Comparison of the gas hydrate plugging potentials of a set of crude oils from the Norwegian...

7
Comparison of the gas hydrate plugging potentials of a set of crude oils from the Norwegian continental shelf using chemometric decomposition of GC–FID data James R. Gasson a , Tanja Barth a , Georgi Genov a,b,n a Department of Chemistry, University of Bergen, Alle´gaten 41, NO-5007 Bergen, Norway b Department of Physics and Technology, University of Bergen, Alle´gaten 55, NO-5007 Bergen, Norway article info Article history: Received 7 May 2012 Accepted 14 January 2013 Available online 14 February 2013 Keywords: chemometrics petroleum gas-hydrate plugging GC–FID crude oil abstract Potential complications in the exploitation of subsea oil-fields due to gas hydrate plugs in the transportation lines lead us to explore the suitability of two routinely performed and thus easily available chemical analysis methods coupled with chemometric analysis to separate oils with high plugging potential from dispersion oils. Genov et al. (Organic Geochemistry 39 (2008) 1229–1234) successfully developed a method to separate these based on FT-IR spectroscopy coupled with chemometric analysis. Using whole-oil GC–FID data, we are able to present successful separation of these same oils in this paper and to deliver estimated molecular weight ranges showing significant impact on the plugging potential of different crudes analysed. We have localised retention time areas both in the chromatographic peak and the UCM data that strongly influence hydrate plug inhibition in some of the oils from the sample set. The results point to naturally occurring branched molecules in the nC 9 nC 13 range which, coupled with our prior results based on functional group allocation, suggest that ester or other carbonyl moieties in these components inhibit clathrate hydrate plug build-up. & 2013 Elsevier B.V. All rights reserved. 1. Introduction 1.1. Background Minor constituents in crude oil can be a determining factor for the interactions between the oil and other phases, i.e., gas, water or solids. Such interactions are important in the pipeline transport of oil–water–gas mixtures and in oil recovery operations. A special type of interaction occurs when gas hydrates nucleate and grow in the fluids during long-distance sub-sea pipeline transportation of oil, where low temperatures and high pressures support stable hydrate formation. Clathrate hydrates are ice-like solids, where small guest molecules are entrapped in the cages of host clathrate lattices made of hydrogen-bonded water molecules. Hydrate formation is a first order phase transition in a mixture of water and hydrate-forming molecules or compounds under low temperature and high pressure conditions. The guest molecules are crucial to stabilise the water cage-structure by means of van-der-Waals forces. If the guest molecule is a gas, the resulting solid is called gas hydrate (Sloan, 1998). Hydrates may grow either into solid plugs or into small crystals, which are dispersed in the fluid. Solid hydrates can potentially cause pipeline blockages and damage equipment, while disper- sions are harmlessly transported within the flow (Kelland, 2011). Extensive testing has shown that solid plugs form in some oils, while in others, only dispersed crystals are produced. This beha- viour has been explained by the possible presence of natural inhibiting compounds (NICs) in the dispersion-forming oils (Bergflødt et al., 2004; Høiland et al., 2005). Gas hydrates formed in pipelines and other oil or gas produc- tion equipment have been recognised as a costly problem as early as 1934 (Hammerschmidt, 1934). Today, the oil industry spends roughly 5–8% of the total product plant cost on measures to inhibit gas hydrate formation (Chandragupthan, 2011). The Deep- Star consortium estimated an average cost for the replacement of a mile-long section of a hydrate-damaged pipeline in deep waters at one million dollars (Chandragupthan et al., 2011). In the last 25 yr, the industry has focused on the development of three classes of hydrate inhibitorsthermodynamic inhibitors (TDIs, shifting the hydrate phase boundary), kinetic inhibitors (KIs, delaying hydrate nucleation and crystal growth) and anti- agglomerants (AAs, hampering the agglomeration of small into larger crystals). TDIs are used most frequently today, but the process requires very high concentrations. This increases the cost enormously and invokes environmental concerns (Frostman et al., Contents lists available at SciVerse ScienceDirect journal homepage: www.elsevier.com/locate/petrol Journal of Petroleum Science and Engineering 0920-4105/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.petrol.2013.01.010 n Corresponding author at: Institute of Physics and Technology, University of Bergen, Alle ´ gaten 55, NO-5007 Bergen, Norway. E-mail address: [email protected] (G. Genov). Journal of Petroleum Science and Engineering 102 (2013) 66–72

Transcript of Comparison of the gas hydrate plugging potentials of a set of crude oils from the Norwegian...

