You are on page 1of 23

SPE-195806-MS

Formation Fluid Microsampling While Drilling: A New PVT and Geochemical


Formation Evaluation Technique

Julia Golovko, Christopher Jones, Bin Dai, Michael Pelletier, Darren Gascooke, Peter Olapade, and Anthony Van
Zuilekom, Halliburton

Copyright 2019, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in Calgary, Alberta, Canada, 30 Sep - 2 October 2019.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
Phase behavior characterization (PVT) and geochemical compositional analysis of petroleum samples
play a crucial role in the reservoir evaluation process to help determine producible reserves and the best
production strategy. Openhole samples are the most valuable types of samples for PVT and geochemical
analysis. Unfortunately, traditional openhole sampling methods are costly and limited to ten to twenty
samples, thereby restricting the scope of characterization in a well section. This study summarizes a new
microsampling technique for logging while drilling (LWD) and a corresponding wellsite technique to
provide compositional interpretation, contamination assessment, reservoir fluid compositional grading, and
reservoir compartmentalization assessment. This microscale approach allows fast analysis with a field or
near-field deployment of the analytical tool, providing fast turnaround time for analysis. The results inform
planning for wireline sample retrieval, if necessary.
The microsampler used in the downhole tool is capable of collecting reservoir fluid in small quantities,
suitable for compositional analysis. Because of its small size, the microsampler can gather multiple fluids at
various reservoir depths, while PVT sampling requires larger volumes and has more constraints. However,
when used in combination with conventional PVT-grade samples, the microsamples can provide significant
chemical profiling. The quantity of 40 microliters (μl) provides the opportunity to collect many more
samples than the conventional PVT sample size of 200 to 1,000 milliliters (ml). Additionally, 40 microliters
provides more than enough of a sample for a complete chemical analysis using a liquid chromatograph
or gas chromatograph coupled to either a mass spectrometer for biomarker analysis or a flame ionization
detector for a complete assay. Isotope analysis is also possible.
Recovery to surface of fluid samples collected at reservoir temperature and pressure allows for analysis
with an automated gas chromatograph (GC) deployed in the field, providing reduced labor and rapid
analysis. The unique injection chamber of the GC is designed with the injection port and valve configured
to withstand pressure up to 5,000 psi, which is approximately five times higher than standard GC injection
valves. This allows for injection of the microsample with a solvent carrier as a single-phase fluid so that
analysis can provide composition and fluid properties, such as gas to oil ratio, without a flash. The GC
has two detectors including a flame ionization detector (FID) for hydrocarbon components and thermal
2 SPE-195806-MS

conductivity detector (TCD) for inorganic gas components, such as carbon dioxide, nitrogen, and hydrogen
sulfide. The system can quantify hydrocarbon components from C1-C36 and perform contamination studies
of oil samples with drilling fluids.
This study provides a new technique for reservoir engineers to characterize a reservoir completely,
without limit to the number of acquired samples. In combination with conventional PVT samples, it is
possible to extrapolate the PVT properties to all pump-out stations, and conduct a complete geochemical
profile of the reservoir.

Pressure/Volume/Temperature Phase Behavior


The chemical and physical properties of fluids in a petroleum reservoir are highly variable and directly
affect the commercial value of the reservoir asset (McCain 1990; MacKay 1994; Fox and Martiniuk 1994;
Fernandez-Lima et al. 2009; Hunt 1996; Hammond 1991). Traditionally, a laboratory will recover and
analyze the fluid samples. However, the time, expense, and tendency of samples to change after sampling
by means of phase separation, precipitation, loss of volatiles, etc., make it desirable to determine at least
some of the fluid properties in situ through instruments located directly in the reservoir formation (Mullins
1999; Hashem et al. 2007). Some important reservoir fluid properties are physical in nature—temperature,
pressure, density, viscosity, radiation, resistivity, etc.—for which in-situ sensors already exist (Hashem et
al. 2007; Ellis 2008). Other important properties, such as component concentrations, require measurement
of fluid chemical properties. Data providing chemical composition are vital for determining the nature of a
reservoir, governing the techniques necessary to recover its contents, and helping establish the economics
of production. Conventional instruments, known as PVT(X) systems, control and independently vary the
pressure, volume, temperature, and composition of reservoir fluids to determine the associated physical
properties of the petroleum fluid mixtures as a function of composition. The resultant data enables equation-
of-state (EOS) modeling, which is vital for reservoir simulation modeling (Zhuan et al. 1996; Linsky et al.
1987; Tekáč et al. 1984; Jacoby and Yarborough 1967; Bridgman 1931; Francis 1959). This instrumentation
has the ability to inject gas or liquid components and dynamically mix the resultant composition by
circulation, to achieve homogenization of a single phase or multiphase constitutions. The sample volume of
the common petroleum PVT(X) is 200 ml with a typical lower limit of 50 ml. Lower sample-volume systems
on the order of 5 ml have been reported (Larter et al. 1997), but systems such as these are not commercially
available. Further, because multiple PVT tests require fresh samples for analysis, typical laboratory stock
required for PVT testing is 200 to 1000 ml of fluid. To complete an entire suite of analytical testing requires
less than 1 ml of fluid and generally only microliters for individual tests. Therefore, the entire demand of
downhole sample volumes is for PVT testing, not analytical testing.

Geochemistry
Organic petroleum geochemistry is widely used in the petroleum industry to rank prospects, improve
the understanding of an asset, and understand production and production problems (Hunt 1996; Larter
et al. 1997; Behar et al. 1997; Dong et al. 2007; Elshahawi et al. 2008). Fundamentally, the chemistry
of petroleum is dependent on the biological source material, and all generation and alteration processes
that the material undergoes in becoming petroleum trapped within a reservoir (Behar et al. 1997). Using
sensitive analytical techniques to uncover the detailed chemistry allows a reconstruction of the processes
that produced the oil, particularly compared to the source rock material at differing levels of maturity
(Hunt 1996; Behar et al. 1997). Much information, particularly regarding the reservoir architecture, and
including interconnectivity or compartmentalization of the reservoirs, is available directly from the oil,
without regard to the source material (Hunt 1996; Larter et al. 1997; Dong et al. 2007; Elshahawi et al.
2008). In fact, study of the petroleum at different levels of maturity can allow some simple extrapolation to
source-rock kerogen properties (Behar et al. 1997). The analytical geochemistry techniques used to provide
SPE-195806-MS 3

the detailed chemical description of the petroleum generally include bulk composition, biomarker analysis,
and isotopic analysis (Hunt 1996). Bulk composition is usually analyzed with chromatography techniques,
including gas chromatography, liquid chromatography, and thin-layer chromatography (Hunt 1996). The
bulk fingerprint is useful for reservoir architecture, analysis, and production geochemistry, such as flow
assurance studies (Larter et al. 1997; Abouie et al. 1997). The bulk fingerprint is also useful for equation-
of-state modeling of the phase behavior characterized with the PVT analysis (Pedersen and Christensen
2007; Abouie et al. 2017; Tekáč et al. 1984). Chromatographic analysis usually requires only microliters
of petroleum fluid, although some competing techniques, which provide similar information, use a few
milliliters. Biomarker analysis usually combines a gas chromatograph with a mass spectrometer to quantify
molecular, chemical fossils of the biological source material. This technique, termed as biomarker analysis,
provides detailed reconstruction of the generation and alteration processes (Hunt 1996; Behar et al. 2007).
This analysis usually requires only nanoliters of petroleum. The biomarker analysis also contributes to
further study of the reservoir architecture of a field (Hunt 1996). Stable isotopic analysis measures carbon
12 vs. carbon 13, hydrogen vs. deuterium, oxygen 16 vs. oxygen 18, sulfur 32 vs. sulfur 34, and nitrogen
14 vs. nitrogen 15 constituents of the oil. Dependent on the analysis, the resolution necessary, and the
type of equipment used, tens to hundreds of microliters can be necessary. The isotopic analysis provides
much the same type of information as the biomarker analysis, but perhaps with a coarser resolution. The
isotopic analysis has the advantage of usually being widely applicable, even in situations wherein specific
biomarkers are not present. Reservoir architecture within the field can also be uncovered with isotopic
analysis (Hunt 1996). Biomarker analysis is robust against filtrate contamination, but contamination affects
both bulk compositional fingerprint analysis and isotopic analysis. Therefore, it is important to obtain
samples of low contamination and determine the contamination level accurately to help mitigate its effects
upon the analysis. New analytical techniques provide ever more information with less sample, usually
requiring only a few microliters (Fernandez- Lima et al. 2009; Mullins et al. 2006). Fortunately, it is possible
to derive the contamination estimation directly from the bulk compositional analysis. Because only very
small quantities of fluid are necessary for analytical geochemistry, a small microsample volume from the
subsurface reservoir is all that is necessary for geochemical logging.

