Anda di halaman 1dari 25

PETROLEUM SOCIETY

CANADIAN INSTITUTE OF MINING, METALLURGY & PETROLEUM

PAPER 2003-181

Using Decline Curves to Forecast Waterflooded Reservoirs: Fundamentals and Field Cases
R.O. Baker, T. Anderson, K. Sandhu
Epic Consulting Services Ltd.
This paper is to be presented at the Petroleum Societys Canadian International Petroleum Conference 2003, Calgary, Alberta, Canada, June 10 12, 2003. Discussion of this paper is invited and may be presented at the meeting if filed in writing with the technical program chairman prior to the conclusion of the meeting. This paper and any discussion filed will be considered for publication in Petroleum Society journals. Publication rights are reserved. This is a pre-print and subject to correction.

ABSTRACT Decline analysis is the most used technique for forecasting reserves. Although decline analysis for gas fields has been shown to have a strong theoretical background, decline analysis for multiphase situations is less clear. This paper suggests when it is appropriate to use decline analysis for waterfloods and what is happening physically when decline trends develop. It also shows field situations which follow clear trends and allow us to use decline techniques to diagnose field behaviour. This paper presents field cases for waterfloods and suggests a theoretical analysis that ties in well with the observations from these field examples. Field cases as well as analytical analysis/simulation(1) generally support harmonic or hyperbolic decline for late stage waterflood behavior. In other words, reservoir factors generally lead to hyperbolic or harmonic decline late in the waterflood life. However, that is not to say that

exponential (b=0) or super exponential decline (b<0) never occur. When they do occur, usually non-reservoir factors are involved. This paper also shows how incremental oil recovery can be calculated using decline methods accounting for changes in fluid and injection rates. The waterflood decline correlation period should have the following criteria: a. the watercut should be greater than 50% b. the voidage replacement ratio should be close to one c. well count should be relatively constant d. injection and fluid production rates should be relatively constant e. the reservoir pressure should be relatively constant

f. producing well pressures should be constant g. the GOR should be relatively constant h. the volume of water injected should be greater than 25% of the hydrocarbon pore volume. In order to properly access future waterflood performance, we need to estimate what is controlling the oil decline rate. After substantial water breakthrough has occurred, the oil rate profile is usually controlled by: a. relative permeability b. changing volumetric sweep c. water handling constraints d. fluid rates handling constraints e. permeability/injectivity in the near wellbore regions f. well positions. Most successful waterfloods are hopefully in reasonably good continuity reservoirs with moderate to high permeabilities, therefore well interference is very likely. However, because of well interference in such moderate to high permeability reservoirs, individual well decline analysis is to be used with caution. We would therefore recommend using an aggregate analysis of a group of wells as a more realistic scenario rather than a sum of individual wells. INTRODUCTION Diagnosis of the character of producing wells is the key skill that petroleum engineers in an operating environment require. This characterization and corresponding reservoir forecast/diagnosis task is most often performed using the technique of decline analysis. Because of its ease of use, decline analysis has a huge appeal as a forecast method and is accordingly the most widely used technique. The pioneering work on production decline analysis was performed by Arps(2) in 1944. Equations for pressure decline were formulated empirically based on statistical analysis of production data obtained from non-fractured reservoirs. Production decline was considered to be proportional to pressure decline through assumptions of

constant wellbore pressure and constant productivity index. Many others have made contributions to the techniques of decline analysis(3-6). In particular, Fetkovichs 1973 paper(7) showed that exponential decline is the long term solution of the diffusivity equation with constant wellbore pressure. Using Fetkovichs techniques, exponential, hyperbolic, and harmonic decline curves can be identified based on Arps analysis. Fetkovich used advanced decline analysis (ADA) to describe these new techniques. Fetkovichs work generally dealt with singlephase flow, small constant compressibility systems with radial outer boundaries, or linear systems with hydraulic fractures. Fetkovich type curves are valid for circular bounded reservoirs with a well in the center. As a result, the work is very applicable to high pressure gas systems or under-saturated liquids. What made Fetkovichs type curve approach work very useful is that it allowed one to identify reservoir properties from the analysis; furthermore, this work showed that decline analysis has a solid basis in reservoir engineering fundamentals. Fetkovich curves also allow a consistency check on reservoir parameters versus forecast. Thus, Fetkovichs work allows a coupling of physical parameters to decline analysisat least for single phase systems. Fetkovich developed decline curve analysis that could be applied using type curves, so that production decline in hydraulically fractured reservoirs, stratified reservoirs, and the effect of changing back pressure could be readily analyzed. Thus, there is a very strong tie between decline analysis and physical parameters, at least for single phase, single well gas systems (i.e., where there are no interference effects). The purpose of this paper is to try and extend advanced decline analysis techniques to multiphase waterflooded/water drive and multiwell systems. This topic was touched on by Laustsen, who presented a practical overview of methods and use of decline curve analysis(8). He identified common misinterpretations and rules to avoid these misinterpretations. He stated two important conclusions or rules of thumb for general decline analysis, stressing that:

