SPE 109831

More Rapid and Robust Multiple History Matching With Geological and Dynamic
Uncertainties: Heavy-Oil Case Study
J. Poncet, G. Vincent, M. Inizan, P. Henriquel and P. Jannes, Total

Copyright 2007, Society of Petroleum Engineers

This paper was prepared for presentation at the 2007 SPE Annual Technical Conference and
Exhibition held in Anaheim, California, U.S.A., 1114 November 2007.
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, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officerf, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the Society of Petroleum Engineers is
prohibited. Permission to reproduce in print is restricted to an abstract of not more than
300 words; illustrations may not be copied. The abstract must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, Texas 75083-3836 U.S.A., fax 01-972-952-9435.

Abstract
The generation of reservoir simulation models that match
field production data has been and is still a long-time industry
challenge, not only for the time spent on history matching
studies but also because of the non-uniqueness of the solution.
This paper presents a new approach called "Hybrid Models" to
accelerate this process and get more realistic history match
models. Hundreds of stochastic possible geological models are
produced and tested in regard to the dynamic data. The Hybrid
model is a composite geological model, not only constrained
by the initial well data but also with selected parts of the first
realizations matching around some wells. This technique
allows a relatively quick history matching process and results
in a series of matched geological models.
This process was applied in part of a heavy oil field (14
horizontal wells in fluvial reservoirs were considered), after 3
years of production. The objective was to explain and
reproduce the high water-cut, oil rates, GOR and bottom-hole
pressures in this part of the field. A complete uncertainty
workflow was applied with sedimentological and
petrophysical uncertainties as well as fluids and dynamic
uncertainties. Results showed that static uncertainties were
essential to get a coherent match and "Hybrid Model"
technology was applied with success.
The Hybrid model technique gives several matched
geological models. All models have been carried out through
forecasting keeping the present development plan, evaluating
the potential impact of remaining static uncertainties.
Dynamic uncertainties were also considered on one geological
matched model. Several combinations of dynamic parameters
have been computed to keep a match. Corresponding models
have been transferred through forecasting. Final conclusions
were that at fixed development plan, dynamic uncertainties are
more to be considered and combined for the forecast than
static ones.

The use of the "Hybrid models" technique and the

integration of static and dynamic properties as matching
parameters have been shown to be efficient to produce
accurate multiple production history matched models. From
those models, it has been possible to quantify the remaining
uncertainties in terms of future production and to propose new
developments.
Introduction
The field considered in this study produces 8.5API
gravity of Extra Heavy Oil (EHO) with a viscosity at reservoir
conditions between 1800- 3500cP. The EHO is upgraded to
market of high quality 32 API synthetic crude oil. The first
phase of development is completed and includes more than
300 horizontal wells.
The reservoir section, Middle Miocene in age, is
subdivided in two main intervals. The lower part is mainly
stacked unconsolidated sands deposited in a braided meandering fluvial system. The upper part corresponds to
sands encased into a shaly sequence associated to a fluviodeltaic system with tidal influence.
Approximately, 80% of producing wells are drilled in
fluvial sands characterized by water production risks. In
contrast, deltaic sands represent 20% of total oil production
with little or no water risk. Due to extreme viscosity contrasts,
after breakthrough the water cut in many wells increases
rapidly.
In order to get a good understanding of the production
mechanism and then propose further development, it was clear
that the history match phase was essential. This history match
was not easy to reach, as the main parameter driving the
reservoir dynamic behavior was the location of shale barriers
within the model. Instead of modifying the geological model
on a cell by cell basis (without keeping the geological
coherency) a new approach called Hybrid model has been
developped in order to get a relatively quick history match
preserving the entire geological coherency.
Static and Dynamic Uncertainties
Before describing this hybrid model technique, it is
essential to come back to the geological and dynamical models
and their associated uncertainties. Those uncertainties will be
assessed and ranked with respect to their relative impact on
the history matching process, before planning the way
forward.

