CA2550406A1 - The blending of high vapour pressure diluents for reducing the viscosity of heavy oil - Google Patents

Abstract

It is postulated that the blending formula for the viscosity of two products such as bitumen and diluent is not a continuous function of the diluent amount, as has been proposed by all theories to date, but is instead a discontinuous series of at least three log-linear functions, each representing a different phase, as the fluid transitions through the three different primary phases;
gas/liquid, liquid/liquid, and solid/liquid. The log viscosity of a blend of a high vapour pressure natural gas liquid (HVP-NGL) and a heavy oil can be expressed as a simple sum of the relative mole weighting of the log viscosities of the two individual materials, as measured between the phase change boundaries. The liquid/liquid phase over the middle range of diluent % mole defines the primary range of utility in pipelines when utilizing heavy oil, as this oil usually requires at least 15 mole % to reduce the viscosity to the specification, and one must avoid asphaltine deposition beyond 85 mole % for operational reasons.
This is based on the observation that sharp discontinuities in viscosity are usually caused by phase changes. Where a phase is constant, the graph of log viscosity vs. blend mole % composition is approximately linear. It is also postulated that this curve fit is similar across a range of diluents from ethane to pentanes and possibly heavier products all the way through to light cycle oil, implying that (say) 50 mole % propane provides a similar viscosity reduction as 50 mole % pentane for any particular oil.

Description

Title: The blending of high vapour nressure diluonts for reducing the viscosity of hesvv oil The initial e,as loading phase from 0% to 15 mole % defines a different zone of economic utility for oi1 pipelines, as a flow enhancer for a stream that is within about 300% of the pipeline viscosity specification. The viscosity reduction is greatest over this initial loading range, however the range is not that large, only 1.5 /fl by mass with methane. The slope of the curve describing log viscosity vs. mole %
diluent is linear over this range as well, again running between the points at the phase/change boundaries.

In addition to knowing the viscosity of the heavy oil and diluent, and as a necessary precursor to any calibration of this basic rule, in order to establish the slope of the straight line segment in the area of interest, one must measure the viscosity near the phase change boundaries. For the liquid/liquid phase, central zone of interest, the log viscosity of the pure undiluted heavy oil (adjusted for the initial 1-2% by mass gas phase injection of the HVP-NGL) defines the left intercept of the line. The other end point occurs at an "inflection point", where the slope of the log viscosity vs. mole % diluent changes. This is Irypothesized to occur at that location just prior to where the asphaltines begin to precipitate, as this viscosity intlection appears to occur prior to the onset of precipitation.
This point is a measure of the onset of a phase change to the fluid, and it is expected that the viscosity slope changes in eonjunetion with this phase change. An accurate viscosity predicdon can then be inade once this inflection point is established (measuring viscosity at approximately 65% - 75 mole % propane in the Saskatchewan Winter Oil case used in the example below). The log viscosity line is straight between the areas of phase change of gas/liquid at about 15 mole % and liquid/solid at about 85 mole %.
Any accuracy better than about plus/minus 30% to 50% in the viscosity prediction would exceed current methods. While this error band sounds large, viscosity is an exponential function, and small changes in the slope parameter lead to large deviations in results.

The solid/:liquid phase change occurs with less propane than with hexane or heavier hydrocarbons (propane is a more active precipitator of asphaltenes), however it is a slightly less viscous diluent. It is expected that these two effects will offset and a similar mole weighting of eitller heavier or lighter HVP-NCrL would be required to acliieve a similar viscosity reduction effect, over the central range of interest.
As the visaosity of the various diluents is quite different, to see little change in the results of blending with propane or kerosene would imply that viscosity of the diluent is not that iniportant to the result.

Using this viscosity prediction technique, wherein the mole % of various light hydrocarbon diluents is similar for a similar viscosity reduction of any given heavy oil, the economic utility of operating an oil pipeline composed of H'VP-NGL and bituInen is described. The liistoric reason against the use of these diluents is that the higher pressure systems have higher costs for storage, handling and operating pressure. It is demonstrated that the use of HVP-NGL will be superior to the conventional low vapour pressure diluents utilized cun=ently for viscosity reduction, as its volume and mass is less (because a mole of propane weiglis less and takes up less space than a mole of heavier diluent). This allows for a greater proportion of heavy oil to be transported by the pipeline, all else bcing equal, and has economic utility in eircumstances where the diluent is either dead freight or carries a cost.

l3ackgroond Heavy oil and/or bitumen are used herein to include all oils witli a viscosity higher than 350 centipoise (cP) at the minimum operating temperature of the pipeline, approximately 8 degrees C in Canada. The pipeline transportation of heavy oil requires that the viscosity of this heavy oil be reduced prior to the introduction of the material into the pipeline. Several methods have been proposed historically to reduce the viscosity of the heavy oil. These include:

1. Heating the oil and insulating the pipeline such that the oil reinains hot throughout the pipeline length.

2. Thermally processing the oil to reduce the viscosity (cracking or chemical reformation).

3. Removing the asphaltene portion of the heavy oil through a solvent extraction process. As the asphaltenes are known to have a disproportionate impact on viscosity, their selective removal reduces the viscosity of the remaining heavy oil disproportionately.

4. Blending with a ligliter diluent such as naphtha or gas well condensate (a blend of hydrocarbon paraffins from C5 to CIO). The design specification for virtually all existing pipelines is that the bubble point pressure of the blend be less than atnsospheric pressure at a temperature of about 38 degrees C. This has historically precluded the use of any of the high vapor pressure fluids such as carbon dioxide, ethane, propane, ethylene and propylene. In addition, this limits the use for any volumes of butane to less than about 4% by volume.

The first 3 steps all result in an economic loss or cost, either through the shrinkage of the final sales product, or the cost of implementing the proposed solution. The 4il' step, blending with condensate or light hydrocarbons with a bubble point pressure less than or equal to atinospheric pressure, has been pursued almost -exclusively in the western Canadian basin, as a ready supply of gas well condensate exists and is already being transported out of the basin. As all the condensate available is now being used, alternative mechanisms of viscosity reduction must be developed.

The basic art of blending two dissimilar products (one being a gas at STP) for purposes of viscosity reduction was revealed by Juris Vairogs in US patent # 3,618,624, wherein the dilution effect of C02 dissolved in oil was analyzed. While no art was developed using any other diluent or any proposed mixing rules, this resulted in a series of claims (now expired) of which the first one is the most applicable to the subject matter herein.

"Therefore I claim a method for transporting a hydrocarbon material which normally exhibits a viscosity sufFtcient to inhibit flow through a pipeline comprising: a) introducing a miscible fluid, whicll is gaseous under standard conditions oftemperattire and pressure into the hydrocarbon nzaterial and, b) maintaining the temperature and pressure conditions of the pipeline so as to prevent the formation of a gaseous phase within the pipeline." With this definition, all HVPNGL's would appear to be covered, even though there is no reference to them in the patent, and little in the way of prior art applicable to this problem.

Thc key omission from patent # 3,618,624 was the explanation of economic utility of the concept, as the high vapour pressure diluents carry additional pipelnie and storage related costs over and above the low vapour pressure diluents. This omission occured because no correlation formula is described which would allow a comparison between high vapour pressure diluents and low vapour pressure diluents. As no practical application of this idea has come forward since its invention in 1971, the economic utility argument is seen to be crucial, as was the serendipitous capture of intellectual property over a broader range than was demonstrated or suggested by the art.

There is no formula that will predict the viscosity of a single heavy oil fluid based on known, nieasured parameters (that do not include at least one measurement of viscosity). "All such equations available to date however have been generally unreliable, rarely predicting better than 15%
and often in error by more than 30%

Wltile the blending of heavy oil and condensate has been ongoing for over 30 years, there exists no agreed upon formula that will clearly describe the resulting viscosity from blending a diluent with heavy oil. The predictive methods proposed in the past have ranges of uncertainty that are too large to be acceptable. "Useful correlations for calculating the viscosity of bitumen-solvent mixtures are very limited. Most correlations developed with crude oil data are of little utility since errors of 30% to 50%
are common." (2) The only accepted meahanism is to actually run experimental tests with actual fluids.at various temperatures to establish a base-line measurement. This can then accurately be interpolated over the range of temperatLue using siandard viscosity I temperature relationships.
The sensitivity with liquid/liquid mixtures to pressure is small, and can usually be ignored or simply approximated, so long as the pressure of the fluid exceeds the bubble point pressure of the niixture with a margin.

