You are on page 1of 23

IPA17-105-E

PROCEEDINGS, INDONESIAN PETROLEUM ASSOCIATION


Forty-First Annual Convention & Exhibition, May 2017

HANDIL DEEP ZONE GEOMODEL: HOLISTIC APPROACH TO OPTIMIZE WELLS


LOCATION AND MINIMIZE THE RISK FOR THE FUTURE DEVELOPMENT IN A MATURE,
MULTI LAYER AND HIGHLY COMPARTMENT RESERVOIRS

Rachman Phasadaon*
Gany Gunawan*
Hudan W. Alam*
Iswahyuni F. Hayati*
Sinto Yudho*

ABSTRACT comprises over 550 hydrocarbon accumulations


from depth ~200mSS to 3300mSS in structurally
Handil Deep Zone (HDZ) is the backbone of gas stacked and compartmentalized fluvio-deltaic sands.
contributor in Handil field, characterized by multi
layered, highly compartmented and low permeability The structure of the field is a simple 4 way dip
reservoir. Limited well data, risk of sand occurrence, anticline divided by a major East-West fault. Handil
depletion, and productivity are the main challenges reservoirs were divided into several zones, namely
for HDZ development. Hence, it is important to Shallow, Main, Lower, and Deep zones. Shallow
develop a robust geomodeling methods to minimize zone to Main zone are known as Oil Pool, while
these uncertainties. the bottom of Main zone down to Deep zone are
known as Gas Pool.
Sedimentological concept is essential for predicting
sand distribution orientation due to the absence of The Gas Pool are located below marker R16-5 at
reliable seismic data. Dynamic data then used ~2200mSS down to marker ~R30-4 at ~3080mSS
extensively to validate sand distribution, (Figure 1a). Reservoirs are deposited in deltaic
hydrocarbon accumulation, and reservoir environment during the Middle Miocene time.
compartmentalization. Effective porosity are varying by depth, in general
the Lower zone ranges from 13 to 22 p.u, while in the
Furthermore, understanding behavior of depletion in Deep zone ranges from 5 to 14 p.u (Figure 1b).
low-permeability HDZ is crucial. Introducing a
method to estimate pressure distribution as a function Most wells in Handil Field were drilled in the zone
of distance active producer is one of the key elements of normal hydrostatic pressure. However, one well
in HDZ reservoir pressure prediction. encountered high overpressure, H-B1. The top of
overpressure in this well is located at depth around
Burial and compaction may lead to the productivity 9500 ft (2896mSS) near to marker R30 where there
issue in HDZ. Perforation result statistic is used is an abrupt transition into very high overpressure
along with the kriging method to create map to zone, close to lithostatic gradient all the way to TD.
highlight the area based on reservoir flow probability.
Such flow prone map can be used as guidance on GEOMODELING DATA
locating the best location for drilling new wells.
Seismic Data
Using all these approaches integrated in the latest
geomodel enable us to locate the good pressure, good Since its discovery in 1974, Handil field was only
productivity and high probability of sand occurrence covered with 2D seismic data which were combined
area so-called sweet spot for the future wells from several surveys from 1973 to 1984. Sparse
location. spacing between 2D seismic lines (~600m), poor
vertical resolution and lack of reflection continuity
INTRODUCTION on 2D seismic cause poor correlation between
seismic interpretation and well data. In 1998, 3D
Handil field, discovered in 1974, is an oil and gas survey was acquired over Kerbau area located in the
field located in internal axis of modern Mahakam eastern part of Handil field. It covers approximately
Delta in East Kalimantan, Indonesia. The field 180km2 including the eastern part of Handil anticline.

