Capturing the correct reservoir heterogeneity in a geological model is critical for designing and accurately forecasting expected production benefits from improved/enhanced oil recovery processes. In simple terms, reservoir heterogeneity is often considered as a measure of statistical variation of static properties, such as porosity and permeability. The Lorenz coefficient and Dykstra-Parsons coefficient are two such measures of reservoir heterogeneity that account for these static effects. These measures are considered simplistic because the spatial distribution and arrangement of these properties is more critical for reservoir characterization than its statistical variation. The dynamic Lorenz coefficient is one such measure that accounts for the spatial distribution and arrangement of porosity and permeability. The dynamic Lorenz coefficient cannot be directly measured in the field but can be implicitly inferred from tracer data. This inferred heterogeneity will be influenced by the spatial distribution of static properties and can also be significantly influenced by multiphase flow effects, such as viscous fingering and gravity over/underride, which often arise from the differences in the viscosities and densities of the different phases involved. The dynamic Lorenz coefficient interpreted from a tracer test response lumps both static and multiphase (dynamic) effects into one measure of heterogeneity. During geological model construction, the dynamic Lorenz coefficient is also used to rank the model in terms of heterogeneity to select the most representative model. However, the geological models are ranked using fast single-phase/streamline simulations and are devoid of complications caused by adverse mobility or density effects. This creates a disconnect between the heterogeneity measures used to characterize a geological model and those available to characterize a reservoir. Thus, calibrating a geological model to field data is a laborious task.

In this paper, we present a workflow that bridges this gap by decoupling the effect of adverse mobility to obtain an approximate measure of heterogeneity that can be cross-checked against the geological realizations to select the most representative model. The workflow developed in this paper is for inverted, seven-spot patterns and is mainly focused on water-wet reservoirs.

在地质模型中捕获正确的储层非均质性对于设计和准确预测改善/增强的采油工艺的预期生产效益至关重要。简而言之，储层非均质性通常被视为衡量静态性质（例如孔隙度和渗透率）统计变化的量度。Lorenz系数和Dykstra-Parsons系数是说明这些静态影响的两种这样的储层非均质性度量。这些措施被认为是简单的，因为这些属性的空间分布和排列对于储层表征比其统计变化更为关键。动态洛伦兹系数就是其中一种测量孔隙度和渗透率的空间分布和排列方式的方法。动态洛伦兹系数不能在现场直接测量，但可以从示踪剂数据中隐式推断出来。这种推断的异质性将受到静态特性的空间分布的影响，还可能受到多相流效应的影响，例如粘性指法和重力超驰/超驰，这通常是由于所涉及的不同相的粘度和密度不同而引起的。 。从示踪剂测试响应解释的动态洛伦兹系数将静态和多相（动态）效应都归纳为一种异质性度量。在地质模型构建过程中，动态Lorenz系数还用于根据异质性对模型进行排名，以选择最具代表性的模型。然而，使用快速的单相/流线模拟对地质模型进行排序，并且避免了由于不利的迁移率或密度效应而导致的复杂性。这在用于表征地质模型的非均质性度量与可用于表征储层的非均质性度量之间造成了脱节。因此，根据现场数据校准地质模型是一项艰巨的任务。

在本文中，我们提出了一种工作流程，该工作流程可通过消除不利迁移率的影响来获得一个近似的非均质性度量，以弥补这一差距，该度量可与地质实现情况进行交叉核对，以选择最具代表性的模型。本文开发的工作流程适用于倒置的七点模式，主要集中于水湿储层。