Methods for Petroleum Well Optimization: Automation and Data Solutions

By Rasool Khosravanian and Bernt S. Aadnoy

Drilling and production wells are becoming more digitalized as oil and gas companies continue to implement machine learning andbig data solutions to save money on projects while reducing energy and emissions. Up to now there has not been one cohesiveresource that bridges the gap between theory and application, showing how to go from computer modeling to practical use. Methodsfor Petroleum Well Optimization: Automation and Data Solutions gives today’s engineers and researchers real-time data solutionsspecific to drilling and production assets. Structured for training, this reference covers key concepts and detailed approaches frommathematical to real-time data solutions through technological advances. Topics include digital well planning and construction,moving teams into Onshore Collaboration Centers, operations with the best machine learning (ML) and metaheuristic algorithms,complex trajectories for wellbore stability, real-time predictive analytics by data mining, optimum decision-making, and case-basedreasoning. Supported by practical case studies, and with references including links to open-source code and fit-for-use MATLAB, R,Julia, Python and other standard programming languages, Methods for Petroleum Well Optimization delivers a critical training guidefor researchers and oil and gas engineers to take scientifically based approaches to solving real field problems.

• Bridges the gap between theory and practice (from models to code) with content from the latest research developments supported by practical case study examples and questions at the end of each chapter
• Enables understanding of real-time data solutions and automation methods available specific to drilling and production wells, suchas digital well planning and construction through to automatic systems
• Promotes the use of open-source code which will help companies, engineers, and researchers develop their prediction and analysissoftware more quickly; this is especially appropriate in the application of multivariate techniques to the real-world problems of petroleum well optimization

• Language: English
• Publisher: Elsevier Science
• Release date: Sep 22, 2021
• ISBN: 9780323902328

Rasool Khosravanian

Rasool Khosravanian has worked as a postdoctoral fellow sponsored by Equinor and Aker BP, in the Department of Energy and Petroleum Engineering (IEP), University of Stavanger, Norway, since 2019. His focus has been on implementing digitalization in a drilling and wells organization. He holds MSc and PhD degrees in industrial engineering from the Iran University of Science and Technology in optimization techniques in the petroleum industry. Rasool received his BS degree in drilling and mining engineering from Kerman University. He was a faculty member and an assistant professor at Amirkabir University of Technology (Tehran Polytechnic) from 2011 to 2018. His research interests include large-scale optimization, data mining, artificial intelligence (AI), megaproject management, engineering economics, and risk and uncertainty analysis. He has published over 27 papers in international journals and 40 conference papers, with 10 years of drilling experience working both in academic research and with the petroleum industry. He has six years of professional experience from EPD companies and has also been a strategic planner in the implementing of business strategy for largesized companies. He is a member of the Society of Petroleum Engineers (SPE) and Tekna in Norway.

Book preview

Methods for Petroleum Well Optimization

Automation and Data Solutions

Rasool Khosravanian

Bernt S. Aadnøy

Table of Contents

Cover image

Title page

Copyright

Preface

Acknowledgment I

Acknowledgment II

Chapter One. Introduction to digital twin, automation and real-time centers

1.1. Digital twin technology

1.2. Drilling automation

1.3. Real-time centers

1.4. Summary

1.5. Problems

Chapter Two. Petroleum well optimization

2.1. Mathematical optimization

2.2. Petroleum well optimization

2.3. Summary

2.4. Problems

Nomenclature

Chapter Three. Wellbore friction optimization

3.1. Elementary models for wellbore friction

3.2. Advanced models for wellbore friction

3.3. Application of friction models to wells

3.4. Design of oil wells using analytical friction models

3.5. Summary

3.6. Problems

Nomenclature

Chapter Four. Wellbore trajectory optimization

4.1. Introduction

4.2. Constraints potentially affecting the optimal well trajectory

4.3. Well path optimization

4.4. Well trajectory optimization for preventing wellbore instability

4.5. Summary

4.6. Problems

Nomenclature

Chapter Five. Wellbore hydraulics and hole cleaning: optimization and digitalization

5.1. Hydraulic optimization

5.2. Hole cleaning

5.3. Real-time assessment of the hole cleaning efficiency

5.4. New methods for drilling hydraulics

5.5. Summary

5.6. Problems

Nomenclature

Chapter Six. Mechanical specific energy and drilling efficiency

6.1. Introduction to mechanical specific energy

6.2. Mechanical specific energy: next-generation digital drilling optimization

6.3. Rock drillability assessments

6.4. Drilling system energy beyond mechanical specific energy

6.5. Summary

6.6. Problems

Nomenclature

Chapter Seven. Data-driven machine learning solutions to real-time ROP prediction

7.1. Introduction

7.2. Data piping in real time

7.3. Drilling rate of penetration optimization workflow

7.4. Statistical and data-driven rate of penetration model

7.5. Summary

7.6. Problems

Nomenclature

Chapter Eight. Advanced approaches and technology for casing setting depth optimization