www.petroman.ir

SPE 109831

Geological Uncertainties
Part of the initial geological grid was extracted in order to
be representative of the IA and IB well clusters and the
surrounding ones.
Three types of uncertainties will be considered in the
geological workflow, starting from the sedimentology, then
petrophysics and finally fluids and the associated flushed
zone.
Two facies are defined in the model: shale and sandstone.
They correspond to lithological log types based on a cut-off on
PHIE and Vsh logs. They are defined on all the wells (vertical,
slanted and horizontal). Horizontal trends for the shale
proportions were extracted from full field maps for each
stratigraphic interval. On the studied clusters of the field, the
fluvial system evolves from a braided system to a meandering
system in 4 main stratigraphic cycles (each of them showing
an increasing shale proportion upward). The cycles are
naturally separated by floodplain argillaceous levels locally
eroded by the overlying channels. These vertical incisions are
one of the key parameters to understand the dynamic behavior
of the reservoir. Differents uncertain parameters have been
identified in this representation of the reservoir, starting from
the facies proportion to the shale heterogeneity dimensions
and stratigraphic repartition.
After a complete data analysis on the available data, the
petrophysical workflow began with the porosity simulation,
horizontal permeability derived from porosity and then
vertical to horizontal permeability ratio derived from the
horizontal permeability but with a low correlation. Note at this
stage, that the Net To Gross was set deterministic, 0 for shales
and 1 for sands.
In both IA and IB clusters, as well as in most of the fluvial
sands, an area has to be defined on top of the water-oil contact
(WOC). This region called flushed zone is considered to be
swept by water so that, only water can flow. The top position
of this zone was uncertain and has then been considered in the
study. Deterministic water saturation was calculated
depending on porosity ranges and position relative to the
flushed zone and OWC.
Considering those described uncertainties, many
geological realisations have been generated and are passed
through the dynamic simulator. The same set of dynamic
parameters will then be considered. In this first approach, the
chosen set will correspond to the most likely values. The
spread of responses due to the static uncertainties is very large.
This is particularly true for the cumulative water cut, where
the response can vary from no water to a case where the well
is producing more than 50% water-cut after only 1 year of
production. In most of the case the historical points are within
the static uncertainty range, meaning that more water is
produced in some realisations and less water is produced in
some others. It seems to be reasonable to think at this stage,
that some models can be found fitting the water-cut trend. But
the number of runs allowing a match is reduced.
Dynamical Uncertainties
Meanwhile working on the static uncertainties, various
dynamic uncertainties have been defined and characterized by
their own range of variation (i.e. minimum and maximum
values).

Forteen different uncertain dynamic parameters have been

defined and used in the present work:
- Horizontal and vertical permeability enhancements;
- Rock Compressibilty;
- Top Flushed zone;
- Water saturation and relative permeability in the
flushed zone;
- Oil viscosity;
- Aquifer volume;
- Relative permeability end points (Critical gas
saturation, Gas to oil, oil to gas, water to oil, oil to
water relative permeability);
- Mobile oil range.
In order to study the dynamic uncertainties impact on the
history match responses, the experimental design technique
has been applied in order to be systematic and exhaustive. It is
clear at a first glance that the variability obtained with
dynamic parameters is very large in terms of water-cut. As for
the static uncertainties impact, the range of variation goes
from no water runs to full water ones. It could be pointed out
that the range of impact is similar to the static ones, even if it
appears that more runs present a very low water cut. For the
bottom-hole pressure, the range of variation is large too, even
if it is a little bit more difficult to define this range. However,
the historical points are often within the range of variation and
the dynamic impact is similar in terms of range to the static
one.
Looking at the results in more details, it appears that the
well responses for a given parameter vary always in the same
direction. That means that, changing any dynamic parameter
from its minimum to its maximum value will impact the
history match response in the same direction (reduction or
increase). The impact of each dynamic parameter is global. In
conclusion, if the starting point of the dynamic history match
process (i.e. the geological model) is not right enough, it
would be quite impossible to reach a coherent match model
just playing with dynamic uncertain parameters only. static
uncertainties are essential to get a match.
Hybrid Models Description
As mentioned previously, the range of variation obtained
with the static uncertainties is very large and quite always
covers the historical data points. However, the percentage of
success for each of the well is very low (only 4% in average of
the geological realisation are giving a match for a chosen
well). Letting the static uncertainties without constraints will
decrease this probability. With 14 wells to be match, the
chance to get a match for all the wells in the same time should
be (0.04)14 = 2.68 10-20 representing less than one chance over
37 1018. It is clear at this point that this solution could not be
reached alone and that more constraints have to be put on the
models to get a match. That is the reason why the hybrid
model methodology has been implemented.
A Hybrid Model is a geological model or realisation, not
only constrained by the well data but also with previously
chosen geological inputs. Those geological inputs (parts) have
been selected on the initial multi-realisations to fit dynamic
data near the corresponding well. In order to complete this
hybrid model, the methodology to be applied would be the
following:

