Abstract

In this paper we demonstrate the power of numerical simulation to match pressure data of complex horizontal wells of different types and reservoir heterogeneities. Analyzing large number of horizontal wells has shown that often there are cases that cannot be satisfactorily interpreted using analytical techniques. It requires more detailed investigation to handle several theoretical and practical challenges that are not normally encountered in analyzing vertical wells. The most important of these challenges are (1) the number of flow regimes in case of homogeneous infinite-acting reservoir, are 3to 4 namely; early radial, linear, late radial and/or sometimes hemiradial and(2) the reservoir heterogeneity encountered while drilling horizontal wells(i.e. layering, faults/fractures, composite systems, ...)

Finding the optimum grid settings necessary to model the flow regimes is by far the most critical aspect of the numerical simulation. Without an optimum grid, pressure response that represents numerical artifact, rather than physics of the problem, can be generated.

Three field cases are presented in this paper; a horizontal well intersecting a finite conductivity fracture, a horizontal power water injector well at the edge of the reservoir and a horizontal close to a finite conductivity fault. An integrated approach was used to match the pressure responses.

Introduction

The risk in reservoir management decisions has been increasing with increasing reservoir complexities. It is generally agreed that well tests are the only tool to predict reservoir properties under dynamic conditions. It has been seriously challenged to meet the new needs as an integral part of the reservoir description and characterization tools. The numerical modeling approach has recently been used to model wide range of complex reservoirs systems, unlike conventional analytical solutions which that are limited and simple.

There is a growing demand to analyse well tests performed in wells with complex geometry drilled in heterogeneous reservoirs that have faults, layers, permeability and fluid property variations. This often requires a numerical approach. Numerical models that use a reservoir simulator have flexibility in terms of handling spatial variation of properties and multi-phase effects. The full power of simulation is accessible only through grids capturing well and structural geometries, physical properties and flow patterns.

Our work focuses on a few practical situations at the prolific Ghawar field of Saudi Arabia where numerical well testing techniques were effectively used. The highlights of our experience have been documented in this paper.

There are ten different hydrocarbon-bearing stratigraphic horizons the Ghawar field. The field cases provided here are all from the Arab D reservoir. Geologically, the Arab D reservoir is divided into several layers: layer 1, at the top of the reservoir, is relatively tight with very low porosity. Layer 2A, directly below Layer 1, is mostly skeletal oolitic limestone with scattered vugs and local super-permeability zones. Then, Layer 2B which commonly include sdolomite and cladocoropsis based super-permeability. The reservoir quality decreases in Layer 3 and becomes extremely tight in Layer 4.

Numerical Simulator and Gridding System

The work in this paper was performed using a well test analysis package that automatically generates a simulation model and runs a reservoir simulator in the background.

A variety of gridding techniques have been successfully used in reservoir simulation(1). We use the Control Volume Method where the control volumes are Voronoi (or PEBI) cells in most of the area. This has been combined with non-Voronoi fit-for-purpose local refinements around wells. This is because Voronoi grids are ideal for honoring complex 2D geometry and can be generated automatically but they cannot be used easily in high aspect ratio situations such as refinement around fractures(2,3).

This content is only available via PDF.
You can access this article if you purchase or spend a download.