8.1. Introduction

8.2. Problem statement

8.3. Mathematical approach: casing string placement optimization under uncertainty

8.4. Multiple criteria approach: casing seat selection method

8.5. Real-time approach: casing seat optimization using remote real-time well monitoring

8.6. Technological approach: reduced number of casings using unconventional drilling methods

8.7. Summary

8.8. Problems

Nomenclature

Chapter Nine. Data mining in digital well planning and well construction

9.1. Data mining techniques

9.2. Data mining application in digital drilling engineering

9.3. Summary

9.4. Problems

Chapter Ten. Well completion optimization by decision-making

10.1. Basic concepts

10.2. Well completion optimization by decision-making

10.3. Summary

10.4. Problems

Nomenclature

Chapter Eleven. Monte Carlo simulation in wellbore stability optimization

11.1. Basic multivariate statistics

11.2. Uncertainty assessment of wellbore stability

11.3. Numerical examples

11.4. Summary

11.5. Problems

Nomenclature

Chapter Twelve. Case-based reasoning (CBR) in digital well planning and construction

12.1. Basic concepts

12.2. Application of case-based reasoning in digital well construction planning

12.3. Summary

12.4. Problems

Nomenclature

Index

Copyright

Gulf Professional Publishing is an imprint of Elsevier

50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States

The Boulevard, Langford Lane, Kidlington, Oxford, OX5 1GB, United Kingdom

Copyright © 2022 Elsevier Inc. All rights reserved.

No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions.

This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein).

Notices

Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary.

Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility.

To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein.

Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library

ISBN: 978-0-323-90231-1

For information on all Gulf Professional Publishing publications visit our website at https://www.elsevier.com/books-and-journals

Publisher: Joe Hayton

Senior Acquisitions Editor: Katie Hammon

Editorial Project Manager: Madeline Jones

Production Project Manager: Kamesh Ramajogi

Cover Designer: Christian J. Bilbow

Typeset by TNQ Technologies

Preface

In today's world, there are two paths: navigating to a new digital future or being engulfed by exponential competitive change. For this reason, many companies have started to focus on digitalization, optimization, and tight control of their core operations while at the same time innovating.

Optimization and decision-making during the planning and execution of drilling, and afterwards during well operations, is challenging due to subsurface uncertainty, limited availability of measurements, and the need for interaction and collaboration between different disciplines. However, digitalization has driven a radical shift in how we can manage and run well operations remotely. It is a growing force in the offshore oil and gas industry too. Its potential to optimize operations, including increasing safety and quality, and reducing risk, is a strong driver for an industry with ever-rising costs.

The coronavirus (COVID-19) crisis has accelerated the pace of digitalization beyond anything we could have imagined. Many companies have already boosted their digital transformation with the use of cloud computing, big data, artificial intelligence (AI), and the Internet of Things (IoT). This has made a significant contribution to the transformation toward fully connected and automated systems that will result in high-performance operations in different industries, including petroleum and energy.

Data mining, metaheuristic optimization algorithms, multiple-criteria decision-making (MCDM), case-based reasoning (CBR), Monte Carlo Simulation, and machine learning (ML) are attractive tools in this age of artificial intelligence. ML algorithms, such as deep learning, could form one of the fundamental pillars for prediction and optimization in petroleum well operations across the industry. This book, the first of its kind, presents a unique, understandable view of optimization, machine learning, and the other available tools, using many practical examples.

Models do not deliver enough value if there is no direct path to code production, yet, up to now, there has not been one cohesive resource that bridges between theory and application, showing how to go from models to code. Therefore, in this book, we have focused on giving today's engineers and R&D teams real-time data solutions specific to drilling and production assets.

In this book, you will learn how to translate an executable model of your application into running code. Information on how to access the relevant open-source code has been given at the end of most chapters.

The availability of open-source code seems to be helping to promote digitalization. Also, open discussions about the particular advantages and drawbacks of specific code or software will help companies, developers, and users view the trade-offs between the different software.

Digitalization is the way forward, and we believe it is time for MSc and PhD programs in universities' petroleum and energy departments to focus on the knowledge and skills required to tackle the oil and gas industry's most challenging problems using digitalization. Such a program could be an interdisciplinary course, encompassing a range of updated petroleum engineering fundamentals, which would produce technically well-prepared graduates with a sound knowledge of the industry. This book provides support for such a course, filling the gaps between theory and practice in earlier text books.

Finally, we think the way forward for companies is neither to adopt a wait-and-see strategy to get a better picture of how digitalization develops before implementing it themselves nor to pursue a conservative digitalization strategy. The authors of this book propose that researchers, and oil and gas companies, their CEOs, managers, and engineers should understand the significant impact that digitalization can have and accelerate its integration into their business's core priorities.

Stavanger, July 2021

Rasool Khosravanian

Bernt S. Aadnøy

Acknowledgment I

I would like to express my very great appreciation to Professor Bernt Sigve Aadnøy for his support, guidance, and collaboration during the preparing of this book. Without his experience, knowledge, and support, this book would not exist.

I would like to acknowledge my gratitude to Equinor (Academia Program) and Aker BP which sponsored my postdoctoral position and provided me with the opportunity to further my research into oil well optimization.

I would also like to thank Øystein Arild, Head of the Department of Energy and Petroleum Engineering, for his support during my postdoctoral studies at the University of Stavanger.