Summarv of Exfstinn Proposed Viscosity Blending Fonnulae "Gragoe (1933) proposed to use a viscosity blending equation based on a term called "liquidity" which varies inversely as the viscosity"(3). Cragoe's mixing rule for two different fluids involves using the weight fraction of each component times the liquidity value (log viscosity) for each component, plus adjustments.

"Shu (1982) proposed the use of an extended form of the Lederer equation (1933) combined with the classical Einstein equation (1956) for suspension rheology."M Shu's mixing rule for two different fluids involves using the volume fraction of each component times the log of the viscosity of each cotnponent, plus adjustments.

Mehrotra and Svrcek (1984/1985) proposed the use of the geometric mean of the mol fraction and the weight fraction, times the viscosity of each fluid, to predict the viscosity of the final fluid. The geonietric mean of a 20% weight fraction and a 60% mole fraction is the square root of 0.2 times 0.6 or 0.346. As the geometric mean of the other component is 0_566 (square root of 0.8 times 0.4), it can be seen that the sum of the geometric means is not unity, thus the line on Graph # I is not linear. This formula also requires adjustments. (5) A paper by Gateau et al in 2004 called "Heavy Oil Dilution" utilired the Lederer modification to the original classic Arrhenius expression. This modification uses the volume weigllting technique but adjusts the oil volume with an empirical colistant based on the relative density and viscosity of the unblended parameters. The effect is to reduce the volume % of the oi]
component to about 0.4 of what is actually used in both the numerator and denominator. (11) Other mixing formulae have been mentioned in the paper cited as reference (10). ln the following summary, taken from that reference, the value of x (the weighting) is the mole %.

"From the different expressions existing in the literature, the foiiowing equations reiating viscosities of binary mixtures as a function of those of the pure components were seiect.ed:

McAllister (1960) (for three body interactions):

la (+-M) gN (vtMl)+x32lsi(vzM2)+3xi X3 la(V12 M12)+3x1 xsin(v31 M21) 11ds = (2M; + M j)!3 where:
In the above equations M, and ol are the molecular mass and the kinematic viscosity of the ith componenk, respectively. ~lp and 021 are adjustable parameters. M is the mean molecuiar mass of the mixture computed as M = x, M, + xz Ms.

k:atti and Chaudri (1964):

1S1 (v IVL] X I]t1 (v i M1) + X2ln( v z M2) 'F' X 1 X2 PIT
where W,1, is the interaction parameter.

Grumberg and Nissam (1949):

}117i -SllriT11 +X2 lri'rla +SI S2 d where d is an interaction parameter that is a function of the nature of the components and temperature. This parameter has been regarded as a measure of the strength of the interaction between the components.' (10) Mehrotra summarices the entire background of viscosity rules for both single fluids and binary mixtures.
There are 24 different systems of rules described, from 1960 to the present.
(12) All formulae represent continuous fiuictions, as far as 1 could detennine.

The Saskatchewan Research Council (SRC) paper (2005) is the major recent source for a propane experiment and it describes the state of the art for all viscosity mixing rules. (6) "For the heavier solvents such as propane, n-butane and i-butane, even less information is available.
Jacoby (19$6) measured the effects of ethane, propane and butane on viscosity (this is incorrect, Jacoby measured only the e,~''ect of propane on viscosity, thc: ethane and butane tests were done on extractirtg asphalt from the bitumen) ...... Although some other sources of similar types of data may exist, we are not aware of them, and they are certain to be few in number. It appears that the available information is tlot sufficient to develop correlations for the desired properties: '(6) They curve fit their viscosity data using the Josse-Steil-Thodos correlation and states that they achieved a reasonable fit to viscosity. A variation of this model is available on the internet at:

tto.Jfekofisk stanford edu/software/gem html#VISCOSITY CORRELATION SPECII'I
ATIO
N (OP'I'IOA1ALl This internet model utilizes the mixing rule of Herning and Zipperer (Gas und Wasserfach, v. 79,49 (1936) and then adjus[s it with the .losse-Steil-Thodos formula (A1ChE J, v.
8, 59 (1962)).

All of the above formul.ae involve the use of additional numerical parameters (some dependent on temperature, some imputed as constants) to better fit the formulae to the real world tests for specific oils and diluents over specific ranges of temperatttre, blend % and pressure, such parameters sometimes not being based on any particular sciejitific theory, aside $om the need to develop a better curve fit. The use of these additional parameters, and the implied lack of understanding of the basic science of viscosity, makes extrapolation of the rules very risky beyond the ranges and components actually measured.

Ilefirtinm the Problem In order to determine the economic utility of the concept of utilizing HVP-NGL
as a heavy oil diluent, it is important to establish a formula which will accurately predict the viscosity resulting from the blending of a heavy oil and HVP-NGL. The costlbenefit must be better than with dilbit (condensate/heavy oil blend) to create economic utility. As the cost of an NOL
based system will undoubtedly be higher because of the higher pressure handling systems around the pipeline, there must be an offsctting benefit to provide utility. Using the mole % blending rule described herein, the volume and mass of HVP-NGL diluent would be reduced as compared to heavier condensate to achieve a similar viscosity reduction effect, implying utility. The area of utility as regards diluent % will be bounded on its lowest % diluent limit based on the maximum allowed viscosity on the oil pipeline at its operating temperature. It will also be bounded on some upper limit. This will either relate to the tendency of the diluent to cause asphaltenes to precipitate out of the bitumen leading to issues of plugging and fouling or it will relate to the maximum allowed bubble point pressure on the pipeline (as grcater amounts of NGL diluent leads to higher bubble point pressure and lack of utilization of the full pressure drop potential between stations). The asphaltene precipitation limit for Athabasca tar is known to occur with a sudden onset above about 30% by mass of 14VPNGL diluent ( approx 85 mole %) and is dependent on the bitumen (higher asphalt levels in the basic bitumen has a sudden onset with less propane) and the diiuent (lighter hydrocarbons cause a more sudden onset of precipitation). (14) These upper and lower limits are known to those skilled in the arts, thus it is the viscosity calculation, and it alone, which prevents fiuther developinent of the art. Most historic work in this area takes as a starting point that the viscosity/temperature relationship of each of the two fluids is known, thus the resulting viscosity depends primarily on developing a mixing rule that can accommodate a wide range of heavy oil viscosities (from 10,000 centipoise to 10 million centipoises at 10 degrees C for example) and a wide range of diluents (from ethane to kerosene for example).

The range of temperature over which the invention provldes utility depends on the bitumen. As is shown in Graph 10, some bitumens achieve a pipeline viscosity specification without any diluent at about 50 degrees C. Other heavier bitumens require a temperature of about 90 degrees C.
The comparison between viscositv and tem =re. over the range of econonric utility between about 10 degrees C and 90 degrees C and fior all the components of interest, describes an approximate straight line on a log (viscosity) vs. temperature graph (actually a slight curvature is shown in Graph 10 but is immaterial to this general discussion), and most of the different lines for the different components of interest arc roughly parallel.

The bitumen line and the diluent line can be placed on the same graph as shown in graph 1. The viscosity of the blend is expected to be approxiatated by a third line, parallel to (or not depending on other factors) and sornewhere betwe n the individual component lines. Tlie location between the two extremes will depend on the amount of the two relative components. Therefore, all viscosity blending formulae at their most basic level follow the general model that the log of the viscosity of the blend is equal to the amoun.t of component 1 times the log of its viscosity plus the amount of component 2 times the log of its viscosity. This is shown in gaph 1 and the following formula.

log (viscosity-mix) _ [bitumen % * log (viscosity-bitumen)] plus [diluent % *
log (viscosity-diluent)]
Graph 1- Viscosity vs Temperature (various blends of bitumen and diluent) 1000000.00 - -=
100000.00 ~-.. ~
a 10000.00 - ~--~
~ 1000.00 --N -e nd7s ci 100.00 - ~ ~- --~1 ~ Blend50 10.00 ---- ~ -"sc - Blend25 0 --0-Diluent j 1.00 0.10 X
0.01 ___- -- - ~_-_ _-- - -----Temperature C

There are several questions that have not been clearly answered with the prior art.

