The Use of Model Predictive Control for Asset Production Optimization: Application to a Thin-Rim Oil Field Case

Abstract

Abstract Increasingly the upstream oil & gas industry is using active flow control (e.g. feedback loops) or passive flow control (e.g. passive ICD) technologies to optimize asset production. They are used, for example, to commingle production, stabilize production in case of water or gas breakthrough, and to mitigate the effect of slugging in wells. While the merits of such flow control technologies are becoming clear, so do their limitations. One main limitation of using traditional flow control technologies is that, in the presence of complex and changing process dynamics, it is difficult to come up with a controller design that optimizes asset production. Advanced control, specifically Model Predictive Control (MPC), is widely used in the industry to optimize complex downstream processes. In this paper we will start to explore the merits of MPC for upstream applications by means of a realistic test case of a thin oil rim that suffers from gas breakthrough. In the paper three aspects of MPC will be explored into detail. Firstly, we take a look at the dynamic process models, which lie at the basis of MPC. In MPC the dynamic model is used to compute control actions that optimize the future behavior of the production process. In order to do this, the model has to be sufficient accurate and needs to have a computational burden that allows for real-time optimization. We will discuss how a mixture of physical modeling and data driven models may be used for this purpose. Secondly, we give concrete examples of such fast dynamic models for the production process of the thin oil rim case. The models describe the dynamic behavior of the gas cone in the reservoir, as well as the effect that gas breakthrough has on well performance. Using such a model, we will evaluate 2 different thin oil rim production strategies that are based on downhole and topside flow control. Thirdly, we develop a production strategy using the MPC framework. In this approach the dynamic well-reservoir model is used to determine optimal settings of wellhead and ICD valve positions that optimize production. This advanced production control production strategy is benchmarked in terms Net Present Value (NPV) and barrels produced, and compared to cases that use only conventional feedback wellhead flow control. Production improvement in oil production over the first 1.5 years is 17% in the test case. Since beginning of the 1990's, starting in refineries and base chemical plants, MPC has become the standard advanced control methodology in downstream industries, putting more process knowledge (models) into production control and giving rise to optimal production. A similar development may be expected in the upstream industry. As the paper shows MPC has the potential for bringing real-time production optimization a step closer.