Non-Linear Geostatistics for Reservoir Modelling
|Prof.Dr. Stephen Tyson and Dr.Ing. Sebastian Hörning (Universiti Teknologi Brunei and The University of Queensland)|
|2 to 5 days|
|Geology – Geological Modelling|
|10 to 25 CPD points|
SPATIAL COPULA PYTHON RESERVOIR MODELLING PERMEABILITY GEOSTATISTICS UNCERTAINTY
The course will show the attendees how to test for linear spatial dependence and introduce the concepts of non-linear geostatistics. Attendees will develop an excel spreadsheet and a python notebook which can be used for spatial data analysis and non-linear stochastic simulation.
Existing geostatistics algorithms based on the kriging matrix can be shown to underestimate the connectivity of extreme values because they assume a linear spatial dependence model. Moreover, the estimation of uncertainty based on these techniques uses the kriging variance, which is not dependent on the values of the spatially distributed variable. It can also be shown that these uncertainty estimate are often implausible. This course will explain the reasons why most spatial variables in geoscience do not have a linear spatial dependence, even after monotonic transformations, and what the impact of this in the estimation of petrophysical properties.
The course will show the attendees how to test for linear spatial dependence and introduce the concepts of non-linear geostatistics. Attendees will develop an Excel spreadsheet and a python notebook which can be used for spatial data analysis and non-linear stochastic simulation.
Upon completion of the course participants will be able to;
1. Assess whether spatial dataset has a linear spatial dependence
2. Determine the dependence structure between two or more variables (i.e. go way beyond linear correlation)
3. Understand what different correlation measurements mean
4. Understand the importance of asymmetry in spatial modeling
5. Run spatial interpolation and simulation using copulas.
1. Demonstration of the shortcomings of traditional geostatistics based on simple and clear examples (Assumptions, Symmetry, Uncertainty…)
2. Brief repetition of basic statistics, multivariate statistics, and traditional geostatistics
3. Introduction to rank-order geostatistics and copulas
4. Different copula models and spatial copulas
5. Copula-based measures of dependence (Rank correlation, different types of asymmetry functions)
6. Spatial interpolation using copulas
7. Spatial simulation using copulas
Practical part (Exercise based on Excel and IPython)
1. Empirical data analysis
2. Variogram and Covariance functions (just to repeat the basics)
3. Empirical copula densities
4. Rank-correlation function
5. Different asymmetry functions
6. Spatial interpolation/simulation using copulas
This course is designed for petrophysicists, geomodellers, geologists, anyone with an interest in spatial modelling, volume estimation, preparation of simulation models and uncertainty analysis.
Some knowledge of Excel and some basic statistics.
Bárdossy, A.: Copula Based Geostatistical Models for Groundwater Quality Parameters.Water Resources Research 42 (2006). W11416, doi:10.1029/2005WR004754
Bárdossy, A. und J. Li: Geostatistical Interpolation using Copulas.Water Resources Research 44 (2008) W07412. doi: 10.1029/2007WR006115
P. Guthke ; Non-multi-Gaussian spatial structures: process-driven natural genesis, manifestation, modeling approaches, and influences on dependent processes
C. Haslauer ; Analysis of real-world spatial dependence of subsurface hydraulic properties using copulas with a focus on solute transport behaviour
S. Hörning ; Process-oriented modeling of spatial random fields using copulas
About the instructors
Steve Tyson is the Chair Professor of Petroleum Engineering at the Universiti Teknolig Brunei. Previously he was Chair of Subsurface Modeling at the Centre for Coal Seam Gas and Director of the Centre for Geoscience Computing in the School of Earth Sciences. He has worked in reservoir characterization and modeling in the oil industry for more than 30 years in both conventional and unconventional reservoirs. His current research interests are in model validation, verification and acceptance criteria for both static and dynamic models, upscaling, uncertainty modeling and non-linear geostatistics and spatio-temporal reservoir analytics
Sebastian Hörning heads the Spatio-Temporal Reservoir Analytics group in the Centre for Geoscience Computing at the University of Queensland. His research covers spatial statistics problems over a range of scales; he has worked on basin scale spatial dependence and on the spatial models from micro CT images. In all cases he has found compelling evidence indicating that linear spatial dependence is rare. Tests for linear dependence are simple, even with conventional geostatistical software. However new techniques had to be developed to estimate spatial variables with more complex spatial dependence structures. Dr Hörning joined the University of Queensland in 2016 from the University of Stuttgart where he worked with Professor Andras Bardossy on the development of non-linear geostatistics.
Explore other courses under this discipline:
Instructor: Dr Dave L. Cantrell (Cantrell GeoLogic and Stanford University, USA)
Reservoir quality prediction has long been the ultimate goal of industry geologists, yet few have achieved this in a truly quantitative fashion. This workshop presents a new approach to reservoir quality prediction that involves the integration of a variety of modeling techniques to understand, quantify and predict the geological processes that control reservoir quality. Since initial reservoir quality is established at the time of deposition, numerical process models are used to predict initial reservoir quality; diagenetic process models are then used to modify these initial results and ultimately produce a quantitative and geologically-based prediction of present-day subsurface reservoir quality.