* TOTAL E&P INDONESIE


A new 3D Handil seismic survey was acquired in some crevasse splays or mouth bars. Channel and
December 2011. The survey covered approximately bar/crevasse splay are identified based on the log
91.2 km2 area of Handil field with high density data pattern, approximate thickness and correlation to
sampling (bin size 12.5m x 12.5 m) which aims was surrounding wells. Fluvial channels are commonly
to improve 3D structural interpretation mainly in the indicated by a massive and thick (5-30m) sand body
Shallow zone and to get interpretable images of the with sharp base at bottom (erosional surface) while
reservoir bodies. Seismic data processing was carried the distributary channels usually have fining upward
out in 2012 for Time migration and continued in at the top, sharp base at bottom and maximum 20m
2013 for Beam PSDM processing. thick for individual sand body. Bar sands are
typically coarsening upward pattern, with relatively
Well Data poor reservoir quality compared to channel with
maximum thickness of 10m for 1 individual bar sand
Most of the old Handil wells were targeting the Main (Figure 5).
zone interval and fewer wells drilled until penetrated
the Lower and Deep zone. In general, numbers of Facies modeling aims to create a realistic
wells penetration are becoming less as the interval representative of depositional environment. The
getting deeper (Figure 2). 171 wells are used as input model should capture the heterogeneity of lithology
wells in geomodel, including the most recent wells and act as guidance to model petrophysical
drilled in 2015. properties. Sedimentologist interpreted limit of the
channel fairway based on well log correlation with
STATIC MODELING
analogue to present day Mahakam delta and the most
In general, the geomodeling workflow of Handil recent study on the sand geometry. The interpreted
Lower and Deep Zone Geomodel can be divided into boundaries then are used as basis to create sand body
several phase sequentially: structural and property, namely channel fairway, bar/crevasse splay
stratigraphic modeling; well log up-scaling; and background shale. Different parameter will be
sedimentological review and facies modeling; applied to each Sand Body property during facies
petrophysical modeling, contact and region modeling. Figure 6 illustrates the Sand Body
definition for initial and current condition, pressure property modeling on layer R21.
& flow probability modeling and finally volumetric
calculation integrating static and dynamic data. The second step of facies modeling then performed
after Sand Body properties have been established for
Structural and Stratigraphic Modeling each marker. Sequential Indicator Simulation (SIS)
algorithm was used to distribute the facies log within
One major E-W direction normal fault and two each sand body. Variogram range of channel is
seismic horizons (R19 and R28-3) were used as input 1500*500m with azimuth derived from channel
during VBM structural modeling of Lower and Deep fairway azimuth map, while for bar the range is
zone (Figure 3). 1200*1000m with orientation perpendicular to major
The upper and lower boundary of the Handil Deep direction of channel fairway (Figure 7). Vertical
Zone, consecutively marker R16-5 and R30, are Proportion Curve (VPC) was used to guide the
characterized by thick shale interval represent the 3rd algorithm to honor vertical distribution trend of each
order maximum flooding events in the middle facies proportion as seen in the well data.
Miocene. Log pattern indicated a regressive stacking
pattern from R30 up to the marker R21-5, followed Petrophysical Modeling
by transgressive pattern up to marker R16-5. Series
of 3rd order regressive-transgressive cycles are Sequential Gaussian Simulation (SGS) algorithm
observed between R30-R26, R26-R20 and R20-R16- was used to perform petrophysical modeling of NTG
5, average thickness of those cycles are between 100- and PHIE NET for Handil Lower and Deep zone.
300m. (Cibaj, I., 2009). Smaller scale sequence is Variogram range and azimuth are similar with the
defined as deltaic cycle (4th or 5th order), with one in facies modeling.
average thickness ~ 20-30m are used as input well
tops marker during structural modeling (Figure 4).
a. NTG Modeling
Facies Modeling
Net sand cut off was defined based on PHIE and
In Handil field, sedimentations are mainly channel VSH. Table 1 show the cut offs definition based on
deposits in fluvial to delta plain environments with study conducted in 2009.

Pressure test acquisition during drilling is frequently Then the initial Bg log is distributed throughout the
done in Handil wells. Cross plot of pressure test 3D grid by using Gaussian Random Function
result vs. PHIE coming from 82 wells (644 good Simulation without any specific variogram
pressure test result) that penetrate the gas pool zone differentiation for facies.
(R20 R30-5) show that minimum porosity that still
able to flow during pressure test is +5% (only 0.3% DYNAMIC DATA INTEGRATION
of data (or 2 out of 644) shows good pressure below
5% PhiE) (Figure 8). Contact and Region Definition

NTG log is calculated by using cut-off from PHIE Basically, fluid contact in Handil Lower and Deep
and VSH log. The calculation resulted in beta shape zone model can be divided into 2 categories. The first
distribution of NTG well log. Beta distribution fluid category is defined based on log data; this type
transformation was applied in data analysis to get the has the highest confidence level. When the fluid
better match between model histogram and up-scaled contact was not observed in the log data, then CGIP
data histogram (Figure 9). data is used as additional parameter when applying
minimum value between GDT +1h and Midpoint
b. Effective Porosity (PHIE NET) between GDT-WUT as fluid contact.

