Abstract

A pilot water flood was carried out in the Marrat reservoir in the Magwa Field. The main aim of this pilot was to allow an assessment of the ability to sustain injection, better understand reservoir characteristics. A sector model was built to help with this task.

An evaluation of the injectivity in Magwa Marrat reservoir was performed with particular attention to studying how injectivity varied as injected water quality was changed. This was done using modified Hall Plots, injection logs, flow logs and time lapse temperature logs.

Data acquisition during the course of the pilot was used to better understand reservoir heterogeneity. This included the acquisition of pressure transient and interference data, multiple production and injection logs, temperature logging, monitoring production water chemistry, the use of tracers and a re-evaluation of the log and core data to better understand to role of fractures.

A geological model using detailed reservoir characterization and a 3D discrete fracture network model was constructed. Fracture corridors were derived from fractured lineaments interpreted from different seismic attribute maps:

A sector model of the pilot flood area was then derived and used to integrate the results of the surveillance data. The main output is to develop an understanding of the natural fracture system occurring in the different units of the Marrat reservoir and to characterize their organization and distribution. The lessons learned from this sector modeling work will then be integrated in the Marrat full field study.

The work described here shows how pilot water flood results can be used to reduce risk related to both injectivity and to reservoir heterogeneity in the secondary development of a major reservoir.

Introduction

The Marrat reservoir in the Magwa Field, part of the Greater Burgan field in Kuwait, was discovered in the mid 1980's. To date it has been produced by primary depletion with limited aquifer support.

Kuwait Oil Company (KOC) is considering water-flooding the reservoir to increase recovery. There have been concerns about the potentially harmful effects of asphaltene deposition in the reservoir. These can be avoided by keeping pressure above the asphaltene onset pressure. In 2007 a pilot water-flood was initiated to demonstrate the feasibility of water-flooding the reservoir at high pressure and to help better understand injectivity and reservoir heterogeneity. This pilot was concluded in early 2012.

The pilot has provided valuable information about the operation of a high pressure water injection system, injectivity into the reservoir and reservoir characterization. In the first two to three years of pilot operation the emphasis was predominantly on operational issues. This paper describes how extending the pilot allowed valuable additional information to be acquired and overcame some limitations in the information acquired during the early stages of the project.

A short history of the water flood pilot is given. This is followed by a description of an analysis of the early performance of the pilot (2007 – 2010) and the analyses that were carried out in this period. These analyses gave an unduly optimistic impression of water flood performance. A more detailed analysis of injectivity is then discussed. This lead to decisions to give greater emphasis to understanding injectivity based on multiple injection logs and to perform a tracer study. These are described. Additionally a fracture study and simulation study of the pilot area were carried; these are also described briefly here.

History of the Pilot

The water flood pilot was intended to address operational issues related to providing high pressure (WHP pressure up to 10000 psi) finely filtered aquifer water, to establish the ability to sustain injection and to a better understand reservoir heterogeneity.

Figure1 shows an outline of the map of the WFP area and Figure2 gives two cross -sections through the pilot area. Two of the wells M-PROD-A and the injector, M-INJ-A, were newly drilled to create an approximate 5-spot pattern. Both of these wells were cored extensively over the productive Middle Marrat section. The production and injection histories of the wells during the pilot are shown in Figure3. A log of the injector M-INJ-A over the zone of interest and the perforated intervals is shown in Figure4.

Fig.1

Location Map of Pilot Water Flood

Fig.1

Location Map of Pilot Water Flood

Fig.2

Correlation profiles along N-S (a) and E-W (b) directions

Fig.2

Correlation profiles along N-S (a) and E-W (b) directions

Fig. 3a

Calendar Day Production: M-PROD-A

Fig. 3a

Calendar Day Production: M-PROD-A

Fig. 3b

Calendar Day Production: M-PROD-B

Fig. 3b

Calendar Day Production: M-PROD-B

Fig. 3c

Calendar Day Production: M-PROD-C

Fig. 3c

Calendar Day Production: M-PROD-C

Fig. 3d

Calendar Day Production: M-PROD-D

Fig. 3d

Calendar Day Production: M-PROD-D

Fig. 3e

Daily Water Injection: M-INJ-A

Fig. 3e

Daily Water Injection: M-INJ-A

Figure 4

Log of M-INJ-A

Figure 4

Log of M-INJ-A

Within the first few months of injection an interference test was conducted. Apart from this the emphasis on surveillance activity in the period 2007 – 2010 was on monitoring injection performance, in particular water quality and injection rates and pressures, and on relatively standard monitoring of production in the pilot producer wells Following a review of production performance in 2010 it was recognized that (i) due to mechanical problem, much of the injection may have been "out of zone" and (ii) that it was not possible to determine whether injected water had broken through at the producers based on the available water analyses.

