Vous êtes sur la page 1sur 8

HISTORY MATCHING

EXERCISE
3 CASES COMPARISON OF RESERVOIR
RELATIVE PERMEABILITIES

PRIMA ADHI SURYA


PETROLEUM ENGINEERING
PERTAMINA UNIVERSITY
101316114
INTRODUCTION

This paper reports an attempt to match production result in simulation against real production
histories of a field. In the petroleum industry, this is commonly called as history matching. It is usually done to
forecast future productions to decide actions that needs to be taken to maximize the profitability of the field.
In this case, the objective that the students need to achieve is to look for simulation result that has error of
less than 10% (<10%) between the real field data. The student will report 3 cases of production simulation
with different relative permeability models. The method to match production simulation with real production
data will be done through trial and error. And finally, the simulation result is discussed.

THE SIMULATED FIELD

The field model has been prebuilt by the lecturer. It is a field A with 2536 m x 4271 m x 174 m in size
and gridded to corner-point type 50x50x4 (10.000 blocks total) for the simulation. The field model starts at
3161 m depth from the top. The shape is a dual anticlines field with slightly concave bottom. The field is a 2-
phase flow system (oil & water) that has radial porosity distribution on each anticline top, with stronger mean
porosity values in one of them. The field permeability from layer 1 to 4 (top-to-bottom) is 200 mD, 100 mD,
50 mD, and 200 mD uniformly distributed for X and Y direction and 40 mD, 10 mD, 20 mD, and 40 mD
respectively for Z direction.

Initial condition shows that the field pressure is 27000 kPa (3916 psi) at 3050 m with bubble point
pressure at 6500 kPa (942,75 psi). The oil-water contact of the field happens at 3080 m below surface.

Picture 1. Left picture shows permeability values from layer 1 to 4 in field A. Each layer has uniform distribution of 5 times its Z
permeability in X and Y direction. The second picture shows field A 3D view with radial porosity uniform to Z direction and higher
mean of radial porosity distribution in one of the anticlines.

THE SIMULATION

In attempting to match production history, the variable that will be controlled for simulation is relative
permeabilities of the rock-fluid interactions. The simulation uses IMEX in CMG Suite 2015.10. It is an adaptive
Implicit-Explicit simulator suited for black oil and unconventional reservoirs. Although the model selects 3-
flow systems for the simulation, but since Oil-Gas contact happens way beyond the model (at 1.980 m, less
than 3.161 m where the top grid starts at), the Gas-Liquid relative permeability has no effect and will not be
taken into account for the relative permeability adjustment.
3 CASES OF RELATIVE PERMEABILITIES

The student has used the following 3 relative permeability cases that has been tested and matched
quite closely with the real production data. The initial relative permeability model is also shown prior to
modifications.

Picture 2. Oil-Water Relative Permeability Model of Each Cases.

The correlation value are also provided and some values are not shown since it is used to simulate
when gas is taken into account. The generalized equation used to generate them are shown on equation (1)
and (2).
𝑁𝑁𝑤𝑤 𝑁𝑁𝑜𝑜𝑜𝑜
𝑆𝑆𝑤𝑤 − 𝑆𝑆𝑤𝑤 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑆𝑆𝑜𝑜 − 𝑆𝑆𝑜𝑜 𝑟𝑟
𝐾𝐾𝑟𝑟𝑤𝑤 = 𝐾𝐾𝑟𝑟𝑤𝑤 𝑖𝑖𝑖𝑖𝑖𝑖 ∗ � � … (1), 𝐾𝐾𝑟𝑟𝑜𝑜𝑜𝑜 = 𝐾𝐾𝑟𝑟𝑜𝑜 𝑐𝑐𝑐𝑐 ∗ � � … (2)
1 − 𝑆𝑆𝑤𝑤 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 − 𝑆𝑆𝑜𝑜 𝑖𝑖𝑖𝑖 1 − 𝑆𝑆𝑤𝑤 𝑐𝑐𝑐𝑐𝑐𝑐 − 𝑆𝑆𝑜𝑜 𝑟𝑟

where:

Krw = Water relative permeability Krow = Oil relative Permeability


Krwiro = Krw at irreducible oil saturation Krocw = Kro at connate water saturation
Sw = Water saturation So = Oil saturation
Swcrit = Critical Water Saturation Sor = Residual Oil Saturation
Soir = Irreducible Oil Saturation Swcon = Connate Water Saturation
Case SWCON SWCRIT SOIRW SORW KROCW KRWIRO KRWIRO Exp KROCW Exp
0 0,21 0,21 0.3 0,3 0,9 0,5 1 1
1 0,26 0,267 0,289 0,362 0,905 0,54 1,57 0,83
2 0,15 0,493 0,31 0,31 0,9 0,54 4 1
3 0,4 0,47 0,27 0,27 0,87 0,5 1,3 0,8
Table 1. Correlation values used to generate relative permeability of oil-water interactions.