The major question is "What basis should be used to predict the % amount of the coinponents; mass, volume, moles or a combination of these elements?" This is a source of major conftision. As an example, none of the works referenced herein aside from the analysis on TAME
(ether) provided sufficient data on the molecular weight and the density of the two fluids so that accurate transcriptions between tlzese three reference scales could be determined with any certainty.
Most of the experimental data was reported on a mass basis.

If the two temperature/viscosity lines are not parallel in the first place (the two components react differmtly to temperature), most historic formulae assume that the averaging technique still applies, at all temperatures. However, this then sometimes requires that additional parameters be introduced to fit the curves. Many blending foxmulae also include these additional paraineters, and in different combinations have been applied to every factor in this "basic" formula to improve the curve fit over whatever range of components and specifications was being investigated.
Ideally, a relationship to temperature for all contemplated HVP-NGL blends could be developed, not requiring 4dditional parameters. In most cases, these additional parameters mainly apply to the tails of this curve, at very high or low temperatures where this departure from a linear line is most apparent. With NGL blends in pipelines, the area of economic utility will be in the 10 - 90 degree C range, thus the tails are less important. Note that the actual log graphs used for establishing the relationship between temperature and viscosity are not the classical log scales used by mathematics. They are derived log scales so that the curving line shown in Gr$ph # 10 is represented as a straight line.

Another question relates to the use of the gas or liquid viscosity for the HVP-NGL as they are different and will give rise to different results. The propane gas viscosity is about one order of magnitude less than its liquid viscosity (approximately 0.01 vs. 0.1 cP at typical pipeline conditions) (9). Gas viscosity increases with rising ternperature as compared to liquid viscosity which reduces with rising teraperature.
Mehrotra et al believes that it is the gas viscosity that is the defining feature of any mixing formula using light hydrocarbon gases at or near the bubble point. (5) US Patent # 3,618,624 describes the swelling factor of using light hydrocarbon gases. "It has been known in the art of pipelining transportation that the solution of ligltt hydrocarbon gases such as methane and ethane in hydrocarbon liquids swells the volume of the resulting solution. In addition to a definite volume change there also is an apparent decrease in the viscosity and density of the resulting hydrocarbon miscible gas solution."

There is very little literature on the impact of using olefins instead of paraffins as the diluent. A footnote on Figure 23-35 of reference (9) states " Olefins are approximat.ely 15% less viscous than the corresponding normal paraffin hydrocarbon." Ethylene has a lower critical temperature than ethane (9.2 vs. 32.2 degrees C), such that the benefit from lower viscosity would be offset by a higher bubble point pressure. The critical temperature of propylene is almost the same as propane (91.8 vs. 96.7 degrees C), such that the benefit of lower viscosity would not liave an offsetting higher cost from a higher bubble point pressure. Butylene is also similar to butane in regards to critical temperature. Propylene and butylene would therefore be preferred diluents, all else being equal.

The unadjusted Peng-Robinson equation of state does not accurately predict the resulting blend density of two liquids. The "volume-adjusted" formula does, based on a correlation developed by Peneloux et al (1982) (15).

The increase in viscosity for solid/liquid slurries is sreater than with pure liquids. In one series of coal/oil slurry experiments, a 50/5 0 (mass) slurry had a viscosity seven times greater than the 100% oil carrier fluid. (19) Historic Ex erimental Rmults No record exists that I am aware of any systematic testing done in the area of bitumen and HVP-NGL
blending, across a rmige of oils and diluents, sufficient to develop a viscosity nlixing rule. The first tests encounterad were done by Jacoby (1986) and were primarily to, detemine if Athabasca bitunien was even soluble in light hydrocarbons (14). The state of the art at the time was summarized by Jacoby as "In 1950, Blair stated that propane could not be used to extract Athahasca tar from its sand matrix because the tar was essentially insoluble in propane". Jacoby's paper then revealed the discovery that at a low propane concentration, the oil was 100% soluble. The paper went on to determine at what point the asphaltene begins to precipitate wlien using light hydrocarbons such as pentane, butane, propane and ethane. I-ie measurexl viscosity at various (high) propane concentrations and noted that there appeared to be a phase change, as illustrated by the viscosity of the blend going through fui inflection point, with two different lines above and below the inflection point on the graph of log viscosiiy vs. temperature. He provided very little viscosity data at the low blending % contemplated herein and did uot provide th.e viscosity of the base Athabasca tar (it could fluctuate by several orders of magnitude), but noted several important features/discoveries:

"At high propane concentrations they (the asphalrines) do not dissolve (the bitumen is nor 100%
soluble) but near 60 weight % tar, they do indeed approach 100% solubility!"

"'Tar is more soluble in butane than propane, and more soluble in pentane than buiane..... Note that the temperature and pressure have much less effect on solubility than mixture composition"

"Propane (for exarnple) may be added to Atliabasca tar up to about 30 weight %
before any solids will precipitate."

"Thus, the threshold of (asphahine) precipitation will vary for each crude, being higher for high degree API crudes and lower for heavier crudes, but in general for low API oils, it is in the neighborhood of 35 - 40 weight % for propane."

"These phase behavior and viscosity data have implications for a surface separation process which might be better than r.h.e present hot water process. The raw tar could be mixed with just enough propane in an enclosed vessel at relatively low pressure (less than propane vapor pressure) to obtain a suitable viscosity for rapid settling out of the sand."

"Still further downstream, propane extraetion might be used as a refining process to sepa.rate product streams having specific properties..... It was observed that the first 20 weight % of the tar extracted bad a bright, clear yellow color, the first 37 % was still transparent and yellow-orange in color. At the 50-55% level of extraction, the tar oil was reddish-brown and at the 69% level, the oil was brownish-black and completely opaque."

"The higher the initial oil viscosity, the more dramatic is the viscosity reduction per mol of propane added."

In regards to the inflection point on viscosity as one passes tluough the asphaltene precipitation inflection point "Although this is indirect evidence, it is inferred that some solid began to precipitate at the discontinuity temperature, causing the (viscosity measuremerrt tool) rolling bali to slow down."

The tests done by the Saskatchewan Research Council in 2005 were with a heavy oil called Saskatchewan Winter Oi1(density 973.1 kg/M3 at 15 degrees C, viscosity of 5880 mPa at 15 degrees C, MW 412). The purpose of the tests was to calibrate a new PVT apparatus, not to advance the art. The analysis used propane for a series of tests at difi'erent blending amounts and at two different temperattares. The density of the propane is not provided, and as most propane contains some volume of ethane and other components (the specification relates primarily to the overall vapour pressure of the propsne), this is an important factor when developing a correlation to density. One paper by Subramanian and Hanson (2205) describes the received propane as containing by weight composition 0.2% (C 1), 5.4% (C2), 93.9% (C3) and 0.5~la (C4).(13) This material, because of the relative impact of ethane, will have a lower density, vapour pressure and viscosity than pure propane. With only one oil and one diluent (and the diluent composition suspect), a robust formu]a cannot be developed and the paper provides little advance in the prior art, aside from accurate measurements under one set of conditions.

Mass weighting in a blending formula implies that it is the mass/size of molecules that gives rise to the underlying physics of viscosity. Volume weighting iniplies that it is the space taken up by the molecules in unblended form that is important. Mole weighting implies that it is an electrica] bonding phenomena, depending in part on the uin nber of molecules and their polarities. As the plienomena is primarily an electrical phenomena with London and van der Waal forces and hydrogen bonding being important (I 1), intuition would prefer a formula where the determining factor is the number of molecules, not the mass of the molecule nor the space it normally takes up. In other words, one would intuitively prefer to use the mole %.

While the differences between using mass, volume and/or mole weighting for these fluids is small when using diluents of a similar molecular weight, it becomes very large when comparing diluents with significantly different molecular weights or densities. This is shown in Table 1 below which looks at a 50150 mole blend of various diluents and provides the blend equivalent % in terms of moles, mass and volume. A weighting formula comparing ethane and decane as a blending diluent for bitumen would weight them the same using moles (50/50), but would weight the decane four times larger (l 9.8% vs.

5.0%) using mass and twice as large (25.2% vs. 12.8 /a) using volume (assuming no swelling or fluffing effect). Because of this large range, finding a fornlula that works well with dccane through to ethane would go a long way towards an tmderstanding of the underlying science of viscosity blending.