My grateful thanks are also extended to Joanne Stone who is a professional native English language editor for her help in editing this book and for her valuable time and suggestions during the preparation of this book. The assistance provided by Elsevier in producing this book was greatly appreciated. Thanks to everyone on the publishing team.

Not least, I want to thank my wife and daughter, Doctor Maryam and Sana, thank you!

Stavanger, July 2021

Rasool Khosravanian

Acknowledgment II

12 years ago I started working with Rasool Khosravanian when he worked on his PhD project on multivariate analysis of casing setting depth in Iran. We have since cooperated over the years, and 2   years ago, he moved to Norway as an invited researcher sponsored by Equinor. We decided then that we should write a book on optimization. Actually, many models exist, but they needed to be put in the right context for optimization analysis.

Rasool has been very dedicated to the book project and has put in a considerable effort. Whenever I visited him, he was working on the book. Luckily we got Joanne Stone, a language professional, into the project and she has improved both the language and the structure considerably.

The oil industry is now working towards digitalization at all levels. We hope that our book will contribute to this development with the ultimate goal of increased efficiency at a lower cost. Since the way the petroleum industry is working is changing, we also plan to implement courses at the University of Stavanger introducing multivariate computerized solutions in well engineering to prepare the students for the digitalized industry.

We are highly appreciative of the positive support we have received from the University of Stavanger's Department of Energy and Petroleum, from Equinor and Aker BP, and from the many individuals with whom we have discussed our work.

Stavanger, July 2021

Bernt S. Aadnøy

Chapter One: Introduction to digital twin, automation and real-time centers

Abstract

The oil and gas industries are now moving toward using digital twin during the life cycle of well construction. Digital twin driven by real-time data helps to design the optimal plan for the operation with a focus on safety, risk reduction, and improved performance.

The objective of the chapter is to describe and present results of using a digital twin in drilling operations (planning and engineering, training, and operational support).

In this chapter, we describe an onshore operations center, from which a skilled team provides planning and engineering expertise, and support to offshore crews. Using IT and communication technology to share information and data sets for analysis, preventive and corrective actions, and problem solving, allows the members of the team to work together to deliver first-class results in terms of operational performance, and also with regard to QHSE standards.

Keywords

Automatic drilling systems; Digital twin technology; Drilling automation; Integrated real-time operation center services; Onshore collaboration center; Well construction

Key concepts

1. A five-dimension model provides reference guidance for understanding and implementing digital twin. We look at the frequently used technologies and tools for digital twin to provide a guide to how digital twin models could be employed in the future.

2. Oil and gas operators and service providers are now undergoing digital transformation to enable them to thrive in a digital environment and to gain a competitive advantage. Drilling generates large volumes of data from many sources, which leads the industry into the world of big data. As technology is advancing rapidly, the industry has started to talk about drilling automation, machine learning, artificial intelligence, and big data analytics. To be able to use these, drillers need employees with the appropriate technical expertise and data that is provided in the correct digital format.

1.1. Digital twin technology

1.1.1. Digital twins

A digital twin is a virtual and simulated model or a true replica of a physical asset. It is a computerized companion of the physical asset and can be used for various purposes as depicted in Fig. 1.1 below.

The model in Fig. 1.1 specifically finds expression through five enabling components: sensors and actuators from the physical world, integration, data, analytics, and the continuously updated digital twin application. These constituent elements are explained at a high level below (Parrott and Warshaw, 2017):

• Sensors: Sensors distributed throughout the manufacturing process create signals that enable the twin to capture operational and environmental data pertaining to the physical process in the real world.

• Data: Real-world operational and environmental data from the sensors are aggregated and combined with data from the enterprise.

• Integration: Sensors communicate the data to the digital world through integration technology (which includes edge, communication interfaces, and security) creating a two-way link between the physical world and the digital world.

Figure 1.1  Visualization of digital twin creation. 

Modified from Offshore Engineer, 2019. Visualization of Digital Twin Creation. https://www.oedigital.com/news/470548-digital-twin-taking-shape-of-the-offshore-ecosystem.

• Analytics: Analytics techniques are used to analyze the data through algorithmic simulations and visualization routines that are used by the digital twin to produce insights.

• Digital twin: The digital side of Fig. 1.1 is the digital twin itself, an application that combines the components above into a near real-time digital model of the physical world and processes. The objective of a digital twin is to identify intolerable deviations from optimal conditions along any of the various dimensions. Such a deviation is a case for business optimization; either the twin has an error in the logic (hopefully not), or it has identified an opportunity for saving costs, improving quality, or achieving greater efficiencies. The identification of an opportunity may result in an action in the physical world.

• Actuators: Should an action be warranted in the real world, the digital twin produces the action by way of actuators, subject to human intervention, which triggers the physical process.

Below we describe the key technologies for digital twin from three perspectives: data-related technologies, high-fidelity modeling technologies, and model-based simulation technologies. Fig. 1.2 presents the technology architecture for digital twin.

Figure 1.2 Technology architecture for digital twin. 

