Catalog of Well-Test Responses in a Fluvial Reservoir System

Abstract

This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 180181, “Catalog of Well-Test Responses in a Fluvial Reservoir System,” by J.L. Walsh and A.C. Gringarten, Imperial College London, prepared for the 2016 SPE Europec featured at the 78th EAGE Conference and Exhibition, Vienna, Austria, 30 May–2 June. The paper has not been peer reviewed. Well-test analysis in fluvial reservoirs remains a challenge because of the depositional environment conducive to significant internal heterogeneity. Analytical models used in conventional analysis are limited to simplified channel geometries and, therefore, fail to capture key parameters such as sand-body dimensions, orientations, and connectivity, which can affect control-fluid flow and pressure behavior. The complete paper aims at a better understanding of the effect of channel content in complex fluvial channel systems on well-test-derivative responses. Methodology Geological Modeling. 3D geological models with a centrally located well were generated and populated with varying fluvial geologies. A 6950-m×6950-m×300-ft geological model was set up that allowed the averaging effects of the heterogeneities and the reservoir boundaries to be visible on the derivative at late times. Modeling the geology of a fluvial system is challenging because of changes in channel amplitude, amalgamation, and other processes through geological times, which yield highly variable distribution and shapes of fluvial deposits. Field X was modeled as isolated elliptical sand bodies and channel bodies, with sand-body dimensions of 105 m (width)×420 m (length)×5 ft (thickness) for the base case. The sand and channel bodies are schematically represented in Figs. 1 and 2. Object-oriented modeling was used instead of stochastic, sequential indicator simulation and Gaussian simulation to retain control over the modeling parameters.  Numerical Simulation. The corresponding pressure and derivative dynamic responses were generated using a proprietary finite-element simulator with a uniform grid and a fine local grid refinement (LGR) around the wellbore. The fluid was black oil at a reservoir pressure greater than the saturation pressure, and the relative permeability to water was low enough to limit water movement within the model. Results and Discussion of Base-Case Model A drawdown of 115 years was simulated for a geological model 6950 m×6950  m×300 ft with a cell size of 50 m×50 m×5 ft in the x, y, and z directions, respectively (total cell count without LGR=1,159,260), with a fine Cartesian LGR around the wellbore to reduce numerical artifacts around the wellbore (total cell count with LGR=1,327,200). The model consists of two facies. All simulations were performed without including wellbore dynamics or mechanical skin.