Instructors: Prof. Dr Stephen Tyson and Dr Ing Sebastian Hörning (Universiti Teknologi Brunei and The University of Queensland)
The course will show the attendees how to test for linear spatial dependence and introduce the concepts of non-linear geostatistics. Attendees will develop an excel spreadsheet and a python notebook which can be used for spatial data analysis and non-linear stochastic simulation. Existing geostatistics algorithms based on the kriging matrix can be shown to underestimate the connectivity of extreme values because they assume a linear spatial dependence model. Moreover, the estimation of uncertainty based on these techniques uses the kriging variance, which is not dependent on the values of the spatially distributed variable. It can also be shown that these uncertainty estimate are often implausible. This course will explain the reasons why most spatial variables in geoscience do not have a linear spatial dependence, even after monotonic transformations, and what the impact of this in the estimation of petrophysical properties. The course will show the attendees how to test for linear spatial dependence and introduce the concepts of non-linear geostatistics. Attendees will develop an Excel spreadsheet and a python notebook which can be used for spatial data analysis and non-linear stochastic simulation.
Instructor: Prof. Shiyi Zheng (London South Bank University)
This course on well test analysis will provide attendees basic knowledge on the subject. The course will start with a review of an oil field life from exploration, appraisal, development and abandonment to define the role of well testing. The history and clear definition of the subject will also be given as the summary of this introduction. The course will cover classic, modern and numerical well testing concepts and techniques.
Instructor: Dr Bjorn Wygrala (Schlumberger)
The term “Petroleum Systems” and the technology “Basin and Petroleum Systems Modelling” will be introduced by showing applications in areas with critical exploration challenges, including salt basins and thrustbelts. Technical breakthroughs in the last 10-15 years have been the extension of the technology from 2D to 3D, and the ability to perform multi-phase petroleum migration modelling using different methods in high resolution geological models. This enables temperature, pressure and petroleum property predictions to be made with higher levels of accuracy and in the most complex geological environments such as in the sub-salt or in thrustbelts. Case studies will be used with live software presentations to illustrate key points. Applications of the technology will range from frontier exploration in which large areas with only sparse data are screened, to detailed assessments of exploration risks in structurally complex areas, to petroleum resource assessments of yet-to-find oil and gas.
Instructor: Prof. Dr Richard Swarbrick (Swarbrick GeoPressure Consultancy)
All wells drilled require a pre-drill prediction of pore fluid and fracture pressures which defines the drilling window. This course explains the objectives, methods and uncertainties of prediction, based on extensive global experience. The necessary understanding of the geological/geophysical context of abnormal pressures leading to standard algorithms will be provided. Part of the challenge is terminology and contrasting display methods of geoscience and operations/drilling groups. Both approaches are necessary and investigated in the interactive exercises which will form an essential component of the course.
Instructor: Dr Dirk Nieuwland (NewTec International)
Unconventional hydrocarbon systems require unconventional approaches to decide on drilling locations and development techniques. The information contained in natural fracture systems can be used to support the drilling and well stimulation technique for the development of unconventional hydrocarbon systems such as shale gas. This short course is based on geomechanics as a technique that can be used to understand and to develop unconventional hydrocarbon systems such as shale gas systems, and fractured crystalline basement, where conventional logging and seismic systems are inadequate.
Instructor: Dr Tim Wynn (AGR-Petroleum Services)
Reservoir modelling for field development planning is a well-accepted process but its application to fractured reservoirs requires specific considerations which are less commonly known. This course describes a practical methodology for building 3D static (“geocellular”) reservoir models for naturally fractured reservoirs using standard modelling software, covering such considerations. The issues addressed include the integration of log, core and seismic data, the sourcing and application of in situ stress data, the process of defining and building the static reservoir model itself, and the creation of output in a form appropriate for dynamic modelling using dual porosity reservoir simulators where appropriate. More complex workflows using discrete fracture networks will also be summarised, as will general issues of fracture description, uncertainty-handling and volumetrics.
Instructor: Prof. Dorrik Stow (Heriot-Watt University)
Sandstones deposited in deep marine environments form important hydrocarbon reservoirs in many basins around the world. Interbedded mudstones can be important as source rocks, as well as acting as barriers, baffles and seals. Deepwater reservoirs are currently the principal target for oil and gas exploration, with over 1600 existing turbidite fields and plays.
Instructor: Dr Vasily Demyanov (Heriot-Watt University)
Reservoir prediction modelling is subject to many uncertainties associated with the knowledge about the reservoir and the way they are incorporated into the model. Modern reservoir modelling workflows, which are commonly based on geostatistical algorithms, aim to support development decisions by providing adequate reservoir description and predict its performance. Uncertainty about reservoir description needs to be accounted for in modelling workflows to quantify the spread of reservoir predictions and its impact development decisions. The course aims to build awareness of the impact the modelling choices on the reservoir predictions and their relation to the way uncertainty is incorporated into reservoir modelling workflows. The course addresses the problem of tying the workflow with the expected geological vision of a reservoir subject to uncertainty. This is associated with one of the common issues, when standard assumptions of a workflow are not consistent with the model geology or do not reflect possible variations due to existing uncertainty. The course demonstrates the implementation of geostatistical concepts and algorithms in geomodelling workflows and the ways uncertainty is accounted for in reservoir description and predictions. The course includes an overview of the state-of-the art conventional techniques and some novel approaches, in particular machine learning for reservoir description.