PHIE NET values represent the effective porosity of Since Handil reservoir are mostly associated with
the net sand body that has passed the cut offs during channel facies and within single marker/zone there
the NTG modeling. The 3D property was distributed might be more than one discrete side bar
with co-kriging to PHIE model (Figure 10). PHIE log accumulation. Region property was created as a
was used as input for porosity modeling without property in facies modeling module to distinguish
applying any cut offs, hence the value has wide any different accumulations which imply on several
ranges between ~1% to ~30%. Normal distribution fluid contacts within single zone.
transformation is used to direct the histogram
distribution in 3D grid honoring up-scaled well log Hydrocarbon region/Reservoir Pressure Unit was
data histogram distribution defined based on observation on static and dynamic
parameter. Beside well correlation, it is necessary to
c. Water Saturation (Swe) integrate hydrocarbon contact and pressure evolution
as basis to distinguish region (Figure 11). In general,
Water saturation (Swe) from log interpretation has pressure is used more extensively in defining region
high uncertainty due to variability of water salinity, while contact evolution is less used because its
being a mixture of connate water and low salinity evolution in Handil Gas Pool is not significant which
compaction water (Walgenwitz, 2003) and global indicates weak to none aquifer support.
increase of salinity vs. depth in Handil (Salze, 1987).
Therefore, Swe is defined as a function of PHIE. The Other constraint taken into account is in place
latest study in Swe function was conducted in 2016. volumetric comparison between IGIP and CGIP
Since Handil Gas Pool is dominated by gas channel comes from P/Z observation. The iterated region
reservoir then it is considered not necessary to allows us to highlight the area with high possibility
discriminate Swe function based on fluid type and of remaining potential.
facies.
Case Study: Static and Dynamic Iteration to
d. Bg at Initial Condition Define New Region in R21-5

Initial Bg log was calculated as function of reservoir In the previous model R21-5 was considered as
pressure. Two formulas are used in property single region (all reservoirs are connected). This
calculator based on the pressure value. The Bg initial layer located in Handil south compartment and has
value derived from calculation has been calibrated been produced by WELL-3 & WELL-2, where
with PVT data base. WELL-3 located in the top of structure while WELL-
2 located in the more down-dip position.
If Pressure < 2890 Psi Bgi = 1000 * 7.943 *
(Psi^-0.917) / 1000 WELL-2 died in 2000 due to integrity issue, while
WELL-3 died in 2010 due to water out. Consider this
If Pressure >= 2890 Psi Bgi = 1000 * 0.218 / as single region reservoir (based on old model), it
(Psi^0.466) / 1000 means no remaining potential left due to water

contact move up to top structure. But based on new analysis formula (Exponential Integral) into model
approach where dynamic-static iteration is the key to the predicted pressure value in the nearby wells as a
redefine new region, and one of those method is function of distance can be calculated (Figure 15).
using pressure to define compartmentalization inside The calculated pressure then will be distributed
the reservoir, its confirm in reservoir A consist of 2 throughout region using kriging algorithm to create
regions instead of 1 region, region 17 & region 18. region pressure map (Figure 16).
The dynamic input, in this case P/Z method can be
used also to control shape and volume of the new Productivity Index (PI)
channel/region, this is very useful in specific case
where static data is limited (Figure 12). Productivity Index is introduced to give an idea
where the area with high possibility to have good or
New region definition, may open new possibility for flowing perforation result based on previous wells
future development in reservoir-A (region 18), where perforation. The index value used for mapping is 1
the actual contact remain as initial and remaining for flowing reservoir and zero for non flowing
potential still economic to be produced with current reservoir. Definition of flowing reservoir is reservoir
technology. that has initial gain after perforation >= 0.5MMscfd,
the cut off 0.5 MMscfd is derived from pre-frac
Integration between static and dynamic data in successful realization, 4 hydraulic fracturing wells
geomodel allows faster screening and evaluation to have pre-frac initial rate above the cut off. Index
locate sweet spot area of remaining potential not only values are distributed inside the HC accumulation
in term of net pay but also reservoir pressure and limit by using make surface tool in petrel with
productivity index. Two dynamic driven maps kriging method, PHIE-NET map use as trend surface
therefore are introduced in this new model. in pre-processing tab. Index values are distributed
inside the HC accumulation limit by using make
Region is treated as facies once the limit is set by surface tool in petrel with kriging method, PHIE-
analyzing the numbers of different initial contacts NET map use as trend surface in pre-processing tab.
found in one marker. The iteration between dynamic Below is the example of Productivity index map
and static is performed to confirm that the limit is set generated in R28. R28 have been perforated by 5
correctly in term of the volumetric. After the wells which the initial gain as shown in the table 2
boundary is created in 2D map and value is given for below.
each boundary, the region properties are then created
in facies modeling step by using assign value from Index value is identified as such then distributed
surface option (Figure 13). inside the region boundary to create the PI Map
(Figure 17)
Reservoir Pressure Model
RESULTS
Pressure Map is introduced to visualize the risk of
highly depleted pressure regime (< 0.4 Deq) in each The blind test was performed on 6 HDZ wells drilled
region that cannot be produced anymore. Remaining from 2012 to 2015: where 5 wells located in North
potential is defined based on current contact depth compartment and only one well in the South
and current reservoir pressure >= 0.4 Deq. compartment. The blind test is divided into 2
categories:
The example below describes how to evaluate risk of
depletion in layer 27-5 region 1. WELL-4 is main Static parameter consists of net sand and net pay.
producer in this region where its production impacts
on region depletion is found at several locations, Dynamic parameter consists of pressure and
observed by recent drilled wells (WELL-6, WELL-5 productivity index.
and WELL-7) after the well stop producing, no
contact movement is detected because of low water Static Parameters Blind Test
production which proves typical dry gas behaviour
(Figure 14). The depletion rate can be seen over a. Net Sand Prediction
distance at observer wells from the source of
depletion, the further distance to depletion the less Total actual net sand thickness is only 0.92% (-3.3m)
depletion occurred exception for WELL-7 where is less than prediction (351.6m actual vs. 354.9m
situated at isolated mouthbar that has poor model). More variability is observed if we look in
transmissibility. By applying pressure transient more detail, 75% of net sand prediction > 5m falls

