Copyright 2005, Society of Petroleum Engineers This paper was prepared for presentation at the 2005 SPE Annual Technical Conference and Exhibition held in Dallas, Texas, U.S.A., 9 – 12 October 2005. This paper was selected for presentation by an SPE Program Committee following review of information contained in a proposal submitted by the author(s). Contents of the paper, as presented, have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material, as presented, does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE meetings are subject to publication review by Editorial Committees of the Society of Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to a proposal of not more than 300 words; illustrations may not be copied. The proposal must contain conspicuous acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O. Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435114

Abstract Reservoir compartmentalization is a major cause of production underperformance in the oilfield. A well drilled into a compartmentalized reservoir will see only part of the hydrocarbon in place over the production time scale. The obstruction to free flow can be sealing faults, fault baffles, pinching out layers, sand lenses or low permeability areas. Although it is recognized that a comprehensive approach to compartmentalization must make combined use of all available rock and fluid data, the case studies in this paper show Downhole Fluid Analysis (DFA) to be one of the most effective techniques in recognizing compartmentalization via fluid signature comparisons and fluid density inversions. Corroboration with geological, petrophysical, reservoir engineering and production data confirms the strength of the DFA technique

Fluid compositional variations must be considered in order to acquire fluid samples representative of the reservoir at large and to devise optimal production strategies. In addition, fluid compositional variations can be utilized as a tool to identify compartmentalization because different compartments are likely to be filled with different fluids. The limitation on use of such techniques in the past has been the need to rationalize the use of wireline sampling to collect and analyze only the necessary samples. In this paper, DFA is shown to be the “missing link.” It provides the information necessary to optimize the sampling process and to decide in real time on where sampling is needed without necessarily having to bring all samples to surface.

Introduction Reservoir heterogeneity at a variety of scales can be caused by structural complexity, stratigraphic stacking patterns, or diagenetic alteration of pore system continuity. This heterogeneity commonly causes barriers or baffles to fluid flow. Such heterogeneity, often manifested as compartments at the reservoir scale, tends to hinder. The consequence of not recognizing flow compartmentalization is generally to anticipate more efficient drainage than is actually achieved. As a result, facilities are improperly sized, and reserves, production, and cash flow models suffer from inaccuracies. Various techniques exist to assist operators in detecting compartmentalization. In deep water and similar high cost operating environments, the traditional methods, drill stem tests (DST) and extended well tests (EWT) often become impractical, with costs approaching the costs of new wells and with emissions becoming increasingly undesirable. Thus, compartments often have to be identified by some other means. Looking for pressure continuity is one widely used method, but it is important to keeping mind that pressure communication is a necessary yet insufficient criterion to establish fluid communication particularly in normally-pressured basins. In other words, reservoirs in pressure equilibrium are not necessarily in thermodynamic equilibrium and/or flow communication, and being able to fit a single linear pressure gradient to a number of small compartments does not necessarily mean that they are in flow communication. One can easily imagine two large sand bodies with an intervening shale layer. If the shale layer has a few baffles or leaky faults connecting the sands, there will be pressure communication over geologic time, but little or no flow communication. Geochemical and other surface analytical methods can be applied to hydrocarbon fluid samples to help uncover compartmentalization. The problem is that without a priori knowledge that fluid complexities exist, the cost of comprehensive multiple sample acquisition is prohibitive. In addition, due to frequent discrepancies involved in sample acquisition, transfer and lab analysis, redundant samples are routinely processed further adding to the cost. What is needed is a new, cost effective technology to reveal fluid complexities, one for which costs are commensurate with the value of the information.

SPE 94709

The Missing Link—Identification of Reservoir Compartmentalization Through Downhole Fluid Analysis H. Elshahawi, SPE, and M. Hashem, SPE, Shell Intl. E&P Inc., and O.C. Mullins, SPE, and G. Fujisawa, SPE, Schlumberger Oilfield Services