Page 1: Comparison of the gas hydrate plugging potentials of a set of crude oils from the Norwegian continental shelf using chemometric decomposition of GC–FID data

Journal of Petroleum Science and Engineering 102 (2013) 66–72

Contents lists available at SciVerse ScienceDirect

Journal of Petroleum Science and Engineering

0920-41

http://d

n Corr

Bergen,

E-m

journal homepage: www.elsevier.com/locate/petrol

Comparison of the gas hydrate plugging potentials of a set of crude oilsfrom the Norwegian continental shelf using chemometric decompositionof GC–FID data

James R. Gasson a, Tanja Barth a, Georgi Genov a,b,n

a Department of Chemistry, University of Bergen, Allegaten 41, NO-5007 Bergen, Norwayb Department of Physics and Technology, University of Bergen, Allegaten 55, NO-5007 Bergen, Norway

a r t i c l e i n f o

Article history:

Received 7 May 2012

Accepted 14 January 2013Available online 14 February 2013

Keywords:

chemometrics

petroleum

gas-hydrate

plugging

GC–FID

crude oil

05/$ - see front matter & 2013 Elsevier B.V. A

x.doi.org/10.1016/j.petrol.2013.01.010

esponding author at: Institute of Physics an

Allegaten 55, NO-5007 Bergen, Norway.

ail address: [email protected] (G. Geno

a b s t r a c t

Potential complications in the exploitation of subsea oil-fields due to gas hydrate plugs in the

transportation lines lead us to explore the suitability of two routinely performed and thus easily

available chemical analysis methods coupled with chemometric analysis to separate oils with high

plugging potential from dispersion oils. Genov et al. (Organic Geochemistry 39 (2008) 1229–1234)

successfully developed a method to separate these based on FT-IR spectroscopy coupled with

chemometric analysis. Using whole-oil GC–FID data, we are able to present successful separation of

these same oils in this paper and to deliver estimated molecular weight ranges showing significant

impact on the plugging potential of different crudes analysed. We have localised retention time areas

both in the chromatographic peak and the UCM data that strongly influence hydrate plug inhibition in

some of the oils from the sample set. The results point to naturally occurring branched molecules in the

nC9–nC13 range which, coupled with our prior results based on functional group allocation, suggest that

ester or other carbonyl moieties in these components inhibit clathrate hydrate plug build-up.

& 2013 Elsevier B.V. All rights reserved.

1. Introduction

1.1. Background

Minor constituents in crude oil can be a determining factor forthe interactions between the oil and other phases, i.e., gas, wateror solids. Such interactions are important in the pipeline transportof oil–water–gas mixtures and in oil recovery operations.A special type of interaction occurs when gas hydrates nucleateand grow in the fluids during long-distance sub-sea pipelinetransportation of oil, where low temperatures and high pressuressupport stable hydrate formation.

Clathrate hydrates are ice-like solids, where small guest moleculesare entrapped in the cages of host clathrate lattices made ofhydrogen-bonded water molecules. Hydrate formation is a first orderphase transition in a mixture of water and hydrate-forming moleculesor compounds under low temperature and high pressure conditions.The guest molecules are crucial to stabilise the water cage-structureby means of van-der-Waals forces. If the guest molecule is a gas, theresulting solid is called gas hydrate (Sloan, 1998).

ll rights reserved.

d Technology, University of

v).

Hydrates may grow either into solid plugs or into small crystals,which are dispersed in the fluid. Solid hydrates can potentiallycause pipeline blockages and damage equipment, while disper-sions are harmlessly transported within the flow (Kelland, 2011).Extensive testing has shown that solid plugs form in some oils,while in others, only dispersed crystals are produced. This beha-viour has been explained by the possible presence of naturalinhibiting compounds (NICs) in the dispersion-forming oils(Bergflødt et al., 2004; Høiland et al., 2005).

Gas hydrates formed in pipelines and other oil or gas produc-tion equipment have been recognised as a costly problem as earlyas 1934 (Hammerschmidt, 1934). Today, the oil industry spendsroughly 5–8% of the total product plant cost on measures toinhibit gas hydrate formation (Chandragupthan, 2011). The Deep-Star consortium estimated an average cost for the replacement ofa mile-long section of a hydrate-damaged pipeline in deep watersat one million dollars (Chandragupthan et al., 2011).

In the last 25 yr, the industry has focused on the developmentof three classes of hydrate inhibitors—thermodynamic inhibitors(TDIs, shifting the hydrate phase boundary), kinetic inhibitors(KIs, delaying hydrate nucleation and crystal growth) and anti-agglomerants (AAs, hampering the agglomeration of small intolarger crystals). TDIs are used most frequently today, but theprocess requires very high concentrations. This increases the costenormously and invokes environmental concerns (Frostman et al.,

Page 2: Comparison of the gas hydrate plugging potentials of a set of crude oils from the Norwegian continental shelf using chemometric decomposition of GC–FID data

Table 1Selected oil sample propertiesa.

