Prediction of CO2 leakage risk for wells in carbon sequestration fields with an optimal artificial neural network

Carbon Capture and Storage (CCS) is a key climate mitigation technology. Leakage of the injected CO2 is one of the major environmental concerns. The potential for CO2 leakage from wells is one of the critical risks identified in geological CO2 sequestration. The objective of this study is to develop a computerized statistical model with the neural network algorithm for predicting the probability of long-term leak of wells in CO2 sequestration operations. Well design and operation data for over 500 CO2 exposed wells were generated from the West Hastings oil field and Oyster Bayou oil field in southern Texas, USA. The well integrity conditions were assessed by analyzing the well attribute data (well type, well age, CO2 exposed period, well construction details and materials), well operation histories and regulatory changes. Leakage-safe Probability Index (LPI) was assigned to individual wells. A computerized statistical model with network algorithm was developed based on data processing and grouping. Comprehensive training and testing of the model were carried out to ensure that the model was accurate and efficient enough for predicting the probability of long-leak of wells in CO2 sequestration operations. The accuracy of the trained neural network for well leakage prediction was also verified by the field operation in the Cranfield Field, Mississippi, USA. The developed neural network model can improve the efficiency of the storage operations by predicting the risk of CO2 leakage in the current exposed wells. In addition, it can also contribute in developing best practices standards by proposing recommendations for well construction in future wells.