Modified from Liu, M., Fang, S., Dong, H., Xu, C., 2021. Enabling technologies and tools for digital twin. J. Manuf. Syst. 58 (Part B), 3–21. https://doi.org/10.1016/j.jmsy.2019.10.001.

1.1.1.1. Data-related technologies

Data are the basis of digital twin. Sensors, gauges, RFID tags and readers, cameras, scanners, and other equipment should be chosen and integrated to collect total-element data for digital twin. Data then should be transmitted in a real-time or near real-time manner. However, the data required for digital twin are usually of big volume, high velocity, and great variety, and such data are difficult and costly to transmit to the digital twin in the cloud server. Thus, edge computing is an ideal method for pre-processing the collected data to reduce the network burden and eliminate the chances of data leakage, and real-time data transmission is made possible by 5G technology. Data mapping and data fusion are also needed to understand the collected data.

1.1.1.2. High-fidelity modeling technologies

The model is the core of digital twin. Models of digital twin comprise semantic data models and physical models. Semantic data models are trained by known inputs and outputs, using artificial intelligence methods. Physical models require comprehensive understanding of the physical properties and their mutual interaction. Thus, multi-physics modeling is essential for high-fidelity modeling of digital twin.

1.1.1.3. Model-based simulation technologies

Simulation is an important aspect of digital twin. Digital twin simulation enables the virtual model to interact with the physical entity bi-directionally in real time.

Kritzinger et al. (2018) classified three uses of the term digital twin, based on the level of data integration between the physical asset and digital representation in the described digital twin: digital model (DM), digital shadow (DS), and digital twin (DT). When there is no automatic real-time data communication between the physical asset and the digital representation, as in Fig. 1.3, then the described digital twin is classified as a digital model. When there is automatic real-time communication from the physical representation to the digital twin but not from the digital representation to the physical asset, as in Fig. 1.4, then the described digital twin is classified as a digital shadow. Only when there is automatic real-time communication both from the physical asset to the digital representation and from the digital representation to the physical asset, as in Fig. 1.5, is the described digital twin classified as a proper digital twin.

1.1.2. Five-dimension digital twin model

The five-dimension digital twin model can be formulated as Eq. (1.1) (Tao et al., 2018a).

(1.1)

where: PE are physical entities; VM are virtual models; Ss are services; DD is digital twin data; and CN are connections. The five-dimension DT model expressed in this formula is illustrated in Fig. 1.6.

Figure 1.3 Data flow in a digital model. 

Modified from Kritzinger, W., Karner, M., Traar, G., Henjes, J., Sihn, W., 2018. Digital twin in manufacturing: a categorical literature review and classification. IFAC-Pap. Online 2018 51 (11), 1016–1022.

Figure 1.4 Data flow in a digital shadow. 

Modified from Kritzinger, W., Karner, M., Traar, G., Henjes, J., Sihn, W., 2018. Digital twin in manufacturing: a categorical literature review and classification. IFAC-Pap. Online 2018 51 (11), 1016–1022.

Figure 1.5 Data flow in a digital twin. 

Modified from Kritzinger, W., Karner, M., Traar, G., Henjes, J., Sihn, W., 2018. Digital twin in manufacturing: a categorical literature review and classification. IFAC-Pap. Online 2018 51 (11), 1016–1022.

Figure 1.6 Five-dimension digital twin model.

1.1.2.1. Physical entities in digital twin

DT is used to create virtual (digital) models of physical entities to simulate their behaviors digitally (Tao et al., 2018b). The physical world is the foundation of DT. For the purposes of DT, the physical entity may be a device or product, physical system, activities process, or even an organization. These entities act according to physical laws and are subject to uncertainty in their environments. Physical entities can be divided into three levels, according to their function and structure: unit level, system level, and system of system (SoS) level (Tao et al., 2019).

1.1.2.2. Virtual models in digital twin

Virtual models should be faithful replicas of physical entities that reproduce the geometries, properties, behaviors, and rules of the original. The three-dimension geometric model describes a physical entity in terms of its shape, size, tolerance, and structural relationships. Based on physical properties (e.g., speed, wear, and force), the physical model reflects the physical phenomena of the entity, such as deformation, delamination, fracture, and corrosion. The behavior model describes the behaviors (e.g., state transition, performance degradation, and coordination) and response mechanisms of the entity to changes in the external environment. The rule model endows DT with logical abilities such as reasoning, judgment, evaluation, and autonomous decision-making, by following the rules extracted from historical data for the physical entity or supplied by domain experts.

1.1.2.3. Digital twin data

Twin data is a key driver of digital twin. DT deals with multi-temporal scale, multi-dimension, multi-source, and heterogeneous data. Data can be acquired in the following ways: obtained from physical entities, including static attribute data and dynamic condition data; generated by virtual models, which reflect the simulation result; obtained from services, and describe the service invocation and execution; and in the form of knowledge provided by domain experts or extracted from existing data.

1.1.2.4. Services in digital twin

With product-service integration now taking place in all aspects of modern society, more and more enterprises have begun to realize the importance of service. Service is an essential component of DT in light of the paradigm of Everything-as-a-Service (XaaS). First, DT provides users with application services related to simulation, verification, monitoring, optimization, diagnosis and prognosis, prognostic and health management (PHM), etc. Secondly, a number of third-party services are needed to build a functioning DT, such as data services, knowledge services, algorithms services, etc. Lastly, the operation of DT requires the continuous support of various platform services, which can accommodate customized software development, model building, and service delivery.