Oil GH plugging

potentialb

Biodegradation

levelc

Density

(kg/m3)

Viscosity at 15 1C

(Pa s)

B1cd High 2 938.6 0.51867

B2c Low 5 925.6 0.17503

B4a Low 8 893.3 0.04407

B4c Low 2 891.3 0.03267

S2b High 0 843.0 0.00563

S3c High Low 850.7 0.0428

S3b High Low 851.3 0.0099

S5b High 0 826.9 0.0133

S7a High Low 847.8 0.03453

S7b High Low 849.6 0.04362

Test High 8 931.2 0.4506

a The presented data is an excerpt from the oil properties originally given in

Genov et al. (2008).b Evaluation of plugging potential carried out at Statoil ASA, Norway and

communicated to us.c Values after Barth et al. (2004) unless specied otherwise; biodegradation

levels are according to Peters–Moldowan scale; biodegradation levels assigned as

Low were communicated to us by Statoil ASA, Norway.d Generally heavily biodegraded but is mixed with small amount of non-biode-

graded oil, bringing overall biodegradation level down to 2 (Statoil ASA, Norway).

J.R. Gasson et al. / Journal of Petroleum Science and Engineering 102 (2013) 66–72 67

2003; Kelland et al. 2011). KIs and AAs are both low dosagehydrate inhibitors (LDHIs), and therefore may be a more costeffective and environmentally friendly method of hydrate inhibi-tion. NICs appear to work by a mechanism similar to KIs or AAs,but the precise mechanisms are yet to be fully explained(Bergflødt, 2001; Fadnes, 1996; Gaillard et al., 1999; Høilandet al., 2005). Identification, isolation, and subsequent addition ofactive NICs to plugging oils could turn these into dispersion-forming oils. This could prove to be an inexpensive and environ-mental friendly way of counteracting potential hydrate plugs.

1.2. The approach

The assessment of hydrate plugging potential is a complex,cost intensive method, which is realised using high pressure flowsimulators. These are typically found in both saphire cell (Fadnes,1996), or wheel-shaped loop setups (Hemmingsen et al., 2008).

Although a wide range of analytical procedures aimed at char-acterising oils on a chemical basis are routinely carried out within oilrecovery, only marginal efforts have been undertaken to try thesemethods to classify oils with high and low plugging potential with thehelp of multivariate methods or detect potentially present NICs(Bergflødt et al., 2004). The wide range of largely standardisedapproaches within the oil industry has prompted us to select twowidely used analytical methods, which we believe may be suitable toseparate high and low-risk plugging oils in combination withchemometric analysis. A successful classification approach basedon the two methods of choice, Fourier transform infra-red spectro-scopy (FT-IR) and whole-oil gas chromatography (GC) could beused to create a database approach to classify newly analysed oilsin terms of their plugging potential. The results of the latter arepresented in this paper. In addition, we hope that these methodswill aid in the identification of inhibiting components within thedispersion-forming oils.

For this, a set of 10 oils from the Norwegian continental shelfand one additional sample from the coast of Africa were analysed.The selected oils comprise a suitable sample set, with a largevariation of biodegradation ranks and include 3 dispersion-forming and 8 oils with high gas hydrate plugging potential.

In our previous work, infra-red spectroscopy data of this set ofcrude oils were used to compare biodegradation ranks andplugging potentials both on the basis of a classical spectroscopyapproach as well as principal component analysis (PCA) (Genovet al., 2008). Both approaches effectively distinguished biode-graded from non-biodegraded samples. All samples with lowplugging potential were to some extent biodegraded. One effectof biodegradation is an increase of acidity of the oil (Barth et al.,2004; Behar and Albrecht, 1984; Meredith et al., 2000; Wengeret al., 2002; Yemashova et al., 2007). Experimental evidencesuggest that adding the acidic component extracted from non-plugging oil into plugging oil can turn the plugging oil from aplug-former into a dispersion-former (Hemmingsen et al., 2007;Høiland et al., 2005). Using the acquired FT-IR data, differentia-tion between low and high plugging potential oils within a subsetincluding only biodegraded samples could only be achieved bymulticomponent analysis. The results suggested that oils havinghigh plugging potential contain predominantly longer and morelinear compounds in comparison with the low plugging potentialoils within the sample set. However, this trend, which is asso-ciated with the degree of biodegradation, is not the singledetermining factor as to whether a given oil will prevent theformation of hydrate plugs, as can also easily be concluded bycomparing the plugging potentials with biodegradation ranks inTable 1.

To complement the physical and bulk property information on theoils given by the FT-IR measurements, whole-oil GC analysis was

chosen to gain further insight into individual components of these oilsthat might inhibit plugging. Due to the vastly complex chromato-grams and differentiating samples, chemometric evaluation was usedto extract potentially underlying information. Combinations of differ-ent GC techniques with chemometric evaluation have been used indifferent fields ranging from food to the petroleum industry and haveproven to aid in extracting crucial information from the chromato-grams (Bartolome et al., 2007; Berlioz et al., 2006; Bodle, 2007;Camara and Arminda Alves, 2006; Cajka et al., 2010; Christensenet al., 2004, 2005; Ventura et al., 2011). Applications range fromproduct classification to fingerprinting, tracing of origin and forensics.The whole-oil GC analysis used here allows insight into the morevolatile components of the crude oils. The characteristic patternscommonly seen when analysing crude oils on GC originate fromthe major straight-chained hydrocarbons. Using a standard non-biodegraded oil as a reference, allocations of several componentswithin different samples is possible using pattern and retentiontime comparison, without the requirement of using a detector thatidentifies individual components.

