Similar documents
Identification of natural fractures and in situ stress at Rantau Dedap geothermal field

GEOMECHANICALLY COUPLED SIMULATION OF FLOW IN FRACTURED RESERVOIRS

Geological Evaluation of the Waringin Formation as the host of a Vapor-Dominated Geothermal Reservoir at the Wayang Windu Geothermal Field

IN SITU STRESS, FRACTURE AND FLUID FLOW ANALYSIS EAST FLANK OF THE COSO GEOTHERMAL FIELD

Analysis of stress variations with depth in the Permian Basin Spraberry/Dean/Wolfcamp Shale

J.V. Herwanger* (Ikon Science), A. Bottrill (Ikon Science) & P. Popov (Ikon Science)

Simplified In-Situ Stress Properties in Fractured Reservoir Models. Tim Wynn AGR-TRACS

Integration of Uncertain Subsurface Information For Reservoir Modeling. Ali Ashat Institut Teknologi Bandung Wednesday, March 5, 2013

Modeling pressure response into a fractured zone of Precambrian basement to understand deep induced-earthquake hypocenters from shallow injection

Reservoir Geomechanics and Faults

DISCRETE FRACTURE NETWORK MODELLING OF HYDRAULIC FRACTURING IN A STRUCTURALLY CONTROLLED AREA OF THE MONTNEY FORMATION, BC

Geomechanical controls on fault and fracture distribution with application to structural permeability and hydraulic stimulation

4D stress sensitivity of dry rock frame moduli: constraints from geomechanical integration

Practical Geomechanics

Overview of Microseismic in Sustainable Monitoring of Geothermal Reservoirs in Indonesia*

Critical Borehole Orientations Rock Mechanics Aspects

WAMUNYU EDWARD MUREITHI I13/2358/2007

Optimising Resource Plays An integrated GeoPrediction Approach

Applicability of GEOFRAC to model a geothermal reservoir: a case study

DETAILED IMAGE OF FRACTURES ACTIVATED BY A FLUID INJECTION IN A PRODUCING INDONESIAN GEOTHERMAL FIELD

Role of lithological layering on spatial variation of natural and induced fractures in hydraulic fracture stimulation

FRACTURE REORIENTATION IN HORIZONTAL WELL WITH MULTISTAGE HYDRAULIC FRACTURING

Discrete Element Modeling of Thermo-Hydro-Mechanical Coupling in Enhanced Geothermal Reservoirs

Summary. Introduction

Gas Shale Hydraulic Fracturing, Enhancement. Ahmad Ghassemi

ractical Geomechanics for Oil & Gas Industry

Heat (& Mass) Transfer. conceptual models of heat transfer. large scale controls on fluid movement. distribution of vapor-saturated conditions

Tu D Understanding the Interplay of Fractures, Stresses & Facies in Unconventional Reservoirs - Case Study from Chad Granites

OVERVIEW OF THE WAIRAKEI-TAUHARA SUBSIDENCE INVESTIGATION PROGRAM

Strength, creep and frictional properties of gas shale reservoir rocks

AADE 01-NC-HO-43. result in uncertainties in predictions of the collapse and fracture pressures.

GeothermEx, Inc. GEOTHERMAL RESERVOIR ASSESSMENT METHODOLOGY FOR THE SCIENTIFIC OBSERVATION HOLE PROGRAM, KILAUEA EAST RIFT ZONE, HAWAII TASK 1 REPORT

An Investigation on the Effects of Different Stress Regimes on the Magnitude Distribution of Induced Seismic Events

URTeC: Abstract

Overview of Indonesian Geothermal System

Geologic Considerations of Shallow SAGD Caprock; Seal Capacity, Seal Geometry and Seal Integrity, Athabasca Oilsands, Alberta Canada

Background. Developing a FracMan DFN Model. Fractures, FracMan and Fragmentation Applications of DFN Models to Block & Panel Caving

11th Biennial International Conference & Exposition. Key Words: Geomechanical, Strain, Discrete Fracture Network (DFN), Slip stability

SPE DISTINGUISHED LECTURER SERIES is funded principally through a grant of the SPE FOUNDATION

HIGH TEMPERATURE HYDROTHERMAL ALTERATION IN ACTIVE GEOTHERMAL SYSTEMS A CASE STUDY OF OLKARIA DOMES

Effect Of The In-Situ Stress Field On Casing Failure *

Abstracts ESG Solutions

Geomechanics for reservoir and beyond Examples of faults impact on fluid migration. Laurent Langhi Team Leader August 2014

Evaluation of Structural Geology of Jabal Omar

Introduction and Background

Focal Mechanism Analysis of a Multi-lateral Completion in the Horn River Basin

Geothermal Application of Borehole Logging in New Zealand

3D Finite Element Modeling of fault-slip triggering caused by porepressure

NOTICE CONCERNING COPYRIGHT RESTRICTIONS

An Enhanced Geothermal System at Coso, California Recent Accomplishments

QUANTITATIVE INTERPRETATION

Colleen Barton, PhD Senior Technical Advisor Baker Hughes RDS GMI. HADES - Hotter And Deeper Exploration Science Workshop

Geothermal Systems: Geologic Origins of a Vast Energy Resource

Microseismic Geomechanical Modelling of Asymmetric Upper Montney Hydraulic Fractures

PROCEEDINGS, INDONESIAN PETROLEUM ASSOCIATION Thirty-Ninth Annual Convention and Exhibition, May 2015

Iwan Yandika Sihotang, Tommy Hendriansyah, Nanang Dwi Ardi

Production-induced stress change in and above a reservoir pierced by two salt domes: A geomechanical model and its applications

Brittle Deformation. Earth Structure (2 nd Edition), 2004 W.W. Norton & Co, New York Slide show by Ben van der Pluijm

Fault Reactivation Predictions: Why Getting the In-situ Stresses Right Matters

Activity Submitted by Tim Schroeder, Bennington College,

Yusuke Mukuhira. Integration of Induced Seismicity and Geomechanics For Better Understanding of Reservoir Physics

The San Andreas Fault Observatory at Depth: Recent Site Characterization Studies and the 2.2-Km-Deep Pilot Hole

PETROLEUM GEOSCIENCES GEOLOGY OR GEOPHYSICS MAJOR

Constrained Fault Construction

Earthquake and Volcano Clustering at Mono Basin (California)

Induced microseismic fracture prediction

Geomechanics, Anisotropy and LMR

GEOTHERMAL WELL TARGET APPROACHES IN THE EXPLORATION STAGE

Egbert Jolie 1, James Faulds 2, Inga Moeck 1.

