Msc Petroleum Engineering
Reservoir Simulation Course
History Matching of Reservoir Simulation With the Eclipse
Reservoir Simulator
Student Name: Longos Konstantinos
Registry Code: 2014/028/017
Date: 29/6/2015
Introduction
Reservoir simulation gives insight into dynamic rock and fluid properties for evaluation of
past reservoir performance, prediction of future reservoir performance, and reserve
estimation. Reservoir history matching is a tedious and time-consuming exercise,
undertaken to reduce the differences in performance between a reservoir simulation model
and its historical field performance. To replicate the reservoir’s performance, dynamic and
static data from the reservoir is required to characterize the reservoir model. This enables it
to reproduce a model very close to the historical performance
SIMULATION MODEL
In the current project the following reservoir model was given and the purpose was to run
different scenarios – simulation runs- thus by affecting the characteristics of the reservoir in
order to history match available production.
This project concentrates mainly on manual history matching carried out on field above. It
involves making changes to pre-selected sensitive parameters using production data. Main
concerns involving selecting changes in transmissibility, within blocks proved to be the main
parameter of concern as their perturbations significantly affected the outcome of the final
match.
Nomenclature
COMPDAT Well Completions Specification Data, Eclipse100 Keyword.
P-BRH - Production well
P-A1H - Production well
P-A2H - Production well
P-A39A - Production well
P-A17 - Production well
FGPR Field Gas Production Rate
FPR Field Pressure
Injec Injection
MLT Multiplied
MULT Multiplier
MULTZ Transmissibility Multiplier in Z-direction (downwards)
NTG Net to Gross ratio
WOPR Well Oil Production Rate
WOPRH Well Oil Production Rate History
WOPT Well Oil Production Total
WOPTH Well Oil Production Total History
WWCT Well Water Cut
WWCTH Well Water Cut History
HISTORY MATCHING OVERVIEW History matching is the process of adjusting the reservoir geological model to match the
model from field production data. Reservoir production performance greatly determines the
economic feasibility of oil and gas recovery and also the future of production operations.
Thus, for efficient reservoir management, a thorough analysis of past, present and future
reservoir performance is required, and history matching is a very handy tool for this. History
matching aids in updating the current reservoir model, matching it with past production, and
optimized future prediction.
The main reason for history matching is not just to match historical data, but to enable the
prediction of future performance of the reservoir and thus production optimization with
regards to economy and oil and gas recovery by improved or enhanced methods. A
combination of Schumberger’s Eclipse100, Office and Floviz were used for the history
matching process. While the Eclipse100 was used for simulation runs, Floviz was used to
view the reservoir in production
Model Description
In our given model we had to proceed with history matching based on available production
data and in order to do it we had to choose 4 producers as a reference base to proceed with.
Thus we have chosen to proceed with history matching based on the PA17 ,PA39A, PA2AH
producers. And thus we have to match our reservoir model to the following production
history of the aforementioned producers. Those producers were chosen based on the fact
that they were more close one to another, forming a “block” and after many trials done it
was decided to match our production history based on production of them in order to
producing more satisfying and realistic results.
Before proceeding to explain the basic steps that were accomplished during simulation it is
worthwhile to estimate our “initial situation” that is the oil production through chosen
producers of our initial model.
Graph.1: Cummulative Oil production vs time that should be used as a reference for history
matching.
The initial FOPRH and FOPR vs time is been giving to the following diagram
Graph.2: FOPRH and FOPR vs time is been giving to the following diagram
Graph3: FOPRH and FOPR vs time is been giving to the following diagram
Before proceeding with trials and changes in the code we should have a reference on
how we are going to proceed based on a first estimation of how far the history of
each production well is from our current – initial model. This will be used as a
reference for the optimization.
graph3: Initial “Matching” of model - Cummulative Oil production vs time
From our initial model we can see that P-A17 & P-A1H are the producers that are closest to
the production history thus our first consideration is the optimization of the rest of the
producers.
We can see that producer P-A17 is the most acceptable matched so we have a reference
point to further increase the later producers by using appropriate boxes in the run code.
After running the aforementioned code to the Eclipse Reservoir Simulation we ve used the
GeoQuest post-processors FloViz. During this course FloViz is used for 3D displays of the
models created in order to further asses and optimize the matching. By running step-wisely
the flow simulation we can visualize incrementally the flow from the injectors to the
producers.
From the grid visualization we can see the flow from injections to producers and estimate
which blocks should be used in order to improve our model. Thus after several attempts to
history match the producers we saw that we had firstly to reduce substantially the
transmissibility towards the PA1- whereas it was in the “sense” of manual history-matching,
impossible to match all producers thus we had to attempt a first option of isolating the flow
towards the one P35 who further more according to the schedule details has less significant
impact to the production cause it’s been induced later in time. There were made several
trials in order to proceed with optimizing production one by one each well. In the following
the process that we followed will be analyzed (figure3). Firstly we have induced a block in
the area near the producer P35 and use accordingly the MULT X, MULT-Y , MULT Z in our
code in order to reduce the flow towards this producer.
