Screening CO2 Storage Wells To Identify Corrective Actions with a Machine Learning Model

Summary In this study, we develop a model to help operators and governments with engineering and science-backed guidelines for identifying high-risk wells and prioritizing corrective actions. The proposed model uses a series of risk factors that affect wellbore integrity. A total of 10 attributes were evaluated. Each risk factor was scored as a product of a severity score (SS) and a weighting factor (WF). The SS represents the severity of different categories of each attribute, while the WF represents the importance of each attribute compared with others in screening wells for corrective actions. To assign WFs and SSs, we analyzed and applied machine learning to a public data set that included 48,033 wells. Among 10 attributes, there were three that were not included in the public data set; therefore, the estimated leakage rate was based on experimental work, the locations of perforation with injection reservoirs were based on practical work, and stimulation status was based on other studies. The model was validated by comparing it with other existing studies. Results from the model show that the type of wells, estimated leakage rate, cement status, and relationship between perforations and injection reservoir have a high WF; wellbore trajectory, the existence of intermediate casing, and stimulation status have a medium WF; and surface casing depth, well age, and well status have a low WF on the screening of wells for corrective actions. A worst-case scenario is an inactive horizontal gas/disposal/injection/coalbed methane well that was completed before 1930, has no casing, has perforations that are closed to the underground water zone, has multiple perforations or treatments, and was perforated in reservoirs, which are injection and high-pressure reservoirs. The model was used to simulate the risk of 36 wells because of a carbon dioxide (CO2) plume from a CO2-storage well located in a hybrid area of review in the San Juan Basin. This work holds potential utility in the selection of high-risk wells, estimation of leakage rates of wells, and giving recommendations for wells for which corrective actions should be conducted. The development of such a screening model is significant because it enables operators and regulators to quickly identify high-risk wells, prioritize corrective actions, and ensure compliance with regulatory requirements for safe and effective CO2 storage.