www.petroman.ir

SPE 109831

1) Determination of the matching models. Upon all the

various realisations simulated during the static
uncertainties impact work, the matching realisations for
each of the well will be kept for future work. Not only one
single realisation will be kept in order to keep uncertainty
in the matching area.
2) Determination of the area of influence. The whole
matching realisations kept for each individual well are not
to be introduced as constraints in the final hybrid model.
Only a part around the corresponding well has to be
conserved. The aim of this part is to determine what part is
really influencing the match.
3) Implementation of the new constraints. All the retained
constraints will be introduced in addition to the well data.
4) New geological realisations. Geological realisations will
then be realised, fitting all constrained data and letting the
uncertainties in the other part of the reservoir. The idea at
this stage is to keep the geological input in those
realisations.
The coherency of the new realisations is kept as the
geostatistical simulations keep the same modeling techniques
and the same rules. For some cases, the added constraining
data are not compatible with each others, so that the
percentage of really constrained data is less than 100%. In
such cases, a new constraining realization has to be found to
ensure the geological coherency. The percentage of constraint
data is a very important criterion to check to be sure that the
methodology can be applied without risk.
Applications
Considering the various models obtained using this hybrid
model technology, the methodology is not too much
constraining; the geological input and variability far from the
wells are kept in all those realisations. This method looks
appropriate.
For most of wells, the spread of water-cut and static
pressure is really reduced and the match has been
characterized as good, keeping the geological model coherent.
At this stage, hundreds of realizations were simulated and
were keeping the match. It was then very easy to use such
realizations in order to propagate them in a forecast exercice
with the current developpement plan (no additional wells). In
a second time, it was decided to focus on one medium
realization and to use dynamic uncertainties in order to get
multiple matches (with the same static realization but with
other combinations of dynamic parameters). Use the
experimental design technique with the already defined
uncertain dynamic parameters, hundreds of dynamic
combinations of parameters were found as matching ones for
different wells. Those realizations could then be transferred to
forecast and compare with previous ones. The impact of
dynamic parameters in terms of forecast is much larger than
the static ones. As it was clear that the match was essentially
static, the forecast appears to be essentially dynamic.

match. Letting Monte-Carlo draw reaching a match was not to

consider since the chance of getting a right image was very
low.
A new technique called Hybrid Model had been
developed and tested on this field. This technique is looking at
finding geological realizations matching the dynamic data,
determining the area of influence, i.e. the area necessary to
keep a match but small enough not to constrain too much data,
adding those new data to the well constraining data and
generating new geological realizations keeping the same
workflow and geological rules. This technique has been
considered to be very useful to accelerate the matching
process.
Hundreds of various coherent geological models were then
obtained with this technique. Those models were then used to
forecast production and also study new wells locations and
corresponding reserves.
Dynamic uncertainties were also studied on a single
matching geological realization. Their impact in terms of
forecast was large and reserves are still linked to such
parameters.
Acknowledgements
We thank Bernard Corre, Jean-Michel Guemene, Vincent
Marlot and Pierre Biver (Total) for support and helpful
discussions.
References
1.
Den Boer L., Poncet J., Biver P., Henriquel P.,
Marlot V.:A New Technique To Achieve History
Match Using a Probabilistic Approach SPE 102871
2.
Periera E., Vincent G. and Ichbia J.M.: Controlling
Water Risks in Extra-Heavy-Oil Environment SPE
97708-MS International Thermal Operations and
Heavy Oil Symposium 2005.
3.
Inizan M., Henriquel P., Vincent G., Di Cuia R.,
Lapointe P. , Vieban F.:Reducing Uncertainties and
Ensuring Robust Forecast Using Multiple
Production History Matches EAGE Uncertainties
& Production History Match Workshop 2004
4.
Vincent G., Corre B., Lendre P., Charles T.
:Optimisation of the incremental development
policy for a mature waterflood project SPE 78579
5.
Corre B., Thore P., De Feraudy V. and Vincent
G.:Integrated Uncertainty Assessment For Project
Evaluation and Risk Analysis SPE 65205
6.
Charles T., Guemene J.M., Corre B., Vincent G. and
Dubrule O.:Experience with the Quantification of
Subsurface Uncertainties SPE 68703
7.
Vincent G., Corre B., Thore P.:Managing
Structural Uncertainty in a Mature Field for Optimal
Well Placement SPE 48953
8.
Alabert F., Corre B.:Heterogeneity in a complex
Turbiditic Reservoir: Impact on Field Development
SPE 22902

Conclusions
Starting with static and dynamic uncertainties in order to
help the history match process for the two studied clusters, it
rapidly appeared that the choice of static realizations
(geological model) was essential in order to get a coherent

www.petroman.ir