GC provides not only data in the form of the chromatographicpeaks, but also information that can be gained from the unre-solved complex mixture (UCM) ‘‘hump’’ in biodegraded oil, whichis indicated in Fig. 1. This unresolved ‘‘hump’’ can be used as acharacteristic identifier for oils and potentially as an indicator forsome of the bulk properties as well as biodegradation ranks.

Our aim in this paper is to evaluate the classification potentialof whole-oil GC data from a set of crude oils in respect to theirplugging potential. In addition, we aim to compare findings basedon FT-IR data from our previous work with the new informationgained in the approach in this paper. By this, we try to identifymarked compositional differences between a set of crude oils inconnection with their plugging potentials.

2. Materials and methods

2.1. Samples

Ten oils from the Norwegian continental shelf were analysedby GC–FID. Four oil samples are biodegraded and are labelledwith a ‘‘B’’ as the first letter in the sample code. The remaining six

Page 3: Comparison of the gas hydrate plugging potentials of a set of crude oils from the Norwegian continental shelf using chemometric decomposition of GC–FID data

Fig. 1. FID traces of the heavily biodegraded oil B2c (top) and the non-biode-

graded oil S5b (bottom) illustrating the spread of sample-set as seen in GC–FID.

The unresolved complex mixture (UCM) area is indicated for B2c.

J.R. Gasson et al. / Journal of Petroleum Science and Engineering 102 (2013) 66–7268

are non-biodegraded and are labelled with an ‘‘S’’ as the firstletter. The second character in the code is a number related to theoil field. The third describes the different sample batches from thesame field, which have near-identical or at least similar proper-ties. Oil code B1c can hence be read as the third batch of oil, fromthe first biodegraded source. Table 1 summarises some selectedoil properties of relevance to this work. A further eleventh oilfrom the coast of Africa is included in the sample set as a testsample.

2.2. Experimental

Sampling of the oils on whole-oil GC was carried out afterheating the stock canisters to 50 1C and shaking to achievehomogeneity. The oils were analysed using a Thermo Finnigantrace-GC equipped with FID. Samples were introduced withoutany further dilution or preparation in 0.2 ml volume via a manualinjection port onto a 50 m Agilent Hp-PONA column with 0.2 mmdiameter. Helium was used as the carrier gas. The initial tem-perature of 30 1C was held for a period of 15 min. The heatingcontinued at a rate of 1.5 1C/min to a temperature of 60 1C,followed by a heating rate of 4 1C/min to a maximum temperatureof 320 1C. This post-run temperature was held for 15 min. Allsamples were analysed in triplicate. A standard reference oil(NSO-1) was used to calibrate and identify components in theoil samples based on retention time (Weiss et al., 2000). Nofurther standard was added, as the dominant characteristicpatterns of straight-chained hydrocarbons was deemed sufficientto identify and allocate the important reference components inall oils.

2.3. Data acquisition and pretreatment

The data acquisition programme (Dionex Chromeleon 6.0)records the electronic signals from the FID and later saves theseas compressed non-equidistant data. The underlying algorithmenables the programme to recognise areas in the chromatogramwith small and large alterations, as given for example by com-pound peaks, and thus alters the time-step size between saveddata points, enabling a compact data format, whilst still allowingsufficient peak resolution.

To enable chemometric analysis all analysed data sets shouldhave matching time-tags. This is, however, not always immedi-ately the case and can have the following reasons: (a) theconstituents of the analysed oils can differ greatly and as a result,the stored data has very different sampling time. (b) The chro-matographic peak of the same component may appear at variableretention times in different runs, also of the same sample. This is

due to different factors such as minor temperature variations inthe heat chamber through the different runs, causing some minorvariations in retention times. Typical implementation issues,problems, efforts on how to deal with these, and the subsequentimpact on results have been well described by e.g. Johnson et al.(2003) and Mason et al. (1992).

A number of successful algorithms dealing with these issueshave been developed in recent years, of which some are alsocommercially available. Currently, the most common algorithmsare based on ‘‘time warping’’, e.g. see Nielsen et al. (1998). Fastmatching algorithm developments for the implementation of gaschromatographic data for chemometric analysis have further beendocumented by Johnson et al. (2003). In this work, two approacheswere considered: (a) convert the time-non-equidistant datasetsinto equidistant ones by interpolation, and (b) implement algo-rithms for peak matching. The first approach was dismissed due tothe impractical sizes of the resulting data files. The single stepsdescribing the second simple time-binning approach we havechosen are listed below:

1.

Every chromatogram is normalised so that its integral equals1. This is done to provide concentration independent data,hence, making the different runs comparable.

2.