Drilling Fluid
The hydrostatic pressure generated by a mud column keeps the formation fluids from invading the wellbore.
As such, there is a net driving force for liquid to enter the formation. Clay particles, which build on
the surface of the well into a filter cake, compress over time to form a low-permeability barrier, which
prevents further loss of fluid into the formation (Allen et al. 1991). The process results in a near-wellbore
zone invaded by mud filtrate. Typically, this zone extends from 8 to 32 in. (Phelps 1995). The organic
components of a drilling fluid can be petroleum distillation fractions, such as diesel or mineral oil, or
synthetic components, such as olefins, esters, or ketones (Bloys et al. 1994; Burke et al. 1995; Growcock et
al. 2011). For organic oil-based mud (OBM), the filtrate is highly miscible with petroleum formation fluid
and, as such, there is no laboratory technique to separate them without disturbing the inherent petroleum
composition (Hadibeik et al. 2013).
A piston displacement model is useful for describing the invasion process (Allen et al. 1991; Phelps 1995;
Donaldson and Chernoglazov 1998; Lane 1993; Civan 1994; Riazi 1996; Malik et al. 2009). Even though
more complicated, fingering displacement mechanisms exist, piston displacement is the common invasion
model used because of its simplicity (Peeters et al. 1999; Wu et al. 2001; Wu et al. 2005; Malik et al. 2007;
Angeles et al. 2008). For piston displacement, the mud filtrate displaces the formation fluid much like a plug
from the near wellbore region. As shown in Fig. 1, three zones form around the wellbore: (1) the flushed
zone of only drilling fluid filtrate; (2) the transition zone comprising a composition intermediate to that of
the filtrate and formation fluid; and (3) the uninvaded zone containing the native formation fluid. Together,
4 SPE-195806-MS

the flushed zone and transition zone comprise an invaded zone. All three zones influence formation-probing
sensors, including resistivity, nuclear, acoustic, and NMR sensors (Allen et al. 1991).

Figure 1—Illustration of a cylindrical section of a fluid-saturated formation centered on a wellbore containing


drilling fluid filtrate. Filtrate drives into the formation, displacing the formation fluid in a near-wellbore
region. The invasion rate is proportional to the differential pressure, and inversely proportional to the
thickness of the mud filter cake. Finite element models can simulate filtrate invasion profiles (Wu et al.
2001; Wu et al. 2005; Malik et al. 2007; Angeles et al. 2008). The transition shown by the embedded graph
is a generic illustration of the typical formation fluid profile in the three zones. The graph illustrates the
transition from 0% formation fluid in the near-wellbore region to 100% formation fluid in the uninvaded zone.

Formation Testing and Sampling


Wireline or LWD formation testing provides petroleum asset evaluation and risk reduction information for
field development. The analysis of openhole formation-fluid samples is a primary means to provide the
fluid properties necessary to evaluate a reservoir. The results enable simulation of production strategies,
completions design, surface facility design, anticipation of flow assurance production issues and associated
operational expenses, and ultimately the financial decision as to whether an asset should be developed
(Hashem et al. 2007 Abouie et al. 2017; Elshajhawi et al. 2007).
Accurate compositional measurements of a reservoir petroleum fluid is necessary to help ensure a well
is safely drilled, identify a new discovery, evaluate the production potential and value of that discovery,
optimize the capital investment necessary to the produce petroleum, and to design a field management
system for multiple reservoirs in a field (Elshajhawi et al. 2007). There are three primary methods for
obtaining chemical information of a petroleum fluid contained in a reservoir: mud gas analysis, drill-stem
tests, and bottomhole sampling by formation testing. Mud-gas analysis is qualitative and only provides the
gaseous portion of the reservoir fluid, and drill-stem tests, which provide a surface phase-separated sample
SPE-195806-MS 5

of oil and gas, are often infeasible (Ramaswami et al. 2012; Kyi et al. 2014; American Petroleum Institute
2003; Michaels et al. 1995). Bottomhole sampling acquires samples directly from a reservoir. The device
has pumps designed to extract petroleum from a precise location along the wellbore and place that sample
into a pressurized container, which is then sent to a laboratory for analysis (Pedersen and Christensen 2007;
Elshajhawi et al. 2007; Hashem et al. 2007). Typically, a single wireline sampling run collects only three
to nine samples of 200 to 1,000 ml each, although in special circumstances more samples or larger samples
are possible (Badry et al. 1993; Proett et al. 2001). A typical sampling run usually takes from one to three
days but can take up to two weeks. The laboratory analysis of fluids can take as little as three weeks, but
when considering transport and lead time, it can take several months before sample analysis can begin,
especially in remote locations (Hy-Billiot et al. 2002; Freyss et al. 1999). Usually, by the time laboratory
analysis results are available, the well section that yielded the samples is cased and cemented and, often,
the drilling rig moved to another location.
Unfortunately, wireline formation testing is often the last activity before casing and cementing a zone.
Only one opportunity is available to acquire samples, without the benefit of a laboratory analysis to mitigate
poorly acquired samples, or to determine if those samples were not from the best locations. However,
if LWD formation sampling is available, there may be a second chance to obtain samples with wireline
testing if sample analysis results are available in time. Rapid analysis of LWD samples can also enhance
future wireline sampling capability. LWD has an advantage in that invasion time is shorter and, therefore,
the time to obtain the cleanest sample possible is shorter. Unfortunately, often, the mud cake is not cured
(i.e., not thick or compacted) and therefore the reservoir is often still undergoing active invasion during
LWD sampling. Therefore, it is even more important to determine the quality of samples immediately
after an LWD sampling operation, to determine if wireline sampling is necessary. Rapid surface analysis
determines drilling fluid filtrate and formation fluid sensor responses for downhole fluid sensors, including
resistivity, capacitance, density, viscosity, bubble point, compressibility, refraction index, speed of sound,
and compositional sensors (Elshajhawi et al. 2007; Badry et al. 1993; Proett et al. 2001; Dong et al. 2008).
In-situ instruments for some aspects of downhole chemical analysis are of particular use (Mullins et al.
2006; Crombie et al. 1998; Andrews et al. 2008; Dong et al. 2008).
Three critical questions define the success of any openhole sampling program, specifically, where, when,
and how to sample (Elshahawi et al. 2008; Elshajhawi et al. 2007; Ahmed et al. 2016; Dai et al. 2017).
Samples must come from the best locations along the wellbore, in order to define fluid trends adequately.
The timing of sample acquisition must coincide with a point during pump-out at which the contamination is
sufficiently low to achieve the goals of sampling. Lastly, the samples must be representative of the formation
fluid and must remain so in transit from the downhole reservoir to the laboratory (Ahmed et al. 2016).

Where to Sample
Reservoir compartmentalization and fluid column compositional grading within a reservoir compartment
are the two primary defining factors in the choice of sample location. Conventional wireline log data
and formation pressure-test data can provide some actionable information but the inherent fluid properties
provide the most direct assessment. Even when not ambiguous, conventional log data is generally only
indicative of compartment barriers and provides no information about compositional grading. In addition,
when of sufficient quality, pressure test data is necessary but not sufficient to define all compartments; only
in the most fortunate of circumstances can the pressure test data provide hints of compositional grading.
Laboratory compositional analysis performed with contamination correction can provide an assessment of
reservoir compartmentalization and compositional grading with a good degree of certainty. Geochemical,
statistical, and thermodynamic methods may also provide an assessment of continuity (Elshahawi et al.
2008).
6 SPE-195806-MS

It is generally desirable to obtain at least one sample from every reservoir compartment, and up to
three samples from reservoirs that exhibit sufficient compositional grading (Hy-Billiot et al. 2002; API RP
44 2003). This assumes prior knowledge of the compartmentalization and compositional grading, which
may not be available without taking samples. To escape this circular dependency, sampling locations are
usually determined from conventional logs and sometimes pressure-test data. Depending on the number
of compartments, the number of desired sampling locations can exceed the number of samples available.
Significant advancement has been made toward using downhole fluid analysis data for the purpose
of compartmentalization assessment and compositional grading determination (Elshahawi et al. 2008;
Venkataramanan et al. 2006; Zuo et al. 2008; Andrews et al. 2008; Dong et al. 2008; Kyi et al. 2014).
However, few real-time examples exist, with most examples clearly executed after the formation-testing
run has ended and augmented with additional information. Fundamental to the issue is a paradigm in which
a rapid assessment follows immediately after LWD sampling. If immediate assessment of a microsample
indicates the variation along the wellbore, then by using geochemical techniques, specifically with the GC,
the circular dependency is resolved, enabling better planning for subsequent wireline sampling.