Exploration of Geothermal High Enthalpy Resources using Magnetotellurics an Example from Chile

Hijiori HDR Reservoir Evaluation by Micro-Earthquake Observation

NUMERICAL MODELING STUDY OF SIBAYAK GEOTHERMAL RESERVOIR, NORTH SUMATRA, INDONESIA

Numerical Simulation of Devolution and Evolution of Steam-Water Two-Phase Zone in a Fractured Geothermal Reservoir at Ogiri, Japan

Tensor character of pore pressure/stress coupling in reservoir depletion and injection

A fresh look at Wellbore Stability Analysis to Sustainable Development of Natural Resources: Issues and Opportunities

Integration of Surface and Well Data to Determine Structural Controls on Permeability at Salak (Awibengkok), Indonesia

URTeC: Summary

Mapping the Preferential Flow Paths within a Fractured Reservoir

Search and Discovery Article # (2015) Posted April 20, 2015

Chapter 6. Conclusions. 6.1 Conclusions and perspectives

Targeting of Potential Geothermal Resources in the Great Basin from Regional Relationships between Geodetic Strain and Geological Structures

The Deep Fault Drilling Project, Alpine Fault Getting Inside the Earthquake Machine

Activity Pacific Northwest Tectonic Block Model

Comprehensive Wellbore Stability Analysis Utilizing Quantitative Risk Assessment

Earthquakes. Earthquakes are caused by a sudden release of energy

Instituto De Ingenieros De Minas Del Peru

INFLUENCE OF LOCAL PERTURBATION ON REGIONAL STRESS AND ITS IMPACT ON THE DESIGN OF MAJOR UNDERGROUND STRUCTURE IN HYDROELECTRIC PROJECT

PRESSURE PERTURBATIONS IN TWO PHASE GEOTHERMAL RESERVOIR ASSOCIATED WITH SEISMISITY

Taller de Geotermica en Mexico Geothermal Energy Current Technologies

Integrated Geophysical Model for Suswa Geothermal Prospect using Resistivity, Seismics and Gravity Survey Data in Kenya

Microseismicity applications in hydraulic fracturing monitoring

Quantitative Interpretation

Microearthquake (MEQ) Investigation Reveals the Sumatran Fault System in Hululais Geothermal Field, Bengkulu, Indonesia

Kinematic inversion of pre-existing faults by wastewater injection-related induced seismicity: the Val d Agri oil field case study (Italy)

IN-SITU STRESS ESTIMATION IN OFFSHORE EASTERN MEDITERRANEAN WITH FINITE ELEMENT ANALYSIS

Determination of Calcite Scaling Potential in OW-903 and OW-914 of the Olkaria Domes field, Kenya

RESERVOIR-SCALE FRACTURE PERMEABILITY IN THE DIXIE VALLEY, NEVADA, GEOTHERMAL FIELD

Automatic Moment Tensor Analyses, In-Situ Stress Estimation and Temporal Stress Changes at The Geysers EGS Demonstration Project

Instantaneous Spectral Analysis Applied to Reservoir Imaging and Producibility Characterization

A circular tunnel in a Mohr-Coulomb medium with an overlying fault

Transcription:

PROCEEDINGS, Fourtieth Workshop on Geothermal Reservoir Engineering Stanford University, Stanford, California, January 26-28, 2015 SGP-TR-204 Structural Permeability Assessment Using Geological Structural Model Integrated with 3D Geomechanical Study and Discrete Facture Network Model in Wayang Windu Geothermal Field, West Java, Indonesia Asrizal Masri 1, Colleen Barton 2, Lee Hartley 3, Yuris Ramadhan 1 1 Star Energy Geothermal Indonesia, 2 Baker Hughes Reservoir Development Services, 3 Amec Foster Wheeler Star Energy Geothermal Indonesia, Wisma Barito Pacific, Star Energy Tower, 11 th flr. Jl. Let. Jend. S. Parman Kav.62-63, Jakarta 1410, Indonesia Asrizal.Masri@starenergy.co.id, Colleen.Barton@bakerhughes.com, Lee.Hartley@amecfw.com, Yuris.Ramadhan@starenergy.co.id Keywords: Wayang Windu, Geothermal, Permeability, Critically Stressed Fractures ABSTRACT This paper presents the application of a structural permeability assessment using a Geological Structural Model integrated with a 3D Geomechanical Study and Discrete Fracture Network (DFN) Model in order to obtain a better understanding of the subsurface permeability distribution in the Wayang Windu geothermal reservoir. In this study, we assess critically stressed, hydraulically conductive fractures that control permeability and fluid flow through the interaction between the in situ stress state and the character and orientation of fractures or discontinuities. A multidisciplinary data set has been used comprising geology (rock types, fault structures, alteration, thermal manifestation, etc.), geophysics (MT, gravity, MEQ, image logging data, etc.), geochemistry (chemistry trends, tracer test results, etc.), production (temperature, pressure and spinner data), geomechanical properties and drilling information to construct a geological structure model, to analyze the in-situ state of stress (geomechanical study) and to develop a DFN model as well as deterministic hydro-structure planes interpreted from MEQ events constrained by stress and permeable fractures from the wells. The model results have been integrated to assess the potential structures and fracture permeability zones that may associate with fluid pathways. Over the reservoir structure, the predominant stress sensitive fracture intensities appear to align to form permeable fracture corridors or hydro-structures. The study results will be utilized to guide development and to improve output from the Wayang Windu Field through a combination of new well drilling and treatment initiatives of select wells. The study workflow is proposed to become the standard process to assess subsurface structure permeability distribution. 1. INTRODUCTION The Wayang Windu Geothermal Field is situated within the Southern Volcanic Complex in Java Island, located 40 km south of Bandung and 190 km south of Jakarta Indonesia (see Figure 1). The field is surrounded by high topography geothermal systems such as the Kamojang, Darajat, Patuha, Salak and Karaha Bodas fields. The Wayang Windu Field has been in operation since 2000 with a total installed capacity currently at 227 MWe and it generates the electricity supply for the Java-Bali grid system. The power plant is supported by 22 production wells and 3 injection wells. Java Island lies on the convergent tectonic margin between the Indian-Australian plate and the southeastern margin of the Eurasian Continental plate called the Sundaland (Hamilton, 1979, Setiadji, 2010). The regional tectonic setting has been of a volcanic island arc and fore arc basin complex since the early Tertiary age. The Quaternary age Wayang Windu volcanic complex in West Java, is a high temperature geothermal system associated with the active volcanic arc. The field is dominated by major WNW, NW, and NE faults and fracture trends. Consistent with regional tectonic setting of Java Island, the present-day stress orientation in the southern part of the island is thought to be NNE to NE, relatively perpendicular to the plate boundary inboard of subduction zone (Tingay et al, 2010). The Wayang Windu Geothermal Field is interpreted to be transitional between vapor dominated and liquid dominated systems (Bogie, 2008). It is assumed that fluids are stored within the porosity in the rock matrix, while fracture and or fault permeability (and not the inherent connectivity between pore voids in the rocks matrix) provides conduits for fluids to move in the reservoir. As indicated in many cases of fracture dominated geothermal reservoirs, permeability and fluid flow depends on an interaction between in situ stress state and the character and orientation of the fractures or discontinuities. Fractures which are critically stressed under ambient stress conditions are assumed to be hydraulically conductive and control permeability and fluid flow (Barton et al., 1996). Therefore, assessment of the fractures distribution and their relative stress sensitivity can be of significant benefit to understanding reservoir performance and exploiting natural fractures to enhance production. 2. METHODOLOGY The methodology to assess the subsurface permeability distribution of the Wayang Windu field integrates three field-scale models derived from a broad multidisciplinary set of data. A geologic model provides the structural framework, lithology and reservoir 1

