Waterflood Tomography: Mapping High Contrast Permeability Structures Using Injection/Production Data

The paper presents a novel method to detect the existence and determine the orientation of high permeability channels between injection and production wells in a waterflood. We apply the concept of transmission tomography and model a waterflood reservoir by considering water injection rates as inputs and measured production rates as outputs. We solve the inverse problem, in which the goal is to determine the existence and location of high permeability channels in the field by measuring the lag time in response to the variation in the injection rate. The main advantage over other methods (e.g., tracer testing) is that this technique can be applied without significantly affecting daily operation. We show that the lag times at the producers can be estimated by monitoring production rates, given that known time-varying injection rates are applied. We propose a mixture model to characterize the lag time for each injector-producer pair, where the system is initialized by assuming that multiple candidate fractures exist. Our algorithm iteratively modifies the length, orientation and location of each high permeability candidate in order to match the measured response time between wells. It is well known that the solution may not be unique if we only have limited measurement data. Thus, in order to choose between possible models, we propose to use the “total length” of high permeability channels as regularization metric. Our method will select the model that fits the measured lag time with minimum total high permeability channel length. It is also possible within our framework to adjust the regularization so that it takes into consideration other type of prior information, e.g., preferred orientation known to exist in a given field. To validate our approach, we use a commercial simulator to test a synthetic line drive and a 5-spot waterflood. In the first case, we test a five spot with hydraulic fracture located in the central production well. In the second case, we test a single fracture with 45 degree orientation located between rows of producers and injectors. Our results show that our method can provide very accurate estimates of fracture orientation, with decreased uncertainty as the well density in the field increases.