(1) An understanding of the principles of decline analysis, depletion mechanisms, and rock and fluid characteristics is essential to establish reliable decline interpretations; and (2) Subtle changes to curve fits within the engineering accuracy of the data can result in large differences in estimated reserves. There are a number of sub-objectives for this paper; these are to show: 1. How important an understanding of the waterflood reservoir mechanism is to correctly forecasting waterflood recovery; 2. How to identify, with production data, what are the controlling factors in a waterflood; 3. The applicability of Lijek and Masoners work to real field cases; and, 4. How supplemental techniques such as Recovery Factor (RF) versus Hydrocarbon Pore Volume Injected (HCPVI), log(WOR) versus cumulative oil produced (Qo or Np), and log cumulative fluid produced (Qo+Qw) versus Qo can be used to increase ones confidence level in decline analysis. MISUSE OF DECLINE CURVES Although decline analysis is, as indicated above, overwhelmingly one of most used techniques, it is, in our opinion, unfortunately one of the most misused techniques as well. The main misuse of conventional decline analysis stems from a lack of understanding of its limitations; its ease of use means that one may overlook the premise that it should only be applied in situations of non-changing conditions. The behaviour of the reservoir is therefore tacitly assumed to follow a trend that can be depicted by simple analytical equationseven when in fact the reservoir may be undergoing or subject to changing conditions that make its description much more complex. The parameters of decline analysis equations are nevertheless easily matched with production data, and the calibrated equation is then expected to accurately forecast reserves. Unfortunately, with this empirical technique, we can completely bypass the reservoir model

and the associated petrophysical, geological, and reservoir parameters(9). According to Slider(10): Decline-curve analysis may be one of the most misused and at the same time, one of the most neglected reservoir engineering techniques. Declinecurve analysis can only be used as long as the mechanical condition and reservoir drainage stay constant in a well and the well is produced at capacity. Their limitations lead to much misuse. On the other hand, the more theoretically inclined petroleum engineer may not appreciate decline-curve analysis and fail to use it to augment or backup his theoretical prediction. Decline curves can be characterized by three factors: (1) initial production rate or the rate at some particular time (qoi), (2) curvature of the decline (b, which is the Arps exponent), and (3) rate of decline (Di). For single phase systems, oil rate is simply given by:
k (k ro )h DP Bo m r e ln + S r w

q=

....................................................... (1)

In single phase oil systems, since the terms permeability (K), oil relative permeability (kro), pay thickness (h), skin (S), and outer drainage radius (re) are usually constant, the main variable controlling decline rate is reservoir pressure drawdown (P) and more specifically reservoir pressure. If pressure depletion is the only varying factor, in a liquid expansion drive system, before reaching the bubble point (which is assumed to be a situation of constant compressibility), we would expect to see an exponential type of decline (i.e., b = 0). This is because, in such a situation, the change in oil rate Dq o is proportional to reservoir pressure which is proportional to cumulative oil withdrawals; therefore, qo (N-Np). The oil formation volume factor and the oil viscosity also change with depletion and therefore can control decline rate. In solution gas drive systems, the compressibility of the system changes and, as a result, according to Fetkovich(7), typically solution gas drive systems have hyperbolic type declines with b 0.3.

Generally, oil rate decline type is a function of two main reservoir effects: (1) decreasing reservoir pressure and (2) decreasing oil saturation due to water invasion or increasing gas saturation. Oil rate production signatures are a function of many parameters such as: 1. 2. 3. Transient effects; Pressure depletion effects (resulting from a decrease in average reservoir pressure); Changes in fluid properties such as Bo, Bg, mo, and mg with pressure depletion; 4. 5. 6. 7. 8. 9. Operational effects (change in back pressure) and skin buildup; Relative permeability effects; Changes in drainage area/well interference effects; Interface movement (water-oil/gas-oil contact movements); Flood front movements in injection processes; and, Pro-rationing of oil rates.

Because of the number of parameters, advanced decline analysis (ADA) is not an easy problem to deconvolve into individual parameters. Fetkovichs work extensively covers the first three points due to its focus on single-phase assumptions. The literature generally does not address related injection processes, however; in these processes, relative permeability and flood front movement can become controlling factors. For the purpose of this paper, we will assume reservoir pressure is maintained and that drainage area is constant. Also, we will assume no changes in skin or operational parameters. Most waterfloods are deployed in medium to high permeability (Keff > 1 md)/low compressibility reservoirs, so transient effects are relatively short lived (< 5 days); therefore, neglecting transient behaviour is generally a valid assumption. Most wells are in a pumped off condition, so assuming a constant bottom hole pressure is also generally valid. Waterflood Decline Analysis In a waterflood or water drive system, the parameters controlling decline rate result in a more complicated