An algorithm for separating peaks from the background is run.This step results in the generation of two separate datasets,one containing only the background/UCM of each chromato-gram, and one containing the chromatograms corrected forbackground. Since, sometimes, up to 99% of the total chroma-togram integral (consequently possibly also a large degree ofthe information contained in the chromatogram) is containedin the UCM, the UCM data was subjected to a separatechemometric analysis.

3.

A peak search algorithm is run on the UCM-corrected chro-matograms. This algorithm yields the position (retention time)of the maximum of all detected peaks (above a certainpredefined threshold of intensity) and their integrals.

4.

To correct for peak mismatching due to ‘‘breathing’’, timebinning with bins of adequate time size (i.e., here 0.5 min) isintroduced. All peak integrals from all samples are thendistributed in their respective retention time bins. The binneddata set is ready for chemometric analysis.

PCA was performed using ‘‘The Unscrambler X 10.1’’ software(Camo Software AS, Norway). Mean centred data were analysedby PCA using a non-linear iterative partial least squares (NIPALS)algorithm to evaluate similarities and differences amongchromatograms.

3. Results

3.1. UCM analysis

The score plot of the first two principal components of theUCM data, (Fig. 2) shows a relatively good clustering of samples,taking into consideration the low resolution data from a smoothedbaseline. One replicate of the sample Test was removed as anoutlier due to irregularities in the data acquisition. The firstprincipal component describes 92% of the variance among thesamples, the second contributes 6%. The oils in the left half of thediagram have low biodegradation rank, whereas those on theright are highly biodegraded. Oils classified as having a lowplugging potential cluster in the top right quarter of the plot,whereas the oils B1c and Test, which are highly biodegraded andhave high plugging potential, cluster in the bottom right quarter.

Page 4: Comparison of the gas hydrate plugging potentials of a set of crude oils from the Norwegian continental shelf using chemometric decomposition of GC–FID data

Fig. 2. Score-plot illustrating the spread of oil samples based on the first two

principal components of the UCM data analysis. Both PCs together describe a total

of 98% of the variance within the data (PC-1: 92% and PC-2: 6%). B2c, B4a and B4care the non-plugging oils.

J.R. Gasson et al. / Journal of Petroleum Science and Engineering 102 (2013) 66–72 69

The loadings (not shown) show a positive correlation for low riskplugging biodegraded oils with higher values for the UCMbetween 84 min and 110 min retention time. This indicates thatsome of the compounds eluting within this retention time spanmay contribute to the inhibition of the build-up of solid hydrateplugs.

Fig. 3 shows the score plots for the possible combinations ofthe first three PCs describing the chromatographic peak data afterbaseline (UCM) subtraction. The clustering of the different oilsamples shows good reproducibility of the chemical analysis anddata treatment. The three PCs describe a total of 87% of thevariance within the data set.

Fig. 3b shows a clear separation between biodegraded andnon-biodegraded samples. The biodegraded oils occur on the leftside of the plot, whereas the non-biodegraded oils are on the rightside. This separation is mainly described by PC-1 (60% of variancedescribed). The loading line plot for PC-1 is shown in Fig. 4. Thepositive loadings for PC-1 are dominated by peaks representingn-alkanes. This is in accordance with the preferred microbialdegradation of straight-chain relative to branched components(Barth et al., 2004).

The separation between high risk plugging and non-pluggingoils is not entirely distinct. Whereas oils with low pluggingpotential cluster in the top left quarter of Fig. 3b, samples B1cand Test also plot close to the non-plugging oil B2c. For identi-fication of contributing components that inhibit plugging, asubset analysis of the apparently closely related plugging oils inaddition to the non-plugging oil B2c in the bottom left quarter ofthe plot might allow a clearer distinction of the seemingly minordifferences between these samples.

The score plot illustrating PC-2 (19%) vs. PC-3 (8%) in Fig. 3cshows an equally good clustering of the sample oils which indicatesthat systematic variation is described. However, the positioning ofthe samples in the plot does not lend itself to immediate inter-pretation regarding biodegradation or plugging potential.

In Fig. 3d the clearly dominating first principal component isrelated to the biodegradation, as discussed for Fig. 3b. Again, amixed group of B1c, B2c and Test plot closely together in thebottom left hand of the plot, whereas the other two non-pluggingbiodegraded oils B4a and B4c plot in the top left quarter ofthe plot.

The subset of the three biodegraded oils, plotting in thebottom left hand quarter of Fig. 3b and d, was reexamined tofind the differences between the non-plugging oil B2c and theplugging oils B1c and Test. The score plot is given in Fig. 5, whichshows that the differentiation between the plugging and non-plugging oils is described by PC-2 for this subset.

The loading line plot in Fig. 6 shows the signals contributing tothe second PC in the score plot of the subset of oils in Fig. 5. It canbe seen that the describing components are not necessarily themost dominant signals in the chromatograms. Positive correla-tions between the plugging potential and abundance of com-pounds situated in the range nC14–nC20 can be seen, whereascompound signals in the range nC9–nC13 are negatively correlatedwith the plugging potential and thus possibly describe activeplugging inhibitors. None of the given peaks coincide with any ofthe straight-chained components which could be identified withthe help of our reference oil.