1.1.2.5. Connections in digital twin

Digital representations are connected dynamically with their real counterpart to enable advanced simulation, operation, and analysis. DT has six pairs of connections between physical entities, virtual models, services, and data: physical entities and virtual models (CN_PV); physical entities and data (CN_PD); physical entities and services (CN_PS); virtual models and data (CN_VD); virtual models and services (CN_VS); and services and data (CN_SD) (Tao et al., 2018a). These connections enable information and data exchange between the four parts.

Through its integration with mobile internet, cloud computing, big data analytics and other technologies, DT is potentially applicable to many fields which require the mapping, fusion, and co-evolution of the physical and virtual spaces. The applications of DT are illustrated in Fig. 1.7.

Figure 1.7 Different application fields of digital twin. 

Modified from Qi et al. (2021).

1.1.3. Value of digital twin

Building on a report from Oracle, eight value additions of digital twin have been identified (Rasheed et al., 2020):

• Real-time remote monitoring and control: Generally, it is almost impossible to gain an in-depth view of a very large system physically in real time. A digital twin can be accessed anywhere. Not only can the performance of the system be monitored but it can also be controlled remotely using feedback mechanisms.

• Greater efficiency and safety: It is envisioned that digital twinning will enable greater automation, with humans interfering as and when required. This will allow dangerous, dull and dirty jobs to be allocated to robots, with humans controlling them remotely. In this way humans will be able to focus on more creative and innovative jobs.

• Predictive maintenance and scheduling: Comprehensive digital twinning will ensure that multiple sensors monitoring the physical assets will generate big data in real time. Smart analysis of that data will allow faults in the system to be detected early or to be anticipated before they occur. This will enable better scheduling of maintenance.

• Scenario and risk assessment: A digital twin, or to be more precise a digital sibling of the system, will enable what-if analyses, resulting in better risk assessment. It will be possible to perturb the system to synthesize unexpected scenarios and study the response of the system as well as the result of the corresponding mitigation strategies. Using a digital twin is the only way to perform this kind of analysis without jeopardizing the real asset.

• Better intra-team and inter-team synergy and collaborations: With greater autonomy and all the information at their fingertips, teams can better utilize their time in improving synergies and collaborations, leading to greater productivity.

• More efficient and informed decision-making: The availability of quantitative data and advanced analytics in real time will enable more informed and faster decision-making.

• Personalization of products and services: With detailed knowledge of historical requirements and the preferences of the various stakeholders, as well as evolving market trends and competitive environments, demand for customized products and services is bound to increase. For the factories of the future, a digital twin will enable faster and smoother gear shifts to address changing needs.

• Better documentation and communication: Readily available information in real time combined with automated reporting will help keep stakeholders well informed about the drilling operations, improving transparency.

1.1.4. Modeling basis used in digital twin development

The basis of the digital twin system is established from the first principles of physics. The digital twin uses all available real-time drilling data, both surface and downhole. These are linked to and combined with real-time modeling to supervise and optimize the drilling process. The real-time data, well configuration data, and other relevant data are used to visualize the wellbore and the drilling process status in 3D in real time. The digital twin is composed of the following (Mayani et al., 2020):

• An advanced and fast integrated drilling simulator, integrating transient hydraulic, thermal, and mechanical models: The integrated drilling simulator models the different drilling sub-processes dynamically. The interactions between the sub-processes are also modeled in real time. This simulator is used to perform forward-looking simulations automatically and can be used for planning revisions on the fly (what-if) as well.

• Automatic correction and quality checking of the drilling data: This ensures the data are suitable for processing by computer models.

• Algorithms that monitor the drilling process in real time by means of a combination of time-based drilling data and real-time modeling of data results.

• Algorithms for diagnosis of the drilling state.

• A 3D visualization (Virtual Wellbore), with dynamic visualization of the downhole process.

• Data flow and computer infrastructure.

1.1.5. Monitoring of the drilling wells using digital twin

Digital twin was first used in drilling monitoring with the introduction of advanced monitoring, which is the most recent evolution in drilling monitoring. Automated monitoring offers real-time simulations, comparing simulation results with measurements in real time, detecting diagnostics automatically as well as manually, and self-detection of problems. In the next stage of this evolution drilling operations will move toward real-time optimization. It will be possible to perform forecasting simulations, forward-looking simulations, what-if simulations and predictive analysis.

The digital twin virtual well consists of an advanced mathematical model which includes a complex mechanical model as well as a hydraulic model. Examples of the parameters that these models can calculate include:

• pressure

• SPP (standpipe pressure)

• ECD (equivalent circulating density)

• temperature

• choke pressure

• pore and fracture pressure

• pit gain

• cuttings concentration

• rate of penetration

• wellbore stability

• torque and drag

• torque on the bit

• torque at the surface

• hook load

• static, axial, and rotational friction

• string elongation

• block speed

Real-time mathematical models utilize real-time drilling data sent from the rig. The models compare real-time downhole measurements with the modeled parameters to monitor downhole conditions during drilling and casing operations. This allows the early detection of symptoms of downhole deterioration, which are shown in the automatic diagnostic messages provided by the model. Thus, the use of digital twin helps to improve drilling and casing performance based on downhole conditions (Figs. 1.8 and 1.9).

