Model-Based Adaptive-Predictive Control and Optimization of SAGD under Uncertainty

Abstract

When it comes to SAGD optimization, two of the biggest challenges are controlling subcool to achieve conformance (a uniform growth of the steam chamber along the complete length of the well pair), and maximizing an economic performance measure, such as net present value (NPV); both desirable outcomes are not necessarily associated with the same values of the operational parameters (e.g., injection rates). Overcoming these challenges is necessary for achieving optimum SAGD performance, but this may be difficult through common operating policies, e.g. injecting steam at a empirically specified constant rate, considering well-known features of SAGD processes, e.g., complex dynamics (nonlinear, slow, high order, time varying, potentially highly heterogeneous reservoirs), operational constraints, model uncertainty and measurement noise. In the context of this work, the aforementioned challenges are both formulated as optimization problems of adjusting injection rates in order to optimize a particular objective function (e.g. minimizing subcool error or maximizing NPV). To address these challenges, this paper presents a nonlinear model-based adaptive-predictive control approach, alternative to the classical Proportional-Integral-Derivative (PID), for subcool control and NPV optimization of SAGD processes under uncertainty. Using case studies with an idealized heterogeneity pattern (subcool control) and multiple geological realizations based on logs from the Orinoco Belt region (NPV optimization), the proposed approach was compared with a decentralized PID for subcool control, in terms of response speed (mean square error and reaching time), steady state behavior (settling time, measured subcool mean and standard deviation) and control energy spending. While the proposed approach offered a slower response (not a critical issue in terms of oil recovery), it significantly outperformed the PID control during steady state and in control energy spending. On the other hand, the effectiveness for NPV optimization under uncertainty was demonstrated against constant steam injection strategies considering mean NPV, steam injection, water produced, and SOR, and when modeling alternative risk aversion scenarios.

Optimum design of CO2 storage and oil recovery under geological uncertainty

“…Gharbi [26] presented an EOR expert approach to optimize reservoir management and CO 2 floods in carbonate reservoirs. Several studies have focused on optimizing NPV of EOR projects under uncertainty [27][28][29]. These studies only focused on optimizing oil recovery rather than CO 2 storage.…”

Real‐time feedback control of SAGD wells using model predictive control to optimize steam chamber development under uncertainty

“…In another work, Guevara et al used nonlinear neural network‐based ARMAX models in MPC for subcool and compared its performance with a decentralized PID control strategy. The authors concluded that while their approach offered a slower response, it outperformed the PID control overall in terms of steady state behaviour and control energy spending.…”