2 SPE 94709

In this paper, several examples of fluid comparisons are shown to establish compartmentalization or the lack thereof. In particular, the question of medium-scale compartmentalization has been addressed by using multiple probe tools with one probe located below a possible flow barrier and the other above. Downhole fluid analysis is then performed on the corresponding fluids without an intervening tool reset to better understand compartmentalization. Compartmentalization Petroleum reservoirs can consist of flow units or compartments that may span the whole range from massive to quite small. This heterogeneity does not describe the overall size of the reservoir but it does strongly impact our ability to drain the reservoir. It is instructive here to relate reservoirs to everyday objects. A kitchen sponge is a well-connected open-cell system; the individual cells are connected so water can flow easily throughout the sponge. In reservoir lexicon, the sponge is a single compartment. A single hole or well placed in the center of the sponge could drain the entire fluid contents of the sponge. On the other hand a spool of bubble wrap is a closed-cell system. Fluids cannot flow from one bubble or cell to another. Indeed, if a knitting needle penetrated through the spool of bubble wrap, only those cells that are penetrated would drain. The spool of bubble wrap is highly compartmentalized, again in oil field jargon. Reservoir compartmentalization is a problem for E&P companies because it is a major cause of production under-performance. There are no established methods for generating compartment size distributions, but it is logical to expect that such distributions would be similar to distributions of the natural process that generate compartments in the first place. Faults are one example of such processes, and so explorationists may benefit from what structural geologists have known about fault (and fracture) distributions for a long time. Analysis of fault populations has shown that faults exhibit many characteristics of fractals. This means that some fault properties (e.g. the length/throw ratio) are relatively independent of fault size and that there is a frequency vs. size relationship between fault size (length or maximum throw) and the number of faults with a certain size (few large faults, many small faults). For fractal fault populations, this relationship can be used to predict the number of sub-seismic faults. Outcrop and 3D seismic data show that single faults are approximately elliptical and bounded by a tip-line of zero displacement. The ratio between fault length and maximum throw varies (Fig.1-a) between 15 and 150. This relationship can be used to predict whether, on 2D fault maps, long faults should be split into separate faults of smaller length. Faults tend to develop in families in which orientation varies over a limited range only (30 - 40). Histograms of fault frequency versus orientation often exhibit one or two normal distributions (Fig.1-c). A power-law relationship exists between the number of faults and fault size, i.e. many small faults and few large faults. This relationship is used to predict the number of small, sub-seismic faults (Fig.1-b). On the other

hand, fault throw gradients (along strike) vary over a limited range, generally <0.1 (Fig.1-d). Relationships between fault length and maximum throw and between fault size and fault frequency support a fractal model. It is instructive to look at the Gulf of Mexico (GOM) for clues about compartmentalization, as it is the most mature deepwater province in the world. In fact, July 2005 marks the 30th anniversary of the first field discovered in the deepwater, defined here as water depths of greater than 600 feet. In 1975, Cognac, the first commercial field, was discovered by Shell. Between then and 2002, over 21.2 billion barrels of oil equivalent were discovered and approximately 9.2 TCFG and 2,200 MMBO were produced [2].

There have been approximately 232 commercial fields discovered during the period from 1975 to the end of 2003. It is interesting to note that initial well flow rates in the GOM deepwater did not exceed 10,000 BOPD until 1995. Only three years later, individual well rates at Ram Powell and Troika exceeded 20,000 BOPD. The current mean size of these discoveries is 94.3 MMBOE, with the two largest fields found to date being Mars at 750 MMBOE and Thunder Horse at 1,000 MMBOE [1]. These statistics in themselves point to a wide distribution of compartment sizes. There are currently three major exploration plays in the deepwater province: the Flex Trend play, the Mini-Basin play and the Fold Belt play. In addition, the "sub-salt play" is simply any of these plays obscured by salt. Some of the larger flex trend discoveries were Lena, Zinc, Pompano and Green Canyon 18. Wells generally targeted "bright spots" on 2-D seismic. Most discoveries in the play were fields with small reserves, discontinuous sands and wells with fairly low flow rates. The mini-basin fields including Auger, Mars, Diana, Genesis, Troika and Europa had better than expected aquifer support and needed fewer wells to develop them. [1]. The Tahoe field is a classic example of compartmentalized reservoirs. This Shell-discovered and operated field is located 140 miles east-southeast of New Orleans in the GOM Viosca Knoll Blocks 783, 784 and 827 in water depths ranging from 1,200 ft to 1,600 ft. The main M4.1 reservoir is a Late Miocene sand located approximately 10,000 ft below sea level. It was formed by turbidite flows that entered an unconfined slope setting and deposited a NW-SE trending channel-levee system, draped over an anticlinal dome and cut by normal faults. Pressure analysis shows multiple pressure cells within the M4.1 indicating compartmentalization. High seismic amplitudes parallel to structural contours in the southeastern portion of the field correspond to a hydrocarbon-water contact. Amplitudes decrease as the levees thin laterally to the southwest and northeast, and this is interpreted to record stratigraphic pinch-outs. Faulting forms the up-dip trap in the M4.1 and enhances the compartmentalization of the reservoir. Similar Faults and stratigraphic pinch-outs play a key role in the compartmentalization and trapping of hydrocarbons elsewhere in the GOM [1].

