Documente Academic
Documente Profesional
Documente Cultură
Steamflooding
by C.S. Matthews, SPE
C.S. Matthews is senior petroleum engineering consultant with Shell Oil Co. in Houston. After
receiving a as degree in chemical engineering and a PhD degree in chemistry from Rice u., Mat-
thews joined Shell Development Co. in 1944 as an engineer in San Francisco. He transferred to
Houston 4 years later to work on production research and became chief reservoir engineer for
Shell's Technical Services Div. in 1956. In 1965 he became manager of exploitation engineering in
New York City. Later he served as director of production research with Shell Development and as
manager of engineering in the Production Dept. Matthews, with D.G. Russell, wrote SPE's first
Monograph, Pressure Buildup and Flow Tests in Wells. They were 1967-68 Distinguished Lec-
turers on that topic. Matthews also chaired the Reservoir Engineering Technical Program Commit-
tee for the 1969 Annual Meeting and the Lucas Gold Medal Committee during 1972-73. He
received the Lester C. Uren Award in 1975 for distinguished achievement in petroleum engineering
technology. His current interests include enhanced recovery of oil and gas, tar sands, and
geothermal energy.
Introduction
Steamflooding has become an established recovery Venezuela and the U.S. Most of the oil produced by
technique within the last 20 years. This overview steam from 1960 to 1970 was by this process.
discusses its evolution, methods for selecting and Typically , steam is injected for several weeks and then
designing steamfloods, constraints, and possible the well is closed in (soaked) until the steam has
improvements. condensed. A pump then is run and the well is placed
The tenn steamflooding is used here in a general on production. When oil production falls to a low
sense. The discussion includes steam soak (cyclic level, the cycle is repeated.
steam injection) and steam drive. For additional Thennal production in Venezuela is still almost
infonnation the reader should refer to Farouq Ali and entirely (95 %) from steam soak. Compaction of these
Meldau. I thick, unconsolidated sands together with solution-gas
drive is leading to satisfactory oil recovery at very
Early Developments: Steam Soak high ratios of oil recovered per barrel of steam used.
There are records of steam injection into a Texas oil
reservoir as far back as 1931 .2 Steam drive on a sus- The Netherlands. One of the first large-scale steam
tained basis, however, did not begin until 1959-60 drives began at the Schoonebeek field in The
when Shell affiliates undertook steam flood pilots in Netherlands in 1960. 3 Recovery of the moderately
Schoonebeek, the Netherlands,3 Mene Grande , viscous oil (180 cp) was quite successful. This
Venezuela,4 and Yorba Linda, CA.5 Steaming encouraged additional field trials of steam drive in the
operations still under way at Schoonebeek and Yorba U.S. and in Venezuela. Coring showed that steam was
Linda are discussed later. achieving low residual oil saturations of the order of
8 %. The residual oil in the zone swept only by hot
Venezuela. The steamdrive pilot in Mene Grande led water was about 35 %. Later tests at Schoonebeek were
to development of the steamsoak process. 4 While carried out at high pressure and high temperature.
attempting to relieve the fonnation pressure by Under these conditions C02 and H 2S are generated in
opening a steam injector to production, oil was situ . Consequently, new alloys had to be developed.
produced surprisingly at a rate of 100 to 200 BID. Thus, Schoonebeek led to pioneering along both
This was the first steamsoak well. This process reservoir and metallurgical lines.
