Receding horizon optimisation of petroleum reservoir waterflooding using sequential quadratic programming

In this presentation, optimization of reservoir waterflooding process is studied using three different strategies: static optimization, and two forms of receding-horizon control (RHC), which are moving-end and fixed-end RHC. The MATLAB Reservoir Simulator (MRST) from SINTEF was used for reservoir simulation. The objective function to be maximized is the net present value (NPV) of the venture, while the control variable is the water injection rate. A sequential quadratic programming (SQP) algorithm was applied to solve the optimization problem. The SQP solver was integrated with a genetic algorithm (GA) solver to search for the globally optimal starting point (GOSP), so that the static optimization is not tripped at a local minimum. The GOSP obtained was then used for the RHC algorithms. It was found that the moving-end RHC gave the highest NPV, with an increase of 15.07% over the static optimization case. The improvement is as a result of early accelerated oil production and drastic reduction in water injection for subsequent years, which was followed by a total shut-in after 6.07 years. The increase in NPV obtained by the fixed-end RHC is just about 2.83% compared with static optimization. However, it is noted that the effectiveness of the moving-end RHC can be further appreciated if a variable optimization window is used with a stopping criterion, such as a point of zero cash-flow, instead of the fixed 20-year production period. The next step of the research will focus on online control design to implement the RHC strategy.