A workflow for risk analysis and optimization of steam flooding scenario using static and dynamic proxy models

Abstract Heavy oil reservoirs are full of uncertainties because of the difficulties in fluid and core sampling as well as well testing operations. Therefore, making any decision on development plan of heavy oil reservoirs under strong uncertain conditions needs risk analysis. Different thermal processes like steam injection have been used for the recovery of heavy oil. Because of high steam generation costs, it is necessary to optimize the process. But both risk analysis and optimization are very time consuming and expensive tasks as they both need too many simulation runs. Creating a proxy model, which replaces the simulator and emulates simulator outputs very fast, seems to be a good solution to this problem. Different static proxy models have been used to-date, which can optimize the process only at one certain time of simulation and they are not valid for other times. In this study for the first time dynamic or time dependent proxy models are used for uncertainty analysis and optimization. The term dynamic or time-dependent proxy model is a response surface of desired objective parameters, which is valid for the whole time interval of the process. This study demonstrates the application of artificial intelligence for optimization of steam flooding using dynamic proxy models. A new time-dependent artificial neural network is introduced as a dynamic response surface. By coupling this response surface with genetic algorithm, optimum injection conditions such as steam injection rate, steam quality, and also optimum injection time are obtained in a no-dip layered heavy oil reservoir. The proposed workflow is a rapid and cost-effective tool for risk analysis and optimization of steam flooding in heavy oil reservoirs.

[1]  Khalid Aziz,et al.  Fourth SPE comparative solution project: Comparison of steam injection simulators , 1985 .

[2]  John Williams,et al.  Optimizing Cyclic Steam Oil Production With Genetic Algorithms , 2005 .

[3]  Swapan Das,et al.  Application of Thermal Recovery Processes in Heavy Oil Carbonate Reservoirs , 2007 .

[4]  K. C. Hong,et al.  Effects of steam quality and injection rate on steamflood performance , 1994 .

[5]  Louis J. Durlofsky,et al.  Application of Statistical Proxies to Speed Up Field Development Optimization Procedures , 2008 .

[6]  Denis Igorevich Zubarev,et al.  Pros and Cons of Applying Proxy-models as a Substitute for Full Reservoir Simulations , 2009 .

[7]  Shahab D. Mohaghegh,et al.  Reservoir Simulation and Uncertainty Analysis of Enhanced CBM Production Using Artificial Neural Networks , 2009 .

[8]  Louis J. Durlofsky,et al.  Optimization of Nonconventional Well Type, Location and Trajectory , 2002 .

[9]  R. F. Stoisits,et al.  Production Optimization at the Kuparuk River Field Utilizing Neural Networks and Genetic Algorithms , 1999 .

[10]  Roland N. Horne,et al.  Reservoir Development and Design Optimization , 1997 .

[11]  Rick Chalaturnyk,et al.  A Simplified Methodology on Selection, Operation, and Optimum Design of Steam Drive Reservoirs , 2005 .

[12]  Guohe Huang,et al.  Development of an artificial neural network model for predicting minimum miscibility pressure in CO2 flooding , 2003 .

[13]  Shahab D. Mohaghegh,et al.  Quantifying Uncertainties Associated With Reservoir Simulation Studies Using a Surrogate Reservoir Model , 2006 .

[14]  L. B. Cunha,et al.  Uncertainty Assessment Using Experimental Design and Risk Analysis Techniques, Applied to Offshore Heavy Oil Recovery , 2005 .

[15]  Swapan Das,et al.  Application of thermal processes in heavy oil carbonate reservoirs , 2007 .

[16]  Hector H. Perez,et al.  Experimental Design: Three Contrasting Projects , 2009 .

[17]  Salvador Pintos,et al.  Surrogate modeling-based optimization of SAGD processes , 2002 .

[18]  Denis José Schiozer,et al.  Risk Assessment of Oil Fields Using Proxy Models: A Case Study , 2007 .