analysis because parameters four to eight above generally control oil production rate. This is true even though reservoir pressure is constant. An idealized waterflood oil response profile is shown in Figure 2 and Figure 3. After primary depletion, waterflood begins; with water injection, reservoir pressure generally increases rapidly and gas collapse may occur. As a result, oil rate may rise rapidly because oil relative permeability increases (due to oil banking). The peak oil rate will be a strong function of injection rates, pattern configuration, volumetric sweep considerations, and permeability heterogeneity. During these early waterflood stages, volumetric sweep changes very rapidly. Oil rate production signature is controlled by volumetric sweep efficiency and fill up considerations. Because of changing fluid rates, changing injection rates, as well as rapidly changing volumetric sweep in the early time period (i.e., until watercuts exceed 50%), the application of decline analysis may lead to erroneous results. In other words, decline analysis on waterfloods requires that volumetric sweep efficiency has stabilized, or at least is changing very slowly (in effect, attained pseudo steady state conditions). Obviously, the point when water breaks through to producers is a function of mobility ratio (the ratio of displacing fluidwaterto displaced fluidoil), phase mobility (the ratio of relative permeability to viscosity for that phase), and permeability heterogeneity. Often in waterfloods that are not facility limited, fluid rates and injection rates will change as water breakthrough occurs as shown in Figure 3. After oil production peaks, water breakthrough occurs. Some time after that, volumetric sweep changes much more slowly. In this period, oil rate profile is strongly controlled by mobility effectsmost prominently, by relative permeability. To repeat, after breakthrough, (late stage) the waterflood oil rate signature is mainly controlled by relative permeability effects. In the early time period until watercuts are greater than 50%: because of changing fluid rate, injection rate, and rapidly changing volumetric sweep, using decline analysis is futile. In other words, decline rate analysis on waterfloods requires that the volumetric sweep efficiency has stabilized. Thus, there are a number of criteria for the selection of waterflood decline correlation periods; these are: (1) watercuts

greater than 50%, (2) the voidage replacement ratio (VRR) should be close to one, (3) constant well counts and pattern configuration, (4) relatively constant injection rates and fluid production rates, (5) relatively constant reservoir pressure, (6) constant producing well pressures, (7) GORs should be relatively constant, and (8) the volume of water injected needs to be high, >25% HCPVI. These conditions basically ensure that we have a relatively constant pressure situation in which the decline signature is dominated by relative permeability rather than by volumetric sweep considerations, consistent with a relatively mature waterflood. REVIEW OF WATERFLOOD DECLINE ANALYSIS; THEORETICAL LITERATURE In the literature, there has been some excellent work done on understanding waterflood decline characteristics. This literature implicitly assumes, during late stage waterflood decline, that drainage area and swept area are quasi-constant. Ershaghi and Omoregie(11) derived a straight line relationship between the factor x = ln(WOR) -1/WOR -1 and cumulative production Qo. Startzman and Wu(12), Timmerman(13), and others have shown that a plot of log(WOR) versus Qo can yield a straight line relationship and has the advantage of being simpler to understand than the Ershaghi x-plot technique. Lo et al.(14) showed that the slope of the log (WOR) versus cumulative oil plot is equivalent to Ershaghis xplot technique at high water-oil ratios. A rough rule of thumb for using the above techniques is that watercut must be greater than 50%. These methods have been shown by both analytical and numerical modeling to hold. In addition, there are numerous field examples that confirm this straight line behaviour of log(WOR) versus cumulative oil, and this is especially true for waterfloods in medium grade oil reservoirs. The Ershaghi x-plot technique and the log (WOR) versus cumulative oil technique were shown to have a strong physical connection to 1D relative permeability characteristics by Ershaghi(15), Lijek(16), Lo et al(14), and Baker(17). Lo et al.(14) showed that the slope of log(WOR) versus Qo could be used to determine swept volume if relative permeability characteristics of the reservoir were known(1).

Lijek(16) and Baker(17)showed that oil rate decline was hyperbolic (i.e., 0 < b < 1) or harmonic (i.e., b = 1) depending upon whether fluid rates were constant or changing. All of the above techniques are only applicable when volumetric sweep is quasiconstant; if so, then relative permeability controls decline rates. Lo(14) et al. showed that, for some layered systems or for gravity override systems where the volumetric sweep continues to change, the plot of log(WOR) versus cumulative oil was not a straight line. We have also noted for floods that were neither strongly gravity nor viscous dominated that the log(WOR) versus cumulative oil (Np) plot was not a straight line, and have indicated that the lack of straight line behaviour was due to the fact that quasi-steady-state volumetric sweep was not reached until very late times(17). The above references are based on the observation that a large part of a plot of log(krw/kro) versus Sw is linear. The above techniques rely on a water-oil relative permeability relationship that is given by:
k ro water production = Ae -Sw ....................................... (2) k rw oil production

Timmerman states that generally large portions of the relative permeability curves can be approximated by the above equation for some reservoirs(13). In recent years, Masoner, using a different relative permeability relationship, has shown that hyperbolic decline is likely to occur(19). Masoner shows that waterflood decline is likely to be hyperbolic with an Arps exponent (b) ranging from 0.25 to 0.8. Masoner used a Corey equation type approximation to oil relative permeability to yield a hyperbolic decline as described in Equation (3)
b k ro = a (S oD ) ................................................................... (3)