The digital twin of the well is visualized in both the 2D and 3D real-time views during drilling. The 2D view uses all available real-time drilling data including surface and downhole data in combination with advanced monitoring models to monitor and provide advisory for more optimal drilling. The various drilling models interact, and the measured values and the calculated results are visualized in real time in a graphical user interface. The 2D virtual well contains the wellbore geometry, tubular properties, drill string, temperature profile, pressure profile including pore and fracture profile, and risk messages icons. ECD variations and the comparison with pore and fracture pressure are also visible in the 2D illustration.

The diagnostic technology is combined with the 3D visualization to create a virtual wellbore. The 3D visualization together with diagnostic updates and virtual gauges can provide a better understanding of well conditions throughout the drilling operation. Bit depth, ECD values and all other information can be monitored using sensors. This data can be included and updated in the visualization.

Fig. 1.10 shows a typical 2D illustration, and Fig. 1.11 shows a typical 3D visualization. The 2D and 3D views of the digital twin can be used during the whole life cycle of the drilling operation as well as during post analysis, training, and experience transfer. For experience transfer and learning the whole operation can be replayed and displayed.

Figure 1.8 Sketch of a well with a digital twin illustrating how estimated pressure values can be extracted everywhere along the flow path. A similar extraction can be done for all modeled variables (such as flow rates, densities, and ECD). 

Modified from Gjerstad, K., Bergerud, R., Thorsen, S.T., 2020. Exploiting the full potential in automated drilling control by increased data exchange and multidisciplinary collaboration, SPE Annual Technical Conference and Exhibition, Virtual Event Held 26–29 October 2020.

Figure 1.9 Real-time mathematical models utilize real-time drilling data. 

Modified from Bergerud, R., August 2016. Powerpoint Presentation: Drilltronics Drilling Process Automation – Statfjord C, Ronny Bergerud Operation Manager Drilltronics. https://docplayer.net/191389103-Drilltronics-drilling-process-automation-statfjord-c.html.

Figure 1.10 2D visualization of digital twin during drilling in auditorium. 

(Aker BP’s flexible & multipurpose Onshore Collaboration Center, Photographer: Bjørn O. Bådsvik).

Figure 1.11 Real-time 3D visualization on video wall in rig-room. 

(Aker BP’s flexible & multipurpose Onshore Collaboration Center).

Figure 1.12 Digital twinning for well construction. 

Modified from Germain, O., McMullin, D., Tirado, G., 2018. Using an E&P digital twin in well construction. In: Embracing the Digital Twin for E&P an iEnergy® eBook. Halliburton Landmark, pp. 27–35.

The properties of the wellbore, drill string, and reservoir are set using field development planning and well construction software that create the basis for the intended design and modeled physical constraints. A high-level view of this complete system is shown in Fig. 1.12. By integrating all the modeling parts, available data from past wells and live access to real-time data, it is possible to create the richest unified representation of well construction preparation or execution. The unified set of models are a combination of plans and actuals, including their respective interpretation and execution uncertainty–thus helping to achieve the highest possible level of fidelity based on the latest information. Models are continually updated, and plans are kept live; consistency is achieved due to their unified representation (see Fig. 1.12).

Digital twins provide us with the ability to investigate the future by combining historical data, real-time data, and physics-based models. There are many applications in the oil and gas domain that apply the data to physics-based models to make predictions about the various drilling processes, systems, and associated equipment.

1.1.6. The concept of digital twinning for well construction

Well construction involves a multitude of physical processes, measurements, control applications, analyses, and decision loops, spanning a range of temporal resolutions and response times or delays. For simplicity, our discussion divides resolution and delay into four groups (see Fig. 1.13): sub-second, sub-minute, intermediate (minutes), and long (hours).

This range of requirements for data resolution and delay in system response indicates that it is likely that closed-loop, sub-second response needs will be met by a locally controlled rig-specific system. Sub-minute, intermediate, and long responses will cover the broader process, in which drilling data are applied automatically or manually to influence or make decisions about the ongoing drilling process, as is believed to be the norm today. It is suggested that a new term, control-time, be introduced for this.

Figure 1.13 Typical timescales in drilling-process management. 

Modified from Thorogood, J., Aldred, W. D., Florence, F., Iversen, F., 2010. Drilling automation: technologies, terminology, and parallels with other industries. SPE Drill. Complet. 25 (4), 419–425. https://doi.org/10.2118/119884-PA, SPE/IADC Drilling Conference and Exhibition, Amsterdam, The Netherlands, 17–19 March 2009. https://doi.org/10.2118/119884-PA.

