EQUATION PREDICTS SHRINKAGE OF HEAVY OIL/CONDENSATE BLEND

12/05/1994 A study found that, within the limited set of heavy oils and condensates examined, a third-order polynomial equation can determine the volumetric shrinkage of an oil and condensate blend solely from the condensate fraction. Volumetric shrinkage of Canadian heavy oil blended with condensate is relatively insensitive to the source or to either the crude oil or condensate densities. To allow conventional pipeline transportation, viscous heavy oil of western Canada is normally blended with low-density condensate. Mixing of hydrocarbons of different composition is seldom additive. Usually the blended volume is less than the sum of the individual component volumes. Because metering is usually by volume, shrinkage can be an important source of volumetric error, of particular importance in the custody transfer of large oil volumes. The new experimental data give different shrinkage factors than predicted by the new APT Chapter 12.3 correlation. The magnitude of the difference depends on the blend components, but in most cases measured shrinkage is higher than predicted by the APT correlation.

PREVIOUS CORRELATIONS
A correlation for predicting volumetric shrinkage during blending was developed in the early 1960s, referred to commonly as APT 2509C.1 Although this publication was last re-affirmed without modification in 1987, a completed revised document, including a new correlation, has recently been prepared and submitted for approvals. The new correlation, to be designated APT 12.3, generally predicts lower shrinkage values than APT 2509C, particularly for blends with greater than 25% of the lighter component. Both correlations were derived empirically from laboratory measurements on light crude oil mixtures with condensates. The new correlation incorporates the data used to develop 2509C, as well as results from newer studies. Only part of the data used for these correlations has been published. 2-4 All previous measurements of volumetric shrinkage have been with conventional light crude oils. Canadian heavy oils are compositionally very, different from light crudes, particularly because they contain large quantities of asphaltenes. This group of large complex molecules is believed to be solubilized or dispersed in the lighter fluid by subtle molecular organization, perhaps analogous to micelle formation of surfactants in water. 5-7 Addition of light hydrocarbons, particularly aliphatic solvents, might be expected to alter liquid organization significantly with unpredictable effects on the density of the mixture. Husky Oil operates a heavy-oil blend pipeline near Lloydminster on the Alberta-Saskatchewan border. At various pipeline entry points, blending and custody transfer are simultaneous, with Husky supplying the condensate required to mix with heavy oil from fields operated by Husky and other producers. The heavy oil and condensate are separately metered into the pipeline, and the produced blend is measured at the other end as it leaves Husky's pipeline. Shrinkage is thus one component affecting volumetric balance in the system. A good volumetric balance requires an efficient pipeline operation and credible accurate custody transfer.

MEASUREMENT METHODS

Previous volumetric shrinkage studies have involved either direct measurement of volume changes on mixing, or indirect determination of volume changes by precise density measurements. Work with crude oil blends has mostly been at 15 C. (59 F.), and at ambient atmospheric pressure or 690 kilopascals (kPa) gauge (100 psig).1 Pipeline blending of heavy oils is at an elevated temperature of 40-70 C. (104-158 F.) and pressure of 200700 kPa gauge (29-102 psig). Because Lloydminster oil viscosities at 15 C. are typically greater than 10 Pa-s (10,000 cp), elevated temperature lowers viscosity and thus makes handling and mixing more efficient. For these reasons it was considered desirable to measure at an elevated temperature. However, the blended condensates contain significant amounts of low-boiling components. Thus the measurement method must prevent loss of the more volatile components. With these criteria in mind, Novacor Research developed a modified PVT measurement method. A traditional mercury-filled loop, as in API 2509C, was rejected, partly for safety reasons, and also because mixing heavy oils with mercury often results in emulsion problems. The measurement apparatus, Fig. 1, consists of a blend mixing container (BMC), reservoirs for oil and condensate, two precision syringe pumps for introducing and withdrawing fluid from the BMC, and connecting piping and valves. The reservoirs and BMC were mounted in a temperature-controlled oven, while the pumps were contained within a thermostatted chamber. Portions of condensate and oil were displaced into the BMC by pumping water into the bottom of the appropriate reservoir, while water was withdrawn from the top of the BMC as required to maintain constant pressure. The oven was kept at 50.05 0.005 C. throughout each measurement experiment, and pressure was adjusted to 345 kPa for all volume readings. The BMC consists of two chambers separated by a rubber diaphragm. The upper chamber was completely filled with degassed water. The lower chamber contained the condensate and oil mixture. Condensate was first introduced into the lower chamber in an amount at least sufficient to fill the evacuated dead volume and begin displacing the diaphragm upwards. Oil was then added, further displacing the diaphragm, and the blend was thoroughly mixed with the mechanical stirrer in the lower chamber. The water volume required to move the oil and condensate into the BMC, and the water volume displaced from the upper chamber of the BMC were recorded. Density of each blend was separately measured to four digits using an Anton-Paar Model DMA 512 remote density cell and DMA 46 measurement unit.