There was a risk that the pilot would fail to provide useful information on injectivity and that the information gained on heterogeneity would be limited. The pilot was extended and greater emphasis was placed on reservoir surveillance. Injection logging confirmed that a significant proportion of injection was out of zone. Performing a work over of the injector was considered. This was rejected because of concerns that it may not succeed and about the delays involved in waiting for a work-over. Instead emphasis was placed on trying to better understand injection performance in different reservoir zones using injection logs and on trying to better characterize reservoir heterogeneity. This work was made more difficult by a decision to switch to injecting relatively poorer quality filtered effluent water based on the initial indications of high injectivity.

Early Performance (2007 to mid-2010) and analysis carried out at that time

As noted above emphasis at this time was placed on the safe and efficient operation of surface injection equipment and on the delivery of high quality water to the well-head. The pilot was successful in this respect..

Some emphasis was placed on monitoring injectivity. This was done using "conventional" Hall Plots [Hall, 1963]. Figure5 show plots of injection rate, and well head pressure, together with a Hall plot, for the period up to mid-2010. The rate data show some variation in rate early in the pilot, a period of injection at approximately 9,000 bwpd as the injector was "pulsed" during an interference test and, later, an increase in injection rate up to 10,000 bwpd. The decision to increase rate was driven by a desire to provide better pressure support and to get more information on water movement. The most prominent feature of the Hall plot is a change of slope when the rate changed. There are also some minor "discontinuities" related to periods when injection was shut in. Whilst it was recognized that using Hall plots for reservoirs where shut-in WHP was relatively large may lead to some artifacts these results were taken as indicating broadly satisfactory injection performance, in particular in indicating that there was not a problem with injectivity decreasing over time. This was attributed to the very good performance of the injection plant in terms of delivery of high quality water to the well head. As a consequence it was decided to relax water injection specifications to evaluate the suitability of filtered effluent water for injection into the Marrat. However considerable drop in the injectivity was observed with filtered effluent water

Fig.5

WHP, Injection Rate and Hall Plot

Fig.5

WHP, Injection Rate and Hall Plot

The interference test carried out at this time provided useful information on reservoir properties and connectivity. An analysis of the test [Gazi, 2012] clearly demonstrated good communication between the wells with the exception of communication between the injector, M-INJ-A and producer M-PROD-A. Figure6 shows the pressure response of the producers and Figure7 summarizes the results of the interference test.

Fig6

pressure response

Fig6

pressure response

Fig.7

Interference Test Results

Fig.7

Interference Test Results

In addition to the interference test build up (PBU) and fall off (PFO) data were analyzed for the pilot wells. An analysis of the PBU data for M-PROD-A showed that this was not inconsistent with there being a barrier to lateral from between M-PROD-A and M-INJ-A. The PFO data for M-INJ-A was also analyzed. The match used a radial composite model; this model would suggest a zone of reduced transmissibility close to the well, possible associated with a water swept zone. The later time data would suggest reduced tramsmissibility further away from the well. A simple fault model could not match this but these data were considered consistent with the interference test result. The PFO analysis is shown in Figure8. It was noted that the kh product interpreted from the PFO was significantly higher than would have been anticipated based on core data. A disparity between core based and pressure transient analysis based estimate of kh was also seen on other wells in the reservoir. This raised some concerns that fractures may be playing a role in flow.