within the + 40% tolerance, while for net sand reservoir from lower zone and deep zone By using
prediction < 5m only 38% prediction falls within the 2014-2015 HDZ wells as blind test data, it means
tolerance (Figure 18). The result suggests that in latest wells that have been used to built the pressure
general the model predict better in thick sand (>5m). model in this test is 2013 HDZ wells. Despite
excluding all new data (2014/2015), pressure
Sensitivity analysis was conducted to evaluate the model/methodology still predictive to estimate
impact of variogram for net sand prediction. Three reservoir pressure, with absolute deviation 0.1 Deq
different cases were evaluated (Figure 19): (Figure 21)

Case-1: current variogram is multiplied by 0.5 b. Productivity Index (PI)


(bar: 600*500 & Channel: 750*250).
There are 15 perforation results data that have Gp
Case-2: current variogram (bar: 1200*1000 & realization >1 Bscf that are used as blind test data,
Channel: 1500*500). covering all layer from R20-2 - R28-8, with data
distribution as below:
Case-3: current variogram is multiplied by 1.5
(bar: 1800*1500 & Channel: 2250*750). Category-1: 7 perforations coming from reservoir
with limited perforation data, which mean by
The realization from those different cases shows that excluding 1 perforation data as blind test data makes
there is no significant impact in term of net sand no perforation data left can be used to develop PI
prediction, indicating that the model variogram is not map. As a result, PI map will be based on PI function
sensitive to net sand predictions. Sedimentology from closest marker/layer.
concept seems to play much bigger role in
controlling the accuracy in predicting net sand Category-2: 8 perforations coming from reservoir
thickness. with sufficient perforation data, which mean PI map
can be generated based on perforation result & PI
b. Net Pay Prediction function in those reservoirs.

Actual net pay was calculated from reservoir defined The result showing PI value that has been built using
as primary target and un-prognosed reservoir. The PI map methodology give slightly pessimistic value,
definition for un-prognosed reservoir is the reservoir with relative deviation 40%. Poor PI prediction
that in the previous prediction based on 2012 model mostly comes from Category-1 (reservoir with
was interpreted as Non-gas reservoir or even non- limited perforation data). Having sufficient
reservoir but in the actual result it is found as Gas perforation data in each reservoir is the key to built
reservoir. Therefore the new model was updated by predictive PI Map (Figure 22).
using this latest result, however in the blind test this
data will be taken out. CONCLUSIONS

Total actual net pay thickness is 34% (-39.2m) less New geomodeling methodology in Handil Deep
than prediction (110.8m actual vs. 170m model). ~ Zone (HDZ) has enable screening to identify area
72% of model prediction falls within the prediction that has potential for future development wells in
tolerance + 40%. Model prediction that falls beyond term of pressure, productivity and probability of sand
tolerance are mostly related to reservoir occurrences.
heterogeneity inside channel fairway which
uncertainties cannot be addressed with current The geomodel is quite predictive as it is validated by
modeling methodology (Figure 20). blind test result that showing good coherency
between the prediction and actual well result for both
Dynamic Parameters Blind Test dynamic and static parameters.

a. Reservoir Pressure Model REFERENCES

There are 21 MDT data (pressure measurement) in Cibaj, I. et al., 2007, Stratigraphic Interpretation of
HDZ wells drilled in 2014-2015 that are used as Middle Miocene Mahakam Delta, Deposits :
blind test data, covering both gas and water bearing Implication for Reservoir Distribution and Quality,

Proceedings of Indonesia Petroleum Association. Wirawan, G., & Hidayat, H. K., 2016, Porosity-
31st Annual Convention, Jakarta. Permeability Relationship and Swe-Phie Function
for Handil Lower-Deep Zone, Internal Memo.
Kristianto, A., 2009, Petrophysical Synthesis of Wiweko, A., 2014, Cartoon of Handil Deposition
Handil Shallow and Deep Zone, Internal Memo. Reconstruction Proccess, Internal Report.