Both Lijek16 and Masoner19 address the nature of decline characteristics in situations where relative permeability controls waterflood/EOR behaviour and where there is a constant drainage area. Both authors show that the nature of declinei.e., Arps exponent

bwhen oil rate profile is controlled by relative permeability, is strongly a function of injection (throughput) rates and fluid production. These authors results are slightly different, because they assume different relative permeability functions. The methods of both Lijek(16) and Masoner(19) assume: 1. Buckley-Leverett theory applies; 2. Quasi-steady-state volumetric sweep (i.e., unchanging with time), which implies a constant drainage area; 3. Little pressure variation within the swept zone; and, 4. Reservoir pressure does not change significantly with time. The most restricting limitation is that the volumetric sweep remains constant. Nonetheless, practical experience shows that there are a large number of field cases where harmonic or hyperbolic decline occurs and plots of log(WOR) versus cumulative oil yield straight line behaviour. This is especially true for waterfloods in heavy oil and for waterfloods in heterogeneous fields. Obviously, the initial volumetric sweep controls to a very large extent the water-oil ratio versus cumulative oil (Np) profile. However, in unfavorable mobility ratio and heterogeneous reservoir waterflood situations, water breakthrough occurs relatively early and then volumetric sweep may increase very gradually with time. Summarizing the literature, the waterflood performance of fields and wells can be governed, in some cases, by not only relative permeability but as well by heterogeneity and by gravity and viscous forces. Therefore, theory may provide guidelines, but a pragmatic approach should be taken. Review of Empirical Waterflood Literature on Decline Behaviour Empirically, Wong and Ambastha have shown for Canadian waterfloods that decline is on average hyperbolic(20). In our experience, a very high percentage of waterfloods worldwide have indeed been observed to have hyperbolic decline, but we have definitely observed field cases of exponential and harmonic decline as well. Most waterflood simulation studies show hyperbolic/harmonic decline characteristics. Generally,

heavy oil waterfloods are most likely to have harmonic or high Arps exponent values of b > 0.7 indicating hyperbolic decline. In the case of very light oil in a relatively homogeneous reservoir with piston like displacement, volumetric sweep efficiency will dominate the oil rate profile; decline will be very rapid and decline analysis techniques will not be useful. Laustsen also confirms this conclusion(8). Unfortunately, many statistical empirical studies of decline rates and decline types in the literature do not include information about drive mechanism, well counts, reservoir pressure, or watercut and GORs. Arps showed that, for his study areas, 90% of the fields exhibited hyperbolic decline character; he observed no harmonic declines. Bush and Helanders Oklahoma study showed mainly harmonic/hyperbolic decline types(21). Ramsay and Guerrero showed that hyperbolic and harmonic declines were typical(22). Schuldt et al. indicated that Alaskan waterflooded oil reservoirs are generally expected to follow hyperbolic decline behaviour(23). Campbell (1959) states that most decline seems to follow hyperbolic decline most closely, the value b = 0.25 being a good average of many curves examined. It is seldom that b exceeds 0.6(24). Selection of Waterflood Decline Correlation Period Before substantial water breakthrough (watercut 50%) occurs, decline analysis is unlikely to be a good technique because pressurization and volumetric sweep effects are likely to dominate the oil production signature. Oil rates at this stage may be inclining. However, after volumetric sweep efficiency is in a relatively constant state, oil rate will decline because relative permeability is then the main variable. In order to select a correlation period for waterflood, we would propose the following criteria: 1. Watercuts should be greater that 50%; 2. Voidage replacement ratios (VRR) should be close to unity; 3. Injection and production well count should be relatively constant;

4. Fluid production and injection rates should be relatively constant; 5. Reservoir pressure should be relatively constant; 6. Producing well pressures should be constant; 7. gas-oil ratios should be constant; and, 8. The volume of water injected should be greater than 25% of the hydrocarbon pore volume. This paper will focus on waterflood response; therefore, transient effects and pressure depletion effects will be ignored. Obviously, if well count is changing or infill drilling is occurring, the oil rate will be strongly affected. If reservoir pressure is declining and VRR<1.0, then some of the reservoir energy is being supplied by expansion energy rather than the waterflood. Thus, it is unlikely in such a situation that the decline rate or Arps exponent will be constant. Similarly, if gas-oil ratios are increasing, it is probable that pressure is declining and therefore, expansion energy rather than waterflood drive is dominating. The rule of thumb of watercuts being greater that 50% insures that sufficient water breakthrough has occurred and that relative permeability rather than volumetric sweep now controls oil rate decline as shown in Figure 4. A simulation model was constructed to test the validity of decline methods and timing of applicability for decline methods for field cases. We also examined how relative permeability, pattern imbalance, changing fluid rates, and viscosity ratio changed decline rate behaviour. Field Cases Over 20 reservoirs have been studied in detail using the correlation period selection criteria. Table 3 shows the selected pools and corresponding decline analyses including the Arps exponent b. To supplement the analysis, log (WOR) versus Np (cumulative oil produced), log cumulative fluid (Qo+Qw) versus Np, and RF versus HCPVI have been used to assist in determining the decline type. A complete analysis is demonstrated on two pools, the Provost Lloydminster O pool and Taber South Mannville A pool.

