Revisiting Reservoir Flood-Surveillance Methods Using Streamlines

Abstract

This paper revisits classic flood-surveillance methods applied to injection/production data and demonstrates how such methods can be improved with streamline-based calculations. Classic methods rely on fixed patterns and geometric-based well-rate allocation factors (WAFs). In this paper, we compare conclusions about pattern performance from classic surveillance calculations to conclusions about pattern performance from a streamline surveillance model using flow-based WAFs. We show that very different conclusions on pattern performance can be reached, depending on which approach is used. We introduce streamline-defined, timevarying injector-centered patterns as the basic pattern unit, with offset producers being those to which the injector is connected. Such patterns give a better measure of an injector's true effectiveness because of the improved estimation of offset oil production compared to fixed, predefined patterns.In the second part of this paper, we illustrate how to build a relevant streamline-based surveillance model. We compare WAFs and offset oil production computed from much more laborintensive, history-matched flow-simulation models to that from much simpler surveillance models and illustrate the difference with a field example. As long as offset-well rates are a function of neighboring-well rates-as is typical in many waterfloodscapturing first-order flow effects is sufficient to produce a surveillance model that is useful for reservoir-engineering purposes. Properly accounting for well locations, historical rates, gross geological bodies, and major flow barriers is generally sufficient to produce a useful surveillance model that replicates well pairs and total interwell fluxes that are similar to those of more-complex and more-expensive history-matched models. We believe that this similarity arises because historical well rates already mirror reservoir connectivity, and it is well rates that mainly impact how the streamlines connect well pairs.

Recent Developments in Closed-Loop Approaches for Real-Time Mining and Petroleum Extraction

“…Other assimilation methods employ 'streamline simulation', a computationally efficient reduced-physics simulation method that also allows for rapid assimilation of oil-water fluid fronts; see Datta-Gupta and King (2007) and Batycky et al (2008). Yet another category avoids the explicit modeling of geological parameters and uses a capacitance-resistance representation of well connectivities (Sayarpour et al 2009;Zhao et al 2016), but has not (yet) found widespread application.…”

The Application of Flow Diagnostics for Reservoir Management

“…Introducing a standard single-point upwind (SPU) finite-volume discretization results in a reducible, blockstructured linear system Ax ¼ b, which can be solved very efficiently. To illustrate this, we consider Model 2 from the Tenth SPE Comparative Solution Project (Christie and Blunt 2001). This Cartesian box model consists of approximately 1.1 million grid cells and is used widely throughout the literature as a standard benchmark for new simulation methods.…”

“…To illustrate how the Lorenz coefficient can be used as a flow proxy for computing recovery factors, we will use 2D horizontal layers of the 3D SPE10 model (Christie and Blunt 2001). This model is highly heterogeneous and will (for unfavorable displacements) give flow patterns with strong fingering effects.…”

“…These patterns will in most cases not be optimal for a strongly heterogeneous reservoir. In this example, we will therefore apply our TOF algorithms to optimize the original five-spot pattern for layers of the SPE10 benchmark (Christie and Blunt 2001). That is, we will optimize the well placement, and for each placement, optimize to ensure that the rates are optimal under the constraints of a total rate and a minimum injection per well set to 1/16 of the fieldinjection rate.…”

“…Previous attempts to develop flow-diagnostic tools have focused on streamline-based simulation techniques (Datta-Gupta and King 2007), and resulting diagnostic tools have been applied to rank and upscale of geostatistical models (Idrobo et al 2000; Ates et al 2005), to optimize well rates in waterflooding (Thiele and Batycky 2003;Izgec et al 2011;Park and Datta-Gupta 2011), to perform flood surveillance on a pattern-by-pattern basis (Batycky et al 2008), and to optimize fracture stages and horizontal-well completions in tight gas reservoirs (Sehbi et al 2011). Although streamline-based methods can be very effective for these and similar purposes, they have limitations in terms of both computational complexity and extensibility.…”