Wellbore integrity remains an inadequately constrained parameter for quantifying the risks associated with geologic CO2 sequestration. Understanding the fate of leaky wells is critical because many wells are already present in the best candidate sequestration sites. Potentially leaky wells are abundant and provide direct pathways for fluid to escape from the storage formation to an overlying environmentally sensitive or economically important zone (potable aquifer, oil/gas reservoir, or back to the atmosphere). Because field-scale case studies are sparse, laboratory experiments and numerical models must be used to define key controlling phenomena. Risk assessment models must be built from these observations and must capture not only the probability of a leak but also the conductivity of the leakage path.
A key finding of the work reported here is that effective permeability is not a static property. A leak path evolves as CO2-saturated brine migrates through it due to reaction between wellbore material and advecting fluids. The leak path conductivity decreases as CO2-saturated water is injected into a fractured cement core, and under constant-pressure-gradient boundary conditions similar to those in the field, the leak is self-sealing. Carbonic acid reacts with wellbore cement to release calcium and neutralize the acid, raising the pH. This calcium release does not enhance the conductivity of the leak, because silicon rich material remains behind and prevents widening of the fracture. Subsequent precipitation of calcium carbonate in the open fracture space leads to sealing when the residence time is sufficiently long.
Large-system risk assessment models that combine many aspects of CO2 sequestration require computationally fast models of component processes. Thus, we develop a model for leakage along a wellbore that combines laboratory and numerical observations with a simple one-dimensional analytic reactive transport description. The model captures two key observations from the experiments. The first is the time period (and fluid volume injected) before leak-path permeability begins to decrease. The second is that the leak-path permeability decreases via an asymptotic decay. Results show that straightforward application of experimental conditions to a hypothetical field-scale scenario results in wells that are unlikely to leak significant formation fluid into overlying aquifers. Parameters are then varied to identify the conditions that might lead to wells that do not seal.
Though simple, this is model captures enough of the key phenomena in this reactive transport system to be useful, and it can be easily integrated into risk assessment modelling. The results can be used to rank wells that are likely to continue to leak and must be fixed. The model can be easily modified to incorporate results from new experiments, field-scale observations, and more complex reactive transport simulations.