Fig.8

Early effort to match M-PROD-A

Fig.8

Early effort to match M-PROD-A

As can be seen from Figuresome wells, in particular M-PROD-B produced significant quantities of water during this period. Given that there had some water production prior to the start of the pilot, and also because for some of the wells (including M-PROD-B) there were concerns about the integrity of the cement, the source of this water was not clear. Soon after the start of the pilot a program of water sampling and analysis had been initiated. This gave data on total dissolved solids (TDS) and also some relatively standard chemical analysis. The TDS data for injected water and the produced water from M-PROD-B and M-PROD-C are shown in Figure9. This could be interpreted as indicating some injected water breakthrough but due the limited accuracy, and significant scatter, of the data it was hard to draw any firm conclusions. Also no data were available until there had been over two months of injection.

Fig.9

Injected and Produced water TDS

Fig.9

Injected and Produced water TDS

As of mid-2010 there were a series of concerns about the performance of the pilot. These included concerns that there may have been relatively early water breakthrough, that the injectivity of M-INJ-A was higher that had been anticipated. There were concerns that if fracture flow was important in M-INJ-A that the pilot may not have fully achieved one of its principal goals; evaluating the ability to maintain injectivity. There was also a desire to better understand reservoir heterogeneity. The pilot, which was initially planned to terminate in 2010, was extended, a tracer injection program was initiated and the scope of an ongoing integrated reservoir study was modified to include a review of fracturing and a simulation study of the pilot.

A more detailed evaluation of injectivity

Injectivity data have been analyzed by using modified Hall Plots [Buell et al, 1990]. This approach plots cumulative injection against cumulative flowing BHP minus reservoir pressure. Flowing BHP was based on WHP, water injection rate and tubing performance (calibrated to injection log data). The average reservoir pressure was based on estimates from pressure transient analysis for M-INJ-A (including data from injection logging), shut-in WHP data and shut-in BHP data and Pressure transient data from wells M-PROD-B, M-PROD-D and M-PROD-C which were in good pressure communication with M-INJ-A. The reservoir pressure was assumed to vary in a piecewise linear fashion. Figure10 is a modified Hall plot covering the early period of injection. This shows a very clear break in the slope of the modified Hall plot (at about 120,000 bbls injection). This is associated with a short period of relatively high injection pressure. After this period injectivity is clearly significantly improved. It should be noted that there is no evidence to indicate the flowing BHP would have exceeded initial reservoir pressure. This behaviour could indicate loss of integrity of the primary cement – and possibly connection to open fractures being established. This possibility was investigated, as described below.

Fig.10

WHP. Injection Rate and Modified Hall Plot

Fig.10

WHP. Injection Rate and Modified Hall Plot

It was still considered useful to use modified Hall plots to study injectivity for the entire duration of the injection pilot. Results are shown in Figure11. Three features of the plot are of note. Firstly the abrupt change in injectivity seen in Figure10 is evident. Secondly there is a prolonged period with very close to constant slope. There is no noticeable change in the slope when the injection rate was increased. There are some deviations from straight line behavior between the early abrupt change in slope and the start of treated effluent water injection. The data would support the view that for a prolonged time injectivity remained approximately constant when finely filtered aquifer water was being injected. Thirdly there is an indication of a decrease in injectivity, steepening of the slope of the modified Hall plot, after the start of injecting treated effluent water. There is a difference in density between the aquifer and effluent water; this was accounted for in the estimation of BHP. The increase in WHP at the onset of treated effluent water injection does not, in itself, indicate a change in injectivity.

Fig.11

WHP. Injection Rate and Modified Hall Plot

Fig.11

WHP. Injection Rate and Modified Hall Plot

These results could be taken as indicating that injectivity could be sustained if finely filtered aquifer water was injected. This was one of the main objectives of the pilot. There were, however, concerns that the possibility that much of the water was being injected "out of zone". To address these concerns injection logging was used firstly to investigate the scope for out of zone flow and secondly to better understand injectivity by zone. Figure12 shows the results of injection and time lapse temperature logs. These indicate flow into the formation above the top perforations. This conclusion has been confirmed by running an Oxygen Activation log [see for example Minhas and Eissa, 1997] along with one of the injection logs and also by running a series of temperature logs after the end of injection.