SIMULATION RESULTS AND ERROR EVALUATION

The simulation results are provided both in summary table and graph forms. Estimation of simulation
error uses average error method. Evaluating simulation accuracy for Water Cut Percentage and Oil Production
Rate uses 3 types of errors: Population Mean Error (PME) evaluates the difference of average-over-time water
cut and oil rate data, Samples Error Average (SEA) evaluates simulation precision by looking at the error
average of each time steps, and Cumulative Error (CE) evaluates the accuracy of the simulation by looking at
all time ranges. The plus-minus notation in SEA tells the tendency of the simulation result in a timestep to
underpredicting (-) or overpredicting (+). The result shows that Case 1 is the closest-matched simulation on
average than Case 2 and 3 with net average error of 0,15% PME and 104,125% CE. This suggests that with the
present historical data, Case 1 has closer rock-fluid relationship with field A and can be used for future reservoir
evaluation.

Case 2 is the least matched simulation with 1,455% PME Average and 179,705% CE. Case 3 stands the
middle ground between Case 1 and Case 2. While Case 3 is less Accurate and Precise simulation than Case 1,
it can still be used for alternative evaluation of the reservoir.

Population Average Samples Error


Population Mean Error Cumulative Error
Case Value Average
WC (%) QO (m3/d) WC (%) Qoil (%) Avg (%) WC (%) Qoil (%) WC (%) Qoil (%) Avg (%)
History 49,12 1521,64 - - - - - - - -
1 49,20 1523,75 0.16 0,14 0,150 +0,21 +0,15 119,83 88,42 104,125
2 48,47 1545,86 1,32 1,59 1,455 -1,30 +1,59 169,72 189,69 179,705
3 48,67 1539,20 0.93 1,15 1,040 -0,90 +1,16 145,73 139,93 142,830
Table 2. Error Evaluation of Water Cut Percentage and Oil Rate.

Evaluating the Cumulative Production Error can be done by looking at the present cumulative
production of each cases and then comparing the difference between simulation and real historical data.
Evaluating this is much simpler since Cumulative Production is the end product of Oil Rate and Water Cut
Production.

Case Present Cumulative Production (m3) Error (%)


History 2.781.073,25 -
1 2.784.334,00 0,12
2 2.825.124,00 1,58
3 2.811.753,25 1,10
Table 3. Error Evaluation of Cumulative Oil Production.

Graph of the Simulation Result of each cases are provided below. In Water Cut and Oil Production
plots, case 1 is isolated from case 2 and 3 curves to give better view on how case 1 is simulating the Water Cut
and Oil Production since it is the best matched simulation.
Picture 3. Graph Plot of Simulation Cases.
Picture 4. History matching of Oil Cumulative Production.
MATCHING VS FORECASTING

In the real practice, the matchings done to simulation results are used to forecast the future
production pattern. The common principle when forecasting a production pattern, however, is that good
matching does not equal good forecasting. It often happens to a reservoir engineer that has a good match
with production data ends up with entirely off-mark forecasts. So to make sure that a good matching has a
good forecasting pattern, we also need to look for how real production data may behaves in the near future.
In this case, we simply take a look at how the graph of each cases may behave when forecasted two years to
the future.

Due to limitations of this practice (no uncertainty analysis, no sensitivity analysis), The author simply
had taken an observational approach to look at the forecasts at every category and every cases, and it is
observed that Case 1 doesn’t forecast a good simulation result on Water Cut Percentage. Taking a look at the
production history trend, in fact, tells that it is off the mark by about 25% in the end of the forecasts. So Case
1 may certainly a good match result, but it is not a good forecast result since it is off mark by a lot. Case 3 is
quite off too so we can say that it is not a good forecast result either. Case 2, however, shows the most
minimum deviation after forecasting. So while Case 2 may not have best matching result in history matching
case, it can be relied upon assuming that water cut slowly reduce the rate of decline.

Oil Rate forecasts shows that all rate decline rapidly the moment they go beyond past real production
data timeline. While the author may not be able to determine which case has good forecasts result on the
graph. The author may be able to explain why the sudden drop after year 1996 February, which will be
discussed in the next subtopic.
Picture 5. Forecasts of 3 Cases Simulation for Water Cut (Above) and Oil Rate (Bottom)
Picture 7. Forecasts of Oil Cumulative Production
The author also observed an early
decline in case 3 cumulative production due
to the early drops in oil rate plot. This results
in Case 1 and Case 2 having more oil
production in the future.

THE SUDDEN DROP OF OIL RATE

The author takes a look at the


simulation results and found that most likely
cause of sudden drop in oil saturation in the
field is some early water breakthrough.
However, this is still quite uncertain and
needs to be tested further.

Picture 6. Oil Saturation Changes over Time

CONCLUSION

In this paper the author laid out his matching attempts to not only have a good matching model, but
also a good forecasting model. While Case 1 is certainly the most accurate and precise model, Case 3 can be
used reliably to predict future production pattern, although at the cost of accuracy and precision in present
history matching. The author suggests to conduct Sensitivity Analysis and Uncertainty Analysis post-
forecasting to see true reliable model for simulation forecasts.

Vous aimerez peut-être aussi