underwent considerable development in both
u.S. Commercial steamflooding began in the U .S. in
0149·213618310003·9993$00.25
1960 in the Yorba Linda field, CA. 6 Prior production
Copyright 1983 Society of Petroleum Engineers of AIME in this field had been achieved only through use of
MARCH 1983 465
TABLE 1-SUMMARY OF TYPICAL WELL PERFORMANCE FOR STEAMSOAK OPERATIONS IN VARIOUS CALIFORNIA FIELDS
Additional
Oil
Net Oil Recovered Recovered
Production (BID)
Sand Steam Cycle Per per
Before After Rate hot Open Injected Period Total Barrel of Barrel of
Field Zone Steam Steam' Rate cold (ft) (bbl) (months) (bbl) Steam Steam
--
Huntington Beach TM 15 160 11 40 4,500 15 29,000 6.5 5.0
San Ardo Lombardy 25 360 14 220 14,000 18 50,000 3.6 2.8
Kern River China 3 140 47 22 4,400 6 11,600 2.62 2.5
Midway-Sunset Potter A 10 110 11 250 6,000 5 9,240 1.54 1.29
Kern River Kern River 14 65 4.6 220 6,500 5 4,730 0.73 0.43
Coalinga Temblor 3 52 17.3 107 9,000 5 4,300 0.48 0.40
Midway-Sunset Tulare 5 56 11 240 12,000 6 4,640 0.38 0.31
Midway-Sunset Potter B 5 35 7 250 7,700 4 3,000 0.39 0.29
White Wolf Reef Ridge 30 85 2.4 75 14,000 4 6,750 0.48 0.23
Poso Creek Etchegoin 7 20 3 80 6,700 6 2,660 0.40 0.21
downhole heaters. Primary production had been Other. Steam soaks are also under way in France and
estimated to be 5 % of the 100 million bbl in place. the Congo. Soaks probably will be used in connection
After the success of steam soak in this field became with steam drives under way or planned for West
known, the process spread rapidly throughout the Germany, Trinidad, Sumatra, Argentina, and Brazil.
state. By 1963 there were 29 steamsoak projects in the
state,5 and by 1965 there were 267. Steam injection of Methods for Analyzing Steam Soak
15,000 to 25,000 bbl per well led to peak oil rates of Steam soak was a process for which field application
100 to 200 BID per well and to declining but still outstripped theory and laboratory research. It was
commercial production for a period of about I year. more economical in many cases to apply steam soak in
The injection cycle then would be repeated. It was a particular case than to conduct research to determine
relatively inexpensive to try steam soak, and a number whether it would be applicable. The main expense was
of operators tried. By 1967 the estimated additional oil that of moving in and connecting a portable steam
production in California from steam soak had reached generator.
120,000 BID. Some 408 steam generators were in use Empirical observations were the first guides to
at that time. 5 applicability. Some of these are:
Steam soaking also was tried widely outside 1. Thick homogeneous sand reservoirs give the best
California. Although success was obtained in a few response, particularly when gravity drainage is
cases, the additional production obtained was very effective.
small in comparison with California. In the U.S. 2. Thin (20- to 40-ft) permeable reservoirs also
outside California, oils are normally of low viscosity. respond well for a few cycles until the pressure is
Waterflooding proved the most economical method of depleted; they also respond well if gravity can provide
supplemental recovery, and steam could not compete. cold oil inflow.
For the thick, viscous oils in California, waterflooding 3. Thin reservoirs or thick reservoirs composed of
gave poor recoveries. This opened the door to steam. several sand members respond poorly if pressures
Early attempts to apply steamflooding to U.S. tar are low.
sands were singularly unsuccessful. This was largely 4. Reservoirs producing at high water rate and high
due to unfavorable reservoir characteristics such as low water cut respond poorly.
oil saturation, thief zones, fractures, or extremely high 5. Reservoirs with low permeability or containing
oil viscosity that kept the steam injectivity rate very oils of high viscosity (> 10,000 cp) respond poorly.
low. These observations led to several methods for
mathematically analyzing a steam soak. Boberg 7
Canada. One of the first sustained steamsoak projects related the improvement in oil production to the radial
in Canada began in the Cold Lake tar sand in 1964. distance heated from the wellbore. His analytical
After many years of experimentation, production had method allowed the heated zone to cool by conduction
reached a level of 7,000 BID in 1980. and by removal of hot fluids. For a typical heavy-oil
Steam soak is not as successful in the more viscous field his method predicts a maximum rate enhancement
(1 million cp) Athabasca tar sands. The viscosity at of 3 to 4 times.
