«JOINT INVERSION OF PRODUCTION AND TIME-LAPSE SEISMIC DATA: APPLICATION TO NORNE FIELD A DISSERTATION SUBMITTED TO THE DEPARTMENT OF ENERGY RESOURCES ...»
JOINT INVERSION OF PRODUCTION AND TIME-LAPSE
SEISMIC DATA: APPLICATION TO NORNE FIELD
SUBMITTED TO THE DEPARTMENT OF ENERGY RESOURCES
AND THE COMMITTEE ON GRADUATE STUDIES
OF STANFORD UNIVERSITY
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF
DOCTOR OF PHILOSOPHYAmit Suman March 2013 © Copyright by Amit Suman 2013 All Rights Reserved ii I certify that I have read this dissertation and that in my opinion it is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy.
Dr. Tapan Mukerji (Principal Adviser) I certify that I have read this dissertation and that in my opinion it is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy.
Dr. Khalid Aziz I certify that I have read this dissertation and that in my opinion it is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy.
Dr. Gerald Mavko Approved for the University Committee on Graduate Studies.
iii iv ABSTRACT Time-lapse seismic has evolved as an important diagnostic tool in efficient reservoir characterization and monitoring. Reservoir models, optimally constrained to seismic response, as well as flow response, can provide a better description of the reservoir and thus more reliable forecast. This dissertation focuses on different aspects of joint inversion of time-lapse seismic and production data for reservoir model updating, with application to the Norne field in the Norwegian Sea. This work describes a methodology for joint inversion of production and time-lapse seismic data, analyzes sensitive parameters in the joint inversion, identifies sensitive rock physics parameters for modeling time-lapse seismic response of a field and successfully applies and compares the family of particle swarm optimizers for joint inversion of production and time-lapse seismic data of the Norne field. The contributions from this research include a systematic workflow for joint inversion of time-lapse seismic and production data that can be and has been practically applied to a real field. Better reservoir models, constrained to both data will in turn lead to better reservoir forecasts and better field management.
The first part of this thesis uses Norne field data to analyze sensitive parameters in joint inversion of production and time-lapse seismic data. An experimental design is performed on the parameters of the reservoir and seismic simulator. The results are v used to rank the parameters in terms of sensitivity to production and time-lapse seismic data. At the same time it is shown that porosity/permeability models is not the most sensitive parameter for joint inversion of production and time-lapse seismic data of the Norne field. The parameters selected for study are porosity and permeability model, relative permeability, rock physics models, pore compressibility and fluid mixing. Results show that rock physics model has the most impact on time-lapse seismic whereas relative permeability is the most important parameter for production response. The results of this study are used in selecting the most important reservoir parameters for joint inversion of time-lapse seismic and production data of the Norne field.
It is established that rock physics model is the most sensitive parameter for modeling time-lapse seismic of the Norne field, but there are rock physics parameters associated with rock physics model that impact time-lapse seismic modeling. So it is necessary to identify sensitive rock physics parameters for modeling time-lapse seismic response. Thus, the second part of this thesis identifies sensitive rock physics parameters in modeling time-lapse seismic response of Norne field. At first facies are classified based on well log data. Then sensitive parameters are investigated in the Gassmann's equation to generate the initial seismic velocities. The investigated parameters include mineral properties, water salinity, pore-pressure and gas-oil ratio (GOR). Next, parameter sensitivity for time-lapse seismic modeling of the Norne field is investigated. The investigated rock physics parameters are clay content, cement, pore-pressure and mixing. This sensitivity analysis helps to select important parameters for time-lapse (4D) seismic history matching which is an important aspect of joint inversion of production and time-lapse seismic of a field.