When to Sample
It is desirable to acquire a sample as soon as it can be determined with certainty that the contamination
is below a threshold required for laboratory analysis. Generally, 5% contamination is sufficiently low
contamination for most medium- and light-oil analysis (Hy-Billiot et al. 2002). The viscosity of heavy oil
is highly influenced by even low contamination, demanding more tightly controlled sampling requirements
in this environment (Ahmed et al. 2016). In addition, the phase envelope of volatile oils and condensates is
highly influenced by slight contamination, and less than 1% contamination is often desirable (Elshajhawi et
al. 2007). Various schemes have been proposed for real-time downhole contamination assessment, including
trend fitting (Dong et al. 2008), endmember fingerprinting (Venkataramanan et al. 2006), and equation-of-
state deconvolution (Zuo et al. 2008). Fingerprinting methods are common in the laboratory but have been
difficult to apply generally and ubiquitously downhole. Equation-of-state methods require higher quality
input than is generally available downhole. To date, the most common commercial means of real-time
contamination assessment is trend fitting, originally proposed by Hammond (1991). Trend fitting relies on
two basic assumptions: (1) that, as the instantaneous pump-out fluid grades from filtrate to formation fluid,
a pure formation fluid is asymptotically approached and (2) that the pump-out gradation follows a strict
analytical form that can be sufficiently fit as to determine the asymptote endmembers. For trend fitting,
an asymptotic trend-fit equation applies to a single parameter, linear with contamination, such as density,
a compositional parameter such as methane, or an optical density channel. The asymptotic signal (SA)
indicates the pure formation property. Given the mud-filtrate signal property (SM), contamination at any
time (C%) can be calculated for a sample signal (SS) according to Eq. 1. Often, the starting value for SM is
not observed and is extrapolated with some uncertainty. Therefore, in practice, a third assumption is often
imposed, that the starting monitor value for mud filtrate properties is known.

(1)

The first assumption that the instantaneous fluid grades to a pure formation fluid is not strictly true. In
fact, the instantaneous fluid grades to a steady state contamination level, which can, in many cases, be low,
but in other cases, it can be high. Vertical conduits for far-field contamination can be surprisingly common.
Mud filtrate can cycle from the wellbore through a leaky mud cake. An imperfect formation seal may exist,
allowing filtrate contamination to inject directly past the formation probe. Lastly, if the relative dip between
the bedding and the borehole is high, or there are many fractures in the formation, the grading, although
not strictly steady state, can reach a rate of such slow cleanup that further cleanup is undetectable. Whether
SPE-195806-MS 7

the pump-out reaches low contamination, steady state, or a rate of slow cleanup, stable cleanup estimation
indicates when a sample is as clean as realistically possible. Nonetheless, better absolute contamination
estimation could be more actionable, particularly if it is determined early during the pump- out, indicating
that the asymptotic limit corresponds to a high contamination level.
The bulk composition analysis of an LWD microsample would provide exact endmember properties
with respect to both the formation fluid and the filtrate. Through direct compositional analysis, and
inferred through equation-of-state modeling, chemical and physical properties of both endmembers can be
determined, leading to better estimates of contamination on a subsequent wireline run.

How to Sample
For a sample to be useful in petroleum asset assessment, it must be representative of the reservoir fluid.
Of concern are phase changes, which can fractionate either heavier components from a bulk fluid, or gas
components from a bulk fluid (Pedersen et al. 2007). For gas condensates, it is common to fractionate a
liquid portion from a bulk gas. However, fractionation of light components relative to heavy components can
naturally take place because of differences in mobility, not unlike the effect of chromatography (Hunt 1996;
Jones et al. 2015). Sometimes, the caustic and reactive nature of a drilling fluid filtrate can suppress acetic
components of a petroleum sample (Harrison et al. 1999; Garcia and Lordo 2007). Further, it is desirable
that the sample retains enough pressure to help ensure preservation of a single phase within a gas/liquid
phase envelop and asphaltene phase envelop (Gonzalez et al. 2008). Pressure and preservation of samples
receive the greatest amount of attention with regard to fractionation issues. The asphaltene phase envelop
is complicated with pressure, temperature, and compositional effects (McCain 1990). Once asphaltenes
precipitate from solution, it is kinetically unfavorable for them to attain their reservoir state, even when
the sample returns to reservoir temperature and pressure. Agitation of the sample does little to help on the
microscale. Therefore, it is important to ensure the asphaltenes do not precipitate in the first place (Ahmed
et al. 2016). Other fractionation issues are usually not discovered unless either a drill-stem well test is
performed or until production, at which point, the petroleum asset evaluation has concluded. Microsample
analysis can detect any bulk compositional alteration. This provides a quality control assessment of the
LWD sample, and suggests better sampling procedures to help ensure sample quality on the subsequent
sampling run.

Sampling and Analytical Techniques


The liquid organic phase contains components of at least carbon number C16 to C22 in a Gaussian or
Lorentzian distribution as a function of carbon number (Pedersen and Christensen 2007). Petroleum fluid
often exhibits an exponential decay with carbon number (Hunt 1996; Pedersen and Christensen 2007).
Because these two shapes are distinct, the use of a chromatogram can easily highlight the influence of drilling
fluid filtrate, as shown in Fig. 2. Laboratories can determine the contamination level by the subtraction
method, involving the iterative removal of a fraction of the chromatogram of a filtrate sample from the
chromatogram of a contaminated sample, until the resultant chromatogram shows only an exponential
decline. Alternatively, the skimming process assumes the fingerprint by fitting an exponential decline to
the peaks outside the region of influence from C16 to C22 and using the peak distribution above the fit as
the fingerprint to remove. The techniques are useful but do not perfectly calculate contamination with up to
±3.5 wt% error depending on the nature of the filtrate and formation fluid (Zuo et al. 2017). The methods
introduced in this paper use the chromatographic instrumentation. Rather than using a direct subtraction or
skimming method, a more robust and automated technique known as multivariate curve resolution is used.
8 SPE-195806-MS

Figure 2—An example chromatogram of a drilling fluid filtrate (top) with a largely Gaussian profile,
and a dead oil containing no dissolved gas (bottom), showing a largely exponential profile as
a function of column time, which is proportional to carbon number or molecular weight. Note
the easily recognizable influence on the chromatogram of the 10% continued formation fluid.

Microsampling attempts to acquire a few microliters of petroleum fluid for chemical profiling instead
of the conventional hundreds of milliliters for a PVT sample. Fig. 3 presents a schematic of the sampling
device. The sampler is in the flow line such that the sampling chamber is always open to flow. As the
sampler moves to the closed position, due to a pressure differential between the fixture void and the flow
line pressure, it traps a small amount of fluid between the two O-rings. On the back of the holder, a small
pinhole allows pressure flow. The back of the O-ring is nominally at atmospheric pressure. The fixture on
the back of the sample holder has a void to allow a mitigated pressure increase as the sampler is closed.
An elastomer jacket holds the sample holder in the flow line fixture, and separates the back of the sample
holder from the front.