Table 1- Constant Mole % with Different Diluents - Resulting Mass % and Volume %
Ethane kglkgm Mole % MW Weight kg Wt% kgmlM3 M3 Vol %
C2 5010% 30.07 15.0 5.0% 356.2 0.0001 12.8%
Situmen 50.0% 577 288.5 95% 1002.0 0.0009 87.2 k Calculated 100% 303.5 100% 919.4 0.0010878 100%
Propane kg/kgm Mole % MW Weight kg Wt% kgmIM3 M3 Vol %
C3 50.0 /6 44.097 22.0 7.1% 507.0 0.0001 13.1%
Bitumen 50.0% 577 288.5 93% 1002.0 0.0009 86.9 !0 Calculated 100% 310.5 100% 937.0 0.0010672 100%
Hexane kg/kgm Mole % MW Weight kg Wt% kgm/M3 M3 Vol %
nC6 50.0% 86.177 43.1 13.0% 664.0 0.0002 18.4%
Bitumen 50.0% 577 288.5 87.0% 1002.0 0.0009 81.6 /v Calculated 100% 331.6 100% 939.8 0.001064 100%
Decane kg/kgm Mole % MW Weight kg Wt% kgm/M3 M3 Vol %
n-C10 50.0% 142.285 71.1 19.8% 734.2 0.0003 25.2%
Bitumen 50.0% 577 288.5 80.2% 1002.0 0.0008 74.8%
Calculated 100% 359.8 100% 934.6 0.00107 100%

Until blending viscosity tests are performed, one cannot extrapolate the results achieved with the range of condensate diluents cunently used by industry (density of 650 - 750 kg/lvl3). They do not necessarily apply to propane (507 kglM3) or ethane (375 kg/M3) or their olefin counterparts. The conclusion from this lustoric review is that no acceptable blending rule exists which can accurately prcdict the antount of high vapour pressure fluid that is required for diluting heavy oil to a specific viscosity. Without an understanding of this basic feature, which benefit must drive the economic utility of the concept, using high vapour pressure fluids as dilueiit will not progress.

t19~ sis The data for the following graphs is t.al:en from the Saskatchewan Research Council paper describing a new PVT apparatus, and uses only the graphs as presented in the paper. (6) The estimates are taken from the graphs in the paper using a ruler as the measuring device. The SRC Winter Oil/Propane tests were analyzed in detail in order to provide a single correlation between propane %
and viscosity reduction of the Winter oil utilized in the tests. The bitumen composition is described in detail making the analysis partially eontrolled, the propane composition was inferred by me, as it was not provided.

The tests were done at 15 C and 28 C and with mole % ranging from 20% to 95%.
The results are shown below in Graph # 3 as a function of mo % propane. The viscosity results were closely curve fit by SRC
to a inodified Josse-St.eil-Thodos viscosity blending formula up to about 70 mole % propane and it is this curve fit, not the actual data samples, that are illustrated in Graphs #
3 - S. (It is this author's view that witli a suitable choice of parameters, virtually all viscosity blending rules can be fit to this result so this result is not stuprising). The data points beyond 70 mole % propane came from the actual asphaltene precipitation tests as listed in a table. The viscosity is illustrated using a log scale and the goal is to achieve a linear curve fit on this log-linear scale over the range of expected utility. Using mole % as is done in Graph # 2 shows a straight line curve fit over the 0% - 70 mole %
range. The dotted line on the graphs represent a straight line between the 100% oil viscosity and a viscosity about fourteen times the propane viscosity for reference only (or about 1.4 mPas or 1.4 cP vs. 0.1 cP
viscosity for liquid propane). Graph # 3 extrapolates the log-linear curve to 100% propane and it can be seen that the curve changes slope over the 70% - 100% range, the measured viscosity would be less than predicted by looking just at the 0% to 700/o range. This is confnmed with the series of asphaltene tests run at 28 degrees C beyond 70 mole % propane, illustrated in Graph # 6. As this extended range falls outside of the area of econoniic utility because of asphaltene deposition issues, the part of the range from 0% to 70% is the most important. The SRC experiments with asphalt precipitation show that it occurs with an oiiset somewhere between 80 mole % and 85 mole % propane content.

The same analysis Is illustrated using mass % propane (Graph # 4) and volume %
propane (Graph #5).
The mass correlation depends only upon the molecular weight of the molecules and the mole %, therefore any density fluffing factor does not affect the results of using a mass weighting technique.
Graph # 4 shows that the curve is not linear over the range analyzed, about 2%
to 20% mass propane.
The vol_ume correlation also does not exist, Graph # 5 clearly illustrates this.

These graphs illustrate that, at least for this particular oil at this particular temperature, the relationship between viscosity and % diluent is approximately log-linear with respect to mole weighting over the expected range of utility. The same correlation was found at 28 degrees C, shown in Graphs # 6 - S. The use of two viscosity measurements at 15 and 28 degrees of any fluid is usually sufficient to establish its log-linear viscosity relationship to temperature over a broader range as temperature extrapolation is allowed as long as there are no phase changes.

One small expected deviation from this straight line can be seen in Graphs # 3 and # 6 between 0% and 15 mole %_ If the left int.ercept of 100% oil is reduced by a small factor, and the right intercept at ] 00 lo diluent is increased slightly to about 2.5 cp, the resultant is improved. The lowering of the left intercept could be a result of the oft.en reported phenomena of gas loading and the initial large viscosity drop from this 1- 2 % mass dissolved gas. This is similar to a phase cliange at that point when the injected light hydrocarbon changes from gas loading to liquid loading. The initial benefit does not go away as additional propane is added, therefore one would expect to see a permanent reduction in the left intercept to reflect the initial gas loading.

Graph #Z - Viscosity Winter OiI/C3 vs C3% -15 C
1000 ~'' - -_ =~
;

Asphalt precipitation tests 0.1 Mole % q Graph #3 - Vlscosity Winter QiUC3 vs C3% -1 S C
fioGIOp 1 000 Area of Asphaft on Depositi ~ .

propane liquld 0.1 0 20 40 80 $0 100 Mole % C3 Graph #4 Viscosity Wirrter CillC3 vs C3% -16 C
"iOOO

- ' ti-.. - ---r 1 - - - ~
0_1 0% 10% 20% 300/a 40% 50% 60% 70% 80% 90% 100%
Mass gb C3 Graph #5 -Viscosity Winter 4iUC3 vs C3% -16 C
'loeo -~ -~, - - -- -0.1 0% 10% 20% 30% 40% 50% 60% 70% 80% 9096 100%
Volume % C3 Graph # 8- Viscosity Winter C1i11C3 vs C3% - 28 C
10000 =-- - _ tai 100 ~- - o - - = -0.1 Mole k C3 Graph # 7- Viscosity Winter OiI/C3 vs C3% - 28 C

~ - -CL

~ -N - .~ -1 ~' 0.1 - -0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Mass Jo C3 Graph #8- Viscosity Vlfinter OiI/C3 vs C3% - 28 C

~ ,~ _ -- -10000 ~ - - . _ -a -- 1 -_ - -0.1 0% 20% 40% 60% 80% 100%
Volumre k C3 The results of the above analysis are sumniarized in Table 2 below, where the blending formula is defined as the following formula requiring a single adjustment factor (AF) to the right intercept, 100%
diluent viscosity value:

log viscosity (blend) = [diluent mole %* (log viscosity (diluent) + AF] + [oil mole % * log viscosity (oil)]
TABLE 2- Results of Viscosity Prediction Adjustment Factor 1.3 1.4 Measured Viscosity Deviation Deviation Mole% mPa-s Predicted % Predicted %
Propane 15 C 28 C Viscosity Viscosity 0% 5880 1630 15 C 28 C
10%
15% 1250 580 1691 35% 571 2%
20% 900 450 1116 24% 402 11%
30% 480 225 486 1% 200 11%
40% 220 110 212 4% 99 10%
50% 100 50 92 8% 49 1%
60% 40 20 40 1% 24 22%
70%
100% 0.15 0.09 Reducing the left intercept and increasing the right intercept would improve tlle fit as the deviation in the two series above is consistent on one side or the other, however the inaccuracy in my extracting information from the very small graphs in the various papers is also fairly large to be extracting such detail. With proper refinement, and over a ranse of economic utility, this technique should match other techniques in its predictive aiid correlative ability. This tecluiique appears to be capable of improving on the 30% - 50% error range seen with otlier predictive techniques aiul is cerlainiy much simpler, implying that exn-apolation, wliich is the ultimate goal, oAn be better demonstrated.
It also is based upon a physical explanation for the slope clianges that can be tested and refined.

