Tomsk Polytechnic University

05/19/2023 | News release | Distributed by Public on 05/19/2023 04:55

Backed by Priority 2030, Tomsk Polytechnic University Assesses Behavior of CO2 Stored Underground

Backed by Priority 2030, Tomsk Polytechnic University Assesses Behavior of CO2 Stored Underground

19 May
17:30
3

Researchers from the Heriot-Watt Center at Tomsk Polytechnic University have developed a methodology to predict the dynamics of carbon dioxide sequestration when injected into deep aquifers for long-term storage. The model proposed by the TPU researchers is based on a variety of process parameters and ensures high prediction accuracy. It will facilitate and speed up the evaluation and selection of facilities for storage reservoirs. The research is supported by the program Priority 2030 of the Russian Ministry of Education and Science.

The research findings are published in the Expert Systems with Applications (Q1; IF:8,665).

Geological storage of CO₂ is a promising technology reducing the concentration of anthropogenic emissions into the atmosphere, which is widely used in the world. In geological storage projects, wells are used to inject carbon dioxide into deep formations. Long-term storage of carbon dioxide in such a reservoir is achieved through its interaction with rock particles and dissolution in the formation fluid. Thus, over time, the proportion of free gas in the formation decreases due to its transition into a bound state.

One of the factors for CO₂ geological storage safety assessment is the CO₂ capture efficiency, i.e., the ratio of the proportion of CO₂ that has passed into a bound state to the initial volume of injected gas. At the same time, the prediction of CO₂ sequestration dynamics depends on a large number of parameters, which hinders the full-scale process of modeling.

The TPU researchers have proposed a model to assess the dynamics of carbon dioxide sequestration through interaction with rock and dissolution in formation water.

The system works as follows. First, the model is trained based on the input data fed into it, which is a set of pairs: "input variable" and its corresponding "result". Our model receives 5450 sets of input data. Then, it determines the relationship between the variables and the outcome, and then learns to predict similar dependencies from new data. Once trained, the model can make target predictions with high accuracy,

explains Shadfar Davoodi, research engineer at the TPU Heriot-Watt Center.

Higher prediction accuracy was achieved through a large training sample and a detailed experiment plan. The developed model can be used for initial assessment of geological reservoirs to select CO₂ storage sites.

The researchers continue to work on the model. They plan to further improve the prediction accuracy of the model by optimizing the algorithm settings and using a new raw data preprocessing method.