Computer Application to Estimate PVT Conditions in Oil Wells in the Ecuadorian Amazon

In the study of a reservoir, petroleum engineering techniques are necessary, together with the information provided by geology, geophysics and petrophysics, to quantify the reserves and identify production mechanisms that predict a reservoir's behaviour [1-4].

The PVT analysis reflects the behaviour of the fluids in a reservoir [5]. PVT properties are measured experimentally by laboratory tests; if they are not available, they are carried out through simulations with computer applications or computer systems, determining the fluid's volumetric properties and its variation with pressure and temperature. This characterization is carried out through software packages, using algebraic equations or mathematical correlations that record specific PVT values [2, 6-9]. PVT correlations depend on the sectors or oil fields in different world regions, leading to studies or analyses of particular fields [10].

PVT analysis is necessary for understanding test results, gaining meaningful insights from oil wells [11]. This analysis establishes the correlation between in situ pressure and volume, chemical components, physical properties such as density, compressibility and viscosity, temperature effects, gas to oil ratio (GOR) [12], the specific gravity of the gas, the bubble point's pressure-volume factor [13]. PVT properties are a function of the reservoir's temperature and pressure and vary considerably depending on the concentration of chemical compounds of different hydrocarbon types. For black oil, they are required for reservoir, production and surface facility calculations, which predict the behaviour of oil fields [14].

When experimental PVT reports are not available, mathematical correlations used in decision-making in the oil industry. When a new reservoir considered, for which there is no laboratory PVT report, they performed. Approximations through mathematical correlations: Standing, Glasso, Al-Marhoun, Vásquez and Beggs, Beal, Petrosky and Farshad [2, 8, 15, 16]. The systematisation of empirical correlations brings advantages in terms of time and economic factors in studying a reservoir [17].

PVT simulators predict reservoir fluid's physical properties with varying degrees of precision, which are essential tools during reservoir evaluation. This software's accuracy is based on comparing experimental results from simulations and laboratory PVT tests [3, 18]. Simulations obtained from computer programs are consistent with data extracted from laboratory test fluid samples, determining a faster and more reliable fluid characterization [8]. In the United Arab Emirates black oil reservoir characterization investigations, using the equation of state to determine the PVT behaviour of crude oil, the results show the excellent agreement of the PVT properties using state equation compared to laboratory tests [19]. The estimates of different PVT software packages are generally consistent, as is the case of the PVT Sim software, used to characterize black oil's behaviour from the Norwegian North Sea, presenting a margin of error of 5%, considered an almost perfect predicted data [6]. The study reported by Aplin et al. [20] reflects the use of a PVT simulation that determines the physical properties of fluids such as viscosity, saturation pressure, gas-oil ratio, density, and molar volume; This study marked an overall accuracy in the prediction of the North Sea black oil PVT data. Liu et al. [21] used the PVT Sim software to determine the composition, and physical properties of black oil from the Lunnan Low Uplift sector of the Tarim basin in western China required calculations of compression, density, saturation, pressure and temperature, reporting accurate data in compositional analysis of crude. Other research indicates using an efficient black oil PVT data prediction algorithm based on brine PVT and black oil PVT data [22]. In an analysis of oil reservoirs belonging to the Gulf of Mexico, PVT data sets were generated, using a simple program under the Microsoft Excel office suite, reporting functional results, thus obtaining the study's correlations [7]. Another PVT property prediction model, based on the analysis of oil from wells from different Colombian oil basins, uses the statistical software Statgraphics Centurion XVI, which allows selecting the empirical correlations that best fit the initial conditions of Colombian oil wells [23]. Predictive models for Nigerian crude oil PVT properties were developed based on multivariate linear regression techniques using the R programming language. According to the data obtained by the correlation coefficient for the bubble pressure, volumetric oil factor and GOR properties, the study results are acceptable [24]. In Pakistan, for predicting PVT correlations, artificial neural networks were used to predict the bubble pressure, volumetric factor, and viscosity of the oil, with the lowest error results in calculating the properties' correlations [10]. The same applies to the estimation of PVT properties of oil fields in the Middle East, Malaysia and the North Sea, where they used artificial neural networks to predict bubble pressure and oil's volumetric factor [4]. In an investigation to predict PVT properties of crude oil, an algorithm called IntelliPVT, developed in MATLAB and Python, was proposed, which shows accurate results in reservoir characterisation through PVT properties such as bubble pressure, volumetric factor, GOR, viscosity and compressibility of the oil [25].