Graph 9- Influence of Dilution on the Viscosity of Venezuelan Crude at 20 degrees C (SPE 69711 - Argillier et al - 2001) AAolecular vUeight 100000 Bitumen - so0 Waphtha - 120 Kerosene - 170 a 10000 LCO - 250 _ 1000 -;-Kerosene -a - ~ w-Naphthe '-- ~~
= ~w 0% 20% 40% 60% 80Q/o 100%
Mole %Diluent Graph # 9 provides an analysis of the Argillier et al paper (2001) and it shows that this basic viscosity prediction technique derived with a propane/bitumen blend can also be applied to heavier diluents. In this experiment, they tested the resulting viscosity and asphrilt deposition of a heavy Venezuelan bitumen (base viscosity of 350,000 cP at 20 degrees C) blended with various heavy diluents such as naphthtt, kerosene and an LCO refinery cut. (20) They expressed the data (Figure 5 in their analysis) on a mass % basis. Using mass % as the correlation, there is the expected scatter of results as seen in the SRC
propane tests. When this graph was re-expressed based on mole % shown in Graph # 9, the same features occurred as with the propane in Graph # 3. The log viscosity vs. mole % line is basically straiglit over the interval of interest, it intersects tlte right intercept at about 3 cl', and all three diluents give almost the satne result (i.e. the lines are superinlposed on the graph).
Gas loading appears to be absent which is expected, these liquids are not gases at STP, This gives support to the extrapolation of the viscosity prediction technique to heavier diluents and further implies that using a mole % blending formula is the correct approximation to use. The paper provided little inforination on the density dr the molecular weight of the bitumen and diluents, so Graph # 9 is based on my estimation of the molecular weight of the various products. The Argillier paper did report the viscosity of the diluents as naphtlia (0.6 cP), kerosene (0.94 cP) and LCO (3.7 cP) all at 20 degrees C, and it suggested that the viscosity difference between the diluents gave rise to the (apparent) scatter in the blend viscosity between the three diluents when expressed on a mass % basis. Given the results shown in Graph # 9, it is clear that the viscosity of the diluent is not a large determinant of the resulting blend viscosity (the results are actually ihe inverse of what one would expect if diluent viscosity were an important feature). The mole % content is far more important and the discrepancies seen on a mass % basis will be largely eliminated by changing this viewpoint.

Table # 3- Actual Blending Comparison with Wintar Oil vs. Various Diluents -No Swelling Factor Mole % kg/kgm Weight kg Wt96 kgmlM3 M3 Vol %
G2 40.00% 30.07 12.028 4.28% 356,19 0,0001 11.7%
Winter Oil 60.0 .6 412 247.2 88.0% 973.10 0.0009 88.3%
Calculated 100% 259.228 92% 900.72 0.0010242 100%

Mole % kglkgm Weight kg Wt9'a kgm/M3 M3 Vol %
C3 40.0% 44.097 17.6388 6.611/o 506.99 0.0001 12.0%
Wnter Oil 60.0% 412 247.2 92.1% 973.10 0.0009 88.0%
Calculated 100% 264.8388 99% 916.95 0.0010756 100%

Mole % kg/kgm Weight kg Wt% kgm/M3 M3 Vol %
nC4 40.0% 58.123 23.2492 8.7% 584,01 0.0001 13.5%
Winter Oil 60.0% 412 247.2 92.1% 973.10 0.0009 86.5%
Calculated 100% 270,4492 101% 920.39 0,0010943 100%

Mole % kg/kgm Weight kg Wt% kgm/M3 M3 Vol %
nC5 40.0% 72.15 28.86 10.7% 631.12 0.0002 15.3%
Winter Oil 60.0% 412 247.2 92.1% 973,10 0.0009 84.7%
Calculated 100% 276.05 103% 920.93 0.0011164 100%

Mole % kg/k8m Weight kg Wt% kgmIM3 M3 Vol %
nC6 40.0 i6 86.177 34.4708 11.5 !0 663.83 0.0002 17.0%
Winter Oil 60.0% 412 247.2 82.3% 973.10 0.0008 83.0%
Calculated 100% 281,6708 94% 920.61 0.001019 100%

Mole % kglkgm Weight kg Wt% kgm/M3 M3 Vol %
nC7 40.09i; 100.204 40.0816 14.9% 688.20 0.0002 18.7%
Winter Oil 60.0% 412 247,2 92.1% 973.10 0.0009 81.3%
Calculated 100% 287,2816 107% 919.96 0.001163 100%

Mole % kg/kgm Weight kg Wt% kgm1M3 M3 Vol %
nCe 40.0% 114.231 46.8924 17.0% 706.96 0.0002 20.3%
Winter Oii 60.0% 412 247.2 92.1% 973,10 0.0009 79.7%
Calculated 100% 292,8924 109% 919.12 0.0011868 100%

Mole % kglkgm Weight kg Wt% kgm/M3 M3 Vol %
nCa 40.0% 128.258 51.3032 19.1% 721.87 0.0003 21.9%
Winter Oil 60.0% 412 247.2 92.1% 973.10 0.0009 78.1%
Calculated 100% 298.5032 111% 918.18 0.0012107 100%

Mole % kg/kgm Weight kg Wt% kgm/M3 M3 Vol %
Qecane 40.0% 142.285 56.914 17.8% 734.21 0.0002 23.4%
Winter Oil 60.0% 412 247.2 77.411A 973.10 0.0008 76.6%
Calculated 100% 304,114 95% 917_25 0.0010378 100%

With all propane products actually tested by various investigat.ors, the expectation is that the maximum amount of ethane had been blended into the product at the source, and sold as propane. This is standard itidustry practice. The impact of even 5% ethane is quite significant, in regards to bubble point pressure, where ethane is far more significant than propane, but should have minimal impact on viscosity. Any tests must take this into account. I=iowever, even with this, the range of diluent quality with propane (or ethane or butane) is an order of magnitude less than with "condensate", making extrapolation of a mixing rule simply dependent on the oil, not the oil and a variable diluent.

The basic technique is applied to several Alberta heavy oils and bitumens and is shown in Graph # 10.
IJach bitumen's viscosity is estimated by the use of a formula (derived from historic experiments) which relates viscosity to temperature based on the exponertrial formula: viscosity = 10 ~{(I0~ bl) * (T ~
b2)]. The viscosity result is shown for the bitumens between 10 degrees and 100 degrees C. Using the x axis co-ordinate now as mole % instead of temperature, for the super-heavy bitumen, if the specification is 350 cp at 10 degrees C, find the 10 degree C point, nlove to the left intersect at zero % diluent, and draw a straight line to the right intersect at about 2.5 cp. The new line crosses the specification at 70 mole % diluent. This same technique is used for Winter oil and results in a 40 mole % blend to achieve 350 cp at 10 degrees C. All other bitumens fall in between.

Graph 'I 0- Viscosity Pure Bitumens vs. Temperature (overlaid with predictive technique for C3) 1.E+08 1.E+07 - = Range of Utility --r-Cald Lake 1.E+06 -a- Peace Rluer Q' -U 1.E+06 - -- *-Athabasca .o= Uoydminster ~ 1.E+04 g Wabasca -o- Mildred Lake 1.E+03 oSuper I-feavy 1.E+02 Pipeiine Spec . - - ' - -~- Winter -+-Aitex 1.E+01 _ == ='=
Prediction 40% to 70% C3 "' = r 1.E-h00 degrees C (and mole~'o for prediction) In conclusion, while the SRC paper was sufficiently detailed to reveal a basic structural understandings for a viscosity blending formula based on a single oil and a single diluent across a range of diluent percentage when analyzed in the above fashion, the wort: was insufficient to actually develop the formula across a range of diluents or oils. The SRC work cannot discern which parameters must be added to a mole % weighting to accurately predict reality within the range of utility, however it appears to eliminate either mass or volume % weighting from the options. This patent disclosure postulates that a straight line on the log-viscosity vs. mole % graph will be seen over a range of bitumens and diluents over the area of economic utility. The SRC paper supports the contention for a single diluent propane and a single oil. prom the Argillier et al papar, it is postulated that this same technique can be applied over a broad range of light hydrocarbon diluents and bitumens, and yield siniilar results.

