Correlation of black oil properties at pressures below the bubble-point

Abstract

In the practice of petroleum reservoir engineering we are often faced with the analysis of processes which require the physical properties of the reservoir fluids, but in many cases no laboratory measurements of properties are available. Our only recourse in such cases is the use of empirically derived correlations This study considers the correlation of black oil Pressure-Volume-Temperature (PVT) properties for pressures at or below the bubble-point pressure. Specifically, we address the following correlations: Bubble-point pressure, Pb *Solution gas-oil-ratio at pressures below the bubble-point, Rs ³Oil formation volume factor at pressures below the bubble-point, B0 In this work we provide a detailed analysis of the most popular correlations in the petroleum literature compared to our two databases (P<Pb, 2213 data points; P= Pb, 728 data points). Our analysis shows that all these correlations have limitations in predicting the behavior of Rs and Bo at pressures below the bubble-point. This is somewhat expected as these correlations were developed specifically to estimate fluid properties at the bubblepoint pressure. Our first effort was to correlate Pb- In this effort we tried several relations, but we ultimately used the "Standing-Petrosky" model. Regressing this equation onto our bubblepoint pressure database we obtained an absolute average error of 1 1.7%. Our approach in developing a new correlation for Rs at pressures below the bubble-point considered the use of the "reduced" variables p, and Rsr' This correlation for Rs (P< Pb) gave an absolute average error on the order of 4%. Similarly, correlating these data using our p, correlation we obtained an absolute average error of 8.4%. A similar reduced variables approach was attempted and abandoned for the correlation of B, (P< Pb), primarily due to our inability to correlate low pressure behavior. We instead used Standing's model; this correlation for B,, (P< Pb) gave an absolute average error on the order of 2%. We also developed a material balance procedure for the determination of Bo using R, from our new correlation and a pseudo-liquid density correlation published by McCain and Hill. This correlation also gave an absolute average error of approximately 2%.