4. Discussion

GC analysis of crude oils is primarily used to display hydro-carbon profiles, while NICs that contribute positively as disper-sants and/or anti-agglomerants in forming hydrate dispersionsare assumed to have dipole moments that are active at thehydrate interface. Thus, GC analysis would seem less suited forcharacterisation of oils with regard to their plugging propertiesthan e.g. FT-IR spectroscopy, as previously used. However, wehave seen that relevant information can also be extracted from GCdata. This is especially clear in the UCM based model, where theclassification based on the UCM profile gives a very similarclassification of the oils compared to FT-IR based models obtainedbefore (Genov et al., 2008). This is to be expected, as polarcompounds may be significant contributors to the UCM in thechromatogram of biodegraded oils. Further analysis of the UCMcomposition by more advanced methods is therefore highlyrelevant.

In prior work, multivariate analysis of FT-IR data of the sameoils showed a more branched carbon-skeleton for the biode-graded oils of which the non-plugging oils demonstrated a highintensity carbonyl band in the score plot, separating them fromthe other oils. The quality of differentiation of biodegraded andnon-biodegraded samples is maintained when using GC–FID datain combination with the statistical analysis. The interpretability ofthe loading line plots even allows a more specific identification ofsingle components contributing to this differentiation. Separationof oils based on plugging phenomena is also retained. The loadingline plots also lend themselves to direct interpretation of rangeswithin which single influential components are found. Due to thecomplexity of the samples and the given resolution, a moredetailed separation method aiming at the identification of com-pounds eluting within these retention time ranges are required.FT-IR analysis has indicated that these compounds include ahetero-atomic functionality and are thus polar species. Althoughnot a standard analysis, high resolution mass spectrometryanalysis with electrospray ionisation (ESI) could be a potentialmethod of choice to further this analysis approach (Marshall andRodgers, 2008).

Work by Erstad and Potz, documented in Erstad (2009), showedpromising results using such an approach. Using high resolutionmass spectrometry these show a very complex composition of thepolar fractions of oils with non-plugging properties. The oxygencontent was high, with esters being the most frequently occurringfunctionality. Borgund et al. found that active fractions hadmolecular weight ranges with average molecular weights of 400–600 g/mole as determined by gel permeation chromatography

Page 5: Comparison of the gas hydrate plugging potentials of a set of crude oils from the Norwegian continental shelf using chemometric decomposition of GC–FID data

Fig. 3. Score-plots illustrating the spread of oil samples based on the first three principal components of the peak data. These first three PCs together describe a total of

87% of the variance within the data (PC-1: 60%, PC-2: 19%, and PC-3: 8%). B2c, B4a and B4c are the non-plugging oils.

Fig. 4. Loading line plot of PC-1 of the baseline corrected peak data. Prominent are

straight chained alkanes, as identified by NIGOGA reference chromatograms and

analysed NSO-oil standard (Weiss et al., 2000). These are positively correlated

with non-biodegraded oils.

Fig. 5. Score-plot of the chosen peak data subset of the three closely related

plugging and non-plugging biodegraded oils. Both PCs together describe a total of

92% of the variance within the data (PC-1: 56% and PC-2: 36%).

J.R. Gasson et al. / Journal of Petroleum Science and Engineering 102 (2013) 66–7270

(GPC), and observed a large UCM in the GC–MS chromatograms ofthe polar fractions for biodegraded oils even after derivatisation(Borgund et al., 2009). Amides were observed in significant con-centrations in functional group analysis. Overall, the biodegrada-tion contributes a wide range of polar compounds to the oil thatcould contribute to their hydrate dispersion properties. Suchcompounds are not easily observed as single peaks in the whole-oil GC traces, both due to the low concentrations and their beingless suitable for chromatographic analysis. This can explain why

the UCM analysis gave the most direct separation of plugging andnon-plugging oils in the PCA analysis.

The GC–FID data most clearly depict the lighter hydrocarboncomponents. These are probably not as important for determininghydrate plugging properties of oil, and thus the interpretation interms of individual components is difficult and mostly showsthe expected patterns of biodegradation. However, even in thisrange there may be additional compounds from biodegradationprocesses that could be present in low concentrations, and thus

Page 6: Comparison of the gas hydrate plugging potentials of a set of crude oils from the Norwegian continental shelf using chemometric decomposition of GC–FID data

Fig. 6. Loading line plot of PC-2 from the PCA of the data subset of the three oils

B2c, B1c and Test. Retention time reference-points to some of the labelled

components from Fig. 4 for some of the components are included.