Provost Lloydminster O Pool The Provost Lloydminster O pool has been on production since September 1973 and has produced a total of 5.0 106m3 of oil. This pool is considered a heavy to medium oil with 23.8API gravity, and a density of 911.0 kg/m3. With an areal extent of approximately 1315 ha, and a pay thickness of 3.1 m, the original oil in place (OOIP) for the Lloydminster O pool is 10.1 106m3 (waterflood area plus primary). Primary production occurred until January 1977, when water injection began, with minimal volumes of water being injected. Injection rates were increased dramatically in 1995, where the field averaged 15,000 m3/d of injected water. The recovery factor to date for the Lloydminster O pool is approximately 50% OOIP. The composite plot for Lloydminster O pool is shown in Figure 5. All data for Provost Lloydminster O pool was obtained from the public data source (AEUB)(25). Using the composite plot, it can be seen that the most appropriate correlation period for decline analysis is late 1998 to December 2002 (end of data set). In this time period, the injection rate, injection well count, production well count, and GOR are all constant. The watercut is above 50% and the VRR (Figure 6) has stabilized and is close to unity. Pre-1995, the VRR was quite erratic. Using an earlier correlation period would only represent the effects of decreasing well count and changing total fluid rates on the reservoir. Interpretation of the log (WOR) versus Np (Figure 7) and log(Qo+Qw) (cumulative fluid) versus Np (Figure 8) plots assists analysis as supplemental techniques for determination of decline behavior. The log (WOR) and log (cumulative fluid) plots for Lloydminster O pool are linear in mid to late time, indicating a hyperbolic - or harmonic decline. The recovery factor (RF) versus hydrocarbon pore volume injected (HCPVI) plot for the Lloydminster O pool is shown in Figure 9. The performance of this pool due to water injection is quite good. At 2 pore volumes (PV) of injection, secondary recovery due to waterflood is almost 40%. Break over point occurs at 15% recovery factor and 25% HCPVI, indicating good waterflood performance. As subsequent pore volumes of water are injected, the curve has a slowly decreasing slope.

Setting Arps exponent equal to 0 and 1 to observe the exponential (Figure 10) and harmonic declines (Figure 11), respectively, the resultant decline curves do not fit the data very well. Note that the log (WOR) versus cumulative oil (Np) plot, log cumulative fluid (Qo+Qw) versus Np, and RF versus HCPVI plots all show clear trends that can be easily extrapolated. Also note that, despite the large amount of water pore volumes injected, there is still good waterflood response. Using a harmonic or exponential decline on this pool (b value too high or too low) could result in an over- or under-estimation of recoverable reserves, respectively. Using commercial software to determine the Arps exponent for this pool for the correlation period of 1998 to December of 2002, it was found that a b value equal to 0.36 gave the best fit to the data. In comparison to the exponential and harmonic decline curves, the hyperbolic curve with an Arps exponent b=0.36 matches the data quite well and is representative of the reservoirs decline behavior. TABER SOUTH MANNVILLE A The Taber South Mannville A pool has been on production since December 1963 and has produced a total of 1,800 103m3 of oil. This pool is considered a heavy oil with 18.1API gravity, and a density of 945.9 kg/m3. With an areal extent of approximately 1394 ha, and a pay thickness of almost 8 m, the original oil in place for the Taber South Mannville A pool is 14.1 106m3 (waterflood area plus primary). Primary production was minimal, reaching less than 1% RF. This pool has been under waterflood since January 1966, and has a secondary recovery factor of 13% with 85% HCPVI. Field water injection rate was relatively constant at 4,000 m3/d until 1994, when it was increased to approximately 13,000 m3/d. The composite plot for Taber South Mannville A is shown in Figure 13. All data used in analysis of the Taber South Mannville A pool has been obtained from the public data source (AEUB)(25). From the composite plot, there are two possible correlation periods that would be acceptable for evaluating decline. These time periods are 1989 to 1992 and 1997 to December 2002. Both periods have constant

fluid rates, GOR, water injection rates, and well counts (both injectors and producers). Watercut is above 50% and VRR is relatively stable for both time periods (Figure 14). As the later correlation period has a slightly more stabilized VRR (has been stable for a longer time period) and is more representative of what is currently occurring in the pool, it is used for analysis. The log (WOR) versus Np plot is shown in Figure 15. This plot clearly exhibits two linear regions. The first is between 800 103m3 and 1,000 103m3, and the second correlation period occurs between 1,200 103m3 and 1,800 103m3. The first linear correlation period (800 to 1,000 103m3) has a very similar slope to the second correlation period on the log (WOR) plot. Both linear periods demonstrate a hyperbolic - to harmonic decline behavior for this pool. As in the previous example, the log(WOR) versus Np, log(Qo+Qw) versus Np, and the RF versus HCPVI plots show clear trends, at least before infill drilling programs are implemented. The log cumulative fluid versus Np plot for the Taber South Mannville A pool is shown in Figure 16. The two linear periods can clearly be seen on this plot, and are separated by a disjoint in the slope between the two regions. This disjoint can be partly attributed to changing total fluid rate (caused by an increase in injection and production wells). The late time region of the plot demonstrates a linear slope, indicating a hyperbolic to harmonic decline, depending on how fluid rates are changing. Extrapolating the WOR curve for each linear region to WOR = 25, and translating the slopes of each linear region to the end of the data set, two recovery factors are predicted based on the OOIP. The extrapolation of the first linear portion leads to a recovery of 14.2% OOIP, whereas extrapolation of the second linear region leads to a recovery of 17% OOIP. This gives an incremental recovery of 2.8% OOIP, which corresponds to about 400 103m3. To date, secondary recovery due to waterflood is approximately 13%. Extrapolation of the RF versus HCPVI plot to 1 pore volume gives approximately 16% total RF (waterflood + primary). The RF versus HCPVI plot for Taber South Mannville A pool is shown in Figure 17. In this plot, the curve does not have a constant

