Quantifying Experimental Impacts on Non-Newtonian Foam Characterization for Flow Modeling in Porous Media: Insights From Foam-Quality and Flow Rate Scan Experiments
本研究聚焦于泡沫在多孔介质中作为非牛顿流体的流动特性建模问题,指出传统仅依赖泡沫质量扫描(foam-quality scan)实验数据来拟合模型参数的做法,可能显著低估或高估泡沫的表观黏度,误差高达 62.5%。为此,作者开展了泡沫质量扫描与**流速扫描(flow rate scan)**两类实验,并结合CMG-STARS软件中的隐式纹理泡沫模型(implicit-texture foam model),采用非线性最小二乘法、参数可辨识性分析(profile likelihood)和贝叶斯推断(MCMC)等方法,系统评估了不同实验数据组合对模型参数估计的影响。
研究结果表明,同时采用泡沫质量与流速扫描数据(Approach C)可有效提升模型参数的可辨识性与预测准确性,显著降低模型输出(如采收率、突破时间、压降)的不确定性。在二维非均质油藏模拟中,非牛顿泡沫模型相比牛顿模型表现出更高的波及效率与更长的突破时间,但也带来更高的注入压降(差异达45%),提示现场应用中需权衡操作压力与采收效益。
CMG软件应用情况
本研究使用 CMG-STARS 商业模拟器中的隐式纹理泡沫模型(implicit-texture foam model),主要功能包括:
- 模拟泡沫驱替过程中的气相流度变化;
- 引入泡沫干燥效应(dry-out function)与剪切变稀效应(shear-thinning function);
- 通过调整参数(如 fmmob、SF、fmcap、epcap)拟合实验数据;
- 对比不同建模方法(牛顿 vs 非牛顿)在油藏尺度模拟中的动态响应差异;
- 评估泡沫在非均质油藏中的流动行为与波及效率。
尽管CMG-STARS模型在工业中广泛应用,但本研究指出,仅依赖单一流速下的泡沫质量扫描数据进行参数拟合,会导致模型在不同流速下的预测失真,因此建议未来应结合多流速实验数据进行联合拟合。
研究结论
- 传统单一流速泡沫质量扫描实验不足以准确表征泡沫的非牛顿特性,会导致模型在不同流速下预测误差高达63%。
- 提出的多实验数据联合拟合方法(Approach C)**显著提升了参数可辨识性与模型鲁棒性。
- 非牛顿泡沫模型在非均质油藏中表现出更高的波及效率与更长的突破时间,但也带来更高的注入压降(差异达45%)。
- 现场应用中需综合考虑采收率提升与注入压力管理之间的平衡。
- 实验设计与参数估计方法必须匹配,才能确保泡沫模型在油田尺度模拟中的可靠性与实用性。
作者单位
- 巴西联邦茹伊斯迪福拉大学(UFJF)
- Shell Brasil(巴西壳牌)

Abstract
Foam injection is a prominent technique for mitigating the effects of high gas mobility in porous media for Enhanced Oil Recovery, Carbon Capture, Utilization and Storage and well-stimulation. Experiments reveal that the foam’s apparent viscosity exhibits shear-thinning behavior, meaning its viscosity decreases with superficial velocity. Foam parameters are commonly characterized using only single-velocity foam-quality scan data. However, this approach can introduce significant uncertainty when evaluating foam’s apparent viscosity at different velocities. This study conducts foam-quality and flow rate scan experiments. Characterization of foam, using both single- and multi-velocity data, is performed using computational models and the solution of the associated inverse problems. Identifiability analysis and inverse and forward uncertainty quantification are performed to evaluate errors associated with the different data sets. The results demonstrate that relying only on foam-quality scan data leads to underestimation or overestimation of foam’s apparent viscosity, with errors of up to 62.5%. The impact of these misfitting issues is assessed in a heterogeneous scenario, where differences in production levels, breakthrough time, and pressure drops are analyzed with errors of 2.5%, 14.8%, and 45%, respectively. Therefore, this study underscores the importance of aligning laboratory experiments with parameter estimation methodologies that accurately characterize the non-Newtonian behavior of foams.