(Figure.3) Box near the area of P35 producer.
Following is the results after this very first change in our box inducing.
graph4: Cummulative Oil production vs time for the 4 producers selected.
From the above diagram no significant improvement is achieved in comparison with the
initial one. Only a slightly increase in the producer PA1H. More “Box’s” in the code were
implemented to proceed with more acceptable results.
First of all based on our model we had to reduce the transmissibility over the area where no
producers have been placed. This was done in order to make the flow go from the injectors
to the producers. Thus we induce a box within the area away from producers and
substantially reduce the transmissibility in the layers involved. Many trials have been done
to find “suitable box” that is the relevant (x,y) points and the z-layers which will improve our
model according to history matching. Following is our model after the aforementioned
changes.
Box– No Box – Grid Coordinates MULT X MULT Y MULT Z
1 1 6 3 21 1 10 0.1 0.1 0.1
The above block which sould be admitted it is quite big is been considered because in the
area around, there are no producers - only one injector – thus by minimizing the
transmisibility to it we enhance the flow towards the other direction.
Figure4: First Block Selected
The representative results considering the cumulative oil production following the above selection
are:
Figure 5: Total Oil Production based on the BOX1 chosen.
As we see in the above graph there is quite improvement in one of the producers more
specific PA17. Thus after many trials this box is considered to be acceptable in improving the
history matching procedure. The optimization sequence that have been used in the process
was based on matching the cumulative production incrementally in each well and also try to
match water cut as well. Furthermore “new box” was included.
In order to improve the flow towards production PA1 we increased the transmissibility in the
area that is near producer PA1.
The forthcoming results up to the above procedures are available in the following graph.
Figure 6: Total Oil Production based on the BOX chosen
Following is the respective result where in the same procedure 2 producers were able to match.
Figure 7: Total Oil Production based on the BOX1 – Box2, Box3 chosen
Thus we see after many trial and errors 2 Producers have been able to be matched. As we
see from the above diagram both PA39, PA17 have been substantially matched. Continuing
in the same way with the rest of the producers we induce the following BOXES to our code.
Box– No Box – Grid Coordinates MULT X MULT Y MULT Z
1 1 6 3 21 1 10 0.1 0.1 0.1
3 7 13 3 13 1 12 5 5 0.8
4 2 5 26 38 1 10 1 1 1
Τhus from the above diagrams for different “values” of boxes we were managing to improve
only 2 out of the 4 wells thus more boxes were induced in our code. In the following picture
a visualisazation of the basic blocks that we introduced is figured and further explanation
and reasoning will be given on why the were chosen
Gaining help from flow-vis we needed to lower the flow within producers PA2H , PA39 thus a
box within them was induced with lowering transmisibilities. (BOX7) . Moreover PA1H is
horizontal thus a flow in the above producer was affected by choosing different
transmisibilities within z- direction. That is different layers.
The following Grid shows us the basic blocks that were used up to now and following are the
results that were obtained by these.
Figure 8 : Total Oil Production based on the BOX1-7 chosen
Box 6 is near the edge of our model and no-producer of the once that we were mainly
concerned with have been there thus the flow towards that direction, has been lowered by
lowering the transmissibility accordingly.
After inducing these basic “blocks” to our model we improved our model history matched by
an accountable percent. Until now we managed to have substantial improved our initial
model. This is obvious to the relevant figures.
Figure 9: Cummulative Oil History matched production vs time (pre- final)
Thus from the above diagram we see a substantial increase in matching the oil production
for each of the producers. As we will see bellow further matching will be achieved by
optimizing the water-cut rate on each of them.
Also some more relevant graphs are presented concerning total production & watercuts.
Figure 10: Field Total production vs history matched one.
Thus we see an acceptable degree of match between the production history and our current
model. As we see after 3000 days our model surpasses the current history and gives higher
values as the current one.
Water- Cut Match
Figure 10: Water-Cut production for PA1H - producer against history data
Figure 11: Water-Cut production for PA2H - producer against history data.
Figure 12: Water-Cut production for PA17 - producer against history data
Figure 13: Water-Cut production for PA17 - producer against history data
As we see our model is currently optimized for the 3 productions PA1H ,PA17, PA2H whereas
for the case of PA39 the history is not matched so well. So based on this last one we will try
to optimize this and see if we can further improve our final model.
Figure 12: Total water cut against history matched.
As we see from the above diagram there is a quite good match to the water cut
against the history matched up to the 3000 day. Some deviations is due to the fact that
some producer’s haven’t been optimized to history match but as has been referred
previously optimization was based mainly in 4 chosen producers.
Figure 13: Total Oil production against history matched
As it is obvious from the above diagrams up until now the production history has
been very well matched in the all 4 producers but as it is obvious from the watercut
diagrams some more work is needed to improve matching production to the PA39 and
PA2AH producer.