Control-time describes the resolution required of the system surface data and control system algorithms for the control of drilling-machinery parameters such as pump rate, hook load, pipe rotation, pipe velocity, and rate of penetration (ROP). The resolution of the downhole data available to the control system during normal operations is lower, currently constrained by the bandwidth of the mud-pulse telemetry. However, the introduction of wired-pipe technology will make these measurements also available to the control system with sub-second resolution. There is also a question of the time it takes to perform a measurement. Some measurements are instantaneous, such as motor torque or standpipe pressure, while others take time and may involve applications of models, such as derived fracture pressure from a leak-off test or wellbore friction from a friction test. However, measurements with a long control time are generally not applied as feedback variables in closed-loop control algorithms, although they could trigger action. For example, as a result of taking a leak-off test, the fracture pressure constraint may be updated in the automated system. In another case, friction analysis may be applied to detect potential hole-cleaning issues, with possible triggering of mediating action (Thorogood et al., 2010).

1.2. Drilling automation

Automation is broadly defined as a technology dealing with the application of mechatronics and computers in the production of goods (manufacturing automation) and services (service automation). Automation can also be defined as the replacement of human labor by electronic or/and mechanical devices. This definition has broadened over time. First, it covers many processes: in the case of drilling, for example, not solely the operation of the drill. Second, the human labor that it replaces can be both physical and mental (Iversen et al., 2013).

Reasons for implementing automation are:

1. a shortage of labor,

2. a high cost of labor,

3. to increase productivity,

4. to reduce costs, and

5. to reduce process lead time.

The evolution of automation can be divided into three eras: mechanization, semi-automation, and local automation. Mechanization means replacing human labor by mechanical power which provides more torque and force. Semi-automation means some of the mechanical operations are automated but skilled human operators are needed to feed the automated machines with the required data. Local automation removes the need for a human interface from the semi-automated operation.

There are three basic categories of automation: fixed automation, programmable automation, and flexible automation (Fig. 1.14).

Sheridan (2002) refers to the human-automation system which he divides into two categories, mechanization, and computerization. Here mechanization means replacing human labor by machines that are physically controlled by a human. Computerization means that the process is operated or controlled by a computer, which is itself controlled by a human, thus providing an interface between human and machine.

Sheridan (2002) categorized automation into four types:

1. mechanization and sensing integration;

2. data processing and decision-making;

3. mechanical and information action; and

4. open-loop operation on closed control.

The precise definition of automation varies according to the industry and technology to which it is applied.

Drilling automation is used by drilling engineers and is an example of Sheridan's human-automation system. Computers are used to control and manage the parameters affecting the drilling operation such as flow rate, downhole pressure (DHP), mud weight (MW), pore pressure (PP), fracture pressure (FP), and so on.

Within drilling, the application of automation is expanding to include drilling machinery, sensor technology, control systems, and computer and communications technology. This explosion of technology is leading a change in drilling automation from the machine level to fully integrated operations (Iversen et al., 2013).

Figure 1.14 The three basic categories of automation.

Defining and recognizing automation as a term or/and as a process level is important to identifying how it can be applied to the operations of the different segments of the oil and gas industry, such as contractors, services, and operating companies.

1.2.1. Automation levels

Automation levels range from a fully manual system (meaning no automation) to a fully automated system, with the semi-automated levels in between having varying degrees of manual and automated operations. A semi-automated system contains decision and action options, which are either assigned to the operator or the computer. If the computer is assigned fewer decision and action options than the operator, then the level is closer to the fully manual level; if the operator is assigned fewer decision and action options than the computer, the level is closer to the fully automated level.

1.2.2. Modeling

Making a model is a process of using pre-existing (historical) data and real-time data. Thus, modeling uses the work done and the optimization processes. There are some parameters that affect real-time data and thus affect modeling (Thorogood et al., 2010), such as:

• functionality type

• frequency

• set point

• reaction time

The functionality of drilling operations could be classified as an open-loop system. The exception is if it is imitated by many real-time issues requiring closed loop. The affecting parameters are:

• a flexible and scalable model accepting additional functionality;

• missing data;

• limited data transmission bandwidth;

• diagnostic algorithm effect on bandwidth;

• modal accuracy estimating under abnormal situations such as missing data;

• fast set-point change under sudden parameters change; and

• physical machine response.

There are many drilling models today, including the earth seismic model, the drilling optimization model, and the fluid model, that control drilling operations such as ROP, cement circulation, tripping, wellbore pressure, and drill string vibration. These models currently work independently, but through automation it may be possible to integrate them into a general drilling automation system based on safety, performance, and economics.

Well construction depends on the analysis of formation behavior based on information taken from previous drilling operations or study reports. This information is used to build up the automation models, and the information can be updated manually, although an electronic source is recommended to ensure high-quality automation.

Remote support and decision-making have both a direct and indirect relationship with the drilling procedures and data resolution that are used for estimations to help in decision-making. This requires the parameters to be updated and then fed back into the models for automated optimization. Time-scale analysis is a central item in the updating of parameters that helps in decision-making on how to manage and update the whole automation system.

Data resolution and response time are important factors for the well instruction processes. Resolution and delay are divided into four groups: sub-second, sub-minute, intermediate (minutes), and long (hours) (Fig. 1.13).

The sub-second response works in a specific system while the other responses work in wider operations.