N Jakarta Salak Bandung Legend Wayang Windu Geothermal Field (WW) Wayang Windu (WW) Patuha Kamojang Darajat Others Geothermal Field Travel Distances To WW From Bandung ± 40 Km From Jakarta ± 190 Km Figure 1. (Left) Wayang Windu Structure Map based on the integration of surface and subsurface data (right); topographic relief from the regional Digital Elevation Model. Red star represents the Wayang Windu Geothermal Field location. properties, a 3D geomechanical model the magnitudes and orientations of reservoir stresses, pore pressure evolution and rock mechanical properties, and a discrete fracture network (DFN) model, fracture characterization and distribution. These interrelated models are coupled to spatially assess the structural fracture permeability of the reservoir and delineate potential areas that may be associated with fluid pathways to optimally locate future well targets. 2.1. Structural Assessment and Geologic Modeling The Wayang Windu field is located in an active tectonic zone where major structural faults are dominated by WNW, NW, and NE trends. As the result of combination between regional and local (volcanic) activities (Figure 1). The surface structural analysis is interpreted from image data including Landsat, a Digital Elevation Model (DEM) and aerial photos and then validated using data and information from surface geological structure mapping i.e. slickensides, fault gouge and brecciation, and the occurrence of thermal manifestations such as fumaroles and hot springs that could represent fluid pathways from the reservoir. To interpret and characterize the subsurface geometry, the surface structural information is integrated with subsurface structural data such as permeable zones from PTS surveys and drilling information (losses, steam kicks, and other drilling parameters) and with fracture trends measured from borehole images recorded in 16 wells. Confidence in the location and orientation of observed surface fault trends that correlate with the depth and trend of subsurface permeable zones is higher than that of surface mapped faults without supporting subsurface data. Understanding the geologic framework of a geothermal reservoir, especially the rock type distribution and fault structure are important as they define the porosity in the rock matrix (Masri et al, 2006). An integrated geologic structural model was developed to understand the distribution of rock types using a stochastic approach within the reservoir and to provide a common framework for geomechanical and DFN modeling. Cores from 4 deep core holes, cuttings from 38 production wells, and the analysis of 16 borehole image logs, are used to develop a consistent rock definition. To correlate the lithology distribution vertically and spatially from well to well and beyond the wells, the facies concept of volcano-stratigraphy was applied where rocks are grouped into packets of similar origin facies of a structurally undisturbed andesitic stratovolcano (Bogie and Mackenzie, 1998).The rock units encountered in the Wayang Windu field are classified into 4 facies: Central Proximal facies consisting of Lava and Breccias, Proximal Medial facies consisting of Breccias and Tuff Breccias, and Medial Distal facies consisting of Lapilli and Tuffs (Figure 2). To validate the geologic model facies correlation within the same layers, XRF analyses were reviewed and compared with the rock type and with the evolutionary trend of the magmatic source of eruption centers (Masri, 2006). Interpreting rock types and distributions within a volcanic setting is complicated by the heterogeneity and discontinuity of the rocks. To develop the geologic framework, geostatistical methods were applied to the structural grid built using the JewelSuite software. For the base case model the facies picks of formation tops from the drilled wells were matched to interpreted facies boundaries as a set of welltie cross sections. A total of seven horizons were integrated into the model, four for the description of the overburden, and three for the reservoir interval including the top of the steam reservoir (TOR), the top of the brine contact (TOB) and the bottom of the reservoir (BOR). This information was used to generate a 3D structural volume with a consistent set of rock types within a given facies region. To simplify the modeling process, the rock type proportions were combined for each facies group and then populated geostatistically using SIS (Bahar, 2008), where the distribution is constrained to the proportion of the rock type and its spatial relationship. Multiple models using the Placket Burnham (1946) approach were created to capture various uncertainties in the key parameters. The base case model result (Figure 3) provides the common framework for the subsequent modeling in this study. 2

Figure 2. North-South cross section showing correlation of facies groups (Modified from Asrizal, 2006) and facies model of structurally undisturbed andesitic stratovolcano (modified from Bogie and Mackenzie, 1998) Figure 3. Base case 3D geologic structural modeling. Rock type distribution (left) porosity (right). 2.2. 3D Geomechanical Modeling A complete geomechanical model consists of estimates of the in situ or present day stresses, the rock strength and rock properties, and the pore pressure or formation pressure. The initial phase of 3D geomechanical modeling involves the development of calibrated, 1D geomechanical models based on data from key wells distributed throughout the target reservoir. For the Wayang Windu study, 17 offset wells were utilized to derive 1D profiles of vertical stress, S v, maximum horizontal stress, S Hmax, minimum horizontal stress, S hmin, pore pressure, P p, and rock mechanical properties (UCS, T 0, E, etc.). For 1D geomechanical models, we assume that one of the principal stresses acts essentially vertically, and that the other two principal stresses act horizontally. The methodologies used for 1D modeling are fully described in Zoback, et al., 2003 and Zoback, 2007. The development of these 1D well-specific geomechanical models is an essential step towards development of a 3D geomechanical model as they are used to initiate and verify 3D dynamic geomechanical finite element simulations. Rock mechanical properties were obtained from a suite of laboratory tests including triaxial rock strength, unconfined compressive strength, and Brazilian tests of tensile strength. Measurements were performed of additional rock physical properties including density and sonic compressional velocity. The densities vary from 2.35-2.55 g/cm 3 for pyroclastic rocks and 2.72 g/cm 3 for andesitic lavas. These measurements together with some of the basic wireline log data (sonic and density data) were used to estimate key rock mechanics parameters. Table 1 provides a summary of the rock mechanical analyses results. Density, static Young s Modulus, and Poisson s Ratio values were assigned to each lithology. Once upscaled, these 1D properties are mapped onto the structural grid using inverse distance weighting as the deterministic method for interpolation. The populated structural grid is then used to allocate the nodal properties of a 3D finite element mesh required for 3D dynamic calculations. The mapped geomechanical parameters are shown in Figure 4. 3