SHRINKAGE FACTORS
When condensate is added to heavy oil, the blended volume is less than the sum of the condensate and oil volumes. The volume shrinkage factor of the blend, SF, is defined by Equation 1, where: Vo, Vc, and Vb refer to volumes of oil, condensate, and blend, respectively. Shrinkage factors are normally expressed as a percentage, obtained by multiplying SF by 100. Note that SF is zero when either Y. or Vc is zero because shrinkage is zero for either fluid individually. Shrinkage factors at 50 C. (SF50) were determined from the measured volumes, Y, using Equation 2. Subscripts w-o, w-c, and w-b refer to the measured displaced water volume. Proof of the equivalence of these values to the actual oil, condensate and blend volumes at experimental conditions is detailed in the box showing the conversion. Volume balance and custody transfer calculations are normally performed on values corrected to standard conditions, 15 C. and 101 kPa. Shrinkage factors at 15 C. can be calculated at 50 C. by correcting each volume as shown in Equation 3. Six densities are required: oil, condensate, and blend at both 50 and 15 C. Oil, condensate, and blend densities at 50 C. are available from direct measurements and/or by calculation from other measured values, using Equation 4. 0 Values at 15 C. were determined from the values at 50 C.

with the generalized crude oil correlation, API Table 53A,8 without the hydrometer correction. This correlation for the condensate is consistent with API recommendations for "natural or drip gasoline,"9 and follows local practice. Measured volume and density correction for the change from the measured pressure (345 kPa) to standard pressure was also considered, using either the API10 or the IP11 compressibility correlations. This correction increases the densities by 0.1 to 0.3 kg/cu m, which, when used in Equation 3, give shrinkage factors that are not significantly different from those determined without the correction. This correction has thus been neglected.

SF AT 50 C
Table I lists the origins, measured densities at 50 C. I and equivalent values at 15. C., of the oil and condensate samples used. The oil densities at 15 C. range from 930 to 993 kg/cu m (20.6-11.0 API), whereas the condensate samples had densities of 705.9 and 748.1 kg/cu m (69.0-57.6 API) at 15 C. Table 2 gives compositional data on the condensates. Shrinkage factors at 50 C. (SF50), calculated from the measured volume data using Equation 2, are in Tables 3 and 4 and in Fig. 2a. Shrinkage variations among different oil blends with the same condensate do not correlate with oil density, but show increased shrinkage as the condensate fraction in the blend increases. At condensate fractions greater than about 15%, blends with the less-dense BPU condensate exhibit more shrinkage than those with the Hardisty. Recalling by definition that shrinkage is zero at 0 and 100% condensate, the data plotted in Fig. 2a have been described by two second-order polynomials, one for blends with each of the two condensates. Table 5 gives coefficients for the best-fit equations plotted in Fig. 2a.

DENSITIES
Conversion of the measured shrinkage factors at 50 C. to standard conditions requires density values. Blend densities, db50, calculated from Equation 4 using the measured volumes, and measured oil and condensate densities are compared to the measured blend densities in Table 3. The calculated values are consistently higher than the measured densities. The differences average about 1.5 kg/cu m. Various error sources have been explored to account for this discrepancy. The most probable source of nonrandom error is in the oil densities obtained with the vibrating tube density meter. Fitzgerald, et al.,12 reported that type of instruments read anomalously high densities for fluids with viscosities greater than a few mpa-s at the measurement temperature. They proposed correction functions for specific glass measuring cells, The correction increases to about 0.5 kg/cu m as fluid viscosity approaches 100 mpa-s, and then increases more gradually to about 0.8 kg/cu m at the maximum viscosity studied (3,000 mPa-s). 12 A similar error in densities was found by Booker, 3 who corrected the densities of his oils in proportion to their viscosity, prior to determining shrinkage factors. Discrepancies in densities of viscous solutions have also been reported by Bernhardt and Pauly. 13 Unfortunately, no analogous information is available for the steel densimeter cell used for the measurements in this present work. Viscosities of the specific oil samples were not measured at 50 C., but values of 5001,000 mpa-s have been obtained for similar Lloydminster heavy oils. This viscosity-induced error will be of less significance for the blends, because large viscosity reductions are observed on addition of even a small condensate volume. The effect on the condensate densities will be negligible. If the measured values for the oil densities are too high, then blend densities calculated using Equation 4 will also be too high. Such an error would explain the discrepancy in blend densities in Table 3. Equation 4 can be rearranged to calculate oil densities consistent with both the volumetric data and the measured blend densities (Equation 5).

The oil densities derived from each data set, and the average values obtained for each heavy oil, are collected in Table 3. The calculated average densities for the heavy oils are from 1.3 to 3.5 kg/cu m lower than the measured values. Although definitive reconciliation of the volume data with the density data cannot be achieved with the available data, shrinkage factors at standard conditions have nevertheless been derived, and the consequences of the density uncertainties on the shrinkage factors can be directly examined.