Fig. 12

Temperature profiles from injection logs

Fig. 12

Temperature profiles from injection logs

The results of injection logs have also been analyzed to try to determine injectivity into the different reservoir zones; in particular to try to understand injectivity into the zones that were not influenced by possible out of zone injection. A first attempt at this involved estimating injectivity based on, rates from injection logs, flowing BHP from injection logs and estimates of average reservoir pressure based on shut-in pressure recorded with the injection log. The estimates of reservoir pressure were based on using either pressure transient analyses (where possible) or simple MHD [Miller et al 1950] plots to extrapolate the pressure to a (fairly arbitrary) 100 hour shut-in time. Injectivity was also estimated from a multi rate injection test carried out at the end of the pilot (April 2012). In this case injectivity was estimated from the slope of the plot of injection rate against injection pressur. For this case injectivity estimates are also given, in brackets, where the injectivity is pro-rated to the zonal injection. Results are summarized in Table-1.

Table-1
DateWaterII of top perforations bbl / day / psiII of all other intervals bbl / day /psi
2010 Highly filtered aquifer water 15.2 1.0 
2011 Highly filtered aquifer water 18.4 0.9 
Feb 2012 Filtered effluent water 7.2 2.0 
April 2012 Filtered effluent water 6.1 (6.6) 4.1 (3.6) 
DateWaterII of top perforations bbl / day / psiII of all other intervals bbl / day /psi
2010 Highly filtered aquifer water 15.2 1.0 
2011 Highly filtered aquifer water 18.4 0.9 
Feb 2012 Filtered effluent water 7.2 2.0 
April 2012 Filtered effluent water 6.1 (6.6) 4.1 (3.6) 

The results for the total injectivity index are in line with our observations from the modified Hall plots, and also from pressure transient analyses. There is no indication of a reduction in total injectivity during the period of filtered aquifer water injection. The injectivity index inferred from the 2008 pressure fall off during the interference test was approximately 18.1 bbl / day / psi, close to the values seen in 2010 and 2011. There is a clear indication of reduced total injectivity after filtered effluent water is injected. This is in line with the modified Hall plots and could be explained by assuming near wellbore formation damage / plugging of perforations due to particulate matter in the injected water.

The implied injectivities of the individual zones is harder to rationalize. The reduced injectivity of the upper zone can be explained but there is no clear explanation for the apparent increase in injectivity of the lower zones. From plots of injection rate into the different perforations against injection pressure there was an indication of different reservoir pressures in the vicinity of M-INJ-A in the upper and lower zones. A simple model was used to investigate whether the apparent increase in injectivity of the lower zones could be explained by different reservoir pressures. If steady state flow into both zones is assumed then we would have:

II(total)=II(zone1)+II(zone2)=Q(total)/(PwfPws)ΔP=Q(zone1)II(zone1)Q(zone2)II(zone2)Q(crossflow)=ΔP(1II(zone)1+1II(zone2))

If we assume that the II of the upper zone remains constant at the value implied by the most recent test then a cross flow value of order 1,500 bbl/day would be estimated for the 2010 and 2011 injection logs. This is not consistent with the log data where no cross flow was measured. Based on this it would seem that the injectivity of the lower perforations has, indeed, increased since the start of effluent water injection. The reason is not clear.

The conclusions of the evaluation of injectivity are that, (i) there was no evidence of reduced injectivity during a prolonged period of injecting good quality filtered injection water, (ii) the upper zone had a very large increase in injectivity early in the pilot and injectivity declined significantly with effluent water injection, and, (iii) there is no indication of injectivity decline in the lower zone. Given that the lower zone may be more typical of injection into the reservoir, it is not influenced by the suspected out of zone flow, this result is encouraging.

Tracer Test Results

Given that the analysis of data from water chemistry did not give a clear view of water movement it was decided to perform inter well tracer tests (IWTT). This is described in more detail in [Canti, 2012].The objectives were to determine injected water breakthrough time between the injection well the pilot producers and to understand the pattern of injection and swept volumes.

Preliminary work on design of the IWTT was needed to choose a suitable tracer, determine an appropriate volume to inject and an appropriate sampling frequency. Laboratory screening of candidate tracers involved static and dynamic tests. The static tests examined thermal stability of the tracer, possibility of absorption, degradation and reaction of candidate compounds with formation rock and fluids. These tests were carried out at reservoir conditions and used produced water, formation rock, and reservoir oil. The dynamic tests compared the behavior of candidate tracers with an ideal tracer, Tritiated water, using response curves. They examined the scope for tracer retardation and dispersivity. The selected tracer was Sodium 4- Trifluoromethyl Benzoate, CTHC-4TFM. A dry weight of 250Kg of tracer was chosen based on the reservoir characteristics.