SPE 94709 3

Linking Fluid Analysis to Compartmentalization The role of downhole fluid analysis in understanding compartmentalization can be better understood by examining the use of its better established cousin, organic geochemistry, in the evaluation of petroleum systems. After all, the two techniques are simply different ways of looking at the same fluids. Traditionally, organic geochemistry was regarded as strictly an exploration tool to evaluate source rocks or oils in order to high-grade prospects. Over the last two decades, however, a number of operators have successfully applied oil geochemistry to reservoir continuity assessment in a diverse range of geological settings including a wide range of field sizes, structural environments, reservoir lithologies, and oil types. As demonstrated by numerous published and unpublished case studies, petroleum geochemistry provides an effective tool for identifying vertical and lateral fluid flow barriers in oil and gas fields [16]. Geochemical techniques are especially useful when combined with other reservoir continuity indications such as WFT pressures, pressure decline curves, oil-water contacts, and fault juxtaposition. A gravity column liquid chromatograph is used to separate crude oils by physiochemical processes into compositionally distinct chemical families: saturate, aromatic, resin, and asphaltene fractions. Other processes can be used to separate the saturate fraction into normal, branched, and cyclic hydrocarbon fractions. As crude oil is generated and expelled from the source rock, some alteration of the originally generated oil occurs. This alteration is dependent upon the chemistry of the oil and gas and the chemical and physical properties of the rocks encountered during expulsion and migration to the trap. The differences between source rock extracts and oils are normally obvious when their chemical composition is compared. The differences in oils generated by the same source rock feeding two different reservoirs are quite subtle. Rather than focusing on the primary components of the oil, minor constituents of the oil are carefully compared. These are compounds that are most directly affected by physiochemical processes that result in minor variations in crude oils from the same source. Though somewhat complicated by multiple charging, seal leakages, biodegradation and the like, the ability to identify distinct reservoir compartments based on oil composition is the cornerstone of production geochemistry. Gas chromatography (GC) is also utilized in organic geochemistry to characterize rock extracts and whole oil samples. Gas chromatography separates and detects various compounds present in a simple or complex mixture by vaporizing the volatile components in the sample, e.g., oil, and then forcing the vapors through a long narrow (0.10 to 0.55 mm) fused silica tube coated with chemicals that interact with the vapors. These are tuned to the compounds of interest in order to obtain maximum separation and resolution. As the compounds elute from the opposite end of the column, they are detected usually by a flame ionization detector (FID) or in the case of biomarkers, the ionized particles are detected by differences in their mass and charge using a mass spectrometer

(MS). When oil is injected, the hydrocarbons present in the oil are separated and detected resulting in a “fingerprint” of the oil, which is basically the yield and distribution of hydrocarbons present in the sample. The predominant compounds in normal crude oils are paraffins, with generally the highest peaks in the fingerprint, and aromatic hydrocarbons. In biodegraded oils, however, the resin and asphaltene fractions of the oil predominate the fingerprint and typically yield few resolved compounds and a high unresolved hump of complex composition. Hydrocarbon fingerprinting is a combination of analytical and interpretive techniques that utilizes the fingerprint of oils to determine whether they are part of the same reservoir compartment. The differences seen in fingerprints are not usually related to source rock differences, but are a result of differences in migration pathways to the reservoir compartment. This results in geochromatography of the oil causing very slight differences in oil chemistry that can be discerned through careful analysis of oil fingerprints. Since a GC fingerprint is a representation of the relative concentrations of compounds present in an oil as analyzed in the laboratory setting. It is, in fact, a histogram of the relative yield and distribution of these compounds. A fingerprint is similar to a well log: it is a molecular log of the composition of a specific oil. While fingerprints are directly related to the source rock of the oil, reservoirs in continuous communication have nearly identical fingerprints even in trace components of the fingerprint, the so-called “grass between the trees”, where normal paraffins are the “trees” that dominate most crude oils. A polar plot or “star diagram” is a simple way to illustrate differences in GC fingerprints particularly selected peak ratios used to distinguish oils from distinct reservoirs. These selected peak ratios may be evaluated by statistical means. Star plots constructed in this way maximize the apparent differences between samples. By stripping away what samples have in common and focusing on how they differ, such plots allow discrete groups of samples to be readily, visually identified. Since a relative comparison of unknown peak heights is utilized, oils need to be analyzed at the same time under the same conditions [16]. Typically, in oils which are in fluid communication, none of the several hundred inter-paraffin compound ratios will differ by more than 10% from the corresponding ratios in oils with which they are in fluid communication. In contrast, when a lack of fluid communication exists between two samples, a large number of ratios in a given oil may differ significantly (typically >10%) from the corresponding ratios in the second oil. These ratios will commonly be distributed throughout the C8-C20 range, and will not be restricted to a narrow portion of the chromatogram. Differences restricted to a narrow portion of the chromatogram are a symptom of sample contamination with substances such as drilling additives and typically do not imply reservoir continuity barriers. The analytical reproducibility for ratios of closely spaced inter-paraffin peaks is typically 1-3%. As the number of ratios with significant differences between the samples decreases, the geochemical case for lack of communication becomes less conclusive.

