Improved Oil Recovery through Advanced Control

Abstract

Norway is one of the major suppliers of oil and gas to the world market. Revenues from sales of oil and gas have played an important role in building modern Norwegian society. Oil and gas are trapped in the subsurface formation of relatively thin slabs of porous rock. The oil wells are drilled into the subsurface with an oil rig to extract the oil and gas from the reservoir. The production of oil can be increased by predicting and managing the future performance of the oil reservoir. However, because of the subsurface complexity and limited data, numerous uncertainties are present in oil reservoir characterization. These uncertainties should be considered for better future prediction of oil reservoir performance. In this paper, a simplified 2D control relevant model for a slightly slanting wedge-shaped black oil reservoir as in Zhang (2013) is made more realistic by incorporating model uncertainty. Furthermore, based on this model with uncertainty, a Proportional Integral (PI) controller is implemented to increase the oil production while minimizing the water production. The model in Zhang (2013) is re-implemented in Julia programming language. A standard package called DifferentialEquations.jl, available in Julia, is used to solve the model. A Tsit5() solver under this package solves the ODE problems with variable time steps which reduce the computational time and improve the performance. Furthermore, the uncertainty in the model is predicted via a parallel Monte Carlo simulation. A parallel Monte Carlo simulation is performed using EnsembleProblem in Julia. Finally, a PI controller is implemented in the model with uncertainty to control the valve opening of the Inlet Control Valves (ICVs) in the production well. The implementation of a PI controller improved the total oil production in 1000 days by 1.7873%. However, the effect is not very significant due to the limited capability of a PI controller. In this case, a more effective controller, such as a model predictive controller (MPC) is required.