slope (change in slope is denoted by the arrow), and the varying slope could be due to changes in interpattern flows. Using commercial software to confirm the harmonic decline exponent (b) for the Mannville A pool, both b=0 and 1 were used on the selected correlation period. Using b=0, it is evident that the exponential decline curve does not match the data set (see Figure 18), reinforcing the previous assessment that the reservoir is not exhibiting exponential decline behavior. Use of an exponential decline for this pool could significantly underestimate the total recoverable reserves. As shown in Figure 19, a much better fit to the data can be achieved using a harmonic decline exponent, b=1. The best fit for this correlation period is a super harmonic decline, with b=1.3 (Figure 20). Figure 15 to Figure 17 allow us to clearly identify the incremental oil due to infill drilling. As indicated before, the extrapolation to an economic water oil ratio (30) in Figure 15 and Figure 16 before and after infill drilling, respectively, yields an incremental recovery of 2.8% OOIP, corresponding to about 400 103m3. Masoner TechniqueMatching Fluid Rates for Predicting Decline Type The Masoner(26) (Chevron) method can also be used to determine the decline behavior of a reservoir. This method makes corrections for changing fluid rates, and makes the assumption that the drainage volume, recovery process, and relative permeability remain constant. The technique is valid for multi-phase flow where the decline is dominated by relative permeability effects. Briefly, in using the Masoner method, the engineer regresses on decline parameters to match a correlation period. The Masoner technique of matching fluid production levels was used to confirm the decline behavior for the Taber South Mannville A reservoir. Using fluid rates and selecting a decline period, it can be seen that the Masoner technique also confirms the hyperbolic to harmonic decline behavior (Table 1) for the correlation period of 1997 to 2002. Using this same correlation period, an oil rate plot was generated, which outlines the historical oil rate and the Masoner predicted oil rate (see Figure 21).

From Figure 21, it can be seen that the Masoner predicted oil rate has an excellent fit to the historical data. The excellent match of historical data for the Taber South Mannville A pool demonstrates that the changes in fluid rates are mainly due to accelerated production. Usually, the application of the Masoner technique does not provide this level of accuracy when history matching. The accuracy of the fit obtained during the selected match period of 1997 to 2002 is shown in Figure 21, as demonstrated by the minimal deviation between the Masoner predicted oil rate curve and the historical oil rate curve. This excellent fit to historical data using the decline period of 1997 to 2002 demonstrates that the pool is exhibiting a hyperbolic decline and confirms earlier analysis. SUMMARY OF FIELD CASES Over 50 Alberta waterflooded pools were reviewed for this study, with the selection of the pools being made at random. A total of 21 pools were studied in detail, pools selected for this study and their corresponding decline analysis summaries are listed in Table 2. Alberta pools were chosen because of ease of access to public data. The decline correlation periods for the selected pools were chosen using the criteria of: watercut greater than 50%; VRR close to 1; constant well counts; constant fluid and injection rates; constant reservoir pressure; constant producing well pressures; constant GOR; and, greater than 25% HCPVI. For the majority of the pools studied, it has been found that achieving a correlation period for greater than a fiveyear interval has been difficult. This has been in part due to changing well counts, varying injection rates, and fluctuations in total fluid rates. All the pools were analyzed using the above correlation period on the traditional oil rate versus time plots and supplemental

analysis using the log(WOR) versus Np, log(Qo+Qw) versus Np, and RF versus HCPVI plots was used to determine decline type. Based on the supplemental analysis, it has been found that, for the selected correlation periods, the waterfloods are generally exhibiting hyperbolic to harmonic decline behavior. The hyperbolic to harmonic decline behavior is represented by a linear trend in the slope of the logarithmic curve of the supplemental plots. Using commercial software to confirm the value of the Arps exponent b, it has been found that for the selected pools, the mean value of b=0.68. The frequency of number of pools and Arps exponent are shown in Figure 22. In Figure 22, it can be seen that 14 out of the 21 (approximately 67%) pools selected have an Arps exponent greater than 0 and less than 1 (0<b<1), whereas the remaining pools have a super harmonic decline (b>=1), representing 33% of the selected pools. No pools were found to have b=0, with the exception of Peejay Halfway pool exhibiting a super exponential decline in late stages (not shown in the tabulations). The performance of all pools was strongly affected by changing well counts, changing fluid rates, and changing injection rates, making it difficult to achieve longer than a five-year correlation period. Generally, operators of the waterflooded pools we examined showed aggressive and successful optimization. For our selected waterfloods, the average field exhibited a high hyperbolic - to harmonic decline character. One field exhibited a precipitous drop off in the late time region followed by a hyperbolic decline. Although not demonstrated here, often individual well decline was much harder to analyze because the data was noisy. Simulation studies on pools show that changes in fluids rates indicate that capture efficiency changes between wells. The log (WOR) versus Np, log(Qo+Qw) versus Np, and RF versus HCPVI plots are recommended as supplementary diagnostic tools. These functions are less noisy (i.e., exhibit smoother curves), making them easier to use in analysis. Trends exhibited by these functions assist in the determination of decline behavior.

