Carbon Capture Utilization and Trapping Efficiency Modelling: A Case Study of Gas Flare Site in Niger Delta

Most significant challenges in recent decades have been the rise in global average temperature, which is mostly caused by greenhouse gas (GHG) emissions. Previous research used complex numerical simulations processes that uses iterative procedures to obtain a certain operating condition, which are time-consuming, tiresome, and require expensive devices that are limited in locations where real-time decisions are required. These methods do not produce models that are flexible, generalizable, accurate, or robust. Smart digital technology concepts such as artificial intelligence (AI) and machine learning (ML) are becoming increasingly popular and are being used in a variety of applications including carbon capture. This study primarily aims to create models for predicting CO2 trapping efficiency using AI such as artificial neural network (ANN) and support vector regression (SVR) to address these problems. The models were built using 260 CO2 trapping efficiency data from a gas flare site in the Niger Delta. The performance of the developed models was analysed using the statistical metrics. The goodness of fit (R2), mean square error (MSE), root mean square error (RMSE), and average percentage relative error (APRE) were 0.9993, 9.0×10–4, 9.53×10–3 and 0.18282 for the ANN model, whereas for the SVR model, the R2, MSE, RMSE, and APRE were 0.9766, 3.095×10–3, 5.563×10–2 and 1.4137 respectively. The parametric importance results for CO2 trapping efficiency show that while CO2 mass fraction had the greatest influence (29.70%), temperature, density, emission rate, pressure and activity rate followed with contributions of 17.29%, 12.41%, 11.30%, 11.06%, and 10.58%, respectively, while time had the least effect of 7.62%. The models were explicitly provided to make them easy to implement into software programmes. The explicitness, accuracy, and suggestion for using the models in the field are among the features of the models given in this study for which uniqueness is claimed. The proposed models would eliminate the need for complicated and time-consuming reservoir simulation at the early stage of the Carbon capture and storage (CCS) project, allowing for real-time findings in the field.