At Wayang Windu, the pore pressure for each well includes three subdomains: hydrostatic in the rock above the reservoir, a steam zone within the reservoir with gradient density of 0.017 g/cm 3, and a brine zone with gradient density of 0.87 g/cm 3 below the steam zone. It was assumed that a transition zone of 50-200 m thickness lies between the hydrostatic pressures in the overburden and the subhydrostatic pore pressure values in the reservoir. The contact between the water and the steam was generally assumed to lie at or near the top of the reservoir (TOR) contact, based on temperature-pressure observations. The transition between steam and brine zones within the reservoir is modeled using the top of brine (TOB) positions. The mapping of the pore pressure within reservoir is based on well control for the onset of the pressure decrease, the end of the pressure decrease and the start of the subsequent increase below the brine level (Figure 5). In 1D, overburden stress is computed by integrating the density profile of the directly overlying rock; when this approach is used the overburden at each surface location is computed completely independently from the overburden computed at every other location. In areas of high topographic relief like Wayang Windu, however, this approach can lead to significant errors as discussed below as it does not account for loads carried over a broad area. The average overburden density of ~2.54 g/cm 3 in Wayang Windu is comparable to Bulalo geothermal field in Philippines. Table 1. Rock mechanical properties of rock types in Wayang Windu field. No Lithology UCS (Mpa) To (Mpa) E (Gpa) 1 Andesite Lava average 137.9 8.0 10.0 0.18 1.0 2 Tuff Breccia Average 107.7 6.0 7.5 0.23 1.1 3 Lapilli Tuff Average 101.4 5.0 8.0 0.23 1.0 4 Tuff Average 37.7 3.0 8.0 0.15 1.3 n m DENSITY POISSON S RATIO YOUNGS S MODULUS Figure 4. Rock properties mapped through the reservoir volume. STEAM INTERVAL P p 4 MPa TOP OF BRINE GRADIENT P p 0.9 g/cm3 Figure 5. Top of Brine markers were derived from well data as an increase of the Pp to a gradient ~0.9 g/cm 3 (left). 3D development of the pore pressure model (right). 4

Analyses of drilling induced wellbore failure, including breakout (BO) and drilling induced tensile fractures (DITF) provide the ability to constrain the orientation and magnitude of the horizontal principal stresses. In the 16 wells with electrical/resistivity image data, DITF or tensile regions were identified in 15 wells and BOs were identified in 2 wells. Detailed interactions between the in situ stresses (S v, S hmin, S Hmax ), pore pressure (P p ), rock strength, and drilling conditions including the downhole annular pressure and temperature were modeled through 1D geomechanical analysis using GMI SFIB. The 1D geomechanical models are verified against observed wellbore failure, and against drilling events, cuttings observations, etc. The 1D studies yield a range of S hmin gradients from 10.8 to 15.3 kpa/m, of S v from 23.1 to 25.1 kpa/m and of S Hmax from 23.7 to 29.2 kpa/m throughout the field. These calculations imply that the Wayang Windu reservoir is in a transitional normal to strike slip stress regime (S hmin < S v S Hmax ), consistent with the regional structural system. Based on this detailed stress modeling at the depth of observed wellbore failures the Effective Stress Ratio method (Zoback 2007) is used to calculate the horizontal stress profiles for each 1D analysis (Figure 6 solid curve). For the 16 study wells effective stress ratios vary between 0.3 and 0.47 for S hmin and 1.0 and 1.2 for S Hmax. The azimuth of S Hmax was found to be extremely variable based on the 1D analyses (see Figure 7 left below), from nearly N-S (000 N) to E-W (090 N) in both the northern and southern regions of the field. The complexity of the Wayang Windu reservoir and the documented heterogeneity of the reservoir stress field require execution of a series of FE simulations to establish the current stress conditions in the reservoir. Effective stress ratio variations between 0.3 and 0.47 for S hmin and 1.0 and 1.2 for S Hmax were simulated using variations in the azimuth of S Hmax of 30, 80 and 170 to explore combinations that encompass the range of values determined from 1D analysis of observed wellbore failure. The approach consists of first computing the vertical stress by applying gravity instantaneously to the entire model. Using a set of ESR, pore pressure and S Hmax azimuths the magnitude and direction of the horizontal stresses are then mapped to the undeformed 3D mesh and the stress tensor is computed again to ensure that all stresses reach equilibrium. Other steps added to this workflow simulate the stress and deformation evolution of the reservoir due to pore pressure and temperature changes but these results are beyond the scope of this discussion. The results of each of these simulations were calibrated against 1D models on a well-by-well basis. Given the significant topography of the Wayang Windu field it is expected that 1D calculations of overburden stress from wells drilled at high elevations would overestimate overburden stress whereas wells drilled off structure would provide a reasonable match between 3D FE simulation and 1D modeling. For all trials of the various effective stress ratio combinations, an ESR Shmin = 0.32 and ESR SHmax = 1.0 provides the most consistent result compared with the 1D data (Figure 6, left and center). Figure 6 (right) shows the correlation of observed tensile failure position along the wellbore with the computed failure position based on the 3D stress tensor derived from the dynamic simulation trials of far field stress orientation, 30, 80 and 170. In all cases an S Hmax azimuth of 80 o is a better match to observations. The orientations of the principal stresses are strongly affected by surface topography, a free surface that acts as a principal plane. This effect is less pronounced deeper in the reservoir; far from the surface the principal stresses are modeled to be approximately vertical and horizontal. Variability in stress orientation derived from 1D analyses of wellbore image data (Figure 7 left) does not account for the heterogeneous rotated stresses of the Wayang Windu Field whereas 3D dynamic geomechanical modeling clearly captures this variability (Figure 7 right). Figure 6. The 1D geomechanical profiles (open curves) are poorly matched with pseudo-logs (filled curve) extracted from the 3D dynamic simulation for well MBB-2 located at a topographic high (left) whereas the same comparison for well WWE-1 located away from topography shows a good match between 1D and 3D models (center). Right shows the modeled orientation of wellbore tensile failure based on the 3D stress tensor derived from dynamic geomechanical modeling (red is S Hmax =80 model, brown S Hmax =30 model, and green S Hmax =170 model) compared with observed failure (blue symbols) along the wellbore. 5

