correlative Approaches

Estimation of CO2-Brine Interfacial Tension Based on an Advanced Intelligent Algorithm Model: Application for Carbon Saline Aquifer Sequestration

Machine Learning-Based Interfacial Tension Equations for (H 2 + CO 2 )-Water/Brine Systems over a Wide Range of Temperature and Pressure

Predicting wettability of mineral/CO2/brine systems via data-driven machine learning modeling: Implications for carbon geo-sequestration

Towards efficient and accurate CO 2 sequestration optimization: Integrating hierarchical spatio-temporal information into deep-learning-based surrogate models

Prediction of hydrogen−brine interfacial tension at subsurface conditions: Implications for hydrogen geo-storage

Advancing CO2 Solubility Prediction in Brine Solutions with Explainable Artificial Intelligence for Sustainable Subsurface Storage

Comparative Analysis Of Four Neural Network Models On The Estimation Of Co2-Brine Interfacial Tension

Comparative Analysis of Four Neural Network Models on the Estimation of CO 2 –Brine Interfacial Tension

Solubility and interfacial tension models for CO2–brine systems under CO2 geological storage conditions

Continuum-scale convective mixing in geological CO2 sequestration in anisotropic and heterogeneous saline aquifers

Machine learning - based shale wettability prediction: Implications for H2, CH4 and CO2 geo-storage

Predicting Phase Equilibria of CO 2 Hydrate in Complex Systems Containing Salts and Organic Inhibitors for CO 2 Storage: A Machine Learning Approach

Prediction of Key Parameters in the Design of CO2 Miscible Injection via the Application of Machine Learning Algorithms

Machine learning-based modelling for geologic CO 2 storage in deep saline aquifers. Case study of bunter sandstone in Southern North Sea

Prediction of Vapor Liquid Equilibrium of Binary CO2-Contained Mixtures for Carbon Capture and Sequestration using Artificial Intelligence

Insight to the prediction of CO 2 solubility in ionic liquids based on the interpretable machine learning model

CO2/brine interfacial tension for geological CO2 storage: A systematic review

Reservoir multiparameter prediction method based on deep learning for CO2 geologic storage

Critical review and meta-analysis of the interfacial tension of CO2-brine and H2-brine systems: Implications for CO2 and H2 geo-storage