Abstract

Quantification of uncertainty in reservoir performance is an important part of proper economic evaluation. The uncertainty in our understanding of a given reservoir's performance (e.g. reserves at the economic limit) arises from the uncertainty in the information we have about the variables that control reservoir performance (e.g. permeability, oil water contact, etc.). The problem is complex since the influence of the variables on the reservoir performance is often non-linear. Conventional techniques often ignore this.

This paper describes the application of the method of Experimental Design to this problem. It consists of the following five steps :-

  1. Identify the ranges of the important variables that influence the recovery,

  2. Experimental Design to identify the values of the variables at which the recovery is forecast,

  3. "Conduct the experiments" - use reservoir simulation to identify the recovery for each of the experiments identified in step (ii),

  4. Analyse "the results of the experiments" and establish a generalised multivariate correlation of the recovery as a function of the variables,

  5. Validate the correlation arrived at in step (iv), and

  6. Predict the distribution of the uncertainty in the recovery.

Here the term ‘recovery’ is used in a generic sense; the technique can be employed to any parameter of interest e.g. gas flow rate after ten years of production. The paper discusses an example of the application of the Composite Design to a North Sea gas field.

This technique has potential value in providing input to evaluation of economic risks in a project in the appraisal and development stages.

Introduction

As the oil reserves from the giant oil pools decline, smaller and geologically more complex fields fill their portfolio. Often, these are marginally economic. In order to make sound business decisions on these projects, it is essential to grasp their production characteristics. In particular, it is essential to systematically identify the uncertainty in the production profiles forecast for these pools.

The uncertainty in our ability to predict the production profiles of reservoirs comes from several sources and at various levels (Massonnat 1997). The task of translating these uncertainties to the uncertainty in the production profile is a two step process. The first step involves generation of different reservoir descriptions and the second involves use of reservoir simulators to predict the fluid flow in the reservoir.

Over the last fifteen years, reservoir characterisation tools have made a qualitative change in our approach to incorporate reservoir uncertainties in our analysis. The use of stochastic modelling software have been particularly useful in generating equiprobable reservoir descriptions that honour all available data input into the analysis. In order to translate this information into uncertainty in production profiles, we need to run reservoir flow simulations with these. Unfortunately, running reservoir simulations on a large number of realisations is expensive and time consuming. This problem has been approached from two angles. The first involves the use of stream tube simulation models to speed up the flow simulation. (Tang and Behrens 1991, Thiele et al. 1994), The second involves minimising the number of equiprobable realisations over which detailed simulation work is to be carried out. Ballin etal. (1993) carried out the flow simulation of a chosen number of the equiprobable reservoir descriptions using a "detailed simulator". A "fast simulator" is used to carry out the simulation on all the realizations. The performances predicted by the two simulators are then correlated to obtain the performance projections of all the realizations.

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