In this study, a data-driven method for predicting CO2 leaks and associated concentrations from geological CO2 sequestration is developed. Several candidate models are compared based on their reproducibility and predictive capability for CO2 concentration measurements from the Environment Impact Evaluation Test (EIT) site in Korea. Based on the data mining results, a one-dimensional solution of the advective–dispersive equation for steady flow (i.e., Ogata-Banks solution) is found to be most representative for the test data, and this model is adopted as the data model for the developed method. In the validation step, the method is applied to estimate future CO2 concentrations with the reference estimation by the Ogata-Banks solution, where a part of earlier data is used as the training dataset. From the analysis, it is found that the ensemble mean of multiple estimations based on the developed method shows high prediction accuracy relative to the reference estimation. In addition, the majority of the data to be predicted are included in the proposed quantile interval, which suggests adequate representation of the uncertainty by the developed method. Therefore, the incorporation of a reasonable physically-based data model enhances the prediction capability of the data-driven model. The proposed method is not confined to estimations of CO2 concentration and may be applied to various real-time monitoring data from subsurface sites to develop automated control, management or decision-making systems.
Bibliographical noteFunding Information:
The software and sample data used in this study are available upon request by personal contact with the corresponding author at email@example.com . Financial support for this study was provided by the Korea Environmental Industry and Technology Institute (KEITI) (project title: Environmental Management of Geologic CO 2 Storage, project 2014001810004 ). The study was jointly supported by the Basic Research Program ( 17-3211-3 ) of the Korea Institute of Geoscience and Mineral Resources (KIGAM), funded by the Ministry of Science and ICT of Korea .
© 2017 Elsevier B.V.
All Science Journal Classification (ASJC) codes
- Environmental Chemistry
- Water Science and Technology