SF AT 15 C.
The available volume and density measurements can be combined in more than one way to give shrinkage factors at 15 C. In Method 1, measured oil and condensate densities at 50 C. were used in Equation 4 together with measured volumes to calculate blend densities. Densities at 15 C. were obtained from the values at 50 C. using the API temperature correlation.8 The densities and volumes were then used in Equation 3 to calculate the shrinkage factors at 15 C. Results are in Tables 3 and 4. A second set of shrinkage factors was calculated using the modified oil densities listed in Tables 3 and 4, the measured condensate and blend densities, and the experimentally measured volumes, again applying Equation 3. The values calculated are designated as Method 2. The essential difference in the calculation methods is the oil density. In Method 1, the measured oil densities are used. In Method 2, oil densities are calculated from measured densities and condensate and blend volumes. Tables 3 and 4 show that the values from the two calculation methods are similar, differing by 0.04% shrinkage or less in all cases. Thus, in spite of the lack of consistency between the density and volume data, no significant error is generated in the derived shrinkage factors. The shrinkage factors at 15 C. determined by Method I (Tables 3 and 4) are plotted versus condensate fraction in Fig. 2b. Although the data points for 30% blends with Hardisty condensate are consistently lower than those with BPU condensate, the distinction is not significant in the data from blends of lower condensate fractions. This observation, the lack of data for Hardisty blends at higher condensate fractions and consideration of the experimental error gives little support for treating the two data sets independently. Numerous forms of Mathematical expression were tested, and several polynomial functions adequately described the complete data set. Equation 6 was selected as the best choice because its form (third-order polynomial) is the same as has been used to describe non-ideal behavior in many pure component systems.14 It also fits the data with similar error to other more complex expressions. Data fits from blends with each of the separate condensates are compared in Table 5 to equations of the same form as Equation 6. As anticipated from examination of the data points, data descriptions with two equations do not significantly improve the prediction.

API CORRELATIONS
Shrinkage factors for a few of the blends studied in this work, as predicted by the recently proposed API Chapter 12.3 correlation," are plotted in Fig. 2c.Consistent with the form of the correlation, the predicted shrinkage factors show significant sensitivity to both crude oil and condensate densities. This contrasts with the experimentally derived values plotted in Fig. 2b.

Comparison shows that for blends up to about 30% condensate, shrinkage predicted by Equation 6, derived from our experimental data, is similar to the API 12.3 prediction for Amoco 28 oil with BPU condensate. For all other blends, the experimentally measured shrinkage is larger than predicted by API 12-3.

REFERENCES
1. API 2509C, "Volumetric Shrinkage Resulting from Blending Hydrocarbons with Crude Oils," American Petroleum Institute, Washington D.C., 1967, Reprinted 1985. 2. Childress, H.M., and Grove, M.B., "Volumetric Shrinkage: big problem to pipe Lines blending distillates and light products into crude streams," The Petroleum Engineer, December 1953, P. 35. 3. Booker, D.R., "Volumetric Shrinkage of Spiked Crude Oils," PhD Thesis, Univ. of Exeter, England, August 1989. 4. Ashcroft, S.J., Booker, D.R., and Turner, J.C.R., "Volumetric Behavior of mixtures of crude oils and light hydrocarbons," J. Inst. Energy, Vol. 65, No. 131, 1992. 5. Koots, J.A., and Speight, J.G., "Relation of Petroleum resins to Asphaltenes," Fuel, Vol. 54, No. 179, 1973. 6. Hirchberg, A., et al., "Influence of Temperature and Pressure on Asphaltene Flocculation," SPEJ, Vol. 24, No. 3, 19&,4, P. 283. 7. Leontaritis, K.J., et al., "Descriptive accounts of thermodynamic and colloidal models of asphaltene flocculation," Proc.Sympos. Advances in Oilfield Chemistry, 195th Nat'l Meeting ACS, Toronto, June 1988. 8. Petroleum Measurement Tables, Volume Correction Factors, ASTM D 1250, 1980 (also API 2540, IP 200). 9. American Petroleum Institute, The Use of the Petroleum Measurement Tables-API Std. 2540 (Chapter 11.1), ASTM D 1250 and IP 200, 2nd Edition, June 1981. 10. American Petroleum Institute, Manual of petroleum measurement standards, Chapter 11.2.1M, Compressibility Factors for hydrocarbons: 638-1074 Kilograms per cubic metre range, 1st Edition, August 1984. 11. Institute of Petroleum, Measurement Manual Part Y section 1, Table 3. 12. Fitzgerald, H., Fitzgerald, D., and Jones, G., "An assessment of laboratory density meters," Petroleum Review, November 1992, p. 544. 13. Bernhardt, J., and Pauly, H., "Pycnometric vs. densimetric measurement of highly viscous protein solutions," J. Phys. Chem., Vol. 84, No. 158, 1980. 14. Battino, Rubin, "Volume Changes on Mixing for Binary Mixtures of Liquids," Chem. Rev., Vol. 71, No. 1, 1971, P. 5. 15. API Committee on Petroleum Measurement, Working Group, "Volumetric Shrinkage Resulting from Blending Light Hydrocarbons with Crude Oils," Chapter 12.3 draft copy, March 1994.