A procedure for evaluating the quality and consistency of PVT data is presented. This procedure uses the material balance and the associated error analysis for investigating experimental accuracy, and to identify the sources of erratic data. The use of this method to "correct" erratic data is illustrated for a near-critical or volatile oil and a gas-condensate.
The need for sampling of petroleum reservoir fluids and the subsequent experimental determination of their properties has grown in the last decade with the increased cost and complexity of recovering hydrocarbons. Reservoir recovery mechanisms relying on compositional effects (such as CO2 flooding or gas cycling), and the construction of more elaborate surface processing plants (e.g. LNG and LPG) have created a situation where engineers must now use mathematical models to understand and predict phase behaviour in a whole range of circumstances. It has therefore become imperative that the experimental data used for generating a mathematical description of a particular fluid be of the highest quality. Despite development in the theory and use of equations of state (EOS) over the last twenty years, the commercial experiments performed on fluids to understand their behaviour have hardly changed. The gulf that now exists between the accuracy of some experiments and the accuracy actually required can result in erroneous predictions from equations of state, however good they may be.
Traditionally the more sophisticated experiments carried out on naturally occurring hydrocarbon mixtures have been closely allied to the anticipated behaviour of the originating reservoir as the pressure declines. Reservoir engineers have used this data for global material balance calculations and also for characterizing the fluid with an equation of state (EOS). This EOS characterisation is subsequently used to predict the fluid phase behaviour in numerical reservoir simulations, which can provide insight into the effects of different production strategies. In contrast to this, EOS characterizations in the field of process engineering are often primarily (if not) solely) based on an extended compositional analysis (sometimes up to C40) of the fluid. Considering the number of degrees of freedom in process design, it is not surprising that few experiments are conducted. Nevertheless, because equations of state are relied on so heavily, it should be recognised that a compositional analysis alone is insufficient to furnish a reliable characterization. The experiments of the reservoir engineer are therefore a good source of additional data.
In this paper we review the conventional PVT experiments used by reservoir engineers: the constant composition expansion, the differential liberation and the constant volume depletion. The data obtained from these experiments can be used in many different calculations, and should form the basis of EOS characterizations. Our main aim is to demonstrate an improved "methodology" for evaluating the quality of the data, whether it is needed within or outside the reservoir engineering environment. The method makes use of material balance calculations and an error analysis coupled with a new variation of the black oil PVT representation, recently presented by Drohm and Goldthorpe.
The application of the material balance calculations to experimental PVT for the calculation of fluid properties was first outlined by Reudelhuber and Hinds in 1957. However their treatment never achieved wide acceptance. In 1983 Whitson and Torp presented the work of Reudelhuber and Hinds in a clear and usable manner, using current SPE nomenclature. In addition they coupled this treatment to a method for calculating black oil PVT parameters first suggested by Dodson, Goodwill and Mayer in 1953. However, Whitson and Torp advocated the use of equations of state rather than the complicated sampling procedure favored by Dodson et al.