Prediction of SAGD Performance Using Response Surface Correlations Developed by Experimental Design Techniques

Abstract

Over 80% of the vast reserves of Alberta's Oil Sands can be produced only by using in-situ recovery methods. Among them, one which is likely the most efficient and important is the steamassisted gravity drainage (SAGD) process. Numerical simulation allows the ideal way of predicting reservoir performance under SAGD process during the whole field development cycle. However, in the earlier stages of development studies when it is necessary to make preliminary engineering design, estimate reserves, screen among other SAGD prospects, as well as consider the uncertainty of some reservoir parameters, it may not be feasible to do a detailed simulation study, due to high computational time involved in a SAGD process simulation.Under these circumstances, a method of predicting reservoir performance using a simple statistical model, that can approximate the reservoir simulator over a given range of some important input parameters, is a good approach to provide means of comparison and preliminary predictions without resorting to numerical simulation.The purpose of this work is to use Experimental Design Techniques to develop a response surface that can predict SAGD performance without the expense of doing simulation. A preliminary screening study was done in order to select the most influential variables on the SAGD performance The variables used for that purpose include reservoir rock/fluid properties such as reservoir thickness, porosity, vertical permeability, vertical-horizontal permeability ratio, methane content, rock thermal conductivity, initial oil saturation and bitumen viscosity; along with SAGD design and operating variables including: spacing between injector/producer, operating pressure, preheating period, maximum steam injection rate and SAGD well pattern spacing.In a second stage the influential variables were used to create a statistically significant correlation, by using the experimental design method and response surface techniques. This simple model allows the prediction of the SAGD PETROLEUM SOCIETY CANADIAN INSTITUTE OF MINING, METALLURGY & PETROLEUM2 performance in terms of maximum Net Present Value over 15 years of project life, for a given range of the most influential parameters.

Effects of Reservoir Parameters and Operational Design on the Prediction of SAGD Performance in Athabasca Oilsands

“…The minimization of PRESS leads to an improvement on the power of prediction of the model. Once the model is constructed, it can be used to predict reservoir performance and to optimize controllable variables (Vanegas, 2008). Table 3 presented the result of NPV from applied to the 75 realizations in the range from 2.99 to 224.2 mm$.…”

Connectionist Model to Estimate Performance of Steam-Assisted Gravity Drainage in Fractured and Unfractured Petroleum Reservoirs: Enhanced Oil Recovery Implications

“…A large number of real data points from SAGD experiments were collected to create an ANN–PSO model for predicting the SAGD recovery factor (RF) and cumulative steam-to-oil ratio (CSOR). The experimental results, numerical modeling outputs, and required data were obtained from studies conducted by several other researchers. ,− ,− The ranges of the input and output parameters used to build the ANN–PSO system are provided in Table . This table covers wide intervals of the important data, based on production history obtained from field trials, pilot-plant runs, and experimental tests of SAGD operations.…”

Advanced ML and AI Approaches for Proxy-based Optimization of CO2-Enhanced Oil Recovery in Heterogeneous Clastic Reservoirs

“…The proxy modeling optimization has never been found in the literature regarding the CO2-GAGD process. However, the proxy model has been adopted in various reservoir studies and EOR modelings, such as oil production optimization (Badru andKabir , 2003, Zangl et al , 2006), water flooding (Guyaguler et al , 2000, Haghighat Sefat et al , 2014, gas flooding (Ampomah et al , 2016), steam injection, Alkaline-Surfactant-Polymer flooding (Zerpa et al , 2007), SAGD process (Fedutenko et al , 2013a,b, Vanegas Prada and Cunha , 2008, Yang et al , 2011, well locations (White and Royer , 2003), history matching (Goodwin , 2015, He et al , 2016, Zubarev , 2009, etc. The DoE approaches create multiple computer experiments (realizations) for the problem by combining the levels for each parameter.…”

“…Many DoE approaches have been used in various reservoir simulation studies to build the proxy models. The most common DoE approaches are fractional factorial design (Vanegas Prada and Cunha , 2008), central composite design (Yeten et al , 2005), D-optimal design (Zerpa et al , 2007), and Latin Hypercube Design (Zubarev , 2009). There are many successful examples of using the proxy models in the literature of reservoir studies, such as second-degree polynomial equation (Avansi , 2009, Hassani et al , 2011, Fedutenko et al , 2013b, White and Royer , 2003, kriging algorithm (Fedutenko et al , 2013b, Osterloh , 2008, Zubarev , 2009, and artificial neural networks (Zangl et al , 2006, Zubarev , 2009.…”