4 SPE 94709

Exceptions to these guidelines exist in cases of certain thick, gravitationally segregated oil columns, where reservoir discontinuities must be identified from discontinuities in otherwise gradational changes in composition with depth. The C8 to C20 molecular weight range on the GC is typically the most diagnostic range for reservoir continuity assessments. If differences are found, the oils are not alike, and, therefore, not part of the same reservoir compartment. However, if oils are found to be alike, there are situations where there could be discontinuous reservoir compartments [16]. Geochemical analysis is quite powerful, but it is by no means a silver bullet. Geochemical analysis results are only as good as the samples upon which they were performed. For instance, the distribution of lower molecular weight compounds is usually not compared in HC fingerprinting, since they can be more readily affected by evaporative losses during sample handling. More generally, lab analysis results may not always be representative of formation fluids as a result of improper sample-handling, delayed evolution of components such as H2S or aspartames, pH for water samples, or leaky sample bottles. Even if all these processes were perfect, it is still prudent to validate the sample at each stage in its journey. Downhole fluid analysis provides the first step in this sample validation process and enables a “Chain-of-Custody” concept for hydrocarbon samples akin to the one currently mandated for blood samples in forensic analyses. Downhole Fluid Analysis In the development of deepwater prospects and other capital-intensive E&P projects, it is essential to understand the precise nature of the hydrocarbon fluids in terms of chemical and physical properties, phase behavior, and commingling tendencies to be able to optimize completion and facilities design and reservoir production strategies. Gas-liquid ratios, saturation pressures and viscosities are among fluid parameters that determine whether and how production is profitably pursued. Moreover, phase transitions such as asphaltene onset, wax, gas hydrate, organic scale and diamondoids can have a huge deleterious impact on production. These important concerns are in addition to the problems associated with flow assurance by now familiar to the industry. Consequently, getting the correct answer right from the start is a must. A factor that complicates the mapping of fluid properties in a given reservoir is the existence of large compositional variations of hydrocarbons vertically and/or laterally within the field. [5,9,16] Without proper accounting of these variations, there can be gross inaccuracies associated with predicting hydrocarbon production. Several operators now accept that hydrocarbon fluids should be considered strongly compositionally graded unless otherwise proven. The problem is that compositional grading is not easily modeled, and proper identification and understanding of it requires dense measurements with a large degree of data redundancy. This is because compositional grading can stem from a number of physical origins such as gravity, biodegradation, leaky seals, multiple charges, water stripping, thermal gradients, and

convective mixing. Of course, in production, miscible flooding and production below a phase transition also produce compositional gradients. Most importantly, many of these mechanisms drive the hydrocarbon column away from equilibrium. Consequently, one cannot collect a single sample and model the fluid variations in the column. Without sufficiently dense sampling, there is little hope for being able to model the complex fluid variations along the column. Downhole Fluid Analysis (DFA) refers to a concept rather than a specific tool. It includes a number of measurements, but the basic measurement relies on near-infrared spectroscopy (NIR). Fig.2 shows the two-stretch overtone peak of different chemical groups containing the C-H oscillator. As is true for all mechanical oscillators, the oscillation or vibration frequency depends on the (reduced) mass. Consequently, CH4, -CH3 and -CH2- groups all have somewhat different frequencies enabling their resolution. The details of NIR application for DFA have been described elsewhere.[7,8] The earliest DFA measurement consisted of obtaining GOR by measuring the dissolved methane vs. liquid oil fraction in a single-phase crude oil using the LFA* Live Fluid Analyzer. [11,15] The physical basis of this measurement is described in the literature. [9,13] Recent work has significantly improved the robustness and utility of this measurement. [2,3] A new downhole fluid analyzer, the CFA* Composition Fluid Analyzer has been more recently developed [5,6] with the validation of these measurements being described elsewhere. [7] The CFA provides analysis of C1, C2-C5, and C6+ fractions with CO2 being a potential product. The hydrocarbon analysis is robust yielding reasonably accurate concentrations of methane, other hydrocarbon gases and hydrocarbon liquids (Fig.3). Fig.4 shows a schematic of one possible configuration of the WFT downhole sampling tool. There are two fluid analyzers in this particular tool string, the LFA and the CFA with the LFA upstream (below) and the CFA downstream (above) of the pump module, but there are situations when it is advantageous to switch the position of these two modules. The modular design of WFT’s allows for different tool configurations to match specific acquisition needs. The LFA can measure GOR in the range of 0-2500 scf/bbl, while the CFA is designed to handle crude oils in the range of 1500 - 20,000 scf/bbl. In general, the CFA is more accurate for GOR determination in fluids with GOR>2000 scf/bbl, while the LFA is more accurate for GOR<1500 scf/bbl. Thus, the two tools together provide compositional analysis over a broad range of crude oils and condensates. The LFA monitors contamination utilizing the OCM algorithm and analyzes for gas using a gas refractometer and GOR [11]. The CFA, on the other hand, detects retrograde dew formation and gives GOR and fluid composition. [4,5,6] There is generally a sizable pressure increase across the pump, with less than formation pressure on the upstream side and borehole pressure on the downstream side. Fluids have a residence time within the pump, and so the pump can act as a phase separator. By placing one of the optical modules downstream of pump, one can readily detect different phases that have segregated. In the