TABLE 1

DETAIL CUT-OFF OF PHIE AND VSH FOR SAND TYPE-A, B AND C

Zone Porosity(v/v) Vshale(v/v)


(TVDSS) A B C A B C
0700 0.2 0.15 0.07 0.6 0.6 0.6
7001200 0.2 0.14 0.07 0.6 0.6 0.6
12001700 0.2 0.14 0.05 0.6 0.6 0.6
17002200 0.18 0.13 0.04 0.25 0.35 0.45
22002700 0.14 0.1 0.05 0.45 0.45 0.45
27006000 0.11 0.07 0.05 0.45 0.45 0.45
=LowerandDeepZone

TABLE 2

INITIAL GAIN DATA FROM FIVE WELLS THAT WAS PERFORATED IN LAYER R28, INDEX
VALUE 1 FOR FLOWING RESERVOIR AND 0 FOR NON-FLOWING RESERVOIR

Well InitialGain Status IndexValue


Well8 0.6 NotTight 1
Well9 0.48 Tight 0
Well10 1.59 NotTight 1
Well11 0.27 Tight 0
Well12 0.1 Tight 0

Figure 1 - a. Handil Reservoir Zonation; b. PhiE Distribution of Handil Gas Pool Reservoir.

Figure 2 - Well Distribution of Handil Gas Pool.

Figure 3 - Handil structural framework.

Figure 4 - Example of Detail Stratigraphic Sequence and Marker In Handil Lower And Deep Zone (after Cibaj, I., 2007).

Figure 5 - Example of Several Facies That Commonly Found in Handil Field (modified from Wiweko A.,
2014).

Figure 6 - Sand Body Modeling Input and Result.

Figure 7 - Facies Modeling Input, Parameters Setup In Petrel and Result.

Figure 8 - Handil Gas Pool Phie vs. Good Pressure Test Result.

Figure 9 - NTG Modeling Input, Parameters Setup in Petrel and Result.

Figure 10 - PhiE NET Modeling Input, Parameters Setup and Result.

Figure 11 - Region Definition Example in R21-5, Region Can Be Identified By Looking at the Initial Contact
Variation in Each Marker.

Figure 12 - Dynamic Input by Using P/Z Method is Very Useful to Identify New Region in Marker R21-5
and Validate the Deterministic Channel Geometry in the New Region.

Figure 13 - Region Properties Modeling Input, Parameters Setup In Petrel and Result.

Figure 14 - Layer R27-5 Perforation Result and Pressure Data after WELL-4 Production.

Figure 15 - Illustration of Reservoir Pressure Model at Layer R27-5 Region 1.

Figure 16 - Applying Pressure Cut-Off into the Initial Netpay Map to Define the Remaining Area above Pressure Cut-Off.

Figure 17 - Productivity Index Mapping Process.

BlindtestNSActualvsModel 360

30

25 354.85
355
ActualThickness(m)

20
40%
100%
351.58
15
+40%
PrimaryTarget
10
350
UnprognoseTarget
NonReservoir
Model prediction
5 < 5m zone
Model prediction
>= 5m zone
0
0 5 10 15 20 25 30 345
ModelThickness(m) Actual Model

Figure 18 - Cross Plot Net Sand Thickness Model Vs. Actual.

Figure 19 - Cross Plot between Net Sand Model Predictions vs. Actual Thickness for Each Case.

Figure 20 - Cross Plot Between Net Pay Predictions Vs. Actual, b. Example of Model Uncertainties Related
to Channel Side Bar Limit.

BlindTestPressure
HSA1_LayerA
Well-12
1
Well-9
HD4LayerB
0.9
0.8
Well-35
HV4LayerC
0.7
DeqActual

0.6
0.5
0.4
+0.1deq
0.3
0.2
0.1
0.1deq
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
completecircle ofprediction
DeqPrediction

*complete circle prediction: DEQ, PI and reserves blind test data

Figure 21 - Cross Plot Between Deq actual (2014/15 HDZ wells) vs. Deq Prediction.

Figure 22 - Cross Plot Between PI actual vs. PI Prediction.

You might also like