So in order to further improve the matching of the last well we induce in the code a box very
near to the producer PA2H in order to reduce transmissibility around where’s the relevant
producer, produced more than the data of the history (BOX 8).
Figure 14
However reducing transmissibility on the one had a counter-interactive effect on the other
as it is obvious in the following figure.
Figure 15: Change in history matching process of one Producer in comparison to other
From the above diagram we see that trying to perfectly match PA2H by reducing
transmissibility it has as effect to increase the production of PA1H . So we induced a final
box - around producer PA1-H as to not let the production surpass a lot the history match
process. It should be mentioned that the last 2 steps are taken very locally within a very
small area of the producers only to improve our model. Generally history matching should
be achieved based on fundamentals’ of reservoir engineering taking into consideration wide
areas of the reservoir as to be more accurate and robust to any minor changes in the
attributes. However in our case where the whole procedure has been done manually and we
managed up to now to match substantially our data it could be an acceptable procedure.
Thus the following grid describes us our final model.
Procedure In The History Matching Process – Summary
BOX 1: Reduce Flow where no-producers exist, thus enhance the flow towards next
direction
BOX 3: Reduce Flow where no-producers exist
BOX 4: Reduce Flow within the area.
BOX 5: Even more reduced in flow within these area.
BOX 6: Substantially reduce transmissibility from there onwards as flow to be
within the rest area. (Many trials Used).
BOX 7: Increase transmissibility between producer’s P2AH – PA39
BOX 8 & 9 : “Small boxes” placed after our initial to further optimize the history
matching of production PA39H where we faced many challenges on it. When we
tried to improve history match this, the PAH1 was out of match. So we placed 2
small box’s near those.
RESULTS - Final Model
The final results are obvious in the diagrams bellow. (Final History Matching Model)
Figure 19: Cummulative Oil production vs History matched production (pre- final)
Figure 20: Water-Cut production for each producer against history data
Figure 21: Total Oil Production against history data
Following are the matching of production rates for each producer.
Figure 22 : Oil production rate against History Matched for PA1H producer.
Figure 23 : Oil production rate against History Matched for PA2H producer.
Figure 23 : Oil production rate against History Matched for PA17 producer.
Figure 23 : Oil production rate against History Matched for PA39 producer.
Discusion – Results
After several simulation runs, a final match was obtained. This final match
was tested on: oil rates, their total production and water cut. For most of
the tested scenarios, the final match gave a good match for the oil
production rates and production totals as can be seen. A very good match
was also obtained for the oil production rates as for producers PA1H, PA2H,
PA17. Also the final match gave a good match for total oil production. In the
case of water-cut we managed to match very well 3 out of 4 producers. More
specifically PA1H, PA2H, PA17. In the case of PA39 the water-cut did not gave
so close results as the other cases. The most difficult parameter to match
using manual history matching was water cut. Several extra adjustments had
to be made to the selected parameters for a good match. Floviz software was
really helpful in helping to locate the cells for which transmissibility had to be
increased or decreased. The other parameters for possible perturbation were
kept at their values given.
Manual history matching though very tedious and time-consuming, can be
very helpful in understanding the contributive effects of several reservoir
parameters to oil and gas productivity. For the model given, transmissibility
multipliers, horizontal and vertical proved to be the key factors in manual
history match. The perturbation of these parameters to obtain the desired or
near match is a very difficult and time consuming process. It should be
performed according to fundamental principles of reservoir engineering and
practical values.
The closeness of the final match model to the historical model was an
indication that the manual history matching method used was quite
successful. It also showed that though tedious, history matching can be very
challenging, bringing The closeness of the final match model to the historical
model was an indication that the manual history matching method used was
quite successful. It also showed that though tedious, history matching can be
very challenging, bringing out the best in ones reasoning ability. Making the
several perturbations and by combinations and permutations, a fairly good
match was obtained, suitable enough for future prediction. So if the field is
small, and the reservoir historical data is very accurate, manual history
matching method is still a good tool and not obsolete as some may think due
to the availability of automatic history matching methods. But for big and
complex fields, with relatively large and highly uncertain data, manual history
matching is clearly not practicable, if time, energy, man-power and most
especially money is of the essence.
Appendix –
Following are the steps - boxes that were included
Box– No Box – Grid Coordinates MULT X MULT Y MULT Z
1 1 5 3 20 1 6 0.1 0.1 0.1
2 2 6 26 38 1 10 1.2 1.2 1.2
3 6 14 19 21 1 2 0.1 - -
4 7 13 3 13 1 12 5 5 0.8
5 8 14 3 13 1 12 5 5 0.5
6 6 14 19 21 1 2 0.3 0.3 0.3
8 14 16 10 13 7 12 - - 0.2
9 3 14 5 16 1 6 - - 0.1