Joint inversion of seismic and flow data for reservoir parameter is highly nonlinear and complex. Local optimization methods may fail to obtain multiple history matched models. Recently stochastic optimization based inversion has shown very good results in the integration of time-lapse seismic and production data in reservoir vi history matching. Also, high dimensionality of the inverse problem makes the joint inversion of both data sets computationally expensive. High dimensionality of the inverse problem can be solved by using reduced order models. In this study, principal component bases derived from the prior is used to accomplish this. In the third part of the dissertation a family of particle swarm optimizers is used in combination with principal component bases for inversion of a synthetic data set. The performance of the different particle swarm optimizers is analyzed, both in terms of the quality of history match and convergence rate. Results show that particle swarm optimizers have very good convergence rate for a synthetic case. Also, these optimizers are used in combination with multi-dimensional scaling (MDS) to provide a set of porosity models whose simulated production and time-lapse seismic responses provide satisfactory match with the observed production and time-lapse seismic data.
The goal of the last part is to apply the results of previous parts in joint inversion of production and time-lapse seismic data of the Norne field. Time-lapse seismic and production data of the Norne field is jointly inverted by varying the sensitive parameters identified in previous chapters and using different particle swarm optimizers. At first the time-lapse seismic surveys of the Norne field acquired in 2001 and 2004 is quantitatively interpreted and analyzed. Water was injected in the oil and gas producing Norne reservoir and repeat seismic surveys were conducted to monitor the subsurface fluids. The interpreted P-wave impedance change between 2001 and 2004 is used in the joint inversion loop as time-lapse seismic data. The application of different particle swarm optimizers provides a set of parameters whose simulated responses provide a satisfactory history match with the production and time-lapse seismic data of Norne field. It is shown that particle swarm optimizers have potential to be applied for joint inversion of the production and time-lapse seismic data of a real field data set.
I look back and find myself surprised and extremely indebted to Stanford for all I have gained throughout these years. I am surprised to see myself a totally different and a much better person. I consider myself extremely lucky to be a part of Stanford.
It provided me, an excellent environment for learning, incomparable infrastructure and golden opportunity to interact with genius faculties and students. All these years of PhD studies are full of precious gifts, on professional as well as personal front. People say you may go to heaven after death, but my years in Stanford are no less than heaven. My real satisfactions are always aligned with knowledge, and since Stanford is the Mecca of knowledge, I always felt satisfied and peaceful during my journey in Stanford. I am indebted to a long list of people who played significant roles in whatever I have achieved during my PhD at Stanford.
I want first to thank my advisor Tapan Mukerji for providing me the incredible support and guidance during my research. This research would have been impossible without his consistent support and faith in me. I admire his remarkable ability to motivate and guide in the research as well as in the coursework. I can feel the optimism and energy that he puts in his efforts. His unique teaching style kept me always attentive for my purpose in the research. Tapan is very approachable, friendly viii and open minded for any discussion. It helped me a lot on both professional as well as personal fronts.
I would like to thank Juan Luis Fernández Martínez for providing me guidance and feedbacks throughout my research on a family of particle swarm optimizers.
Thanks Juan for being always cooperative and available for discussions. He has enriched me through his wide and extensive knowledge on stochastic optimization. I admire his never give up attitude towards research. He always motivated me during my research and taught me how to appreciate and make excellent scientific work that helps other researchers to use it.
I also thank the members of my reading committee, Khalid Aziz and Gary Mavko. I consider myself lucky to interact with Khalid on several occasions. His critical comments, feedback and ideas always helped me to shape and refine my PhD.
I admire his sense of flexibility and clarity towards the research. I owe a lot of gratitude to Gary Mavko for providing me new ideas and critical comments on my research. Thanks Gary for always reminding me to maintain a clear sense of purpose in my research. Your critical comments during my proposal provided the base pillars for this research.
I wish to acknowledge the contributions of David Echeverria and Pallav Sarma for providing me valuable suggestions and feedback on my research. I also thank Mike Costello, Paul Niezguski and Robert Dombrowski for the excellent mentorship and guidance during my internships with Shell, where I got considerable help in deciding the future direction of my research.