Prior to tracer injection, the chemical tracer was diluted with injection water and the tracer mixture was injected into the flow line of water injection. For this case the tracer injection rate was 4L/min and tracer mixture volume was 1500 Liter. The total dissolved solid of tracer solution was approximately 20,000 ppm and the tracer solution was filtered to eliminate suspended particles bigger than 2um. The injection rate was in range of 4-5 liters/min at a pressure of 4000 psi.

The sampling frequency was planned based on the breakthrough time estimation by simplified calculation on Darcy permeability law. In general, sampling frequency should be high in the beginning of sampling program to observe the early breakthrough and lower in the late stage.

For this test the sampling frequency selected was 5 samples / week/ well to detect the possible rapid appearance of tracer due to suspected reservoir heterogeneities. The collected samples were analyzed by gas chromatography -Mass spectrometry ( GC/MS). Then tracer concentration was then calculated. The GC/MS procedure is suitable to analyze up to 100 sample /day with detection limit of 0.5-1.0 ug/L.

Data interpretation included determination of tracer breakthrough curves for sampling well, break through time of injection water, mean residence time, swept volume and water flood pattern.

Tracer breakthrough was very rapid, 9 days, to producer M-PROD-B (700 meters from the injector) at very high concentration; 260 ug/L in the peak value. Later, at 112 and 133 days the tracer was detected at M-PROD-A and M-PROD-C at lower concentrations; 20.9 and 7.3ug/L respectively. The tracer breakthrough time reflect qualitatively how fast tracer moves from injector to producers, and therefore relates to the average permeability and porosity of the zone between wells. The zone between the injector and M-PROD-B indicates a significant "speed zone".

The detection of tracer at the three producers is evidence of communication between the injector and three of the producers. The amount of tracer recovery shows how much injected water was distributed to each producer. This revealed that injected water mainly flowed to M-PROD-B (10.3%, of the total). The volumes going to the other producers were small, less than 1%. The remaining injected water may be still in the reservoir. This, despite the early water breakthrough in M-PROD-B, gives a positive view of displacement in the reservoir.

The tracer data interpretation showed the total swept volume of each zone are roughly 178,000 bbl, 3,000 bbl and 100 bbl for M-PROD-C, M-PROD-B and M-PROD-A respectively.

Tracer response curve also contains the information of the layer between well pair. The interpretation results revealed the number of layer, swept volume, mean velocity of each layer, which are useful. Figure13 shows the layer properties between the injector and three of the pilot producers as determined by.

Fig.13

Tracer Test Results

Fig.13

Tracer Test Results

Fig.14

Later efforts to match M-INJ-A PFO

Fig.14

Later efforts to match M-INJ-A PFO

It is notable that there was tracer flow between M-INJ-A and M-PROD-A despite there being no measurable pressure response at M-PROD-A during the interference test.

In view of these results the PFO test on M-INJ-A was re-interpreted assuming a layered model with one very thin high permeability layer. The results are shown in Figure13.

Fracture Studies

As natural fractures can be expected to play a significant role on the flow performance of carbonate reservoirs, assessing their occurrence and dynamic characteristics represents an important task in the study of the Magwa Marrat reservoir. This step helps in the preparation and construction of static and dynamic reservoir models. Therefore, all available fracture-related information available was analyzed. The objectives of these analyses were to identify main types of natural fractures occurring in the reservoir units and assess their impact on fluid flow.

Cores, borehole image logs and seismic data were used for characterizing the natural fracture network present in the reservoir units. Most of the natural fractures observed from cores are of tectonic origin and uncemented which indicates a potential impact on fluid flow. Joints (opening-mode fractures) and some faults (shear fractures) occur in a clustered organization (Figure-15). This suggests a possible relationship with faulting even if diffuse fractures were also present.

Fracture analysis from borehole image logs interpretation confirms the presence of diffuse and cluster-organized fractures (Figure-16). Two sets of natural fractures, a dominant NE-SW and a subordinate E-W, were defined from the borehole images.