CONCLUSIONS 1. Decline analysis and reserves estimation using decline analysis should be fundamentally grounded in a good understanding of what factors control decline. The same decline techniques should not be applied blindly to all fields and all drive mechanisms. Specifically, arbitrarily using an exponential decline approach (log(qo) versus time assumed to be linear) for water drive, solution gas drive, and gravity drainage systems is neither technically nor empirically correct. Late stage waterflood behavior is generally hyperbolic or harmonic in nature, if reservoir factors dominate. 2. The waterflood decline correlation period should have the following criteria: a) the watercut should be greater than 50%

b) the voidage replacement ratio should be close to one c) well count should be relatively constant

d) injection and fluid production rates should be relatively constant e) f) the reservoir pressure should be relatively constant producing well pressures should be constant

g) the GOR should be relatively constant h) the waterflood is mature; i.e., the volume of water injected should be greater than 25% of the hydrocarbon pore volume. 3. In light of the above, the criterion of watercut >50% for waterflood suggested by SPE and Petroleum Society of CIM is not sufficient to properly forecast future production. 4. In our experience, full waterflood decline occurs only after 50% watercut (WOR >1.0) due to a relatively slowly increasing volumetric sweep at that point. Numerous simulation/analytical studies confirm this characteristic decline start. This conclusion does depend upon mobility ratio effects and permeability heterogeneity: the higher the permeability heterogeneity, the faster decline behaviour occurs.

10

5. Another factor which significantly impacts volumetric sweep is infill drilling. Infill well programs often increase volumetric sweep more dramatically by increasing the number of pressure sink/withdrawal points. 6. RF versus HCPVI, log(WOR) versus Np, log(Qo+Qw) versus Np, and Masoner plots are very useful in identifying infill well incremental oil recovery. 7. Exponential decline (b=0), super exponential (ultrafast) decline (b<0) and super hyperbolic (ultraslow) decline (b>1) do occur. Exponential or super exponential decline can occur if there is skin buildup at the injectors. Super exponential decline generally occurs because of rapid watering out of a hot streak such as a natural fracture, induced fracture, or small hot streak layer. Super hyperbolic decline generally occurs because of multiplying or rising fluid/injection rates. Super hyperbolic decline can also occur in a multilayer reservoir. ACKNOWLEDGEMENTS The authors would like to acknowledge the fine contributions of many individuals in the reservoir engineering literature regarding the techniques of decline analysis as well as the Alberta Energy and Utilities Board for providing an excellent database for production data. The authors would also like to acknowledge Eric Denbina, Kent Edney, Bette Harding, Frank Kuppe, and Shelin Chugh for their assistance and many helpful suggestions in making this paper a reality. REFERENCES
1. Baker, R.O, Sandhu, K., and Anderson, T., Using Decline Curves to Forecast Waterflooded Reservoirs: Modeling Results, paper 2003-163 prepared for presentation at the 54th Annual Technical Meeting of the Petroleum Society of CIM, CIPC 2003, Calgary, AB, June 10 12, 2003. Arps, J.J., Analysis of Decline Curves, Trans. AIME, 1945. Muskat, M.: Physical Principles of Oil Production, McGraw-Hill Book Company Inc., 1949.

4.

Sandrea, R. and Nielsen, R.F., Dynamics of Petroleum Reservoirs under Gas Injection, Gulf Publishing Co., 1974. Ershaghi, I., Handy, L.L., and Hamdi, M., Application of the x-Plot Technique to the Study of Water Influx in the Sidi El-Itayem Reservoir, Tunisia, JPT, 1987. Currier, J.H. and Sindelar, S.T., Performance Analysis in an Immature Waterflood: The Kuparuk River Field, paper 20775 presented at the SPE ATCE, New Orleans, LA, September 23-26,1990. Fetkovich, M.J., Decline Curve Analysis Using Type Curves, JPT, June 1980. Laustsen, D., Practical Decline Analysis Part 1 Uses and Misuses, JCPT Distinguished Authors Series article, November 1996. Gringarten, : Evolution of Reservoir Techniques from Independent Methods to Methodology, Impact on Petroleum Curriculum, Graduate Teaching and advantages of Oil Companies, SPE 39713. Management an Integrated Engineering Competitive

5.

6.

7.

8.

9.

10. Slider H.C.: Worldwide Practical Petroleum Reservoir Engineering Methods, PennWell Books, PennWell Publishing Company, Tulsa, OK, 1983. 11. Ershaghi, I. and Abdassah, D.: A Prediction Technique for Immiscible Processes Using Field Performance Data, JPT, April 1984. 12. Startzman, R.A. and Wu, C.H.: Discussion of Empirical Prediction Technique for Immiscible Processes, JPT, 1984. 13. Timmerman, E.H.: Practical Reservoir Engineering, Part ll, Methods for analyzing output from equations and computers, PennWell Books, PennWell Publishing Company, Tulsa, OK, 1982. 14. Lo, K.K., Warner, H.R., and Johnson, J.B., A Study of the Post-Breakthrough Characteristics of Waterfloods, JPT, April 1990. 15. Ershaghi, I., A Method for Extrapolation of Cut vs Recovery Curves, JPT Forum, February 1978. 16. Lijek, S.J.: Simple Performance Plots Used in Rate-time Determination and Waterflood Analysis, JPT, October 1989. 17. Baker, R.O.: Reservoir Management for Waterfloods Part 2, J C P T Distinguished Authors Series article, January 1998.

2.

3.