Cold Lake is some ten times lower than at Athabasca. Boberg's analysis generally underestimated the rate
Cold Lake tar also apparently has enough gas in achieved in a steam soak, probably because his
solution to drive cold oil into steam-stimulated wells. analysis did not account for the increased solution gas
466 JOURNAL OF PETROLEUM TECHNOLOGY
driven out of the crude by the steam and the effect of scaling up results of a pilot to full scale is difficult
low-level heating beyond the very hot zone. Boberg's without some sort of mathematical or physical
method seems to work best when the reservoir reservoir model. Thus, development of predictive
pressure is still relatively high. In such cases the methods for steam drive went hand-in-hand with field
increased drive due to gas release will be small. development.
Owens and Suter 8 presented a very simple model
for calculating the response to steam soak. They Early Prediction Methods. Laboratory studies such as
proposed: those of Willman et al. 9 or Volek and Pryor lO were
useful in showing the potentially high displacement
Rate hot Viscosity hot obtainable with steam. Distillation by steam of the
Rate cold Viscosity cold crude left as residual oil at the steam condensation
front results in very low residual oil saturation in the
In general, this method tends to overestimate the effect steam-swept zone. For most viscous crudes of interest,
of steam stimulation. In a typical case an enhancement the residual oil saturation is 5 to 15 % of PV, well
of 20 to 50 times will be predicted. Oil production rate below the residual saturation attainable by waterflood.
enhancement in partially depleted fields ranges from These early laboratory studies also showed that steam
about 6 times in some thick Venezuela reservoirs to 10 to displacement was a highly stable process. Although
20 times in many California reservoirs (Table 1). steam is of low viscosity, it cannot "finger" through
A number of additional models for steam soak have an oil bank as can a noncondensible gas. Fingers of
been presented. These are generally similar to the steam lose heat rapidly and condense, heating the
Boberg model. All allow an estimate to be made of oil surrounding rock and fluids. This stability of the steam
recovered per barrel of steam. Some early field results front is an important advantage of steamflooding.
are shown in Table 1.5 Laboratory measurements of residual oil left by
Generally speaking, steam soak does not lead to steam are one element in the design of a field project.
high recovery. Recoveries in the range of 10 to 20% Prediction methods in the early 1960's made use of
of OOIP are normal. However, for certain reservoir these measurements, together with analytical solutions
conditions, particularly where gravity drainage can of heat flow and heat loss, to estimate temperature
provide economic production rates, recovery can be distribution, thermal efficiency, and recovery. The
high. In the Yorba Linda Upper Conglomerate zone, a work of Lauwerier ll and Marx and Langenheim 12
recovery of 35 % of OOlP was reported. This excellent were among the earliest analytical heat flow solutions
recovery was a result of gravity drainage in a thick to be used.
(325-ft) shallow zone where wells could be drilled on To use these methods, the rate of steam injection
very close spacing (0.8 acre). Thus, the wells were had to be determined from an injection test. The
much closer laterally than the pay thickness. Reservoir amount of oil displaced by steam then could be
compaction has given good recovery at main Tia calculated for the case of a homogeneous reservoir
Juana, Venezuela. Total recovery by steam soak with a vertical displacement front (no gravity override
through reservoir compaction and solution-gas drive in of steam).
this reservoir is 25 to 30 % . Myhill and Stegemeier 13 extended these earlier
Steam soak continues to be important, accounting analytical methods and presented a simple method for
for almost half of current steam-induced production estimating the oil/steam ratio for a field project.
throughout the world. It will remain important as an
initiating mechanism for drive projects where initial Scaled Physical Models. These models came into use
injectivity is low, and as the major mechanism in for steamfloods in the late 1960's. They accounted for
reservoirs with poor continuity. It also will remain reservoir heterogeneity and gravity override of steam,
important in reservoirs with good gravity drainage or and thus were superior to the previous analytical
reservoir compaction drive. methods. In addition, they showed where additional
wells were needed or where changes in operating
Methods of Predicting Steam Drives policy could improve results.