Figure 3—Microsampler schematic. Initially the sampler is in the open positon, primed with the sampler
O-ring in the first groove. As a sample enters the open sampler and the pressure to the right overcomes
the O-ring resistance, the sampler moves into the holder. The holder is fixed into a body separating the
flow line to which the sampler is initially open, and a low-pressure dead volume to the left. An opening
in the holder allows pressure to escape as the sampler closes. The sampler stays closed because of the
pressure environment, and latches in place via a simple mechanism (not shown). The sample volume is 43 µl.
SPE-195806-MS 9

Fig. 4 illustrates the sample retrieval key used to remove the sample holder from the flow line fixture upon
retrieval. The sample holder has a hard backstop to allow the O-rings to set into the grooves of the sample
holder. The O-ring resistance is 1,000 to 2,000 psi and the pressure in the vesture void never increases over
100 psi. The sampler has a small latch mechanism (not shown) to prevent the sampler from slipping out of
the holder once the pressure is removed. It is then easy to retrieve the sealed microsampler as a unit from the
flow line fixture. The current microsampler holds 43 µl and has been tested at 5,000 psi for 20 days, 10,000
psi for 10 days and 15,000 for five days with no loss or fractionation of fluid composition. Upon retrieval
of the microsampler, the sample holder screws into a dilution vial containing a suitable gas chromatograph
solvent. The diluted mixture drops the bubble point of the formation fluid, allowing for safe injection as a
whole oil into the chromatograph. Because the dilution is approximately 1:20, the bubble point of a mixture
of pure methane will drop below the pressure limit of the GC injection valve at 5,000 psi when using carbon
disulfide. Injection of the whole oil ensures that both gas and liquid components may be determined in one
injection without flashing.

Figure 4—Prototype microsampler and components assembled on the left and with separate components on the right.
The right-hand image shows the housing in the back row, with the micro-sampler in the front row. The middle row
shows a key for retrieval of the micro-sampler from the fixture, and ejection of the inner sampler into a dilution vial.

Wellsite Gas Chromatography


Chromatography is a separation technique applied to the components of a mixture. Chromatography uses a
difference in component affinity for a stationary phase vs. a mobile phase to separate the components as a
function of time and distance. Column chromatography uses a column to retain the stationary phase as the
mobile fluid phase flushes the components past the stationary phase and out the column. If the mobile phase
is a liquid, then the technique is liquid chromatography; if the mobile phase is a gas, then the technique is
gas chromatography. With gas chromatography, the column is usually a capillary column with a diameter of
only a few microns, and the mobile phase is usually deposited along the walls of the column. The column
can be tens to hundreds of meters long. Generally, the longer the column, the better the separation. Heating
the column improves the speed of the analysis. The heating can be varied throughout the analysis to maintain
good separation while improving the speed of analysis. At the outlet of the column, compound-sensitive
detectors may be located in order to quantify the fluid individual components escaping from the column. If
a compound has a greater affinity for the mobile phase then as the compound moves through the column, it
will spend more time in the mobile phase and escape from the column after a short period. If the compound
has a greater affinity for the stationary phase, then the compound will spend more time in the stationary
phase and escape from the column in a longer period. Compounds of a similar chemical affinity will further
have an additional affinity for the stationary phase based on size and shape. Compounds that are longer
10 SPE-195806-MS

and flatter with greater surface area will have a greater affinity for a stationary phase. Therefore, within a
chemical class, the column approximately separates compounds based on their molecular weight. Typical
detectors include a flame ionization detector (FID), thermal conductivity detector (TCD), optical detectors
including infrared (IR), visible (Vis), or ultraviolet (UV) detectors, and mass spectrometer (MS) detectors.
In the current wellsite micro-GC, both a FID and TCD are used. The FID is sensitive, and detects compounds
as an electrical current, ionized in an anoxic hydrogen flame. This is generally sensitive only to the carbon-
hydrogen chemical bond and, as such, only detects hydrocarbons. The TCD detects compounds based on
their ability to cool a platinum wire. It measures the amount of energy necessary to heat the platinum wire
to a constant temperature as compounds escape from the column and blow across the detector. The detector
is universal, but lacks the sensitivity of the FID. However, the combination of the TCD and FID together
provide a very good characterization of petroleum composition. In typical hydrocarbon analysis, a sample
is flashed and introduced separately to different GCs, one designed for gas using a TCD, one designed for
gas using a FID, and one designed for liquid using a FID.
The customized, portable GC used in this study, as shown in Fig. 5, comprises individual closed modules,
which include two parallel analytical columns for liquid and gas analysis, respectively. As a novelty, each
column connects to two detectors, the TCD and FID. The columns are also individually heated in order to
provide rapid temperature control. Modules are independent units connected by means of a set of tubing,
connectors, and valves housed in the injection oven located in the center of the GC unit. The small size of
the modules provides ultra-fast microscale analysis of the gas, fluid, or fluid-gas mixtures. Compared to
conventional GCs, afew nanoliters of sample volume are sufficient for analysis on the micro-GC analytical
columns. Analysis takes only 10 to 20 minutes, as opposed to two to three hours, as is typical for this type
of analysis. Another advantage of the microscale GC is its overall compact dimensions, which make it easy
to deploy in a small space. Limited air space in the unit, compared to conventional GCs, its unique modular
design, minimal need for supporting materials, and its mobility provide a robust solution for field locations.
Liquid components C7 to C36+ of the oil samples are analyzed using a dimethyl polysoloxane phase
analytical column (Column 1) and individual hydrocarbons are detected on a FID. Gas analysis is
performed on a Q-Bond capillary column (Column 2), which provides separation of N2, CO2, and C1 to C5
hydrocarbons, and TCD detector. An analytical syringe directly injects dead-oil samples into the system,
and a hydrogen carrier-gas flow pushes them through the system under low pressure. Pressurized gas or
reservoir fluid samples can be introduced to the capillary columns through the sampling valve installed on
the top of the instrument, connected to the injection port and designed to withstand pressure up to 5,000
psi. Dilution and homogenization of single-phase reservoir fluid occurs inside the solvent chamber, and
the sample is then directly loaded to the sampling valve, and injected for compositional analysis without
preliminary push-out.
Combination of two analytical columns and two detectors provides parallel analysis of the liquid and gas
components present in the reservoir fluid, in one injection. Column modules and detectors are controlled
by the software. Columns are heated at the rate appropriate for hydrocarbon separation as follows:

• Column 1, FID: temperature program from 40ºC to 385ºC at 1.5ºC /sec

• Column 2, TCD: temperature program from 40ºC to 200ºC at 1ºC /sec

Laboratory-produced fluid mixtures, designed to simulate reservoir fluids, have been analyzed using
direct injection under the above described GC conditions. These fluids are discussed in the experimental
design section. Individual components are validated with the external standard and quantified using
chromatographic peak areas.
SPE-195806-MS 11

Figure 5—Micro-GC with the columns compartments located in the left and right corners, injection oven in the
middle, detector modules in the left and right corners, sampling valve located on the top of the instrument.

Fig. 6 presents a schematic of the GC microsample injection system. The microsample screws in place
at the injection point and enters the fluid stream. A high-pressure liquid chromatographic pump pushes the
solvent carbon disulfide past the microsample from a piston accumulator upstream. The solvent sweeps the
microsample into a six-port injection valve. The waste collects in a closed stainless steel piston accumulator
so that it is never exposed.

Figure 6—Schematic of the high-pressure injection system for the microsamples.

Experimental Design
In an effort to determine the capability of the microsampling and GC system and to test the contamination
deconvolution method, a detailed simulation mimicked a formation-tester sampling job with an LWD
formation sampler. A typical LWD formation sampler can acquire approximately 10 samples from an
openhole section. The mud system during a single section is usually relatively constant in liquid-portion
composition. Therefore, the filtrate invasion from the top of the section to the bottom of the section should
12 SPE-195806-MS

conform to a single profile. The fluids within a single section will probably be of the same family, related
through a common source rock, similar generation history, and similar migration history. However, the fluids
accumulate over time within the reservoir compartments and therefore although of a similar base, will have
distinct compositional features. To mimic this profile, a base oil was mixed with unrelated, trace, petroleum-
fluid samples from zero to 10 weight percent. Table 1 presents the API gravity and full sample designation.
The trace addition perturbs the base fluid’s compositional profile to mimic what would be distinct reservoir
fluids along a wellbore column.
Eleven dead oils with API gravity varying from 22 to 38 and one drilling fluid (filtrate) with hydrocarbon
components ranging from C15 to C20 were chosen from the database for this purpose, as shown Table 1.
Before mixing the experiment, gas-chromatography compositional analysis for all individual dead oils and
filtrate was performed using the wellsite micro-GC with a standard injection system. A medium oil with
API gravity 30.6 (Fluid ID A1-77) was selected as the base fluid for the experiment. Ten trace fluids
(prefix B1-B10) were mixed with A1-77 with the targeted concentration range of 90 to 99% oil A and
1 to 10% oil B. Table 2 presents the actual levels. The resultant family of oils is designed to mimic the
variation of concentration that can be encountered in a typical formation- sampling run. The fluids AB1-
AB10 ideally reproduce potential oil-oil mixing from different zones within the reservoir, and represent
reservoir fluid compositional and API variations observed in actual reservoir fluids, because of reservoir
filling and compartmentalization. Filtrate C, representing drilling fluid, was added to each of AB reservoir
fluids targeting 5 to 15% contamination, with the actuals shown in Table 3. The level of contamination was
chosen to evaluate how effectively low contamination can be determined with the proposed method.
After the ABC mixtures were created, compositional analysis was conducted using the wellsite micro-
GC. The fluids were loaded into the microsampler injection system and injected at a pressure close to but
not exceeding the valve injection limit of 5,000 psi. A backing pressure at injection valve prevented flashing
in the valuing lines prior to introduction into the injection port. The raw chromatograms were recorded for
analysis.