J.R. Gasson et al. / Journal of Petroleum Science and Engineering 102 (2013) 66–72 71

the GC–FID based PCA analysis is interesting also in the perspec-tive of identifying potentially active compounds in the lowermolecular range. For such an application the preprocessing of theGC data is critical, to be sure that peaks are reproducibly assigned.The lighter hydrocarbons also comprise the main ‘‘solvents’’ incrude oils, so selective biodegradation may change the solventstrength of the bulk crude and thus influence interactions withheavier components. The data pretreatment and preprocessingdescribed here can thus increase the quality and reliability of theanalytical data to the level required for multivariate ‘‘datamining’’. Even though no definitive identifications of activecompounds were obtained in this study, a good separationbetween samples with low and high plugging potential could beobserved, exemplifying that multivariate analysis is a viablemethod to extract underlying information from such complexchromatograms. Evaluation of the large already existent datasetscould yield the potential for a good classification model todecipher between oils with varying hydrate plugging potentials.

5. Conclusions

Decomposing total chromatographic GC–FID data of 11 crudeoils to a separated UCM data set and a baseline corrected peakdata set proved to increase the accessibility to the thus max-imised amount of data-information. The low resolved UCM dataallowed a surprisingly good separation in the score plot not onlybetween biodegraded and non-biodegraded samples but alsobetween plugging and non-plugging oils. The extractable infor-mation content from the loadings is however limited and canmerely be used as an indicator to localise relevant compoundpeak retention areas from the peak data analysis.

Using the baseline corrected peak data and establishing asuccessful alignment strategy resulted in a clear separation betweenbiodegraded and non-biodegraded samples based on one principalcomponent. The loadings showed that the difference is largelyexplained by the prevailing straight-chain alkyl compounds in thenon-biodegraded samples and their lower abundance in biode-graded samples.

Separation of plugging and non-plugging oils within thebiodegraded samples proved to be possible. The identification ofresponsible compounds could be pin-pointed to non-straightchained inhibiting components in the nC9–nC13 range, whereasplugging promoters are suggested to be situated in the rangenC14–nC20.

The presented work ties in with the observations made in ourprior work (Genov et al., 2008). The implementation of GC–FIDdata to assess the biodegradation as well as plugging potentialcan be seen as a complement to those FT-IR based results.

The combination of data from standard analytical character-isation techniques and multivariate evaluation shows consider-able potential as a screening technique for hydrate pluggingpotentials, for example in the surveillance of compositionalvariation in pipelines transporting oil from several fields orreservoir compartments.

Acknowledgements

The authors express their gratitude to Statoil ASA, Norway, andStatoil’s VISTA scholarship programme, administered by theNorwegian Academy of Science and Letters, for financing thiswork, in addition to providing both crude oil samples and fieldinformation. We further thank Djurdjica Corak, Bent BarmanSkaare and Egil Nodland for the valuable discussions and com-ments which ultimately led to this paper.

References

Barth, T., Høiland, S., Fotland, P., Askvik, K., Pedersen, B.S., Borgund, A., 2004. Acidiccompounds in biodegraded petroleum. Org. Geochem. 35, 1513–1525.

Bartolome, L., Deusto, M., Etxebarri, N., Navarro, P., Usobiaga, A., Zuloag, O., 2007.Chemical fingerprinting of petroleum biomarkers in biota samples usingretention-time locking chromatography and multivariate analysis. J. Chroma-togr. A 1157, 369–375.

Behar, F., Albrecht, F., 1984. Correlation between carboxylic acids and hydro-carbons in several crude oils. Org. Geochem. 6, 597–604.

Bergflødt, L., 2001. Influence of Crude Oil Based Surface Active Components andSynthetic Surfactants on Gas Hydrate Behaviour. Ph.D. Thesis. University ofBergen.

Bergflødt, L., Gjertsen, L., Sjøblom, J., Kallevik, H., Øye, G., 2004. Chemical influenceon the formation, agglomeration, and natural transportability of gas hydrates.A multivariate component analysis. J. Dispersion Sci. Technol. 25, 355–365.

Berlioz, B., Cordella, C., Cavalli, J., Lizzani-Cuvelier, L., Loiseau, A., Fernandez, X.,2006. Comparison of the amounts of volatile compounds in french protecteddesignation of origin virgin olive oils. J. Agric. Food Chem. 54, 10092–10101.

Bodle, E., 2007. Multivariate Pattern Recognition of Petroleum-based Accelerantsand Fuels. Ph.D. Thesis. University of Akron.

Borgund, A., Høiland, S., Barth, T., Fotland, P., Askvik, K., 2009. Molecular analysisof petroleum derived compounds that adsorb on gas hydrate surfaces. Appl.Geochem. 24, 777–786.

Cajka, T., Riddellova, K., Klimankova, E., Cerna, M., Pudil, F., Hajslova, J., 2010.Traceability of olive oil based on volatiles pattern and multivariate analysis.Food Chem. 121, 282–289.

Camara, J.M., Arminda Alves, J.M., 2006. Multivariate analysis for the classicationand differentiation of madeira wines according to main grapevarieties. Talanta68, 1512–1521.

Chandragupthan, B., 2011. An insight to inhibitors. PetroMin Pipeliner July–Sept,50–57.