Taking into account these results, a careful detection of large-scale fractures - seismic and sub-seismic faults was necessary. Seismic sections and seismic attribute maps were used for validating and extending seismic faults as well as detecting and mapping fracture lineaments in the reservoir (Figure-17).

Although there is no evidence of dynamic effect of fracture from oil production data from the Magwa Marrat reservoir, water production appears to be fracture related since there is not obvious relationship between water breakthrough and distance from the perforated intervals to the WOC. Moreover, the water-cut shape and the water-oil ratio of a few wells suggest a fracture response. However, early water breakthrough is not correlated to the well distance to seismic faults.

Dynamic kh values derived from transient pressure analysis were compared to matrix kh (calculated from permeability matched to core data) for fifteen wells. Test interpretation is controlled by productivity index calculation from kh and skin and comparison to test PI (rate / drawdown). Although a few wells present a kh ratio higher than 10, most of them have a test kh slightly higher than matrix kh, indicating that a significant part of the productivity of the reservoir might be related to matrix characteristics. Even if these results could also suggest some fracture contribution to fluid flows in the Magwa Marrat reservoir, the difference between test and matrix contribution is not high enough to make unambiguous conclusions. Uncertainties are associated to the tests interpretation, but also to the effective productive thickness considered for kh estimate.

Test kh values are plotted as a function of well distance to the nearest fault but no obvious relationship was pointed out. Therefore, fractures that impact fluid flows might not be associated to the largest structural objects.

Flowmeter results were compared to matrix kh integrated over the perforated thickness (for three wells) and to fracture description on core (available for two of them). Production profile along the borehole can be directly overlaid with matrix kh in two wells, M-PROD-B and M-PROD-A. Besides, it appears on M-PROD-A fractures log that no fractures were observed in the most productive interval (11340 – 11350 ft MD, see Figure18). Therefore the highest influx can be attributed to matrix properties and fractures do not contribute significantly to the production in this well. The most productive interval corresponds to a highly permeable shoal facies (even if model permeability is slightly overestimated compared to core data). Note that the absence of fracture contribution to fluid production in M-PROD-A could be due to fractures being plugged by the cement during well completion.

According to the injection profile in M-INJ-A, almost all injected water enters the reservoir in or above a very narrow interval around 11680 ft MD at the very top of the perforation. Fractures have been observed on core in this interval, and matrix kh comparison to flowmeter data suggests that matrix alone is not able to explain the injection profile (see Figure19). Consequently, fractures seem responsible for this influx. However, all fractures do not contribute to fluid flow although higher fracture density has been measured in deeper intervals (but they might have been damaged by the cementation, since mud losses have been observed). A highly permeable shoal facies in the injection interval has been identified by further investigation in the core description, although it was not assigned the highest rock-type quality. Therefore matrix properties may have been underestimated and dominant fracture flow may not be as obvious as it first appeared.

To conclude on static and dynamic fracture characterization, fractures might have a local dynamic effect in the production of the Marrat reservoir, but they are not responsible for total fluid flows. A high matrix permeability facies also has a strong dynamic impact. From the available evidence, the preferred interpretation is that of a limited presence of fractures in clusters around faults. Specific attention was paid to the injection well M-INJ-A since core data, borehole imagery, pressure fall-off analysis and PLT are available.

A fracture model comprising diffuse fractures and fracture clusters was built using FracaFlowTM (Figure20). Diffuse fractures were computed stochastically and fracture clusters were modeled deterministically from the seismic faults and sub-seismic lineaments. Fracture conductivities, which after upscaling are converted into fracture equivalent permeability, were obtained through the calibration of the model using KH obtained from well tests (Figure20).

Pilot area simulation model

A sector model of the Marrat reservoir in Magwa field has been built and fluid flow simulation was performed to analyzing the dynamic behavior of the reservoir in the waterflood pilot area. In addition to the preparation of full field reservoir simulation, a major objective was to develop an understanding of the natural fracture system occurring in the Magwa Marrat reservoir through model calibration against production history.

The sector model focuses on the waterflood pilot area (five-spot pattern). However, a peripheral zone including additional production wells is accounted for to impose relevant boundary conditions to the zone of interest. Pore volume adjustments are applied in this peripheral area to provide enough pressure support to match static pressures in the zone of interest. The geological grid includes 50x50 m cells in the main zone of interest and 25 layers. It contains 110,000 active cells and it is directly used in the simulation model. Facies and porosity are based on well data from the 12 wells lying in the sector model area and they are distributed using geostatistical methods. Permeability is calculated from porosity-permeability laws available for each rock-type. Water saturation is set from J-functions defined for each rock-type and calibrated to log saturation.