Table 1—Nomenclature of the samples used in this study. Sample A1-77 was the base fluid used to make
the family of oils. To this fluid, a trace petroleum sample was added (B1 – B10) to slightly change the
compositional characteristics of the individual samples within the family. The filtrate is designated as fluid C.

Fluid ID Type API

A1-77 Base fluid 30.6


B1-106 Trace fluid 22.0
B2-29 Trace fluid 23.8
B3-20 Trace fluid 25.4
B4-109 Trace fluid 26.6
B5-127 Trace fluid 28.1
B6-74 Trace fluid 28.9
B7-13 Trace fluid 29.8
B8-78 Trace fluid 31.3
B9-111 Trace fluid 36.6
B10-21 Trace fluid 38.0
C-BECD Filtrate
SPE-195806-MS 13

Table 2—Fluid A is the base oil used to make the family of oils in this study. Fluid B is the trace concentration of a
secondary petroleum sample used to slightly perturb the characteristics of the base oil in order to mimic the variation
that may be encountered during a sampling run. The amounts of fluid A and fluid B in each mixture are shown in percent.

ID A Fluid A% ID B Fluid B% ID AB

A1-77 82.64 + B1-106 8.91 = AB1


A1-77 86.52 + B2-29 5.70 = AB2
A1-77 81.81 + B3-20 11.42 = AB3
A1-77 87.66 + B4-109 6.32 = AB4
A1-77 84.07 + B5-127 10.77 = AB5
A1-77 90.68 + B6-74 5.22 = AB6
A1-77 82.69 + B7-13 13.79 = AB7
A1-77 91.56 + B8-78 5.49 = AB8
A1-77 87.09 + B9-111 10.94 = AB9
A1-77 92.59 + B10-21 6.39 = AB10

Table 3—Fluids AB are the family of oils in this study. Fluid C is the filtrate contamination
used to mimic the variation of contamination that may be encountered during a
sampling run. The amounts of fluid AB and fluid C in each mixture are shown in percent.

ID AB Fluid AB% ID C Fluid C% ID ABC

AB1 91.56 + C-BECD 8.44 = AB1C


AB2 92.22 + C-BECD 7.78 = AB2C
AB3 93.22 + C-BECD 6.78 = AB3C
AB4 93.98 + C-BECD 6.02 = AB4C
AB5 94.83 + C-BECD 5.17 = AB5C
AB6 95.9 + C-BECD 4.10 = AB6C
AB7 96.47 + C-BECD 3.53 = AB7C
AB8 97.06 + C-BECD 2.94 = AB8C
AB9 98.03 + C-BECD 1.97 = AB9C
AB10 98.97 + C-BECD 1.03 = AB10C

Multivariate Curve Resolution


The multivariate curve resolution (MCR) form described herein follows the form of de Juan and Tauler
(2003). Multivariate curve resolution attempts to decompose a set of linear channel measurements, herein
referred to as spectra, into a set of pure endmember spectrum estimates for a set of spectra containing
different endmember levels according to the bilinear Eq. 2:
(2)
where D is the set of measurement spectra to be decomposed, C is the concentration profile, and S is the pure
endmember spectrum. Spectra in this case refer to chromatograms. E is the residual error to be minimized
in the decomposition. The algorithm decomposes D according to a rank truncated principal component
analysis (PCA) for a known number of factors. An initial estimate of either the endmember spectrum or the
concentration profile is used to estimate D according to a minimization with respect to the alternate variable.
Assuming, for instance, an initial estimate of endmember spectra, the residual error would be minimized
finding an appropriate C for Eq. 3.
(3)
14 SPE-195806-MS

The new estimates of C would be alternatively fixed and used to estimate S according to Eq. 4:
(4)
The algorithm alternately iterates between Eqs. 3 and 4 until convergence between the two estimates of
C and S is achieved and EC=ES in a process known as alternating least squares (ALS).
The algorithm can suffer high rotational ambiguity, meaning that a large range of combinations of either C
or S could satisfy the solution. Two natural inherent constraints are usually applied to reduce the ambiguity
to an acceptable range. Specifically, both the spectra and the concentration profile must contain positive
values at all points. Neither negative spectra, nor negative concentration profiles are allowed. For formation
pump-outs, an additional inherent closure constraint can be applied. Specifically, the volume of filtrate
and the volume of formation fluid sum to the total volume of the system. The ALS approach allows the
imposition of external constraints. These constraints are external to the inversion solution of either Eq. 3
or 4, and are instead reset to a fixed value or range between each ALS iteration. The remaining rotational
ambiguity may be resolved by appropriately using samples from multiple depths with the constraint that
the formation fluid must have an identical character in all cases. An optional constraint that the formation
fluid must have the characteristic of a known profile sampled from the surface may instead be applied if
a sample of the filtrate is acquired and analyzed on the wellsite micro-GC. The MCR technique benefits
from large variation in the composition of formation fluid samples because the filtrate profile can be more
easily recognized in the presence of a diversity in formation fluid samples. That is, the common vector of
variation in a background of very different formation fluid samples is more likely to be the common filtrate
profile. In this study, the formation fluid samples were only slightly varied in composition so that the current
experimental design will be a good stringent test of the MCR technique. An additional constraint, that the
formation fluid composition declines exponentially as a function of carbon number, may also be applied,
but this constraint is not always true. In addition, this requires that the carbon number be integrated prior
to deconvolution, and the current technique is ideally applied to the raw chromatogram with the integration
applied afterwards. For this reason, the exponential constraint is not applied in this study.
The chromatograms of Fig. 7 were simultaneously deconvoluted to obtain the pure component
chromatograms for formation fluid and filtrate and the contamination estimations. Fig. 8 shows an example
deconvolution of AB1C. The deconvoluted filtrate profile is identical for all samples as defined by
constraint. Note that the deconvolution is visually consistent as the influence of contamination across
the formation fluid profile is not prevalent. Fig. 9 shows the contamination determined by the automated
MCR algorithm as a function of added contamination. Note the high degree of linearity between the
methods. The contamination at higher levels was most accurately determined, likely because of the stronger
fingerprint influences of filtrate at these higher concentrations. The MCR analysis in this case was biased
by approximately 1 wt% compared to the reference measurements. This level of bias is well within
typical laboratory accuracy, and is low enough that the measurement is still very useful for real- time
contamination assessment. Further, because the process is automated, an expert need not be present to apply
a skimming or subtraction technique at the wellsite. With the complete set of deconvoluted GC profiles, the
component concentrations may be integrated and used for EOS modeling and reservoir continuity studies.
The EOS modeling can provide formation fluid property estimates such that a wireline formation tester
could guarantee to sample accurate contamination levels in real time during the sampling job. The continuity
study can help validate the most critical sampling stations for a subsequent wireline sampling operation.
SPE-195806-MS 15

Figure 7—Raw chromatograms for the ABC sampled fluid mixtures. Note the filtrate contamination region from
approximately 1000 seconds to 1400 seconds. Outside the filtrate contamination region the fluid largely exhibits
an exponential decay with respect to the baseline. Note that the baseline has not been corrected for the raw data.

Figure 8—Example MCR deconvolution of formation fluid and filtrate AB1C.


16 SPE-195806-MS

Figure 9—MCR deconvolution contamination results.

Endmember Properties from Equation of State and MCR


The cubic equation of state (EOS) provides another potential platform to combine multiple sensor data.
Cubic equations of state, cubic with respect to volume, relate the critical properties of a fluid system to the
density of that fluid for a temperature and pressure. The cubic equation of state was first proposed by van
der Waals in 1877 as a modification to the well-known ideal gas equation shown in Eq. 5.