Figure 7. Comparison of 1D (left) and 3D (right) stress orientation modeling for ambient pressure conditions (Elev. 1000 m). As part of our wider study of the pressure evolution at Wayang Windu, the effects of depletion due to production were also simulated. Based on direct measurements, the pore pressure depletion within the main steam interval zone since the start of production is approximately 13 bars (Figure 8 left). This depletion causes a response of stress and strain above and below the depleted reservoir volume, an effect than can be observed via surface subsidence measurements. The modeled subsidence from the 3D dynamic geomechanical model yields a displacement of 8 cm over 17 years (1995-2012) (Figure 8 right). An independent time lapse microgravity and leveling analysis (Hallinan, 2013, Figure 8, center) reported a significant change from 2002 to 2012 resulting a maximum subsidence of 9 cm over 10 years in very good agreement with the 3D geomechanical model results and provideds a good validation for 3D geomechanical model. Dynamic 3D geomechanical modeling provides a predictive tool for reservoir development planning to assure pressure maintenance is adequate to mitigate problems with surface structures or the stability of new well trajectories. Figure 8. Pore pressure depletion with production (left). Initial pore pressure conditions (1995) are assumed to be 1.3 MPa greater than current ambient condition within the steam interval. Comparison between microgravity leveling survey and FE depletion modeling shows good agreement in the shape and level of the subsidence (right). 2.3 Fracture Characterization The fracture conceptual model is divided into a large scale hydro-structural model (such as faults) that can be described deterministically, and smaller scale structures that are only partially observed in wells or not at all between wells requiring a necessarily stochastic description. The hydro-structural model is interpreted from integration of the surface structural model with the picking of subsurface structures connecting located MEQ induced by hydraulic and thermal disturbances during water or acid injections, augmented by more direct observation of fluid production or drilling loss zones in wells. Smaller scale structures are observed in terms of their presence in well based imaging techniques that provide statistics of their spatial occurrence, intensity, orientation and apparent 6

resistivity. Indications of their effects on fluid flow can be examined through the correlation of imaged fracture occurrence with anomalies in pressure and temperature gradients or spinner logs as well as drilling losses. Images from 16 wells were used to identify fracture orientation, intensity and occurrence. Orientations are important for understanding the alignments of fractures, as they control connectivity and anisotropy of the network of fractures beyond the wells, and the incident stress on the fracture planes determines their likely hydraulic and mechanical properties. The intensity of fracturing determines whether fluids can percolate through a connected network of fractures and the directional permeability and accessible porosity of that network. The distributions of the fracture orientations were divided into four sets and each analyzed using a Univariate Fisher distribution (Fisher 1953). The NE set evident in 15 of the wells has a consistent and concentrated trend; a more diffuse NNW set is evident in 13 wells; a WNW set is present in 10 wells but rotates with adjacent nearby lineaments; and a sub-horizontal set of diffuse orientation is evident in 13 wells (Figure 9 left). Fracture intensity away from the wells was modeled on the basis of measured linear intensity, i.e. fracture count per length along the well calculated from image log interpretation, but corrected for orientation bias of the drilling direction relative to the fracture orientation (Terzaghi 1965). There is a lot of variability in fracture intensity between and within wells and also the relative importance of fracture sets varies between wells. Understanding of any systematic trends in this variability is important for extrapolating likely structural and hydraulic characteristics to support well planning. Fracture intensities were compared with lithology and mechanical properties (overburden, density, porosity) to explore reservoir property relationships (Figure 9 right). The more brittle andesite lava with low porosity tends to have highest fracture intensity compared to the other more ductile lithologies with higher porosity. The developed conceptual model was that Lava has the highest fracture intensity, Tuff-Breccia and Lapilli Tuff have average intensity, Tuff has slightly lower intensity, and Breccia has the lowest intensity. The fracture intensity in Lava is 2.5 times that in Breccia, and within the Breccia and Tuff lithologies, intensity is higher where porosity is less than 6%. Some indication on the ability of individual fractures to conduct fluids was provided by the interpretation of electrical/resistivity logs suggesting between 10-50% of interpreted fractures were in fact sealed. Another observed characteristic of the fracture system is that although fractures are widespread (> 9000 in the 16 interpreted wells, or more than one fracture per metre on average), flow (as evidenced by production or anomalies in PT surveys and drilling information) is dominated by a small number of structures per well. For the wells with image logs these dominant flows coincide with clusters of high fracture intensity. This implies that flow is restricted to a sparse subset of fractures that lie within deformation zones that have sufficient intensity to form extensive interconnected networks. As a rule of thumb such zones are identified from the image log interpretation where the local fracture intensity is more than three times the average, giving a typical spacing of 70m for such structures. By further examining the predominant orientation of fractures within each deformation zone, one can infer the main trends of the likely fluid conducting structures and their relationship to lithology. Flow conducting deformation zones are predominantly NE and NNW in Lava, NNW in Breccia, NE and NNW in Tuff Breccia and Lapilli Tuff, NE in Tuff. Sub-horizontal deformation zones are fewer and mainly in Lava and Tuff Breccia, with the main flowing ones typically in the base of Lava beds. The fracture characteristics described above provide the conceptual framework for numerical 3D DFN modelling using ConnectFlow. Figure 9. Kamb contour stereonet of fracture orientations from borehole image interpretation (left). Fracture intensities correlated with lithology and porosity (right). 2.4 DFN Modeling Numerical 3D DFN modeling is used to implement the interpreted conceptual model for natural fractures, quantify and test models for fracture parameters, and ultimately make predictions of likely structural and hydraulic properties in new areas of drilling. The modeling starts by interpreting the hydro-structural model, which is essentially the hydraulic companion to the interpreted geological structural model, except it only considers those larger scale structures that can be associated with fluid conduits based on inference from microearthquakes (MEQ) induced by injection, or drilling and production information. These structures may correspond to faults, or parts of faults, or to other fluid conducting structures that don t necessarily have a surface or seismic expression. These essentially form a deterministic skeleton for the description of fluid conducting fractures. However, it is only possible to characterize some larger scale structures deterministically and only within the area where data is of sufficient quality and abundance. Between and beyond these structures a stochastic description of the fracture system is necessary based on the DFN conceptual model described above and the 7

