Abstract
Polymer flooding is one of the most broadly implemented chemical EOR processes due to its low injection cost and its success in prolific production increments. This work develops an artificial-neural- network based expert system by utilizing numerically generated training data using a high-fidelity numerical simulation model. Injection-pattern-based reservoir models are structured to establish the knowledge-base that serves for the ANN training. The injection process starts with water injection and switches to polymer injection when the water cut reaches to a threshold value. The chase-water is followed after a prescribed amount of polymer slug is injected. The expert system is generalized in terms of reservoir rock and fluid properties, rheological properties of the polymer solution, and critical engineering parameters to adjust to the complexities exhibited in the polymer injection projects. In developing the inverse model for project screening and design purposes, we have used an engineering design protocol employing inverse and forward-looking expert systems with the goal of exploiting the non-unique nature of the inverse problems. In this work, we employ the expert system as a forecasting and screening tool that is capable to predict time series based response in terms of oil production, water production and injection well sandface pressure data. The validity of the forward-looking expert system is confirmed via extensive blind test applications within a 5.83% error margin. More importantly, we present a project screening protocol that couples the expert system and particle swarm optimization (PSO ) methodology to maximize the net present value (NPV ) of polymer injection projects. In this way, we take the advantages of the fast computational speed of the ANN model to evaluate the finesses of project parameters. In the application of the inverse expert system, we observe that the proposed design protocol is capable to establish a catalogue of polymer injection design parameters that satisfy an expected hydrocarbon recovery performance. The work described in this paper exhibits the robust nature of the proposed expert system in adapting to strong non-linearities encountered in the polymer flooding projects. The coupled utilization of the inverse and forward-looking modules, enables the design engineers to find solutions that are unique to the problem being studied by simultaneously satisfying the imposed constraints effectively. The expert system proposed in this paper is one of the modules of a comprehensive artificial-neural- network based toolbox that includes a large spectrum of EOR processes.