SPE 94709 5

configuration shown in Fig.4, the CFA is deployed downstream (above) the pump. Detection of condensate formation is compromised in this configuration, but it is advantageous in that the residence time of fluids in the pump allows for gravity segregation of the fluids, which tends to improve CFA analysis. On the other hand, the LFA is deployed upstream of (below) the pump module of the MDT. This enables OBM-filtrate contamination monitoring without interference from pump dead volumes. [3, 12] Towards a Continuous Downhole Fluid Log DFA enables the analysis of hydrocarbon fluids at small marginal cost. If a hydrocarbon fluid sample is found to be similar to a fluid elsewhere in the column, then the sample need not be collected for subsequent laboratory analysis. That is fluid can be brought into the tool, analyzed, and discarded if laboratory analysis is deemed unnecessary. On the other hand, if fluid complexity is found that warrants laboratory analysis, then samples are acquired. Thus, DFA reveals hydrocarbon complexities and enables acquisition of important samples thereby improving the efficiency and effectiveness of hydrocarbon analysis. The overall objective is to generate a continuous downhole fluids log.[4] In the same spirit that pressure station measurements can be used to generate a pressure gradient, DFA station measurements can be used generate a quasi-continuous downhole fluids log. A crucial feature of the DFA method is that the DFA analysis program can be increased without logistical constraints to match the complexity of the hydrocarbon fluid column. For complex columns, many DFA stations are needed, often exceeding 15, employing the gargling technique. In these cases, the enhanced understanding of the fluid column is deemed well worth the marginal increases in cost associated with the additional DFA stations.

Sharp fluid property variations are common near critical points, phase boundaries, fluid contacts and permeability barriers. Compositional grading, on the other hand, tend to be less much less abrupt [16]. Therefore, a high vertical density of DFA stations is required at suspected fluid discontinuities while a lower density of DFA stations is sufficient for continuous fluid columns. Thresholds for compositional variations can be set between DFA stations such that if the threshold is exceeded, an intervening DFA station can be obtained. By acquiring multiple DFA stations in a hydrocarbon column, large compositional variations have been found. For example, a ~50% variation in GOR was found in a 100 ft column of oil. [5] This variation was not evident in the pressure gradient data but was confirmed by various means including acquisition of over 40 MDT samples. In addition, this large compositional variation means the pressure gradient is a curved line, not a straight line. This curvature of the liquid pressure gradient caused the modeled gas-oil and oil-water contacts to move apart. This was validated in real time by producing oil low in the column, thus enabling the operating company to book

more reserves. More subtle compositional gradients have also been measured using DFA.[4]

DFA is generally more sensitive to compositional gradients than pressure-derived gradients. Density is an integral quantity and is thus not very sensitive to abrupt point changes. Therefore, to measure fluid properties by measuring pressure gradients is the wrong way around. Instead, we should measure fluid properties to understand the fluid and check consistency with pressures and pressure gradients. Issues such as depth control, number of stations, and height of the column significantly impact pressure gradient accuracy but not DFA accuracy. Direct measurement of the gas-oil ratio, a proxy for hydrocarbon fluid density is much more sensitive. In fact gas-oil ratio per se has the highest possible sensitivity of any physical measurement as it ranges all the way from near zero in dead oils to infinity in dry gases.

DFA can uncover compartmentalization; zones containing separate fluids may reflect compartmentalization with or without density inversions. This is not to be confused with the hydrocarbon density inversions that often exist over large depths scales. The latter are not generally due to vertical compartmentalization. Instead, these density distributions are simply due to an interaction of source rock maturity, successive entrapment and trap leakage, and biodegradation.

Below are a few examples demonstrating the power of downhole fluid analysis in detecting barriers and uncovering reservoir compartmentalization. The examples will demonstrate the benefits of combining DFA with other log data to accurately depict changes along the hydrocarbon column. Example-I A string similar to the one shown in Fig.4 was run here. The oil at times <1000 seconds is one remaining in the tool from a previous sampling station slightly higher in the column and is similar to but slightly higher in GOR than the oil at the new depth as can be seen on the LFA log shown in Fig.5. In the corresponding CFA log, shown in Fig.6, the GOR of the liquid from the previous sampling station again appears slightly higher in GOR. Not that at 3600 seconds, the pressure was intentionally dropped to get a rough idea of the bubble point. Subsequently, the GOR of the liquid is seen to significantly reduce but then it recovers to stable levels with additional pumping. The measurements are consistent with flow communication and normal compositional grading. The lesson to take away here is that caution must be exercised while interpreting DFA results. The interpreter needs to know the sequence of operations quite well and to understand the impact of those operations on the measurements. The different levels of contamination at different stations must also be taken into account before making any conclusions about communication or the lack thereof. There is no better way to accomplish this than to be involved in real-time.

6 SPE 94709

Example-II This survey was run using a two-probe string to sample and test for compartmentalization. Three potential barriers to vertical flow in the were recognized in this reservoir based on well logs: 1. Uncorrelated calcite cemented zones observed in all wells 2. Correlated shale barriers in the start-up and abandonment deposition sequences 3. Dispersed shale barriers in the sand rich packages Unlike the case described in the Example I, the LFA was run here below the pump module and the CFA above. The WFT was run with two probes in the string as indicated in Fig.4. The lower probe was placed below a potential flow barrier while the higher probe was placed above it. Before 5000 seconds the lower zone was being sampled, and after 5000 seconds, the upper zone was sampled. The tool configuration allowed direct comparison of the two fluids to test compartmentalization. Using the OCM technique (3,6,13,14), fluid coloration and GOR were independently used to compute the change in OBM filtrate contamination with time in this sampling station. Both methods predicted approximately 2.5% OBM filtrate contamination, which was nearly identical to the subsequent surface GC measurements