associated parameterization of fracture orientations, fracture intensity relative to 3D models for lithology and geomechanical properties. Expected variations in fracture orientation and intensity across the field are determined by the fracture characterization and association to the 3D geological and geomechanical models, but the locations, orientations and sizes of individual fractures are sampled by a Monte Carlo approach. Locations of micro-seismic events (MEQ) induced by hydraulic injection are the primary means of interpreting the hydro-structural model. The fluid movement created by injection stimulates pressure to rise and the effective stresses on fractures to change, triggering the micro-earthquakes. The time sequence between the injection and event occurrence was used in time lapse to identify the possible fluid pathways, geometric planes through events, and their inter-connections. But since MEQ events can be sparse for a given stimulation, multiple interpretations (i.e. more than one plane) are possible unless additional information is used to constrain the possibilities. Such information can include nearby structural features such as faults or fumaroles, or fracture orientations seen in nearby wells, intervals in wells showing lost circulations or feedzones as additional control points to define planes, moment tensor analysis, gauged pressure communications between wells monitored during drilling, or the results of tracer tests. However, one of the most important constraints is that implied by the 3D geomechanics model since this can provide an estimate of the amount of injected pressure required to make a fracture of particular orientation and location slip. Hence, when choosing between two possible planes joining located MEQ, if one requires much less pressure to induce slip than another, then this is the more plausible, especially if the alternative would require pressures above the measured injection pressure. In summary, the approach was to gradually animate the 561 located MEQ events in time relative to the record sequence of stimulation operations and pump-rates, to fit or extend structural surfaces through these MEQ (augmented by additional locations of flowing zones evident in wells or fumaroles), and review the plausibility of each choice against the 3D geomechanics model and structural information from nearby wells or surface faults. The interpretation had to be done in many steps considering a few MEQ at a time since discrete injection activities were performed in five phases between 1998 and 2014 involving stimulation in more than 14 wells. The resulting hydro-structural model is shown in Figure 10 for the whole reservoir (left), and around two well pads where MEQ and well data could be integrated (right). The majority of interpreted hydrostructures were located centrally to the Wayang Windu region, local to the stimulation performed. Towards the edges of the region, the MEQ events become too sparse to interpret. The major orientation of the hydro-structures was dominated with NE and WNW trends parallel with the approximate strike and location of several faults, including the Bandung Basin Fault, and Pejaten Fault with WNW trends and the Kijang fault, Cibitung Fault, Cibitung 1 fault, Cibolang Besar Fault, and Citawa Fault which have NE trends. In all 97 hydro-structures were interpreted. A scoring system was devised to assign a confidence rating of high, medium or low to each hydro-structure as an aid to risk assessment in drilling plans. This factored, for example: how many MEQ lay on each structure, whether the structure was verified by production or structural information in one or more wells, how close the fracture was to being critically stressed, if there was a parallel moment tensor solution, and if the structure had a surface expression (lineaments or fumaroles). N 0 0.375 0.75 1.5 km Legend Hydrostructure Plane Surface Fault MEQ Event Figure 10. Hydro-structure plane (purple planes) with MEQ event in colored dots and interpreted fault in solid pink lines (left). Details of hydro-structures around 2 well pads with identified deformation zones from image logs shown as discs. As well as providing a deterministic skeleton for DFN modeling, the hydro-structures provide an indication of fracture intensity-size scaling for the larger scale fractures in the system and how this varies with orientation, which is invaluable in the absence of any direct means for determining fracture size distribution. A common interpretation for the spatial distributions of faults is they follow fractal or multi-fractal scaling laws (King, 1983; Davy, 1993). The assumption used here is that fracture intensity-size follows a power-law distribution the exponent of which can be inferred by matching the intensity of fractures seen at the scale of the diameter of wellbores (decimeter) to the intensity-scaling implied by the hydro-structural model and the fault model (kilometers). It should be recognized that the intensity of open fractures is less than that of all fractures, due to sealing, mineralization and mechanics, and that intensity-size distributions for connected fractures can be quite different, as large fractures are more likely to form connected networks than small ones, see Figure 11 (left). In fact, this behavior provides a means to test and calibrate the intensity-size distribution. Initially fitting a power-law exponent through the intensity measured for all interpreted fractures in the wells at the scale of the well diameter, and the 8