Ultimately, PVT simulators predict the physical properties of reservoir fluids with many degrees of precision, based on the nature of the fluids and the prevalence of their conditions [1, 18].

Oil exploitation in Ecuador began at the beginning of the 20th century with Anglo Ecuadorian Oilfields Limited [26]. An oil country since the 1970s and its economy has based on this resource [27, 28]. For this reason, one of the essential computer applications is the one that focuses on the calculation of the PVT conditions in an oil field, for the oil of eastern Ecuador, which allows us to ask ourselves the following question: Is it possible that a computer application, orderly and systematic, allows estimating statistically with an appropriate approximation the properties of PVT, based on initial conditions of a deposit in eastern Ecuador?

In the Oriente basin's oil wells, physical characteristics of fluids and reservoirs such as oil density, bubble pressure, Gas-Oil ratio (GOR), volumetric factor, compressibility, and viscosity of oil gas are measured with some frequency [1, 29]. These properties are obtained from the oil well's initial conditions and through the analysis of empirical correlations. Those that fit the reservoir's real behaviour, selected for a black oil model probably called ordinary oil, it is mainly characterized by the gas-oil ratio of 2 PCN/BN (Normal cubic feet / Normal barrels), gravities up to 45 °API and volumetric factors below 2 BY/BN (Reservoir barrels / Normal barrels).

This work develops a PVTTESTSYSTEM computer application, using an algorithm to estimate PVT correlations in a sample of 10 wells in eastern Ecuador, which helps decision-making in the different petroleum engineering processes. PVTTESTSYSTEM is a software application organised, user-friendly, customisable, and easy to use, which systematises and estimates empirical correlations. It is an interactive, modifiable, applicable, storable and reliable application. Furthermore, it has an information storage platform, which allows the reproduction of empirical correlations applicable to different properties and reservoirs. Accurate readability of information, showing two-dimensional graphs of the relationship between the initial conditions of the well and its different physical properties.

3.1 Description of the PVTTESTSYSTEM program

According to the algorithm used by Lasater [36], the PVTTESTSYSTEM platform developed, as shown in Figures 3-8, a computer application designed for PVT analysis of black oil that is composed of 5 modules that enable: 1) Window for entering the general data of the company and the oil field; 2) Initial data entry window: pressure, temperature, °API, solubility and specific gravity, 3) Selection of empirical correlations for each of the fluid properties; 4) Calculation of the properties of oil and gas fluids through correlations, and 5) Software reports, which depend on the minimum and maximum pressure. In addition to two-dimensional graphical reports of the relationship between pressure and fluid properties.

3.png

Figure 3. The data entry window for companies and oil fields

4.png

Figure 4. Initial input data from oils wells

5.png

Figure 5. Empirical correlation selection window

6.png

Figure 6. Oil and gas properties main results window

7.png

Figure 7. Oil and gas results window

8.png

Figure 8. Dimensional reports window (pressure relationship and fluid properties)

3.2 PVT results in terms of correlation

3.2.1 Bubble Pressure

Table 3 shows the intervals of values of the initial conditions of the ten wells. Table 4 presents the PVTTESTSYSTEM software results concerning the PVT correlations for black oil from the wells' bubble pressure fluid properties.

Table 3. Range of initial conditions of oil wells

Table 4. Empirical correlations bubble pressure properties

The ranges for each of the initial conditions apply to all properties of the fluid. However, the last column of Table 4 presents the laboratory tests' real values for each of the wells mentioned above. Therefore, the% AARE of each of the correlations determined, thus being the correlation with the lowest% AARE for the bubble pressure property of the Dokla & Osman correlation, followed by the Standing correlation.

