Avoiding Misleading Performance Predictions for a Reservoir System with Limited Production Data by the Application of Experimental Design Technology

Abstract

The general aim of reservoir modelling is to provide a tool capable of generating reliable ranges of performance predictions on which to make development and investment decisions in a way that manages the recognised uncertainties associated with the development. The standard approach is to generate a range of static and dynamic models and parameters that capture the uncertainties, based on available sub-surface and surface information, correlations or engineering judgement. Experimental design (ED) techniques are being increasing adopted as a method to consistently combine the uncertainty parameters and to construct a range of reservoir models. The standard ED approach is to adopt global changes to the range of input parameters and seek combinations that best reproduce the historical data. Calibrating the reservoir models to available historical performance data is a vital step in providing confidence on the predicted results. However, in situations with limited performance data, great care is needed to understand the implications of model calibration so as to avoid misleading prediction results.