I thank the Stanford Centre of Reservoir Forecasting (SCRF) for financial support and Smart Field Consortium at Stanford for critical comments and feedbacks on my research. Comments and feedbacks received during SCRF seminars helped me a lot in shaping up my PhD. I would like to thank Jef Caers for the comments and feedbacks on my research during SCRF seminars. I also thank Statoil and Norwegian ix University of Science and Technology for providing me the Norne data set. A special note of thanks to Mohsen Dadashpour for answering my questions on the dataset very patiently. I also thank the Stanford CEES for computational support and specially Dennis Michael for always helping and solving cluster problems on short notice. I wish to acknowledge the support of Schlumberger Company for providing licenses of Petrel and Eclipse.
I met wonderful people and made good friends at Stanford. It was a pleasure to share doctoral studies and life with brilliant people like Zhe Wang, Siyao Xu, Amar Alshehri, Khalid Alnoami and my fellow SCRF members. Thanks to Nishank Saxena, Madhur Johri, Indrajit Das and Sabrina Aliyeva for incredible friendship. I am grateful to them for memorable evening parties, movie nights and outdoor trips. I am also thankful to my seniors from Indian School of Mines (ISM) and their families for always making me feel at home.
Lastly I would like to thank my family for undemanding love and selfless support. A ton of thanks to my wife Priya for being always there for me in highs and lows of my graduate study, thanks for always listening to my problems and providing pillars of strength to me. I am grateful to my parents for teaching me to respect and love everyone and worship God. Whatever is taught to me in my childhood by my parents has the biggest role to play in shaping up my professional and personal life.
I dedicate my thesis to Mummy, Papa and Priya.
TABLE OF CONTENTS
LIST OF TABLES
LIST OF FIGURES
CHAPTER 1: INTRODUCTION
1.2 THESIS OUTLINE
CHAPTER 2: NORNE FIELD DATA AND RESERVOIR MODEL
2.1 SEGMENT-E OF NORNE FIELD
2.2 AVAILABLE DATA
2.3 RESERVOIR MODELING WORKFLOW
CHAPTER 3: SENSITIVITY ANALYSIS FOR JOINT INVERSION OF TIME-LAPSESEISMIC AND PRODUCTION DATA OF NORNE FIELD
3.2 SENSITIVE PARAMETERS
3.2.1 Porosity and Permeability Models
3.2.2 Multi-Dimensional Scaling
3.2.3 Kernel K-Medoid Clustering
3.2.4 Relative Permeability
3.2.5 Pore compressibility
3.2.6 Rock Physics Model
3.2.7 Fluid mixing
3.4 FLOW SIMULATION
3.5 ROCK PHYSICS MODELING
3.5.1 Cemented-Sand Model
3.5.2 Unconsolidated Sand Model
3.6 TIME-LAPSE SEISMIC MODELING
3.6.1 Change in Saturation
xi 3.6.2 Changes in Pore pressure
CHAPTER 4: SENSITIVITY STUDY OF ROCK PHYSICS PARAMETERS FOR MODELINGTIME-LAPSE SEISMIC RESPONSE OF NORNE FIELD
4.2 ROCK PHYSICS MODELING
4.2.1 Facies classification
4.2.2 Sensitivity to Fluid Substitution
4.2.3 Selection of Rock Physics Model
4.2.4 Constant cement model
4.3.1 Flow Simulation
4.3.2 Time-lapse Seismic Modeling
4.4 SENSITIVITY ANALYSIS RESULTS
CHAPTER 5: JOINT INVERSION OF PRODUCTION AND TIME-LAPSE SEISMIC DATAUSING A FAMILY OF PARTICLE SWARM OPTIMIZERS
5.2 PARTICLE SWARM OPTIMIZATION
5.3 FAMILY OF PARTICLE SWARM OPTIMIZERS
5.3.1 GPSO or centred - regressive PSO (t0 = 0)
5.3.2 CC-PSO or centred - centred PSO (t0 = 0)
5.3.3 CP-PSO or centred -progressive PSO (t0 = Δt)
5.3.4 PP-PSO or progressive-progressive PSO (t0 = 0)
5.3.5 RR-PSO or regressive-regressive PSO (t0 = Δt)