Log(P32 [m2/m3]) Masri et al. intensity-scaling for hydro-structures on the scale of a kilometer, the exponent can be fine-tuned by DFN simulation of the connected network such that the frequency of connected fractures as seen in the wells is consistent with frequency of flow conducting deformation zones as apparent in the well PT surveys, about one per 70m. A large exponent (slope) would imply fractures are mostly small, and the frequency of flowing conducting fractures in wells would be too small; a small exponent would imply more large fractures and hence more frequent flow. An example calibrated intensity-size model is shown in Figure 11 (right). Realizations of the DFN consist of the deterministic hydro-structural model and a sampled stochastic model of the fractures between and around these structures based on the conceptual models for fracture orientation, spatial distribution relative to 3D geological and geomechanical properties, and the intensity-size scaling. Care was taken not to double count deterministic and stochastic fractures around inferred hydro-structures. An example DFN model is shown in Figure 13 left. 0-0.5-1 -1.5 NE Hydrostructures NE Power Law NE Image Error NE PTS Error NE HS Error NE Rock Property Driven Model -2-2.5-3 -3.5-4 -4.5-5 -1.5-1 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 Log(Equivalent Fracture Radius [m]) Figure 11. Power-law intensity-size distribution concept (left): connected fractures follow a different distribution as large fractures are more likely to connect. Cumulative intensity-size used for NE set in DFN model (right): open fractures (black), connected fractures (red). The vertical bars indicate inferred intensity at scales of well diameter, deformation zones and hydro-structure. The description of fracture apertures was derived from the analysis of more than 20 well test measurements, including step-rate tests of pressure responses to injection and / or the fall-off of pressure once pumping ceases. By analyzing these well tests, estimates of the permeability-thickness are obtained using both steady state and transient analysis (e.g. Tampubolon, 1989). Hydraulic diffusivity or storativity values are also obtained using the Cooper Jacob method. The well tests are open hole, and so the equivalent hydraulic aperture of individual fractures can be interpreted directly. Still, in wells that also have spinner logs or drilling information indicating the main flow zones it is possible to estimate the permeability thickness for individual structures and thereby an equivalent hydraulic aperture, which has a geometric mean of about 200 μm, but varying between about 100 μm to 900 μm. A useful empirical relationship derived from the transient well test analyses was a function relating the hydraulic diffusivity [m 2 /s], which determines the rate at which pressure changes are communicated, and permeability-thickness [m 3 ], which determines production: log( D ) 0.48 log( kh) 7.3. This is because the rate at which pressure is communicated through the fracture system during injection is also apparent from the located MEQ events through the position and times of events relative to the position and start times of injection at sufficient rate. The hydraulic diffusivity was calculated from the spatial distance, x x 0, between MEQ within the same hydro-structure and the time 2 interval, t, between events as D x x 0 / t (Stretslova, 1988), and then the geometric mean is taken of hydraulic diffusivities of channels within that hydro-structure. Using the above empirical relationship a permeability-thickness, or equivalent hydraulic aperture, is then estimated for each interpreted hydro-structure even when no well intercepts or well tests were available in that structure. In the same way as confidence in the existence of each hydro-structure is ranked, the likely permeability-thickness of each hydro-structure is ranked. Again a ranking system was used combining several types of data: whether one or more high production zones were measured at intercepts with the structure, if a well test had indicated high injectivity in the structure, and if high pressure diffusivity within the structure was evident from the located MEQ events. Hydraulic aperture is also affected by the current in situ stress and the evolution of rock stresses and pore pressure. In the model hydraulic aperture is modified to account for both normal and shear dilation of each individual fracture by determining the normal and shear stress incident on each plane (or on the finite-element discretisation of the plane) implied by the 3D geomechanics model (Moos and Barton, 2008; Barton et al. 2013). The sensitivity of aperture to normal stress being inferred from step-rate tests conducted in several wells, while the sensitivity to shear dilation is inferred from changes in effective stress on the stimulated hydro-structures interpreted from MEQ data and the measured changes in pressure in the injection wells. Figure 12 (left) illustrates the closeness to slip of the hydro-structures in the 3D stress field by plotting the pressure increase required to start to induce slip for a coefficient of sliding friction of 0.5 (negative pressure means fractures such as WNW that are already critically stressed). This suggests most hydrostructures 9

begin to slip at injection pressure increases below 3MPa, as observed during hydraulic stimulation. This model implies a hydraulic aperture distribution in the hydro-structures as shown as shown in Figure 12 (right). A consistent approach for assigning the hydraulic aperture of stochastic fractures is applied with the initial pre-stressed aperture, e [m], assumed related to fracture size by e h 0 7.5 10 4 ( r /1000) 0.25. The parameters defining hydraulic aperture and its coupling to in situ stress were fine-tuned by dynamic simulation of the hydraulic stimulation of one the wells to calibrate against the measured pressure and extent of the induced MEQ cloud as a function of time using the methodology defined by Barton et al. (2013). h 0 Figure 12. Hydro-structures interpreted from MEQ colored by the increase in pressure required for shear to occur for coefficient of sliding friction of 0.5 (left); Equivalent hydraulic apertures of hydro-structures based on integration of MEQ, well tests and in situ stress (right). Having inferred a DFN model from all available data to obtain geometric and dynamic fracture parameters describing hydro-structures down to the scale of those fractures seen in wells, the model can be used to simulate likely permeability encountered in wells. As a further test of the DFN model and to quantify uncertainty in hydraulic properties implied by the model around existing wells, 10 realizations were generated and the permeability-thickness in each existing well calculated and compared to values interpreted from well tests (Figure 13 right). The permeability-thickness prediction from the DFN model can be compared to those interpreted from well test analysis in 21 wells for validation. The model values are subdivided into a contribution from the deterministic hydro-structures (blue), and a contribution from the stochastic fractures, shown as the median (green) and the total kh values for each of the 10 realisations. In the majority of wells, and especially those with larger permeability-thickness, the deterministically specified hydrostructures are a significant part of the kh estimates from the model. Over all the wells and realizations the calibration has neutral bias, but some individual wells and realizations have a bias. When considering both the overall bias, and the mean absolute deviation from the well test values, Realizations 2 and 8 produce the smallest errors and represent the best interpretation currently. Although much data is used to develop and constrain the conceptual DFN model some uncertainty should still be captured to evaluate the model and to test how new information can be used to update and further constrain the model in the future. One of them is the structural uncertainty from hydro-structures resulting from inaccuracies in the MEQ locations. The other is the location and properties of the connected fractures that cannot be linked to observed MEQ, but still control fluid flow in and around the wells only some of which can be detected by image logs and other well based data. 3. MODEL INTEGRATION Integrating the results from our 3D geomechanical grid with the DFN fracture distribution we can determine using Coulomb failure analysis the proximity to frictional failure of each fracture in the DFN. This provides a means to identify geomechanical sweet spots, zones with the highest density of stress sensitive fractures (Barton et al, 2013), for stimulation design. Figure 14 provides an example cross-section through the reservoir at one identified sweet spot. In this figure, DFN generated stress-sensitive fracture intersections ( / n 0.4) are mapped to the structural grid and color represents fracture intensity. A zone of extremely high permeability occurs along the wellbore adjacent to this sweet spot (red arrow) in good agreement with modeling results. 10