Steam drive began to gain importance in the U.S. in Early scaled models required use of heavy pressure
about 1970. (Outside the U.S., it is still relatively vessels around the model to allow an overburden
unimportant.) At that time, steamdrive production in pressure to be applied to the flow chamber.
California was about 30,000 BID. Some 12 years Consequently, the models were bulky. Much time was
later, production by steam drive had increased to more required to pack them with sand, to saturate them with
than 150,000 BID. This section traces the rise of that water and oil, and to deplete them to conditions
production, beginning with a discussion of predictive prevailing at the beginning of steam drive. In these
methods. early models, steam temperatures equivalent to field
In contrast with the lesser value of steamsoak conditions often were used.
performance prediction, steamdrive performance A significant breakthrough in modeling steamfloods
prediction is very important. Much time and money in the laboratory occurred when it was shown that
are required to conduct a drive pilot. Furthermore, properly scaled results could be obtained at low
TABLE 3-FACTORS UNFAVORABLE FOR to duplicate physical model results, 15 lending even
STEAM FLOODING more confidence to the use of mathematical
simulation.
1. Oil saturation less than 40%.
2. Porosity less than 20%. Mathematical Simulation. The partial differential
3. Oil-zone thickness less than 30 ft.
4. Permeability less than 100 md. equations describing the flow of oil, water, and steam
5. Ratio of net to gross pay less than 50%. together with the flow of heat were formulated for
6. Layers of very low oil saturation and high permeability in petroleum reservoirs in the 1950's. At that time,
the oil zone that act as thief zones. however, computers had insufficient memory and were
7. Extremely high oil viscosity.
8. Fractures.
much too slow to allow these equations to be solved
9. Large permeability variations in the oil zone. simultaneously as a steamflood progressed. Over the
10. Poor reservoir continuity between injectors and producers. ensuing 20 years, not only did computer speed and
11. Deep, high-pressure reservoirs and shallow reservoirs with storage increase phenomenally, but efficient
insufficient overburden to permit steam injection without
computation techniques were developed. Coats, Chu,
fracturing.
and Marcum 16 and Coats 17 have described some of
these techniques.
At the current stage of development, it is possible to
simulate the behavior of a representative portion of a
temperatures with subatmospheric-pressure steam. 14 heterogeneous reservoir under steam drive. The
This enabled the packed-bed models to be enclosed by modeled portion can include several injection wells
thin plastic sheets, which, upon imposition of a and their surrounding producers. The reservoir can be
vacuum, became rigid containers. This technique made heterogeneous both vertically and laterally. The effects
the heavy pressure vessels unnecessary. Furthermore, of dip and gravity can be included.
it allowed researchers to see where displacement was Mathematical reservoir simulation is now the fastest,
and was not occurring. most accurate, and most efficient method for
Although scaled physical models can be used to predicting performance of steamfloods. One problem
portray most mechanisms of a steamflood accurately, that arises in such simulation is the difficulty in
they are still time-consuming to pack and operate and visualizing results. The computer output is usually a
often are limited by availability of materials and fluids discouragingly large stack of paper showing a large
to achieve proper scaling for particular oils and sands. number of computed values of oil production,
Considerable time and care are required to pack the temperature, saturation distribution, etc. In some cases
model with the proper sizes of sand or glass beads, to a series of 2D plots of these data also are provided.
saturate it properly with oil and water, and to carry out The voluminous nature of the output is a deterrent to
the steam flood simulation. More recently, development its analysis. Typical output from one run is about half
of efficient computational techniques and increases in a million numbers.
speed and storage of computers have made One recent improvement in this regard is the
mathematical simulation of steamflooding efficient, Dynamic Visual Display method developed by Shell
reliable, and fast. They also have enabled Development Co. 18 By this technique the numerical
consideration of the effect of additional mechanisms, output from a large computer is processed by a smaller
such as the effect of dissolved gas on a steam drive. one to develop 2D arrays of the process variables.
Recently, thermal numerical simulators have been able These digital results are converted to TV signals that