Assisted Design of Polymer-Gel Floods in Naturally Fractured Reservoirs Using Neuro-Simulation Based Models

Polymer gel treatments have been widely used by the industry to improve sweep conformance and enhance recovery from highly fractured reservoirs. The success of these treatments depends on several factors that include various reservoir properties and gel design parameters. This paper presents a pragmatic approach to optimize the design of polymer gel treatments to improve oil recovery in naturally fractured reservoirs using neuro-simulation based models. A full spectrum of fractured reservoir properties and polymer gel treatment design parameters was used to generate base simulation models. Production rate, oil recovery and water cut trends were used as key performance indicators to monitor sweep conformance and evaluate polymer gel design effectiveness. These simulation models were used to construct, train and validate the neural network. The network topology was effectively designed to achieve a good match with the reservoir simulation models. A given set of reservoir properties including porosity, permeability, net pay thickness, water saturation, polymer gel concentration and injection rate can be optimized using the neural-based model to acquire the desired production rate. Furthermore, results show that the injection rate and cross-linking agent concentration are the most sensitive parameters affecting the production performance. The neural model can be used as an effective screening tool for selecting and designing polymer gel projects as it covers a wide range of field parameters. This work capitalizes on the ability of artificial expert systems in generating tractable, robust and computationally efficient solutions for complex reservoir models. In particular, this paper presents proxy models that are uniquely developed for the first time to optimize oil recovery in naturally fractured reservoirs using polymer gel conformance treatments.