11

18. Baker, R.O. and McClernon, L.L., Estimation of Volumetric Sweep Efficiency of a Miscible Flood, JCPT, February 1998. 19. Masoner, L.O., Decline Analysis Relationship to Relative Permeability in Secondary and Tertiary Recovery, paper 39928 presented at the SPE Rocky Mountain Regional/Low Permeability Reservoirs Symposium and Exhibition, Denver, CO, April 5-8, 1998. 20. Wong, K.H. and Ambastha, A.K., Decline Curve Analysis for Canadian Oil Reservoirs Under Waterflood Conditions, paper 95-08 presented at the 46th Annual Technical Meeting of the Petroleum Society, Banff, AB, May 14-17, 1995. 21. Bush, J.L. and Helander, D.P., Empirical Prediction of Recovery Rate in Waterflooding Depleted Sands, JPT, September 1968.

22. Ramsay Jr., H.J. and Guerrero, E.T., The Ability of RateTime Decline Curves to Predict Production Rates, JPT, February 1969. 23. Schuldt, D.M., Suttles, D.J., Martins, J.P., and Breit, V.S., Post-Fracture Production Performance and Waterflood Management at Prudhoe Bay, paper 26033 presented at the SPE Western Regional Meeting, Anchorage, AK, May 26-28, 1993. 24. Campbell, J.M.: Oil Property Evaluation, Prentice-Hall Inc, September 1959. 25. Alberta Energy and Utilities Board, Albertas Reserves of Crude Oil, Oil Sands, Gas, Natural Gas Liquids and Sulphur, AEUB, December 2002. 26. Masoner, L.O., A Decline Analysis Technique Incorporating Corrections for Total Fluid Rate Changes, paper 36695 prepared for presentation at the 1996 Annual Technical Conference in Denver, Co.

12

PVT Bo 1.042 Bno 1 Decline Period Start Date: 01/01/1997 End Date: 12/01/2002 Decline Parameters D 0.0406 b 0.8168 qoi 675.39 qfi 8964.33 2 r 0.9946 Table 1 Parameters for Decline Behavior: Decline Period 1997-2002

Field Pembina Provost Provost Taber South Viking Kinsella Taber South Caroline Jenner Bigoray Gift Medicine River Parflesh Joffre Rycroft Peejay Sunset Little Bow Pembina Grand Forks Rainbow Chauvin Nisku T

Pool

Correlation Period 1999-2001 1998-2003 1999-2001 1997-2003 1997-2003 1975-1986 1994-2003 1997-2003 2000-2003 1999-2003 1997-2000 1998-2000 1977-1983 1993-1997 1975-1979 1973-1978 1990-1995 1998-2000 1995-2003 1999-2003 1974-1983

b 1 0.36 0.47 1 0.8 1 1 0.5 0.44 0.5-1 0-0.5 0.5 0.9 1 0.5-1 0.446 1 0.5-1 1 0.1 0.41

log(WOR) linear linear linear linear linear linear linear linear linear linear linear linear linear linear linear linear linear linear linear linear linear

log(Qw + Secondary HCPVI Q o) RF%(OOIP) %(OOIP) linear linear linear linear linear linear linear linear linear linear linear linear linear linear linear linear linear linear linear linear linear 15 50 11 13 13.5 35 36 24 48 19 9.5 40 19 33 37 17 23 35 51 66 17 36 550 43 85 110 270 80 400 190 30 23 110 38 80 80 35 180 65 835 430 80

Lloydmijnster O Upper Mnvl OOO & Elrs S Mannville A Sparky I Mannville B Rundle A Upper Mannville O Cardium B Slave Point A Basal Quartz B Upper Mannville G D-2 Charlie Lake A Halfway Triassic A Upper Mannville U Ostracod E Sawtooth MM Keg River EE Mannville A

Table 2 Selected Pools, Correlation Periods, and Decline Analysis Summaries

13

Figure 1 Decline Behavior

Figure 2: Typical Waterflood Oil Rate Response

14

Figure 3: Selection of Decline Correlation Period

15

Figure 4: Schematic Areal View of Water Swept Zones in Waterfloods

Figure 5 Composite Plot for the Provost Lloydminster O Pool

16

Figure 6 VRR Plot for Lloydminster O Pool

Figure 7 log (WOR) vs. Np for the Lloydminster O Pool

17

Figure 8 log(Qo+Qw) vs. Np for the Lloydminster O Pool

Figure 9 RF vs. HCPVI for the Lloydminster O Pool

18

Figure 10 Exponential Decline Curve for the Lloydminster O Pool

Figure 11 Harmonic Decline Curve for the Lloydminster O Pool

19

Figure 12 Best-Fit Decline for Lloydminster O Pool (b=0.36)

Figure 13 Composite Plot for Taber South Mannville A


20

Figure 14 VRR Plot for Taber South Mannville A

Figure 15 log (WOR) vs. Np for the Taber South Mannville A Pool

21

Figure 16 log Cumulative Fluid vs. Np for the Taber South Mannville A Pool

Figure 17 RF vs. HCPVI for Taber South Mannville A Pool

22

Figure 18 Exponential Decline for the Taber South Mannville A Pool

Figure 19 Harmonic Decline for the Taber South Mannville A Pool

23

Figure 20 Super Harmonic Decline of Taber South Mannville A Pool

Figure 21 Masoner Predicted Oil Rate for Taber South Mannville A: Decline Correlation Period 1997-2002
24

Figure 22 Distribution of Arps Exponent (b)

25

Anda mungkin juga menyukai