Figure 13. (Left) Wayang Windu DFN model, Green polygons are NE fractures, yellow polygons are sub horizontal (SH), Blue polygons are WNW, and red polygons are NNW trends. (Right) Comparison of permeability-thickness from existing wells with 10 realizations of the DFN. Realization 2 is shown by red dots and realization 8 shown by yellow triangle. Over the reservoir structure, the predominant stress sensitive fracture intensities appear to align to form hydro-structures (permeable fracture corridors). In general, NE and WNW sub-vertical fracture trends have most hydraulic importance as they are the most stress sensitive and hence more likely to be more permeable. In the northern fault block between the Bandung Basin Boundary and Pejaten faults there is a relatively high intensity of NE trending hydro-structures associated with the Banjasari, Kijang, Cibitung and Cibitung-1 faults along with additional NE structures and splays. Therefore, targeting NE trends in this block appears to be the most likely trajectory for productive wells in this volume. South of the Pejaten fault, WNW trends have the highest fracture intensity amongst the hydro-structures with a cluster of sub-parallel structures surrounding the Pejaten faults. NE hydro-structures seem to gradually terminate south apart from those associated with the Benjasari fault in the west and the Cibolang Besar fault in the east. Hence, targeting WNW trends appear to be the most likely trajectory for productive wells in this volume. NW Well X Cross Section SE Legend Wayang Windu Critically Stressed Fracture Density Fault with dip uncertainties Hydrostructure model Fracture model (DFN) TOR Brine Level Major Feedzone N Well X Figure 14. Cross section of critically stressed fracture intensity through the 3D grid. A geomechanical sweet spot is evident at the depth location of a documented permeable zone along the wellbore (major feedzone). 4. CONCLUSIONS In this study the 3D dynamic geomechanical model developed for the Wayang Windu Field is verified against both 1D observations and an independent analysis of microgravity changes (subsidence) due to depletion from production. Fracture architecture, defined stochastically using available deterministic data provides an inner-connected DFN model. The models have been calibrated on available information, checked for consistency with independent observations of pressure and tracer communication, and used to make predictions of permeability in planned wells. The DFN modeling demonstrates how the currently available data and 3D models can be integrated to develop both a deterministic hydro-structural model closely related to many interpreted surface lineaments and a stochastic description of fracture geometries, connectivity and hydro-mechanics between these structures and away from well controls. 11

The study results will be utilized to guide development and to improve output from the Wayang Windu Field through a combination of new well drilling and treatment initiatives of select wells. It is proposed that this workflow become the standard process to assess subsurface structure permeability distribution. 5. ACKNOWLEDGEMENT We thank the Management of Star Energy Geothermal Indonesia for their support of the work and permission to present this paper. Special thanks go to the many colleagues in Star Energy, Baker Hughes GMI and Amec Foster Wheeler for their assistance. 6. REFERENCES Bahar, A.: Geostatistics for Reservoir Characterization, June 23-27, Bali, Indonesia (2008). Barton, C.A.: Reservoir-Scale Fracture Permeability in the Dixie Valley, Nevada, Geothermal Field. SPE Journal/ ISRM 47371, Trondheim, Norway (1996). Barton C.A., Moos D., Hartley L., Baxter S., Foulquier L., Holl H., Hogarth R., 2013. Geomechanically Coupled Simulation of Flow in Fractured Reservoirs. Proceedings, Thirty-Eighth Workshop on Geothermal Reservoir Engineering, Stanford University, Stanford, California, February 11-13, 2013. SGP-TR-198 Barton, C.A, et al: Phase A Geomechanical Model For The Wayang Windu Geothermal Field, West Java, Indonesia. Final Report of Phase A from Baker Hughes GMI GeoMechanics Services. (2013, unpublished version) Barton, C.A, et al.: Dynamic Geomechanical Model Applied To the Wayang Windu Geothermal Field, West Java, Indonesia. Final Report of Phase B from Baker Hughes GMI GeoMechanics Services (2014, unpublished version) Bogie, I., Mackenzie, K.M.: The Application of a Volcanic Facies Model to an Andesitic Stratovolcano Hosted Geothermal System at Wayang Windu, Java, Indonesia. Proceeding of 20th NZ Geothermal Workshop, New Zealand (1998). Bogie, I., Kusumah, Y.I., Wisnandary, M.C., 2008, Overview of The Wayang Windu Geothermal Filed, West Java, Indonesia. Elsevier Geothermic 37 (2008) 347-365. www.elsevier.com/locate/geothermics. Davis, G., Reynolds, S.J.: Structural Geology of Rock and Region. John Wiley & Son: New England, USA (1996). Davy, P., 1993. On the frequency length distribution of the San Andreas Fault System. Journal of Geophysical Research: Solid Earth (1978 2012), July, 1993. 10.1029/93JB00372. Hallinan, S. et al.: Repeat Gravity Data Review (2002-2012) and 2.75D to 3D Modeling. CGG Report. Milan, Italy (2013). Hartley, L., Applegate, D., Baxter, S.: Discrete Fracture Network Modeling of the Wayang Windu Field. Phase2; Assessment of Fracture Permeability. AMEC report 2014, unpublished version). Hartley, L., Applegate, D., Baxter, S., 2014. Discrete Fracture Network Modeling of the Wayang Windu Field. Phase1; Conceptual and Static DFN models, AMEC (unpublished version). King, G. C. P. (1983), The accommodation of large strains in the upper lithosphere of the Earth and other solids by self-similar fault systems: The geometrical origin of b-value, Pure Appl. Geophys., 121, 761 814, doi:10.1007/bf02590182. Laboratory of Geomechanics and Mine Equipment, 2013, Testing Report of Rock Physical and Mechanical Properties. A Laboratory Analysis Report from Bandung Institute of Technology. Bandung, Indonesia 2013 (Unpublished Report). Masri, A., Hadi, J., Bahar, A., Sihombing, J.M.: 2006, Uncertainty Quantification By Using Stochastic Approach in Pore Volume Calculation, Wayang Windu Geothermal Field, W Java, Indonesia, Proceedings 31st Workshop of Geothermal Reservoir Engineering. Stanford University, Stanford, CA (2006). Moos D., Barton C. A., 2008, Modeling uncertainty in the permeability of stress-sensitive fractures. ARMA. 08-312. 2008. Plackett Burman designs NIST/SEMATECH e-handbook of Statistical Methods. Setiadji, L.D.: Segmented Volcanic Arc and Its Association with Geothermal Field in Java Island, Indonesia. Proceeding World Geothermal Congress, Bali, Indonesia (2010). Streltsova T. D., 1988. Well testing in heterogeneous formations. Exxon, Monograph. John Wiley and Sons, New York. Tampubolon, T.: Analysis of well test data, Indonesian and Iceland. UNU Geothermal Training programme. Reykjavik, Iceland, 4 (1989). Terzaghi, R.D.: Source of Error in Joint Surveys. Geotechnicque 15 : (1965) 287-304. Tingay, M., et al.: 2009. Present-day stress field of Southeast Asia. International Journal of Tectonophysics, 482, (2009) 92-104 Zoback, M.D.: Reservoir Geomechanics, Cambridge University Press, New York, (2007). Zoback, M.D, et al.: Determination of Stress Orientation and Magnitude in Deep Wells. International Journal of Rock Mechanics and Mining Science, 40, (2003), 1049-1076. 12