On the assuniption that an equal number of moles of either decane or propane are required to achieve a similar viscosity, the following conipares the mass and volume composition based on a 70 mole %
prapane / sixper-heavy blend from above. As can be seen in Table 4 below, 173.1 kg of bitumen requires either 30_9 kg of propane or 99.6 kg of decane. As the dollar cost/value of the various diluents is roughly equivalent to mass, this implies that using propane involves acquiring diluent at only 31% of the cost, per barrel of bitumen. Using propane, the amount of bitumen transported increases from 56% to 74% of the total barrels transported, an increase of one-third. Eeonomic trrility has been demonstrated through these comparisons. This calculation will apply directionally to all light hydrocarbon diluents as predicting greater utility with fewer carbons in the molecule.

Table 4- Propane vs. Decane as a Blending Material Propane k9/kgm Mole % MW Weight kg Wt% kgmlM3 M3 Vol %
C3 70.0% 44.097 30.9 15.1% 507.0 0.00030 26.1%
Bitumen 30.0% 577 173.1 85% 1002_0 0_00085 73_84fo Calculated 100% 204.0 100% 873.0 0.00115 100%
Decane kglkgm Mole % MW Welqht kg Wt% kgrnlM3 M3 Vol %
n-C10 70.0% 142.285 99.6 36.5% 734.2 0.00050 44.0%
Bitumen 30.0% 577 173.1 63.5% 1002.0 0.00063 56.0%
Calaulated 100% 272.7 100% 884,2 0100113 100%
Va ur Pressurc and Blend Volume Adding I-IVP-NGL to bitumen will cause the bubble point pressure of the mixture to increase. As the pump suction pressure of a line carrying HVP-NGL will be set above the bubble point pressure to prevent pump cavitation, a higher bubble point pressure implies less useful pressure drop allowed between stations. This will reduce the capacity of the pipeline unless additional pump stations are added.
Therefore, while adding NGL lowers viscosity and increases capacity, it also increases the bubble point pressure which reduces capacity. The area of economic utility exists where the viscosity benefit exceeds the cost of the lost capacity fi-om using a lower pressure drop between stations. While the impact of viscosity reduction appears to be somewhat independent of the choice of diluents, the bubble point pressure is not. Ethane will tunz into a gas above about 32 degrees C, the critical temperature, The critical temperature of propFU'ie is 97 degrees C. n-Bntane has a critical temperature of 152 degrees C. At pipeline operating conditions between about 10 and 50 degrees C, the etliane blend will have a significantly higher bubble point vapour pressure than the propane or butane blend. This implies that the upper limit for ethane would be based on bubble point pressure, while propane and butane would likely be limited by something else such as asphaltene deposition.

The SRC paper describes this bubble point pressure factor (measured right at the saturation pressure where a two phase state develops) at two temperatures of 15 and 28 degrees C, with Winter Oil and with the l0% - 70% mole propane diluent. (6). The results from their analysis are illustrated below in graph #
11. In their paper they identified a different resulting bubble point pressure depending on the loading procedure, sbowing different results depending on a rising or a falling pressure loading method. In subsequent verbal discussions, they say they now believe that the rising measurement had a systematic error intraduced, such that it is the falling method that is accurate. The bubble point pressure would be quite different using the three major HVP-NGL products of ethane, propane and butane, because of the large difference in critical temperature between these three products. The bubble point pressure is also dependent on the final operating temperature of the pipeline.

Graph # 11- Saturatton Pressure Winter Oi1lPropane vs. Moie % Diluent 700 ., _..__._..._ --_._._~.. r._____... .-~-.-- ....._...__.- _..-_ .........-_-..._.....

400 - -s-15 C Falling C Rizing 300 2B C Falqng -4&-28 C Rising ro 2p0.
1oa - - -0% 10% 20 3096 40% 50% 80% 70% 80%
AAeIa % propane An operating environment for an HVPNGL/bitunien pipeline might utilize an operating temperature as high as 90 degrees C. This is because of the large reduction in viscosity on heavy oil with higher temperature. Unfortunately, it leads to relatively high bubble point pressures as shown below in Table #
5, where the bubble point pressure is extrapolated for the saturated samples from graph # I 1 at 15 and 28 degrees C, using a linear relationship between temperature and bubble point pressure in the 15 - 50 degrees C range. In this case, a total of 871 kpa of the available 8,000 kpa pressure drop cannot be used, a reduction of 11 % in available.pressure drop between puznp stations (or an increase of 11% in the required number of pump stations). The benefit of viscosity reduction using IIVPNGL as compared to conventional condensate must overcome this economic hurdle_ Table 5- Bubble Point Pressure at 50 degrees C - Winter Oil/Propane % Mole C3 28% 55.50%
Temp deg C Pressure kpa 50 extra olate 469 871 While bitumens are all different and must ultimately be tested to determine viscosity, 1NP-NGI, has a specification that sliould provide a constant product and a known viscosity within a narrow range. Most effects (such as phase transition, compressibility factor, etc.) of a blend of I-IVP-NGL can be understood by looking primarily at the critical temperature of the mix. Knowing this, one does not need to know the individual constituents. However, bubble point pressure is not one of these factors. Two mixes were ]9 analyzed with 50% (mole) decane and either 50% mole propane or 23.1% mole ethane and 26.9% mole butane. The critical teinperatures of these two blends are identical. The bubble point pressure of the etliane/butane mix is about double the pressure of the pure propane n1ix, at both 10 degrees C (620 kpa vs. 300 kpa) and 50 degrees C (1230 kpa vs. 760 kpa). This suggests a further invention to control bubble point pressure, the addition or removal of ethane from the HVP-NGL
blend prior to injection of the diluent:

The SRC paper states that "the densities werc correlated with the volume-transiated Peng-Robinson equation of state, and the correlations are shown as lines in ..... Even when the measurenients of the propane concentration for the same equilibrium pressure were different (the over/under saturated phenomena described elsewhere and since explained as probably measurement error), the data fell on the same density-composition curve." This implies that the Peng-Robinson equations of state, adjusted for volume, can be utilized to predict the vapour pressure of a HVP-NGL /
Bitumen blend.

The Peng-Robinson technique was applied to a propane / decane mixture to determine if the type of oil affects this calculation of bubble point pressure, in addition to the mole %
of the propane. It appears that the bubble point pressure is dependent on the type of oil as the propane/decane blend bubble point pressure is only about 2/3 of the propane/Winter blend oil bubble point pressure, at the same mole %
propane and temperature, shown below in Table # 6. This would imply that the bubble point pressure issue will become more important as the bitumen gets heavier in density, and this will be exaggerated as the heavier bitumcn requires more propane to achieve the viscosity target in the first place.

Table 6- Bubble Point Pressuro - Winter Oil/Propane vs. Decane/Propanc Temp 15 dog C pressure kpa % Mole C3 23% 61%
Winter Oil / C3 200 600 Decane / C3 153 423 Temp 28 deg C pressure kpa % Ma1e C3 13g6 42%
Winter Qii / C3 200 500 Decane / C3 117 393 Gas Phase Loadina The prior art of this area of interest is well established. The OPSA
Engineering Data Book, page 23-43, Figure 23-27 illustrates the viscosity reduction in oils of different starting viscosities from the effect of the first few % of dissolved gases. However, the described figure does not express the data in a fashion that the blending formula for diluent might be ascertained. When this viewpoint was taken, the figure shows that for oil with a viscosity of 100 cP at 100 degrees F and aunospheric pressure, the inje.etion of gas (i.e. I assumed methane although the chart is not selective to any particular gas) reduces the viscosity significantly. At 1.5% mass methane, the viscosity is reduced to 45%
or 45 cP. At 0.75% mass methane, it reduces to 65%. at 0.4% mass methane, it reduces to 80% and at even 0.2% mass methane, it reduces to 91%. (22) Tllis data was then expressed on a log-linear scale in Graph 12 and converted to mole % assuming that the GPSA chart applies to a light oil with a specific gravity of O.S. It illustrates a straight line over the area of utility. T11is effect appears to end at about 1.5% mass when the oil is saturated. With propane, 1.5% niass is equivalent to 10- 20% mole depending on the molecular weight of the oil.