(5)

In Eq. 5 variables include pressure and critical pressure (P and PC), temperature and critical temperature,
(T and TC), moles (n), Volume (V), and the ideal gas constant R. The van der Waals equation correctly
assumes the molecular nature of fluids and that molecules have attractive forces as described by n2a/V2 and
real volume as described by nb. The van der Waals equation of state can be expressed as molar volume and
arranged for pressure, as shown in Eq. 6.

(6)

The cubic equation of state is a thermodynamic equation as van der Waals himself related his cubic
equation of state to the second law of thermodynamics in 1890, for phase equilibria. Unfortunately, the van
der Waals form has proved less useful for complicated mixtures than the more modern cubic equations of
state. The SRK (1972) (66) is shown in Eq. 7.

(7)

The Peng-Robinson (1976) (67) is shown in Eq. 8.


SPE-195806-MS 17

(8)

Equations of state are both more successful for prediction of fluid properties for complicated petroleum
mixtures. The newer equations of state also take into account molecular acentricity ω and use an updated
law of corresponding states for homologue molecules. For complex mixtures, a and b are usually determined
by the van der Waals binary mixing rules for aij and bij i=j as pure component and i≠j as binary mixtures
of components for the mol fraction of component x.

(9)

Statistically, an interaction of three simultaneous molecules is insignificant compared to the interaction


of two molecules and hence the binary interaction coefficients are all that are realistically considered. More
complicated mixing rules have been suggested, but the van der Waals treatment has been found satisfactory
for moderate density fluids over large pressure ranges. The binary interaction coefficients of many pure
components have been published (Lee et al. 1975; Riazi and Daubert 1980). Péneloux volume translation
(1982) was included in both SRK EOS and PR EOS to help improve the density predictions. The volume
translation has no influence on the gas-liquid equilibrium calculation results (Pedersen and Christensen
2007).
The liquid density and compressibility were computed for each of the contaminated fluids mixed in the
laboratory using the EOS under typical reservoir conditions of 178.8°F and 7783.93 psi. The reservoir fluids
were also re-created numerically using the endmembers predicted by the MCR method. The liquid density
and compressibility were then computed for the numerically combined fluids using the EOS. Table 4 shows
the comparison of the liquid density computed for the laboratory mixed fluids and numerically combined
fluids based on the contamination level predicted by the MCR method. There is a good agreement between
the liquid density predicted by the EOS using the laboratory mixed fluids and the endmembers predicted by
the MCR method. The errors are normally distributed suggesting that the limit of accuracy for the MCR-
EOS method was not largely affected by the bias in contamination determination. The error of 0.004 g/cc
suggests that for a typical formation tester pump-out with density contrast between filtrate and formation
fluid on the order of 0.2 g/cc, contamination may be determined to better than 3% for subsequent runs.

Table 4—The comparison between laboratory measured density and the MCR-EOS calculated density. The
comparison is at a temperature of 178.8°F and pressure of 7783.93 psi, typical reservoir conditions. The RMSE
between liquid density predicted laboratory mixed fluids and numerically combined fluids is 0.0042 g/cc or 0.49%.

Fluids Laboratory (g/cc) MCR-EOS (g/cc) Error Error Squared

AB1 0.847399 0.839987 0.007412 5.49E-05


AB2 0.844574 0.840699 0.003875 1.50E-05
AB3 0.841719 0.841399 0.00032 1.03E-07
AB4 0.845011 0.840856 0.004155 1.73E-05
AB5 0.841953 0.839237 0.002716 7.38E-06
AB6 0.841483 0.837944 0.003539 1.25E-05
AB7 0.843713 0.840449 0.003264 1.07E-05
AB8 0.844375 0.841809 0.002566 6.59E-06
AB9 0.847147 0.840838 0.006309 3.98E-05
18 SPE-195806-MS

Fluids Laboratory (g/cc) MCR-EOS (g/cc) Error Error Squared

AB10 0.84406 0.840866 0.003194 1.02E-05


RMSE 0.004177
%RMSE 0.49%

Table 5 shows the comparison of the compressibility predicted for the laboratory-mixed fluids and
the numerically combined fluids using the EOS. The errors are also normally distributed reinforcing that
the limit of accuracy for the MCR-EOS method was not largely affected by the bias in contamination
determination. The contrast between filtrate compressibility and formation fluid compressibility varies
widely as a function of composition, pressure and temperature. The compressibility of a typical drilling
fluid filtrate and formation fluid is 3X10-6 and 15X10-6 respectively. Therefore, an error of 1.45X10-7 g/cc
could allow contamination to be determined to 1.5%. Also, since the contrast for compressibility increases
substantially for lighter oils, volatile oils, condensates and gas, contamination could be determined to better
than 1.5% for those fluids. Other MCR-EOS properties with corresponding downhole sensor measurements
are also possible to predict including GOR, and bubble point.

Table 5—The comparison of the compressibility computed for laboratory-mixed fluids and the numerically combined
fluids under a typical reservoir condition (temperature of 178.8°F and pressure of 7783.93 psi). The RMSE between
compressibility computed using laboratory mixed fluids and the numerically combined fluids is 1.448X10-7(1/psi)

Fluids Laboratory (1/psi) MCR-EOS (1/psi) Error Error Squared

AB1 7.58E-06 7.90E-06 -3.21E-07 1.03E-13


AB2 7.75E-06 7.88E-06 -1.30E-07 1.70E-14
AB3 7.84E-06 7.83E-06 1.20E-08 1.45E-16
AB4 7.69E-06 7.85E-06 -1.59E-07 2.52E-14
AB5 7.88E-06 7.90E-06 -2.15E-08 4.63E-16
AB6 7.87E-06 7.92E-06 -5.08E-08 2.58E-15
AB7 7.78E-06 7.86E-06 -7.89E-08 6.22E-15
AB8 7.82E-06 7.82E-06 5.21E-09 2.72E-17
AB9 7.62E-06 7.85E-06 -2.26E-07 5.13E-14
AB10 7.79E-06 7.85E-06 -5.91E-08 3.49E-15
RMSE 1.45E-07
%RMSE 1.87%

Reservoir Continuity from MCR


Because the MCR provides a clean fluid estimate for all properties, a clean fluid composition can
be estimated, even for moderately contaminated samples. These clean fluid properties can be used in
conjunction with the equation of state algorithm to determine a set of bulk fluid properties and an inverted
chromatographic profile. Use of the MCR cleaned chromatographic profiles may be combined easily with
laboratory data. Fig. 10 shows a multiwell fluid similarity study by a similarity matching algorithm called
k-nearest neighbor (KNN). In this study the KNN was developed using laboratory-based, whole-oil, gas
chromatography, but then the resultant field data was applied. Although the structure of the dendogram is
dependent on the individual openhole samples included in the study, the multiwell study can establish limits
for what is considered typical for single reservoir fluids, and what is considered significantly dissimilar
such that compartmentalization is highly suspected. For the laboratory chromatographic study shown in
Fig. 10, reservoir continuity is suspected if the similarity index is less than 3.0, whereas, above 7.5,
SPE-195806-MS 19

compartmentalization is suspected. As an illustrative example, a k-nearest neighbor study was performed


on MCR-EOS derived chromatograms vs. a conventional laboratory chromatogram, as shown in Fig. 11.
The local k-nearest neighbor dendogram built from the laboratory chromatographic data suggests continuity
between the samples 724, 666, 754, and 618 with a similarity index less than 3.0 and compartmentalization
and subsequent suggested compartmentalization to sample 841 with a similarity index greater than 7.5. In
fact, similarities between 3 and 7.5 can be somewhat ambiguous. But, in this case, it is completely affirmed
because samples 724, 666 are from the same depth and samples 754, and 618 are from the same depth and
the similar index for the same depth samples is the same as the between depth samples for this group. The K-
nearest neighbor for the EOS inverted spectra is nearly identical with the similarity index for samples 724,
666, 754, and 618 not significantly different and less than three, and sample 841 greater than 7.5. In fact,
the top-level similarity index for the two groups is nearly the same for both the inverted chromatograms
and the laboratory chromatograms. This rapid continuity assessment of EOS-inverted, clean-fluid-estimated
samples is an example of how a sequential MCR EOS combination can be used to refine sampling locations
immediately after an LWD sampling run, to best determine positions for subsequent wireline sampling.