The LFA log (Fig.7) shows the individual optical density channels and the corresponding GOR buildup profile. Plugging with sand occurred momentarily while sampling from the lower probe around 3400 seconds into the test causing the pressure to drop below bubble point as evidenced by LFA gas flags (not shown) and the sudden decrease in GOR. Once the sand plug was successfully dislodged, all values returned to normal levels, and the pressure and GOR were allowed to buildup to establish a reference level. Flow was then initiated from the top probe and allowed sufficient time to build up to the same level observed earlier. For the same pumping duration (~2000 seconds), the fluid from below (at 3400 seconds) and above (at 7000 seconds) the potential barrier are nearly identical. It was concluded, therefore, that the two zones were in flow communication. Fig.8 shows the CFA results for the same station. The fluid composition is shown in terms of weight %. The resulting GOR is also shown. Again, the decrease in CFA GOR 3400 seconds into the test is due to dropping below the bubble point. After similar pumping durations, the compositions of the fluids sampled above and below the potential flow barrier look nearly identical, again implying flow communication. Table-1 shows the agreement between the CFA and lab results. It is worth noting that strong fluorescence was observed with the CFA (not shown) and served as an additional indication of cleanup. This is in contrast to a very low fluorescence observed when the tools were seeing only mud early in the test (elapsed time t <1500 sec). Example-III This example comes from the same reservoir described in Example-II. The same toolstring was used to investigate two

formation depths, one above and one below a potential flow barrier. The bottom probe was first used up to 5400 seconds into the test. Then, the top probe was used for sampling from 5400 seconds to ~9000 seconds. Fig.9 shows the LFA results. The differences in GOR between the two-pump periods indicate that the fluids above and below are significantly different. Furthermore, the high-density fluid (lower GOR) is above the low-density fluid in the oil column. As a result, it was concluded that the fluids are not in perfect flow communication due to the presence of the barrier. These results were also confirmed by a pressure gradient change observed above across the barrier and by pressure interference analysis results. The OCM gives ~ 2.5% contamination at the end of both pumping sequences, confirmed by the lab analysis. Therefore, one can be confident that the differences in fluid properties observed are not simply due to different levels of contamination. The CFA results for the same station are shown in Fig.10. Note that the change in GOR and in coloration between the two pumping depths seen on the LFA is also clear on the CFA, providing greater confidence in the results. It is important to reconcile the finding of DFA with all available measurements. Before reflecting on the results of Examples II and III, it should be mentioned that both a main hole and sidetrack were drilled and logged. Rock and fluid samples were collected in each well bore, and a whole core was acquired in the sidetrack, so a wealth of data is available for analysis. The core data showed that the two boreholes contain a number of carbonate cemented sands that probably represent laterally extensive baffles, but not full barriers. The pressure profiles in both the main and sidetrack well did not show indications of vertical discontinuity. The residual salts in cores acquired from the main well bore contain an abrupt 87Sr/86Sr break that was interpreted to represent a fluid barrier associated with the same carbonate straddled in this station. Based on a relatively minor isotopic shift, the presence of a second baffle at the location of the station in Example II was also suspected, but no corresponding feature could be discerned on the logs or from the pressure interference/DFA station. It is now thought that this isotopic excursion may have represented a paleo-oil-water contact, which formed due to an interruption of reservoir filling. There is currently agreement that it does not represent a true vertical barrier. A number of high yield mud gas samples were also obtained from both wellbores. The C1-C3 carbon isotopic composition are similar above and below the carbonate bed tested in this station, which suggests that the bed is not so extensive as to prevent fluid mixing but only to retard it. A full field dynamic simulation model was constructed to replicate the geometry of the tests performed in this example and honor the other available data. The calcite cemented zones and correlated shale barriers were explicitly modeled as zero transmissibility barriers to vertical flow while the dispersed

SPE 94709 7

shale were represented implicitly by an average kv/kh ratio for the sand packages. Several simulation sensitivities were run with varying extents of the calcite barrier in order to assess the minimum length that the barrier would need to be to completely suppress a pressure response at the vertical monitoring probe to the applied perturbations at the sink probe. The results indicated that the calcite barrier has to be approximately 200 ft in extent in order to suppress the pressure response reaching the observation probe in the time scale of the test. This length is consistent with well-to-well correlations and core observations of the calcite length and at the same time still allows for limited fluid communication as observed in the CFA stations. From a lateral continuity point of view, bulk fluid properties and composition data suggested that at least two of the wells in this five-well appraisal were not in complete communication. Complete lateral compartmentalization is not likely, but the carbonate-cemented baffles interrupt the lateral flow continuity. It is interesting to note here that pressures in the five wells did not show any evidence for reservoir continuity within the main sand; but as mentioned before, lack of pressure contrast alone does not provide sufficient evidence for continuity. This example demonstrates that one need not and should not work in a vacuum. Before building fluid models based on DFA measurements, it is essential to crosscheck its conclusions with available measurements/inferences from other sources in order to be able to make an accurate and reliable assessment of compartmentalization. Example-IV In this example well, the logs show several hydrocarbon bearing zones in the range X630 ft to X730 ft (Fig.10). The density-neutron crossover implies a very light hydrocarbon in the top lower reservoir @X640 ft and indeed the DFA data of the fluid at this depth shows a very light hydrocarbon, essentially a gas while in the bottom lower reservoir @X700 ft, a heavy oil is observed (Fig.12). A third hydrocarbon was also observed halfway at X670 feet further complicating the picture. At the temperatures and pressures encountered in this well, it is unlikely that a gas cap could coexist with oil, and such a steep compositional gradient is highly unlikely over a 60-ft column. Thus, vertical compartmentalization was suspected, but the question that remained was how extensive were those compartments?

