A Novel Automatic History Matching Method and Upscaling Study of Cyclic Solvent Injection Process for Post-Chops Heavy Oil Reservoirs

Cyclic solvent injection (CSI) is one of the most promising processes for a post-CHOPS reservoir. This paper summarizes experimental results of nine CSI tests with three physical models with different scales. A typical western Canadian heavy oil sample with a viscosity of 4,830 cp at the reservoir conditions was used in these nine tests. Additionally, numerical simulation models were established to simulate the tests. A modified genetic algorithm (GA) based history matching method (MGA) was validated by history matching three tests. This MGA method was developed by integrating a population database and orthogonal array with the GA to improve the efficiency and effectiveness of the algorithm. Because of the existence of a population database, the running time with the MGA method was significantly reduced by nearly 75%, compared to that with the GA method. In addition, the accuracy of the history matching, evaluated by the minimum value of GlobalObj, was improved with the MGA method, compared with that obtained by three optimization methods in CMOST®. The remaining six tests were employed to conduct a CSI upscaling study. The uncertainties in the upscaling CSI process, such as the relative permeability curve, capillary pressure, reaction rate in the foamy oil model and dispersion coefficient were investigated by numerical simulation. Sensitivity analysis illustrates that adding an appropriate capillary pressure in each test could refine the history matching results between the simulation and experimental data. In addition, the location of the wormholes may affect the magnitude of capillary pressure employed in history matched cases. The CSI operational strategies in a typical western Canadian heavy oil post- CHOPS reservoir (M reservoir) were investigated by numerical simulation. The corresponding uncertainties (liquid relative permeability, gas relative permeability, capillary pressure and dispersion coefficient) were assessed by numerical simulation. The orthogonal array was utilized to define the simulation matrix, and 18 simulation cases, with 7 factors in 3 levels, were run. The oil recovery factor for ten-year production was selected as the response variable. After that, the multiple-linear regression was performed to construct the response surface and the proxy equations were then generated. Three thousand Monte Carlo simulations, in total, were performed to generate the probability distribution functions, which indicated that the P90, P50 and P10 estimates of the oil recovery factors were 14.08%, 14.69% and 15.33% in the ten-year CSI process, respectively. This study demonstrates that through simulating experiments conducted with physical models with different scales, the uncertainties in predicting the field-scale CSI performance can be significantly reduced.

Name thesis:degree_name

Master of Applied Science (MASc)

Level thesis:degree_level

Master's

Discipline thesis:degree_discipline

Engineering - Petroleum Systems

Grantor dc:publisher

Faculty of Graduate Studies and Research, University of Regina

Year dc:date.issued

2014

Author dc:creator

• Zhang, Min

Advisor dc:contributor.advisor

• Zeng, Fanhua

Committee members dc:contributor.committeemember

• Torabi, Farshid
• Gu, Yongan (Peter)

Language dc:language.iso

en

DOI

https://doi.org/10.82465/5117

OAI identifier oai:identifier

oai:uregina.scholaris.ca:10294/5861