Figure 10—Dendogram for similarity by whole oil normalized GC for a multiwell study.
The green line suggests a limit below which compartmentalization cannot be assigned,
and the red line suggests a limit above which compartmentalization is suspected.
20 SPE-195806-MS

Figure 11—K-nearest neighbor to laboratory chromatographic data (Upper). K-nearest neighbor to EOS inverted data (lower).

Conclusion
The primary purpose of microsampling in an LWD formation tester is to determine the quality of associated
PVT samples, and if a subsequent wireline sampling run is necessary. Secondly it is the goal to determine
physical properties of the clean formation fluid to plan subsequent samples. If the microsample is acquired
close in time to the PVT sample from the same flow line, it is assumed that the contamination levels in
the microsample will be representative of the contamination levels in the PVT sample. It has been shown
that the ruggedized gas chromatogram field equipment may be used with the MCR algorithm to both
compute contamination in the associated PVT sample, and further to provide a cleaned chromatogram
from which formation fluid properties may be calculated by equation of state. These equation of state
calculated fluid properties may then be used for subsequent sample job planning. If the microsamples are
not fit-for-purpose, then it can be assumed that the PVT samples are not fit-for- purpose. The tertiary
purpose of LWD microsampling is to help ensure success of the subsequent wireline sampling operation
by determining when to sample, where to sample, and how to sample. Using the EOS and a subsequent
KNN analysis, recommendations for wireline sampling to accomplish this tertiary goal can be achieved. The
sample properties derived by EOS technique to the MCR derived chromatogram are sufficient to calculate
the properties of the formation fluid and identify any sampling concerns with respect to how best to sample.
The KNN compartmentalization analysis shows where best to sample. The current microsampling platform
prototype has been validated. The sampler size can allow at lesat up to 100 microsamples to be acquired
in a single sampling run. Because most geochemical techniques can collectively be analyzed with the
entire contents of a single microsample, a larger set of microsamplers can be used to gather a very dense
representation of formation fluid from along the wellbore.
SPE-195806-MS 21

References
API RP 44, Recommended Practice for Sampling Petroleum Reservoir Fluids, second edition. 2003. Washington, DC:
API.
Abouie, Ali. et al. 2017. Comprehensive Modeling of Scale Deposition by Use of a Coupled Geochemical and
Compositional Wellbore Simulator. SPE Journal, 22: 1225-1241. SPE-185942-PA.
Ahmed, K., Hassan, F., Taqi, F., Ahmad, F., et al. 2016. Real-Time Downhole Fluid Analysis and Sampling with a
New Optical Composition Analysis Sensor: A Case Study from Kuwait Heavy Oil Formation. Presented at the SPE
International Heavy Oil Conference and Exhibition, Mangaf, Kuwait. SPE-184112-MS.
Allen, D., Auzerais, F., Dussan, E., et al. 1991, Invasion Revisited. Oilfield Review 3: 10–23.
An automated high pressure PVT apparatus for continuous recording of density and isothermal compressibility of fluids.
American Institute of Physics, Review of Scientific Instruments 67: 244-250.
Andrews, A. et al. 2008. Fluorescence Methods for Downhole Fluid Analysis of Heavy Oil Emulsions. Taylor and Francis.
Journal of Dispersion Science and Technology 29: 171–183.
Angeles, R., Torres-Verdin, C., Malik, M. 2008. PREDICTION OF FORMATION-TESTER FLUID-SAMPLE
QUALITY IN HIGHLY-DEVIATED WELLS. Presented at the SPWLA 49th Annual Logging Symposium,
Edinburgh, Scotland.
Badry, R. et al. 1993. New Wireline Formation Tester Techniques And Applications. Presented at the SPWLA 34th Annual
Logging Symposium, Calgary, Alberta. SPWLA-1993-ZZ.
Behar, F. et al. 1997. Thermal cracking of kerogen in open and closed systems: determination of kinetic parameters and
stoichiometric coefficients for oil and gas generation. Organic Geochemistry 26: 21–339.
Bloys, B. et al. 1994. Designing and managing drilling fluid. Oilfield Review 6: 33–43.
Bridgman, P.W. 1931. The Volume of Eighteen Liquids as a Function of Pressure and Temperature. Proceedings of the
American Academy of Arts and Sciences 66: 185–233.
Burke, Christopher J. and Veil, John A. 1995. SYNTHETIC-BASED DRILLING FLUIDS HAVE MANY
ENVIRONMENTAL PLUSES. Oil & Gas Journal 93.
Civan, F. 1994. A Multi-Phase Mud Filtrate Invasion and Wellbore Filter Cake. Presented at the International
Petroleum Conference and Exhibition of Mexico, Veracruz, Mexico 10-13 October. SPE-28709-MS. https://
doi.org/10.2118/28709-MS.
Crombie, A. et al. 1998. Innovations in Wireline Fluid Sampling. Oilfield Review 10: 26–41.
Dai, B., Jones, C.M. and Van Zuilekom, T. 2017. A New Multisensor Approach to Downhole Sample Filtrate
Measurement with a Wireline Formation Tester. SPWLA 58th Annual Logging Symposium, Oklahoma City, Oklahoma.
SPWLA-2017-HHHH.
de Juan, A. and Tauler, R. 2003. Chemometrics applied to unravel multiccomponent processes and mixtures: Revisiting
latest trends in multivariate resolution. Elsevier 500: 195–210.
Devereux, S. 2012. Drilling Technology in Nontechnical Language, 2nd Edition. [ed.] Stephen Hill. 2. PennWell: Tulsa,
Oklahoma.
Donaldson, E.C. and Chemoglazov, V. 1987. Drilling Mud Fluid Invasion Model. Journal of Petroleum Science and
Engineering 1: 3-13.
Dong,C. et al., et al. 2008. ew downhole-fluid-analysis tool for improved reservoir characterization. SPE Reservoir
Evaluation and Engineering 11: 1107–1116.
Dong, Chengli et al. 2007. Improved Interpretation of Reservoir Architecture and Fluid Contacts through the Integration
of Downhole Fluid Analysis with Geochemical and Mud Gas Analyses. Presented at the Asia Pacific Oil and Gas
Conference and Exhibition, Jakarta, Indonesia. SPE-109683-MS.
Ellis, D.V. 2008. Well logging for Earth Scientists, 2nd Ed. Dordrecht: Springer.
Elshahawi, H. et al. 2008. Combining Continuous Fluid Typing Wireline Formaion Testers, and Geochemical
Measurments for an Improved Understanding of Reservoir Architecture. SPE Reservoir Evaluation and Engineering,
11: 27-40. SPE-100740-PA.
Elshajhawi, H., Hashem, M., McKinney, D., et al. 2007. The Power of Real-Time Monitoring and Interpretation in Wireline
Formation Testing-Case Studies. SPE Res Eval & Eng 10: 241–250.
Fernandez-Lima, F.A. et al. 2009. Petroleum Crude Oil Characterization by IMS-MS and FTICR MS. Analytical
Chemistry 81: 9941-9947.
Fox, J.N. and Martiniuk, C.D. 1994. Reservoir Characteristics And Petroleum Potential of the Bakken Formation,
Southwestern Manitoba. Journal of Canadian Petroleum Technology 33: 19-27. PETSOC-94-08-02.
Francis, Alfred W. 1959. Pressure-temperature-density relations of pure liquids. Chemical Engineering Science 10: 37–46.
Freyss, Henri, et al. 1989. PVT Analysis for Oil Reservoirs. The Technical Review 37: 4–15.
22 SPE-195806-MS

