Preselecting Well Test Interpretation Models Using External Data

Abstract

The first stage in interpreting a well test is concerned with identifying a well test interpretation (WTI) model for the reservoir that is consistent with all the known data and whose theoretical output response is qualitatively similar to the actual well test response. Conventionally, engineers identify WTI models by using WTI software to plot functions of well test pressure against time and then diagnose components of WTI models from their characteristic curve signatures, together with evidence from external (geological and engineering) data. This paper focuses on the use of external data for preselecting WTI models, which can then be verified using conventional well test analysis techniques. The KADS methodology for developing knowledge based systems (KBS) was adopted and the inference layer structure for the WTI model preselection task was derived. The WTI model preselection task will form one part of a KBS for well test interpretation, SPIRIT, being developed jointly by Artificial Intelligence Applications Institute and Heriot-Watt University.