# Nonlinear Model Predictive Control for Oil Reservoirs

One way to increase oil recovery in mature fields is to inject gas or water, which will then push the oil towards the production wells. The aim of this project is to better control these processes through the use of a technique called Closed-loop Reservoir Management (CLRM) and to implement CLRM in collaboration with the OPTION project partners Welltec and Lloyd’s Register.

With CLRM, a reservoir simulation model continuously makes an estimate of unknown reservoir properties on the basis of real production data available at the given time in order to optimize the water injection process in the reservoir and thereby increase the oil production.

Oil constitutes the world’s leading fuel, covering 32.6% of the world’s energy consumption in 2014 (BP, Statistical Review of World Energy 2015) and while the consumption is rising, the resources are depleted. Methods for improved oil production have a large economical potential as a small increase in the average oil recovery factor will result in a large increase in the absolute amount of produced oil.

Closed-Loop Reservoir Management is an operational feedback strategy where an optimal long-term simulation-based control strategy is updated when new measurements of the reservoir become available. An important part of CLRM is solving optimal control problems, which give a long-term open-loop strategy for how to produce the reservoir.

The objective of this work is to implement CLRM in collaboration with the partners in OPTION which are Welltec and Lloyd’s Register. The project focuses on combining several different types of algorithms in order to have an efficient and effective implementation of CLRM. These include nonlinear and non-convex optimization algorithms, parallel iterative preconditioned methods for solving linear systems of equations which is the key computational part of CLRM, reduced order modeling for improved computational speed.

It is key that the implementations of such algorithms and use of external libraries make use of high-performance programming languages such as C/C++, Fortran or others, as well as be implemented such that it can be run on massively parallel high-performance computers, such as the HPC clusters located at the Technical university of Denmark.