Figure 9 shows the relationship between the different initial pressures of oil wells and the empirical correlations' bubble pressure properties. Thus, the Dokla and Osman correlation is the one that is closest to the laboratory test, followed by the Standing correlation. In contrast, the Lasater correlation is far from the bubble pressure value concerning the laboratory test. All correlations are determined under the same initial well conditions but differ in the estimation parameters, the %AARE, which indicates the degree of approximation. This difference is highlighted by the Dokla and Osman correlation applicability parameters, concerning bubble pressure which uses a range of 590 to 4640 psi and temperatures from 190 to 275 °F, as opposed to Lasater, which uses bubble pressure from 48 to 5780 psi, and temperatures from 82 to 272°F.

9.png

Figure 9. Pressures of oil wells and the empirical correlations' bubble pressure properties. (A) Laboratory test; (B) Standing correlation; (C) Dokla y Osman correlation; (D) Lasater correlation

3.2.2 Gas-oil ratio (GOR)

10.png

Figure 10. Pressure relationship and GOR properties (gas-oil relationship). (A) Laboratory test; (B) GOR correlation of Standing; (C) Glaso correlation; (D) Petrosky & Farshad correlation

Table 5 shows the empirical correlations’ results (Standing, Glaso, Petrosky & Farshad) of the GOR property (oil-gas ratio) and the real laboratory tests’ results, under the initial conditions reported in Table 3. Therefore, the Standing correlation correlates with the lowest percentage of error, and in turn, the correlation with the highest percentage of error is that of Petrosky & Farshad.

Figure 10 shows the relationship between pressure and the empirical correlations of the GOR property, according to the estimation parameters of the correlations. The %AARE indicates the degree of approximation, where 33.54% is the closest approximation (Standing), and 107.11% is the furthest approximation (Petrosky & Farshad). Therefore, the Standing correlation is close to the laboratory test results, unlike the Petrosky & Farshad correlation, which is far from the laboratory results.

3.2.3 Oil volumetric factor

Table 6 shows the results of the correlations reported by PVTTESTSYSTEM. The% AARE of each of the empirical correlations presented, giving the lowest% AARE that of the Al-Marhoun correlation, different from the Glaso correlation with the highest percentage of error.

Figure 11 shows the relationship between the pressure and the empirical correlations of the oil's volumetric factor property, according to the estimation parameters of the correlations. The %AARE indicates the closest approximation (Al-Marhoun, 2.88%) and the furthest approximation (Glaso, 4.04%). Therefore, the Al-Marhoun correlation is close to the laboratory test results, different from Glaso's correlation which is far from the laboratory results.

11.png

Figure 11. Oil pressure and volumetric factor relationship. (A) Laboratory test; (B) Al-Marhoun correlation; (C) Standing correlation; (D) Glaso correlation

Table 5. GOR properties correlation results

Table 6. Correlations volumetric factor properties oil

Table 7. Correlations compressibility properties of the oil

3.2.4 Oil compressibility

Table 7 shows the empirical correlations of oil compressibility and the laboratory tests under the initial conditions reported in Table 3. The Vásquez and Begs correlation correlate with the lowest percentage of error, whereas the correlation with the highest error percentage is Kartoatmodjo and Schmidt.

12.png

Figure 12. Relationship of pressure and compressibility of oil. (A) Laboratory test; (B) Vásquez and Beggs correlation; (C) Ahmed correlation; (D) Kartoatmodjo and Schmidt correlation

The relationship between pressure and empirical correlations of the compressibility property of oil is shown in Figure 12. Based on the %AARE of the correlations and the estimation parameters of the correlations, 44.78% close approximation and 281.07% distant approximation are determined, Vasquez and Beggs and Ahmed, respectively. Regarding the Vasquez and Beggs correlation, there is an approximation to the laboratory test results, different from the Ahmed correlation, with far data from the laboratory results.

3.2.5 Dead oil viscosity

The correlations by PVTTESTSYSTEM are shown in Table 8. The Kartoatmodjo and Schmidt correlation has the lowest error percentage, different from the Beal correlation, which has the highest error percentage.

