A Bayesian Belief Network to Assess Risk of CO2 Leakage Through Wellbores

One of the approaches to tackle climate change is preventing emissions of carbon dioxide via CO2 storage. Leakage out of the storage complex, especially along wellbores which penetrate the caprock, is a concern and difficult to quantify. This paper presents a quantitative approach to a well integrity risk assessment, which examines the likelihood of a microannulus forming between the annular cement and casing or rock formation, and the leakage through that microannulus throughout the post-abandonment period of a CO2 storage site. The structure of our methodology is threefold, comprised of (i) a geomechanical structural integrity model of a wellbore to predict the probability of microannulus formation, (ii) a transport model to estimate CO2 leakage rates through the microannulus, if present, and (iii) a Bayesian belief network (BBN) for leakage risk assessment. The geomechanical and transport models are based on a particular CO2 storage site and well history, with variations in input parameters and operational conditions. The uncertain parameters are treated as probabilistic and sampled from statistical distributions using the Monte Carlo method. The geomechanical model is used to assess if debonding along the well cement interfaces has occurred over the life of the well (completion, production, re-injection) and to estimate the width of microannulus created by debonding. The transport model simulates migration of fluids through the microannulus and estimates leakage rates. The outputs of numerical simulations are assimilated, organized into ranges and used to define conditional probabilities, which are added to the BBN. The BBN allows the user to compare different scenarios by examining the associated probabilities of debonding and leakage. In this manner, the network serves as a decision support tool, to inform the operator of the risks of a certain scenario and to aid in deciding which mitigation or abandonment measures should be taken.