Middle Marrat reservoir in the Magwa field is initially undersaturated, initial pressure is 9,630 psia at reference depth (11,200 ft TVDSS) while bubble point pressure is 3,130 psia. Reservoir temperature is 235°F. Fluid properties are described with a black-oil model where oil formation volume factor is 1.46 rb/stb and oil viscosity 0.36 cP at initial pressure. The solution gas oil ratio is 1,006 scf/stb.

Following on from initial dynamic model response, two main issues were addressed during history match. First, productivity index adjustment of most wells is necessary to achieve historical oil production. Second, global and local modifications are necessary to match water productions. Natural fractures appear to be necessary to connect the bottom aquifer to the main productive zone.

Dynamic kh values were derived from pressure transient analysis and compared to static kh, calculated along the well from log permeability calibrated to core data. Most wells have higher dynamic kh that could result either from static kh underestimation or highlight an additional contribution to fluid flow. Besides, a negative skin has been estimated from 10 of 19 available transient well tests. As an example, a fractured well signature was pointed out on a well and a negative skin value could be derived from the fracture length. Such well response could result from well connection to a natural fracture network. This hypothesis is supported by core observation where high fracture density has been measured in front of the perforated intervals. All these observations strongly suggest that a natural fracture network contributes to fluid flow. This would explain the poor well productivity in the reservoir model; this is the most probable scenario since matrix permeability is fairly matched to core data. At this step, well PIs are enhanced in the simulation model by assigning directly kh and skin from well tests to characterize the connection between the well and the formation.

Two types of water production are observed in Marrat wells: one is related to the natural depletion period and corresponds to bottom aquifer influx while the second one is associated to water injection. Although the fracture characterization stage was not fully conclusive regarding their impact on fluid flows, evidence of dynamic effect of fractures are provided by the initial response of the simulation model. Indeed, in addition to the need for higher well productivity, water production is absent from the initial simulation model.

A conceptual fracture model is integrated in the sector model to connect the aquifer to the main oil production zone since the poor reservoir properties of the matrix medium in the bottom part of the reservoir do not allow significant communication between the aquifer and the oil zone, neither for aquifer pressure support nor for water influx. Diffuse fracturing in the tight, low porosity "lagoon" facies is accounted for through uniform vertical permeability enhancement. Vertical permeability set to 10 mD is high enough to achieve water production in half wells included in the sector model (basically wells with water-cut lower than 10 %).

However, higher water production cannot be explained by considering only the homogeneous permeability enhancement provided by diffuse fractures. Moreover, water production is not related to the distance between the perforated intervals and the water-oil contact, and some wells present sharp increases of the water-oil ratio. Both model response and water production analysis suggest that water production does not only result from a uniform water influx from the aquifer but also from local preferential flow paths of higher permeability (M-PROD-B, M-PROD-A and three other wells). Since no relationship between water-cut and well proximity to seismic faults was found out, local permeability enhancements are associated to fracture corridors. Fracture corridors cannot be interpreted at seismic scale since they are characterized by very low (or even zero) vertical displacement, although they are supposed to cross-cut the whole reservoir. Nevertheless, they may be detected from seismic attribute maps. Therefore their distribution in the sector model is constrained with both dynamic observation and seismic attribute maps.

An equivalent single medium is considered for flow simulation in the sector model: matrix and fracture petrophysical properties are summed-up in each grid block and matrix-fracture exchanges are supposed to be negligible. This assumption is valid when matrix and fracture media are in equilibrium at each time-step of the simulation. Pseudo-curves are assigned to characterize oil-water relative permeabilities in the fracture corridors cells in order to enhance water mobility at lower water saturation. This method is the usual way to account for dual media behavior in a single medium model. Nevertheless, sensitivities on the relative permeability curves shape in the fracture corridors did not prove to affect greatly the water production in the neighboring wells.