Resolution and delay are also called control-time, which deals with resolution and control algorithms to control the drilling operation's various parameters. Control-time is divided into instantaneous and long time, where long time cannot be used as a feedback in the control algorithm.

1.2.3. Data communication

In the past, the data used for monitoring was adjusted by the operators, using normalization/rows, log transformations, and various filters. With the coming of mechanical drilling, some companies started to use data to plan a drilling program as well as for monitoring. Data recovery started with the advent of electronic communication which made it possible to access data via a network connection to be used widely as a planning tool. For an automated system, the data used depend on some conditions such as availability, completeness, and accuracy.

The short-trend operations accepted some incorrect data, but long-trend operations did not because it affected the operations' performance.

Unconditional data exchange creates problems in the system, so it is important to choose the right data exchange by following a standard communication protocol. In general, the protocols and protocol responses should follow the system's requirements and data requirements where some data depict slow changes and other data depict a quickly changing situation.

System integration is one of the complexities facing automation drilling because of many factors such as:

• poor quality of the information about the system available to the operator;

• the need to avoid information overload between the system and the operator, especially when connecting multiple services;

• the need to initiate standard change-management techniques which have an effect on the magnitude of process changes; and

• the system security, which is a challenge for the industry because of the potential for miscommunication between the different parties (operator, contractor, and third parties).

The machine and model interface is an important issue in automation drilling where machines emulate human actions to execute a process, with the help of real-time data from the models' sensors to update the data for the machine action. This type of continuous communication in the system improves the drilling operation and provides standards of automation that keep up with technological advances (Thorogood et al., 2010).

1.2.4. Modes of automation

Automation modes are classified according to the feature level of the operator and the automation system in that mode. Broadly, there are three automation modes, fully manual, semi-automated, and fully automated. Within the semi-automated mode there are five modes, which are differentiated from each other by the responsibility/feature level of the operator and the automation system. There are, therefore, seven modes in total, briefly described below (see Thorogood et al., 2010; and Ornaes, 2010). Some of the terms used need additional explanation, such as envelope protection, closing the loop, multilevel control structures, feedback control, supervisory control, optimization, and autonomy. We will provide this in the following sections.

Mode 0: Direct manual control mode. In this mode, the driller will receive no support at all from the automation system. The driller is presented with raw signals and simple alarms associated with top-side hardware.

Mode 1: Assisted manual control mode. The significant contribution of the automation system in this mode is the introduction of software that analyzes the current situation of the well and presents the information to the driller. This will improve the quality of the decision-making of the driller.

Mode 2: Shared control mode. This is the first mode at which the automation system will start to directly interfere with the operation of equipment. The main feature of this mode will be envelope protection.

Mode 3: Management by delegation. Some of the operator's tasks are delegated to the automation system and are fully automated by a closed-loop controller.

Mode 4: Management by consent. The automation system provides regulated multiple control loops, where models describing the well reach the right control loop. The operator feeds the automation system with the operation to be performed, the operation goals, the chosen variables, and their desired values.

Mode 5: Management by exception. The automation system decides on the next operational mode by additional logic, and the operator's role is to monitor and interfere if the system does not behave as expected.

Mode 6: Autonomous operation. A fully automated system, and the role of the operator is just to monitor the system.

In all seven modes, the operator retains authority over the operation and remains the main decision-maker for the whole operation to address any risks that may arise.

Figure 1.15 Envelope protection automation. 

Modified from Breyholtz, Ø., Nikolaou, M., 2012. Presenting a Framework for Automated Operations. Society of Petroleum Engineers, pp. 118–126. https://doi.org/10.2118/158109-PA.

1.2.4.1. Envelope protection automation

An envelope protection system takes the well conditions into consideration when calculating the boundaries to be implemented at an offshore installation (Iversen et al., 2009). Therefore, the envelope protection may take the following issues into consideration (see Fig. 1.15):

• Envelope protection sets boundaries and limitations depending on the well conditions and information available.

• The protection system will only interfere when the driller/operator exceeds these boundaries.

• Envelope boundaries must be calculated dynamically and updated according to the new well conditions.

• The dynamic calculations require a computational model that is highly expensive.

• Envelope protection reduces the frequency of critical situations arising but does not eliminate them.

1.2.4.2. Closed-loop automation

Closed-loop control is a higher level of automation than the envelope protection system. In this control system, the set point/control value is defined and set by the operator either manually or automatically. If automatically, the operator uses an automated system to find and update the set point. The closed-loop control system uses an algorithm to calculate the deviation in the real-time measurement from the set point and then activates an order or process to return the operation to the set point. This type of control system requires a large amount of data for the multilevel control structure and decision-making for the whole operation, which can be supplied by a data acquisition system. Figs. 1.16 and 1.17 show a closed-loop system.

1.2.4.3. Multilevel control structure

The timescale is the key element in the structure of a multilevel control system. The timescale ranges from zero for the upper level to infinity for the lowest level because of the difference in the timescale between levels. The higher-level co-ordinates with the lower level to reach the goal of the control system. Optimization and decision-making also depend on the timescale/time length, so defining the control level type (higher or lower) is necessary for deciding on the appropriate systematic control hierarchy for an operation such as drilling, for which multilevel