A remarkable observation was made while sampling at X645 ft. The pressure gradient for this sand body was established prior to sampling as shown in Fig.11. The sampling sequence involved pumping out approximately 30 gallons, after which the pressure was again measured. A 50-psi pressure drop was then observed. To verify this large pressure drop and confirm that the pressure gauge did not somehow simply drift during the sampling process, two pressure measurements were then repeated in the interval X620-X630 ft. The pressures and pressure gradient obtained were virtually identical to those before sampling at X645 ft. Thus, gauge drift and similar

sources of error could be excluded. Another pressure point was acquired in the same sand that has been sampled at X634 ft and indeed, the ~50 psi pressure drop replicated here too yet the pressure gradient was the same as that obtained before sampling. The most likely explanation for these observations was that this was a case of pressure depletion. One can estimate the volume of the compartment knowing the volume produced (30 gallons) and the pressure drop (50 psi). Using a rough estimate of the compressibility of this gaseous hydrocarbon, one obtains a compartment size of 6000 bbl or 34 MCF. This is a very small compartment indeed, several-orders of magnitude smaller than the average field size in the GOM. Note that this compartment is not in flow communication with the formation several feet higher in the column as these zones are not even in pressure communication (Fig.12) nor did the depletion pressure drop even register on the pressures re-measured in the formation above. The existence of a 6000 bbl-compartment implies that there could be lots of other tiny compartments in this formation, most of which probably remain unrecognized.

The ability to fully exploit the power of WFT shows the necessity of having experienced specialists monitoring and directing these jobs in real time. Decisions are made in real time based on encountered complexities. The response to a puzzle is to acquire more data necessary for the resolution of the complexity; additional DFA stations, sampling stations and/or pressure stations become necessary. These obviously cannot be performed after the job is completed. This WFT process breaks with previous tradition. In the past, it sufficed to send an excellent engineer to the rig for acquisition of excellent log data. All analysis could be done after the job was completed.

Conclusions Downhole fluid analysis allows the efficient use of sampling tools and enables robust in-situ characterization of hydrocarbon fluids and accurate assessment of compositional grading and compartmentalization. The robust determination of the existence of a few thousand barrel-compartment in a seemingly innocuous formation reinforces that unsubstantiated optimism regarding compartmentalization is fraught with risk. The issue of compartmentalization must be addressed early during the appraisal and development phase. The examples shown in this paper demonstrate the power of integrating DFA data with all other available and relevant geochemical, geological and engineering information such as fault distributions, fault throws, fault shale gouge ratios, lateral changes in reservoir lithology, WFT and DST pressure data, pressure decline curves, oil-water contact depths, etc. In addition, this paper demonstrates that the process of reservoir characterization requires the intimate involvement of experts in the planning and interpretation phases of the surveys, and particularly in the real-time monitoring and execution of critical operations.

8 SPE 94709

References 1- Cossey, S. P.J., “Deep GOM Discoveries Toasted Over

30 Years,” AAPG Explorer, Sept 2004. 2- Dong, C., Hegeman, P., Elshahawi, H., Mullins, O.,

Fujisawa, G, and Kurkjian, A.: “Advances in Downhole Contamination monitoring and GOR measurement of Formation Fluid samples” Paper FF presented at the SPWLA 44th Annual Logging Symposium, Galveston, TX, June 22-25, 2003.

3- Dong, C., Mullins, O.C., Hegeman, P.S., Teague, R., Kurkjian, A., and Elshahawi, H.: “In-situ contamination monitoring and GOR measurement of formation samples,” Paper 77899 presented at the 2002 SPE Asia Pacific Oil & Gas Conference and Exhibition, Melbourne, Australia, Oct 8-10, 2002.

4- Elshahawi, H., Hashem, M., Mullins, O.C., Dong, C., Hegeman, P., Fujisawa, G., Betancourt, S.: “Insitu Characterization of Formation Fluid Samples - Case Studies,” Paper SPE 90932 presented at the SPE Annual Technical Conference and Exhibition, Houston, TX, Sept 26-29th, 2004.