Garcia, J.M. and Lordo, S.M. 2007. CHEMISTRY AND IMPACTS OF COMMONLY USED AMINE-BASED H25
SCAVENGERS ON CRUDE UNIT TOWERS AND OVERHEADS. NACE International Conference and Expo,
Nashville, Tennessee. 07571.
Gonzalez, D., Francisco, V., Hirasaki, G. et al. 2008. Modeling Study of CO2-Induced Asphaltene Precipitation. Energy
& Fuels 22: 757–762.
Growcock, Fredrick B. and Patel, Arvind D. 2011. The Revolution in Non-Aqueous Drilling Fluids. Presented at the
AADE National Technical Conference and Exhibition, Houston, Texas. 2011. AADE-11-NTCE-33.
Hadibeik, H., Kwabi, E., Torres-Verdin, C. et al. 2015. MISCIBILITY EFFECTS OF OIL-BASE MUD AND IN-SITU
GAS ON CONVENTIONAL WELL LOGS. Presented at the SPWLA 54th Annual Logging Symposium, New
Orleans, Louisiana.
Hammond, P. S. 2001. One- and two-phase flow during fluid sampling by a wireline tool. Transport in Porous Media
6: 299-330.
Harrison, J.R., Stansbury, M., Patel, J. et al. 1999. Novel Lime-Free Drilling Fluid System Applied Successfully in Gulf
of Thailand. Presented at the SPE/IADC Drilling Conference, Amsterdam, NE. SPE/IADC 52817.
Hashem, M., et al. 2007. Formation Pressure While Drilling, Wireline Formation Testing, And Fluid Sampling In a
High Pressure/High Temperature Exploration Well Using Oil Based Mud: A Case History. Petrophysics 48: 258-270.
SPWLA-2007-v48n4a1.
Hunt, J. 1996. Petroleum Geochemistry and Geology, 2nd Ed. W. H. Freeman and Company: New York.
Hy-Billiot, J., Bickert, J., Montel, F. et al. 2002. Getting the Best from Formation Tester Sampling. Presented at the SPE
Annual Technical Conference and Exhibition, San Antonio, Texas. SPE-77771-MS.
Jacoby, Robert and Yarborough, Lyman. 1967. PVT Measurements on Petroleum Reservoir Fluids and Their Uses.
Industrial and Engineering Chemistry 59: 48–62.
Jones, C.M. et al. 2017. A Small-Volume PVTX System for Broadband Spectroscopic Calibration of Downhole Optical
Sensors. American Institute of Physics, Review of Scien4fIc Instruments 88.
Jones, C.M., He, T., Dai, B. et al. 2015. Measurement and Use of Formation Fluid, Saturate, and Aromatic. Presented at
the SPWLA 56th Annual Symposium, Long Beach, California. SPWLA-2015-EE.
Jurcic, Hrvoje, Maretic, Srecko, and Cogelja, Zoran. 2012. Petrophysical Parameters Evaluation in Unconventional
Reservoirs by Well Logging and Mud Logging Data Interactive Correlation Method. Presented at the SPE/EAGE
European Unconventional Resources Conference and Exhibition, Vienna, Austria. SPE-150961-MS.
Kyi, K.K., Lynn, C.S., Haddad, S. et al. 2014. Integration of Downhole Fluid Analysis and Advanced Mud Gas Logging
Reduces Uncertainty in Reservoir Evaluation. Presented at the International Petroleum Technology Conference, Doha,
Qatar. IPTC-17485-MS.
Lane, S.H. 1993. Numerical Simulation of Mud Filtrate Invasion and Dissipation. Presented at the SPWLA 34th Annual
Logging Symposium, Calgary, Alberta.
Larter, S.R., et al. 1997. Reservoir Geochemistry: A Link Between Reservoir Geology and Engineering? SPE Reservoir
Engineering 12: 12-17. SPE-28849-PA.
Lee, B. I., Kesler, M.G. 1975. A generalized thermodynamic correlation based on three-parameter corresponding states.
American Institute of Chemical Engineers. AIChE Journal 21: 510–527.
Linsky, D., Sengers, Levelt J. M. H., and Davis, B.A. 1995. Semiautomated PVT facility for fluids and fluid mixtures.
American Institute of Physics, Review of Scientific Instruments 58: 817-821.
MacKay, Virginia. 1994. Determination of Oil and Gas Reserves. [ed.]. Altona, MB Canada: Petroleum Society of the
Canadian Institute of Mining, Metallurgy and Petroleum 1: 41.
Malik, M., Birol Dindoruk, B., Elshahawi, H. et al. 2009. Numerical Investigation of Oil - Base-Mud Contamination in
Condensates: From Cleanup to Sample Quality. Presented at the SPE Annual Technical Conference and Exhibition,
New Orleans, Louisiana. SPE 124371.
Malik, M., Torres-Verdin C., Sepehrnoori, K. 2007. Axially symmetric compositional simulation of formation tester
measurements. Journal of Petroleum Science and Engineering 59: 33-349.
McCain Jr., W.D. 1990. The Properties of Petroleum Fluids, 2nd Ed. Pennwell Books: Tulsa, OK.
Michaels, J., Moody, M, and Shwe, T. 1995. Wireline Fluid Sampling. Presented at the SPE Annual Technical Conference
and Exhibition, Dallas, Texas. SPE-30610-MS.
Mullins, 0. et al. 2006. Oil Reservoir Characterization via Crude Oil Analysis by Downhole Fluid Analysis in Oil Wells
with Visible-Near-Infrared Spectroscopy and by Laboratory Analysis with Electrospray Ionization Fourier Transform
Ion Cyclotron Resonance Mass Spectrometry. Energy and Fuels 20: 2448-2456. American Chemical Society.
Mullins, Oliver C. 1999. Methods and apparatus for determining gas-oil ratio in a geological formation through the use
of spectroscopy. 5,939,717 USA, August 17th, Grant.
Pedersen, K. and Peter, L. 2007.Christensen P. Phase Behavior of Petroleum Reservoir Fluids. CRC: Boca Raton.
SPE-195806-MS 23

Peeters, M., Allen, D., Gomes, R. et al. 1999. INVASION IN SPACE AND TIME. Presented at the SPWLA 40th Annual
Logging Symposium, Oslo, Norway.
Peng, D.Y. and Robinson, D.B. 1976. A New Two-Constant Equation of State. Industrial and Engineering Chemistry
Fundamentals 15: 59–64.
Phelps, G.D. 1995. Computation of Mud Filtrate Invasion Profiles. The Journal of Canadian Petroleum Technology 34.
Proett, M.A. et al. 2001. New Wireline Formation Testing Tool With Advanced Sampling Technology. SPE Reservoir
Evaluation & Engineering 4: 76–87.
Ramaswami, Shyamalan R., Elshahawi, Hani, and El-Battawy, Ahmed 2012. Integration of Wireline Formation Testing
and Well Testing Evaluation-An Example From the Caspian. SPE Reservoir Evaluation & Engineering 15: 300-313.
SPE-139837-PA.
Riazi, M.R. and Daubert, T.E. 1980. Prediction of the Composition of Petroleum Fractions. American Chemical Society,
Industrial and Engineering Chemistry, Process Design & Development 19: 289–294.
Soave, G. 1972. Equilibrium Constants from a Modified Redlich-Kwong Equation of State. Chemical Engineering Science
27: 1197-1203.
Riazi, M.R. 1996. A new method for experimental measurement of diffusion coefficients in reservoir fluids. Journal of
Petroleum Science and Engineering 14: (235-250). 10.1016/0920-4105(95)00035-6.
Stephen, Alan Graham, et al. 2008. PVT Data Quality: Round Robin Results. Presented at the SPE Annual Technical
Conference and Exhibition, Denver, Colorado. SPE-116162-MS.
Tekk, Viktor, Cibulka, Ivan, and Holub, Robert. 1984. PVT properties of liquids and liquid mixtures: a review of the
experimental methods and the literature data. Fluid Phase Equilibria 19: 33–149.
Venkataramanan, L., Weinheber, P., Mullins, O.C. et al. 2006. Pressure Gradients and Fluid Analysis as an Aid to
Determining Reservoir Compartmentalization. Presented at the SPWLA 47th Annual Logging Symposium, Veracruz,
Mexico.
Wu, J., Torres-Verdin, C., Sepehmoori, K. et al. 2005. The Influence of Water-Base Mud Properties and Petrophysical
Parameters on Mudcake Growth, Filtrate Invasion, and Formation Pressure. PETROPHYSICS 46: 14–32.
Wu, Jianghui, et al. 2001. Numerical Simulation of Mud Filtrate Invasion in Deviated Wells. Presented at the Annual
Technical Conference and Exhibition, New Orleans, Louisiana. SPE-71739-MS.
Zhuan, Wenhaog and Kiran, Erdogan. 1996. Zuo, J. Y., Zhang, D., Dubost, F. X. et al. 2008. EOS Based Downhole Fluid
Characterization. Presented at the SPE Annual Technical Conference and Exhibition, Denver, Colorado. SPE 114702-
MS.
Zuo, J.Y. et al. 2017. Advances in Quantification of Miscible Contamination in Hydrocarbon and Water Samples
From Downhole to Surface Laboratories. SPWLA 58th Annual Logging Symposium, Oklahoma City, Oklahoma.
SPWLA-2017-EE. SPWLA-2017-EE.

You might also like