Chandragupthan, B., Nounchi, G., Thirukkumaran, N., Jayakanthan, D., Jegadeesh, N.,2011. Flow assurance—special focus on hydrate blockage. PetroMin PipelinerJan–Mar, 2–16.

Christensen, J., Hansen, A., Tomasi, G., Mortensen, J., Andersen, O., 2004. Integratedmethodology for forensic oil spill identication. Environ. Sci. Technol. 38,2912–2918.

Christensen, J., Tomasi, G., Hansen, A., 2005. Chemical fingerprinting of petroleumbiomarkers using time warping and PCA. Environ. Sci. Technol. 39, 255–260.

Erstad, K., 2009. The Influence of Crude Oil Acids on Natural Inhibition of HydratePlugs. Ph.D. Thesis. University of Bergen.

Fadnes, F., 1996. Natural hydrate inhibiting components in crude oils. Fluid PhaseEquilibria 117, 186–192.

Frostman, L., Thieu, V., Crosby, D., Downs, H., 2003. In: Proceedings of theInternational Symposium on Oilfield Chemistry. Publication SPE 80269.

Gaillard, C., Monfort, J., Peytavy, J., 1999. Investigation of methane hydrateformation in a recirculating tests of efficiency of chemical additives on hydrateinhibition. Oil Gas Technol. 54, 365–374.

Genov, G., Nodland, E., Skaare, B., Barth, T., 2008. Comparison of biodegradationlevel and gas hydrate plugging potential of crude oils using FT-IR spectroscopyand multi-component analysis. Org. Geochem. 39, 1229–1234.

Hammerschmidt, E., 1934. Formation of gas hydrates in natural gas transmissionlines. Ind. Eng. Chem. 26, 851–855.

Page 7: Comparison of the gas hydrate plugging potentials of a set of crude oils from the Norwegian continental shelf using chemometric decomposition of GC–FID data

J.R. Gasson et al. / Journal of Petroleum Science and Engineering 102 (2013) 66–7272

Hemmingsen, P., Li, X., Kinnari, K., 2008. Hydrate plugging potential in under-inhibited systems. In: Proceedings of the 6th International Conference on GasHydrates.

Hemmingsen, P., Li, X., Peztavy, J.-L., Sjøblom, J., 2007. Hydrate plugging potentialof original and modified crude oils. J. Dispersion Sci. Technol. 28, 371–382.

Høiland, S., Askvik, K., Fotland, P., Alagic, E., Barth, T., Fadnes, F., 2005. Wettabilityof freon hydrates in crude oil/brine emulsions. J. Colloid Interface Sci. 287,217–225.

Johnson, K., Jarman, B.W.K., Synovec, R., 2003. High-speed peak matching algo-rithm for retention time alignment of gas chromatographic data for chemo-metric analysis. J. Chromatogr. A 996, 141–155.

Kelland, M., 2011. History of the development of low dosage hydrate inhibitors.Energy Fuels 20, 825–847.

Marshall, A., Rodgers, R., 2008. Petroleomics: chemistry of the underworld. Proc.Natl. Acad. Sci. USA 47, 18090–18095.

Mason, J., Kirk, I., Windsor, C., Tipler, A., Spragg, R., Rendle, M., 1992. A novelalgortihm for chromatogram matching in qualitative analysis. J. High Resolu-tion Chromatogr. 15, 539–547.

Meredith, W., Kelland, S.-J., Jones, D., 2000. In crude oil acidity and carboxylic acidcomposition. Org. Geochem. 31, 1059–1073.

Nielsen, N.-P., Carstensen, J., Smedsgaard, J., 1998. Aligning of single and multiplewavelength chromatographic profiles for chemometric data analysis usingcorrelation optimised warping. J. Chromatogr. A 805, 17–35.

Sloan Jr., E., 1998. Clathrate Hydrates of Natural Gases, second ed. Marcel DekkerInc., New York, USA.

Ventura, G., Hall, G., Frysinger, R., Raghuraman, B., Pomerantz, A., Mullins, O.,Reddy, C., 2011. Analysis of petroleum compositional similarity using multi-

way principal components analysis (MPCA) with comprehensive two-dimensional gas chromatographic data. J. Chromatogr. A 1218, 2584–2592.

Weiss, H., Wilhelms, A., Mills, N., Scotchmer, J., Hall, P., Lind, K., Brekke, T.,

2000. NIGOGA—The Norwegian Industry Guide to Organic GeochemicalAnalysis. Found Online: /http://www.npd.no/engelsk/nigoga/default.htmS-

(22.11.2011).Wenger, L., Davis, C., Isaksen, G., 2002. Multiple controls on petroleum biode-

gradation and impact on oil quality. SPE Reservoir Eval. Eng. 5, 375–383.Yemashova, N., Murygina, V., Zhukov, D., Zakharyantz, A., Gladchenko, M.,

Appanna, V., Kalyuzhnyi, S., 2007. Biodeterioration of crude oil and oil derivedproducts: a review. Rev. Environ. Sci. Biotechnol. 6, 315–337.