In addition to water influx from the aquifer, injected water is produced in the vicinity of the injection well. All available dynamic data resulting from the waterflood pilot wells highlight fluid flow anisotropy in the area with enhanced communication in the East-West direction and limited connection in the North-South direction. Water productions, tracer data, interference tests results and flowmeters interpretation have been integrated.

Water breakthroughs from injector to the neighbor production wells (westward, northward and eastward) have been attributed to a thin high matrix permeability layer (shoal facies located in the upper part of the production zone at the top perforation) clearly identified on M-PROD-B, M-PROD-C and M-PROD-A logs and where core measurements reach up to 500 mD (see Figure20a). This facies may be extended to the injection well M-INJ-A accounting for uncertainties on log interpretation (see Figure20b). Dynamic evidences such as flowmeter measurements in M-INJ-A and simulation model response support this hypothesis. Indeed, grid-block pressure around the well is unrealistically high and proves the model permeability being far too low.

Extension of the distribution of this high reservoir quality facies in the simulation model allowed matching water arrivals in the neighbor wells within a few months after water injection started and decreasing the injection pressure to reasonable ranges. However, tracer tests pointed out faster tracer displacement (a few days to a few months). Although very efficient and direct flow paths must exist to explain this observation, the very low involved swept volumes suggest that these flow paths are not associated to significant fluid amounts. Indeed effective water breakthrough times prove to be delayed compared to the tracer response. Besides, PLT measurements performed one year after injection started prove the main water production zone to be located in the bottom perforated interval (see Figure21). It is not in agreement with a very fast tracer displacement from injector to producer, since vertical diffusion from top to bottom layers in significant amount would be much longer than tracer travel time.

Therefore, although tracer test results point out the existence of high permeability paths in the waterflood pilot area, especially in the East-West direction, most production water before 2010 would rather come from aquifer influx. A matrix origin has been preferred to explain the strong East-West communication in the waterflood area in the simulation model since no structural discontinuity has been seen on attribute maps and a high permeability facies has been identified from the sedimentological analysis. However, fracture hypothesis is not discarded, although diffuse fractures are unusual in high porosity and high permeability facies.

Conclusions

A water flood injection pilot in the Magwa Marrat reservoir has recently been concluded. The basic aims of demonstrating the ability to inject very high quality water at high pressure and to demonstrate the ability to sustain injectivity have been achieved. The extension of the project to better understand injectivity and reservoir characteristics has provided much valuable information. In particular there is greater confidence in our ability to sustain injectivity and a greater recognition of the potential role of fracturing in reservoir flow.

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.

Acknowledgments

The authors thank Kuwait Oil Company for their permission to publish this paper.

Buell
,
R.
,
Kazemi
,
H.
, and
Potterman
,
F.H.
1990
, "
Analyzing Injectivity of Polymer Solutions with the Hall Plot,
"
SPE Reservoir Engineering
,
41
46
,
February
(
1990
)
CANTI (Centre for applications of nuclear Technique in Industry)
.
2012
: "
Interwell Tracer Test in Middle Marrat Water Injection
"
Unpublished report submitted to
Kuwait Oil Company
.
Gazi
,
Naz
,
Ali
Farida
,
Al-Naqvi
Meqdad
,
Dashti
Laila
,
Al-Qattan
Abrar
,
Farhan
Al-Farhan
.
2012
: "
Reservoir Connectivity Using Long Term Interference Testing in a Waterflood Pilot in the Carbonate Marrat Formation of the Greater Burgan Field, Kuwait
"
SPE 159126
SPE Annlual Technical Conference and Exhibition
8–10 October
,
San Antonio, TX, USA
.
Hall
,
H.N.
1969
: "
How to Analyze Waterflood Injection Well Performance
,"
World Oil
,
1963
(
October
) pp
128
130
.
Minhas
,
H. N.
, and
Eissa
,
M.A.
1997
, "
Selective Downhole Flow Measurements of Water Phase Using Oxygen Activation Log in Gulf of Suez Wells,
"
SPE 37737
Middle East Oil Show and Conference
15–18 March 1997
,
Bahrain
Miller
C.C.
,
Dyes
A.B.
, and
Hutchinson
C.A.
,
1950
. "
Estimation of permeability and reservoirpressure from bottom-hole pressure build-up characteristics,
"
Trans. AIME
, (
189
):
91
104
,
1950
.