What is also known is the idea that 1-IVFNGL can be injected as a liquid under pressure, at these relatively small amounts, and they will take up gas phase characteristics once captured within the bitumen matrix so long as the amount is small. This is done routinely with butane, but not with anything lighter to my knowledge. Liquid injection avoids the time, energy and pressure required to load bitumens with gas bubbling through thc matrix and results in a superior fluid with a vapour pressure not much 1--igher than the starking oil. The vapour pressure rises rapidly with even small amounts of typical gases found in reservoirs such as methane, nitrogen or helium. If a pipeline's vapour pressure specification could be expanded, a relatively large reduction in viscosity is available tluough this flow enhanncement tecluiique, even if the oil already meets the pipeline specification. The amount of extra pressure is a function of the chosen material. Current pipeline vapour pressure specifications are maxed out at about 24% butane, <1% propane and basically 0.2 % ethane. This is basically the inverse of the relative vapour pressure impact of the dilueuts.

Wliat is new is that designing a pipeline to a vapour presstue specification that would allow for large amounts of IIVFNGL, including investments for all the additional up front cost, has utility for (particularly) hydrocarbons with either 2 or 3 carbon molecules and additional utility for the butane family. If a blend of any bIVPNGL is transported using the 75% blend conteinplated, then the resulting mix would have passed far beyond the gas phase loading portion of the blend.
Thus, this range would be automatically included in the final result.

Graph 12 - Oil Viscosity (100 degrees F) vs.
% Dissolved Methane (by mass) a ~ ~ .. .
D
Ci 0.0% 0.5% 1.0% 1.5%
%dissolved methane by mass 2]

Exl2criments with TAME -( tert-amv! neethyl etherl A paper by Anhorn and Badakhshan (1992 - reference 23) analyzed the solubility and viscosity of a Cold Lake bitumen in an oxygenate ether that had been proposed as agasolin.e additive called TAME.
Its MW Is 102.18 and specific gravity is 0.764 so it is similar to n-heptane in MW. Its viscosity is 0.53 cP at 10 C. Their Cold Lake bitumen had a MW of 559, an assumed specific gravity of I and a viscosity of 400,000 cP at 10 C. The tests were performed at 3 dilutions of 20%, 25% and 30% by mass and at three different temperatures of 4, 10 and 20 degrees C. They saw no evidence of asphaltene precipitation.

The results in the paper are expressed on mass %, so these were convei'ted to nlole % and illustrated in Graph # 13 and # 14 on a log-viscosity vs. mole % TAME basis. Graph # 13 shows that a straight line drawn between the 0% TAME left intercept viscosity and the approximate mid-point of the three data points at between 57.7 mole % fuld 70 mole% will intersect the right intercept at a point about one order of magnitude above the actual viscosity of 100% TAME. This appears to be the case for alt three temperature experiments. As the slope of the line appears to be curving between the three dilution % (as if these points straddle the inflection point), Graph # 14 shows the data in greater detail. In this grapli, a straight line is drawn between the three points. As shown, this line intersects the right intereept at 100%
TAME at a point about '/s an order of magnitude above the actual viscosity of 100% TAME which would be expected if the data points were near the inflection point. These results are very similar to what was observed with propane, naphtha, kerosene and LCO. This provides evidence that the viscosity predicting technique described herein might apply to ethers and alcohols in a similar fashion. A primary option would be ethanol as it will be manufactured in large quantities in the future, it must be transported to gasoline centers (i.e. refineries) for blending into gasoline, it is soluble in oil, separation is easy and it has a molecular weight almost the same as propane. If this idea is correct, it would behave similar to propane in its viscosity rcduction ability, but at a low vapour pressure. This then implies that methanol would be superior even to ethanol, as it's MW is similar to ethane and it's vapour pressure is al.so low.

While large quantities of inethanol will not likely be manufactared in the near fature, if one is building a conventional system that includes a condensate return line (i.e. re-cycle the diluent), then Table 7 illustrates that one could require only 22% as much mass of methanol as mass of decane, and 29% as much volume, to carry a similar amount of bitumen. Ethanol is almost as good, at 32% of the mass and 39% of the volume. The boiling point of methanol is 65 C, ethanol is 78 C, pentane is 36 C and decane is 174 C. In any separation system using boiling and condensing of the vapour, it appears that the alcohols would have an easier separation than a condensate wliich contained any amount of heptane to decane components (C7 -- C10). Methanol (at these small amounts) is soluble in oil as is ethanol. It should be possible to retrofit existing low vapour pressure pipelines to carry a bitumen/alcohol blend, so long as a return line is included in the system.

Graph 13 - Viscosity vs. Diluent % - TAME and Cold Lake Bitumen (Anhorn and Badakhshan - 1992) 4.0,000,000 - - - -- - - - - - -- --- -- ----- -- - -1,000,00,b- ti ;~ _ _- _-- - -- - -- -a Q 10.000 1,000 ____ ' ' ' -. - ~ , ' '_~.r~ r___. _ ~ ~ 10 C
100 - -- '' -~''-- =20C
~.. -.- -- _ 1 - _ ----T---T--- ---=-._ ~ - -..-.--,.. i 0 _~_ ......_..._~-_. __,._......._......-.....--- ---_ J
0% 20% 40% 60% 80% 100%
% Mole TAME

Graph 14 - Viscosity vs. Diluent % - TAME and Cold Lake Bitumen (Anhorn and Badakhshan - 1992) 0 100 ....d, = ~ ~1oC
-, ~

> ;, -50% 60% 70% 80% 90% 100%
% Mole TAME

TABLI/ 7- Comparison of Various Diluents with Cold Lake Bitumen Weight Mole % kg/kgm kg Wt% kgm/M3 M3 Vol %
Methanol 68.5% 32.042 21.9 11.1% 796.26 0.00014 13.5%
Cold Lake 31.5% 559 176.1 88.9% 1000.00 0_00089 88.5 k Calculated 100% 198.0 100% 972.42 0_00103 100%

Ethanal 68.5% 46.069 31.6 15.2% 793.99 0.00019 18.4%
Cold Lake 31.5% 559 176.1 84.8% 1000.00 0.00085 81.6%
Calculated 100% 207.6 100% 962.06 0.00104 100%

TAME 68.50/a 102.18 70.0 28.4% 764.00 0.00037 34.2%
Cold Lake 31.5% 559 176,1 71.6% 1000.00 0.00072 65.8%
Calculated 100% 246.1 100% 919.23 0.00109 100%

Propane 68.5% 44.097 30.2 14.6% 507.0 0_00029 25.3%
Cold Lake 31.5% 559 176.1 $5.4% 1000.00 0.00085 74.7%
Calcutated 100% 206.3 100% 875.35 0.00114 100%

DecanE 68.5% 142.285 97.5 35.6% 734.2 0.00049 43,0%
Cold Lake 31.5% 559 176.1 64.4% 1000,00 0.00064 57.0%
Calculated 100% 273.6 100% 885.73 0.00113 100%
Conclusion The research to completely determine the parameters of any viscosity equation that attempts to explain the'impact of a I-IVP-NGL diluent would normally have to be performed at the lab scale wliere different fluids (with known viscosity and asphaltene content) would be combined in diff'erent proportions and the resulting viscosity would be measured at various temperatures and pressures.
The following discussion highlights how measuring the viscosity of any particular heavy oil near the two major phase change boundaries for one diluent can be extrapolated over the area between the phase change boundaries, and across a wide range of diluents. These few measurements should be sufficient to determine the viscosity at all blends of economic utility for that particular oil.

It is postulated that sufficient data exists in the public record to be used as a demonstration of a simple formula to predict how an HVP NGL / bitumen viscosity blending formula can be expressed as a simple sum of the relative mole weiahtine of the two con3bining nzaterials, over the defined commercial range of interest which fs up to about 85 mole % HVP-NGL diluent, This proposed blending formula does not currently exist to the author's knowledge. The log viscosity vs. mole % line should be similar for all IIVP-NGL diluents over the liquid/liquid phase portion of the range. Any accuracy better than about plus/minus 30 - 50% would exceed any current formula's predictive capability.