13.png

Figure 13. Pressure relationship of the viscosity of unsaturated oil. (A) Laboratory test; (B) Kartoatmodjo and Schmidt correlation; (C) Beal's correlation. (D) Glaso correlation

The relationship between pressure and empirical correlations of dead oil viscosity showed in Figure 13, thus identifying the Kartoatmodjo and Schmidt correlation (19.56%) as the correlation that is close to the laboratory test results, different from the Beal correlation (556.23%) that is far from the laboratory results.

3.2.6 Viscosity saturate oil

PVTTESTSYSTEM reports the results of the correlations of the viscosity properties of saturated oil in Table 9. The error percentages of each of the empirical correlations presented; the lowest percentage belongs to the Chew and Connally correlation. In contrast, the correlation of Almehaideb has the highest error rate.

Table 8. Correlations viscosity properties of dead oil

Table 9. Correlations viscosity properties saturate oil

Table 10. Correlations gas compressibility factor

Figure 14 shows the relationship between pressure and empirical correlations of the saturated oil's viscosity property. Thus, the correlation of Chew and Connally (10.43%) is close to the laboratory test results, while the correlation of Almehaideb (92-14.25%) is far from the laboratory results.

3.2.7 Gas compressibility factor

The results reported by PVTTESTSYSTEM corresponding to the gas compressibility factor property shown in Table 10. They project the error percentages of each of the empirical correlations, thus obtaining the Mahmoud correlation's lowest percentage. The highest percentage of error is presented by the Brill and Beggs correlation.

Therefore, the relationship between pressure and the empirical correlations of the gas compressibility factor property is shown in Figure 15. An example is Mahmoud's correlation which has a %AARE of 14.96% due to its estimation parameters, the correlation is close to the laboratory test results, while Brill and Beggs' correlation with %AARE of 42.07% is far from the laboratory results. This case occurs because Mahmoud's estimation parameters used characterised by reduced pressures, and Brill and Begs use pseudoreduced pressures.

14.png

Figure 14. Pressure relationship and empirical correlations saturated oil viscosity. (A) Laboratory test; (B) Kartoatmodjo and Schmidt correlation; (C) Almehaideb correlation; (D) Chew and Connally correlation

15.png

Figure 15. Pressure relationship and correlations of the gas compressibility factor. (A) Laboratory test; (B) Mahmoud correlation; (C) Papay correlation, (D) Brill and Beggs correlation

3.3 PVTTESTSYSTEM algorithm's confidence level

The empirical software values report close approximations to the experimental values from the PVT tests. These values have a strong relationship concerning the initial conditions of the oil wells and the laboratory tests, a relationship measured by correlation coefficients. The application results are evaluated with a significance of 95%, thus achieving a confidence level of 85.15% of the algorithm. Table 11 shows the respective parameters that establish the algorithm's confidence interval, significance level, and confidence level [58].

Table 11. PVTTESTSYSTEM algorithm's confidence level.

PVTTESTSYSTEM is a computer application designed using a consistent PVT simulation algorithm, which predicts oil properties' behaviour in an oil field, approaching experimental results. PVTTESTSYSTEM is a computer application designed using a consistent PVT simulation algorithm, which predicts oil properties' behaviour in an oil field, approaching experimental results. In addition, it has a large-volume database for storing empirical correlations, supporting reservoir characterization procedures, contributing to oil exploration decision-making. The study applied data from eastern Ecuador's ten oil wells, obtaining approximation results in the upper quartile between the experimental field values, with values calculated by the computer tool designed.

The difference between the PVTTESTSYSTEM and experimental results, about the correlations, were: 5% for bubble pressure, 26.4% for GOR property, 0.3% volumetric factor of oil, 44% for oil compressibility, 30, 6% for dead oil viscosity, 8.9% for saturated oil viscosity and 0.6% for gas compressibility factor. These results are acceptable for each correlation's AARE percentages, indicating that PVT properties can be accurately estimated using this software.

PVTTESTSYSTEM allows the entry of different initial conditions from other oil basins that adjust to the correlations proposed in the literature or calculated on the software platform. This factor is vital in the calculation criteria of the PVT correlations since the natural environments of oil fields are diverse due to the fluid's conditions and characteristics. Therefore, it is convenient to perform various calculations with different correlations to select the best settings.