Development of an Expert System for Reservoir Fluid PVT Properties Correlations Ahmed Alzahabi, Texas Tech University Ahmed El-Banbi and Mohamed H. Sayyouh, Cairo University Copyright 2013, Southwestern Petroleum Short Course This paper was prepared for presentation at the 60th Annual Southwestern Petroleum Short Course, April 17-18, 2013. 1. ABSTRACT The accuracy of determination of the crude oil PVT properties is essential for solving many reservoir engineering, production engineering, reserve accurate estimates and surface production and operational problems. A large number of PVT correlations for oil exist in the petroleum literature and numerous studies are also present for with data favoring one correlation over the other. In the absence of PVT representative data from laboratory experiments, it is often difficult to choose which correlation to use to calculate different PVT properties. We approached this problem in two ways. First, we developed an expert system that checks the input parameters (e.g. reservoir parameters) against the valid ranges of input data for different correlations as cited by the author of each correlation, and then recommends which correlations to use for specific input parameters. Second, we tested all available PVT correlations for black oil on a database of selected 3500 data points of crudes to develop criteria on which correlations to use for each PVT property for any specific range of input data. These specific crudes were selected to allow testing of those guidelines on a wide range of reservoir input data for black oils. Our database included oils with o API ranging from 17 to 51, gas-oil-ratios of 8 to 7,800 scf/STB, formation volume factor at bubble point of 1.04 to 4.47 bbl/STB, bubble point pressures of 60 to 4,739 psia, and reservoir temperatures of 40 to 270 ◦F. The present work included 14 bubble points, 6 solution-gas-oil ratio, 14 formation volume factors, 13 oil compressibilities, 14 dead oil viscosities, 9 saturated oil viscosities, 10 under saturated oil viscosities ,12 under-saturated densities, 2 total formation volume factors and 2 saturated oil density correlations. An amazing match was concluded due to combining both the developed PVT-Calculator and PVT-Expert System, which made the conclusion more applicable to be tested to different PVT data points in all future applications. Based on this study, guidelines for selecting an appropriate correlation for PVT oil properties and specific guiding ranges are introduced for black oil PVT properties correlations .These guidelines are recommended in programming of PVT correlations regardless of their geographic origin. 2. INTRODUCTION Usually, PVT properties are experimentally measured in the laboratory. When such direct measurements are not available, PVT correlations from the literature are often used. Fundamentally, there are two different types of correlations in literature. The first group of correlations is developed using randomly selected datasets. Such correlations are called generic correlations. The second group of correlations is developed using a certain geographical area or a certain types of oil. Correlations using randomly selected datasets may not be suitable for certain type of oils, or certain geographical areas. Even though the authors of the generic correlations want to cover a wide range of data .Such correlations still work better for certain types of oils. Specialized correlations represent the properties of a certain type of oil or geographical area (for which they are developed) better than the general purpose correlations. The best source of oil property data is the laboratory PVT (pressure-volume-temperature) analysis of a reservoir fluid sample. However, in the absence of experimentally measured properties of reservoir fluids, these physical properties must be estimated from correlations. Many correlations for estimating crude oil PVT properties have been published in the past 50 years. Most of these correlations yield reasonably accurate results when applied at the original limitations. Here our work presents the details of the error statistics for each correlation. For comparison, error analyses were carried out for this study and for some of the more frequently used published correlations in the industry.