5- Fujisawa, G., Betancourt, S., Mullins, O.C., Torgersen, T., O'Keefe, M., Dong, C., Eriksen, K.O.: “Large hydrocarbon compositional gradient revealed by in-situ optical spectroscopy,” paper 89704 presented at the 2004 SPE Annual Technical Conference and Exhibition, Houston, Texas, September 26–29, 2004.

6- Fujisawa, G., Mullins, O.C., Dong, C., Carnegie, A., Betancourt, S., Terabayashi, T., Yoshida, S., Jaramillo, A.R., Haggag, M.: “Analyzing reservoir fluid composition in situ in real time: case study in a carbonate reservoir,” Paper SPE 84092 presented at the 2003 SPE Annual Technical Conference and Exhibition, Denver, Colorado, 2003.

7- Fujisawa, G., Van Agthoven, M.A., Rabbito, P., Mullins, O.C.: “Near-infrared compositional analysis of gas and condensate reservoir fluids at elevated pressures and temperatures,” Applied Spectroscopy, Vol. 56, p. 1615, 2002.

8- Hashem, M.N., Thomas, E.C., McNeil, R.I., Mullins, O.C., April 1999, Determination of producible hydrocarbon type and oil quality in wells drilled with synthetic oil-based muds: Paper SPE 39093 published in SPE Reservoir Evaluation & Engineering, #55959

9- Hoier, L. Whitson, C.H.: “Compositional grading–theory and practice,” Paper SPE 63085 presented at the SPE Annual Technical Conference and Exhibition held in Dallas, Texas, October 1–4, 2000.

10- Malinowski E.R., 2002, Factor Analysis in Chemistry: John Wiley & Sons, New York, (2002)

11- Mullins, O.C., Hines, D.R., Niwa, M., Safinya, K., 1993, “Apparatus and method for detecting the presence of gas in a borehole flow stream,” US Patent #5,167,149.

12- Mullins, O.C., Schroer, J.: “ Real-time determination of

filtrate contamination during openhole wireline sampling by optical spectroscopy:,” Paper SPE 63071 presented at the SPE Annual Technical Conference and Exhibition, Dallas, Texas, October 1–4, 2000.

13- Mullins, O.C. “Optical interrogation of aromatic moieties in crude oils and asphaltenes,” Ch2. in “Structures and Dynamics of Asphaltenes,” O.C. Mullins, E.Y. Sheu, Editors, 1998

14- Mullins, O.C., Beck, G., Cribbs, M.Y., Terabayshi, T., Kegasawa, K.: “Downhole determination of GOR on single-phase fluids by optical spectroscopy,” SPWLA paper M presented at the 42nd Annual Logging Symposium, Houston, Texas, June 17-20, 2001.

15- Mullins, O.C., Daigle, T., Crowell, C., Groenzin, H, and Joshi, N.B.: “Gas-oil ratio of live crude oils determined by near-infrared spectroscopy”, Applied Spectroscopy, 2001b, Vol. 55, p. 197.

16- Ratulowski J., Fuex, A.N., Westrich, J.T., Sieler, J.J.: “Theoretical and experimental investigation of isothermal compositional grading,” SPE Reservoir Engineering and Evaluation, p. 168, June 2003.

SPE 94709 9

PVT CFA Lab C1 wt% 20% 17% C2-C5 wt% 11% 14% C6+ wt% 69% 69% GOR scf/bbl 2200-2400 2300

Table-1. Comparison of lab and CFA results

Fig.1 Fault property distributions

Fig.2 The NIR spectrum of methane, a dead crude oil and a live crude oil.

Fig.3 Comparison between shop tests of the CFA vs. known fluid compositions.

(b)

(c)

(d)

0

0.2

0.4

0.6

0.8

1600 1700 1800 1900

MethaneDead CrudeMethane +Dead Crude

OD

Wavelength (nm)

-CH3

CH4

-CH2-

0

0.2

0.4

0.6

0.8

1600 1700 1800 1900

MethaneDead CrudeMethane +

0

0.2

0.4

0.6

0.8

1600 1700 1800 1900

MethaneDead CrudeMethane +Dead Crude

OD

Wavelength (nm)

-CH3

CH4

-CH2-

10 SPE 94709

Fig.4 WFT Tool Configuration with two Downhole Fluid Analysis (Insitu characterization of formation fluid samples) modules.

Fig.5 LFA log corresponding to Example #1

Fig.6 CFA log corresponding to Example #1

SPE 94709 11

Fig.7 LFA log corresponding to Example #2

Fig.8 CFA log corresponding to Example #2

Fig.9 LFA log corresponding to Example #3

Fig.10 CFA log corresponding to Example #3

12 SPE 94709

Fig.11 Wireline logs corresponding to Example #4

Fig.12 Pressure profile corresponding to Example #4

x

x

x

x

Y Y Y Y Y Y

Evidence of depletion between upper and lower reservoirs

Bottom Lower Reservoir

Middle Lower Reservoir

Top Lower Reservoir

X

X650

X700

DepthFeet(TVD)

X600

X

X650

X700

DepthFeet(TVD)

X600