Toward mechanistic understanding of Fast SAGD process in naturally fractured heavy oil reservoirs: Application of response surface methodology and genetic algorithm

Thermal recovery is one of the most common enhanced oil recovery (EOR) techniques in heavy oil reservoirs. Fast-SAGD method is a novel SAGD approach in which offset wells are drilled for periodic injection and production. This reduces the costs of SAGD pair injection-production wells and results in more production in a shorter time. Since production from heavy oil reservoirs is becoming more and more important, this study focused on thermal recovery of Mond field as one of the heavy oil fields in Iran. In this study, effective parameters of Fast-SAGD process including injection pressure and rate, injection, production, and offset well heights, offset well pressure and its injection and production periods were studied. The main purpose of this research was solving the problem of Fast-SAGD process optimization in terms of time and accuracy, as similar studies were not systematically studied on optimization of this process. In previous studies, effective parameters on the objective function were considered independently for obtaining the optimum conditions. Genetic algorithm was applied for optimization. The objective function was selected as a combination of recovery factor (RF) and cumulative steam to oil ratio (CSOR). To speed-up the optimization, effective variables were converted to discrete values. Optimization results using discrete and continuous variables were compared to each other. Moreover, results were compared to previously optimization technique applied in this field by Hemmati et al. (2013). Another point in this study was sensitivity analysis application. This showed that reservoir parameters are dependent on each other and traditional methods are inefficient for optimization of complex heavy oil reservoirs and thermal EOR processes. These points led to better results in terms of time and accuracy in this model. Results indicated 3% increase in RF and 4.7% decrease in CSOR for one injection cycle approach in comparison to two-cycle case. Response surface methodology (RSM) was introduced to setup a mathematical basis for effective variables and their interactions. Optimization results using analysis of variance (ANOVA) denoted that the injected steam and RF were 10% and 6% less in comparison to the results of genetic algorithm, respectively.