At a very stnall diluent %, somewhere between 0% and 15 mole %, it is likely that the blend viscosity will depart from the slope derived from the liquid/liquid portion of the phase envelope. It is well known that "live" bitumens have a much lower viscosity than the same bitunien when it becomes "dead", where any small remaining residue of volatile gas has evaporated away. Extending the log-linear line on Graph #3 back to the 0% diluent extrapolation falls on the actual bitianen viscosity, therefore the small deviation seen in the curves towards the left intercept does not appear to have ;atny large scale effect above 20 mole %.

To extend the analysis, the principle data point of note bsyond 70 mole %
propane is the certain knowledge that the viscosity will be 100% equal to that of propane at the 100%
diluent point. This implies a curvature or a change in slope to the log viscosity line, between 75 mole % and ] 00 mole %, as shown in Graph # 3. It suggests that if one anchors the straight line at a point about 1.4 order of magnitude (actual viscosity times 14) above the actual viscosity of the diluent (propane in this case), then the straight line segment over the liquid/liquid phase range of utility for this particular oil can be drawn.

The inflection point is postulated to be caused by a similar phenomenon as that which causes asphaltene precipitation, selective binding between different constituents of these coniplicated groups of molecules.
The theory of asphaltene precipitation has the asphaltenes held in suspension by the resins in the oil. The action of any paraff'm, propane being an exceptionally good one, is to break the resin / asphaltene bond and to attract the resin to the propane. When enough resin has been used up, any further propane injection liberates an asphaltene molecule, they flocculate, and grow quite rapidly into micelles. The fluid goes through a phase change, it is now a solid/liquid slurry, with particularly sticky solids. Up to the ir1{lection point, the action of the propane has been to bind to the non-asphaltenes. After this point, asphaltenes are being liberated and different binding mechanisms will be occurring at a molecular level.
This technique suggests that the amount of different diluents required to achieve a certain viscosity will, at their core, bear a large resemblance to Table # 1, where equal mole blending results in a required amount of ethane and propane equal to about %z the amount of required decane, when measured on a volume basis. It is on a$/13b1 or volume basis that any economic utility is expected to be expressed, As Table # I suggests 1/2 the required volume of diluent, and approximately 500,000 BPD of bitumen is currently being transported to market using 150,000 BPD of condensate, the economic utility of using this technique, over the blending % range of interest, in the area of long distance bitumen pipeline transportation, is demonstrated. Without a useful viscosity prediction technique, whether this will be sufficient to justify the increased cost is unknown.

It is not necessary to understand wltat mechanism is involved in these different segments of the blending % range to usefully employ the technology, it is only necessary to be able to accurately predict the outcome. The required parameter that should result front lab tests would be the viscosity near the asphaltene inclination point witlt approximately 65 - 75 mole % diluent that would allow the proper linear line segtnent to be drawn over the 0 % to 75 mole % range, with a minor gas loading adjustment over the 0% - 15% range.

This techniqtie should be valid for ethane, ethylene, propane, propylene, iso and normal butane and the butenes, the alcohols such as methanol, ethanol, propanol and butanol, and even the heavier diluents such as naptha, kerosene and LCO. This technique or fomnula or theory is in some disagreement with all proposed blending rules up to this point in time that I am aware of, of which there are no shortage, therefore it cannot be "prior art" or obvious to one skilled in the prior art.
If it were obvious, it would be employed. All of the other various mathematical models would be thrown out as they can only predict to within 30 - 50 % of actual values. Instead, they are all still in some form of use or are at least well enough known to be updated with new parameters as required.

Bibliography 1. AOSTRA technical report # 2" The theimodynamic and transport properties of bitumens and heavy oils", July 5, 1984, p 3-30 2. AOSTRA technical report # 2" The thermodynamic and transport properties of bitumens aad heavy oils", July 5, 1984, p 3-44 3. AOSTRA technical report # 2" The thermodynaniic and transport properties of bitumens and heavy oils", July 5, 1984, p 3-44 4. AOSTRA technical report # 2" The thermodynamic and transport properties of bitumens and heavy oils", July 5, 1984, p 3-45 5. Mehrotra, Svrcek "Viscosity, Density and Gas Solubility", series of papers 1982-1992, AOSTRA Journal of Research. The actual paper utilized for the gas viscosity calculation is "Viscosity, Density and Gas Solubility Data for Oil Sands Bitumens. Part 1:
Athabasca Bitumen Saturated with CO and C2146", Mehrotra and Svrcek, August 14, 1984.

6. Freitag, Sayegh and Exelby, "A New Semi-Automatic PVT Apparatus for Characterizing Vapex Systems", SPE/PS-CIM/CHOA 97783, 1985, page 1-2

7. Sabbagh, Akbarzadeh, Badamchi-Zadeh, Svrcek and Yarranton, "Applying the PR-EoS to Asphaltene Precipitation from n-Alkane Dilttted heavy Oils and Bitumens", Deeember 7, 2005

8. Akbarzadeh, Alboudwarej, Svrcek and Yarranton, "A generalized regular solution model for asplialtine precipitation from n-alkane diluted heavy oils and biturnens", March 9, 2005

9. Engineering Data Book, Gas Processor's Suppliers Information, Tenth edition 1987, Figure 23-35 "Viscosities of Hydrocarbon Liquids" and Figure 23-37 "Viscosity of Paraffm Hydrocarbon Gases at One Atmosphere

10. Mariano, Camacho, Postigo, Valen, Artigas, Royo and Urieta "VISCOSITIES
AND EXCESS
ENERGY OF ACTIVATION FOR VISCOUS FLOW FOR BINARY MIXTURES OF
TETRAHYDROFURAN WPTId 1-BUTANOL, 2-BUTANOL AND 1-CHLOROBUTANE AT
283.15, 298.15 AND 313.15 K ' from web-site httjp://www.scielo.br/scielo.php?script-sci ar.text&gid=S0104-

11. Gateau, Henaut, Barre and Argillier "Heavy Oil Dilution", Oil and Gas Science and Technology - Rev. IFP, Vol. 59 (2004), No. 5, pp. 503-509

12. Mehrotr$, Moimery and Svrcek "A revicw of practical calculation nlethods for the viscosity of liquid hydrocarbons and their mixtures", Fluid Phase Equilibria 117 (1996) pp

13. Subramanian and Hanson, "Supercritical Fluid Extraction of Bitumens from Utah Oil Sands", (27/08/2005) Internet reference bLtR.-//users.netropolis.net/subE41paparl.htm

14. Jacoby, Colorado School of Mines, "Phase Behavior of Light Hydrocarbon -Heavy Oil or Tar Systems, and its Application to Recovery Processes", lnterstate Oil Compact Commission Committee Bulletin, Vol 27 No. 2, 1986

15. Peneloux, Rauzy and Freze, Fltxid Phase Equil., 8, 7 (1982)

16. Rahimi, Ellenwood, Parker, Kan, Anderson and Dabros, "Partial Upgrading of Athabsca Froth by Asphaltene Removal"No. 1998.074

17. William J Power, Shell Canada, "Froth Treatment: Past, present and Future", Oil Sands Symposium, University of Alberta, May 3-5, 2004

18. S.M.Blair, "Report on the Alberta Bituminous Sands", Gov't of Alberta, Dec 12, 1950

19. Franiciewicz, Craig, Hanson, Hoover, Perry, report to the Alberta Office for Coal Research and Technology, "transCOM Coal Slurry Technology", Feb, 1989

20. Argillier, Barre, Brucy, Dournaux, Henaut and Bouchard, "Influenc:e of Asphaltenes Content and Dilution on Heavy Oil rheology", SPE 69711 (2001) Society of Petroleum Engineers

21. Sheng, Maini, Hayes and Tortike, "Experimental Study of Foamy Oil Stability", Journal of Canadian Petroleum Technology, April 1997, Volume 36, No 4, paghe 31

22. Engineering Data Book- Gas Processors Suppliers Association - 10' edition,

23. Anhorn and Badakhshan, "Heavy oil-oxygenate blends and viscosity models", Eastern Oil Shale Symposium, 18-20 November 1992, Lexington, KY, USA

Claims