Machine Learning and Data Science in the Oil and Gas Industry

1. Introduction

 1.1 Who this book is for
 1.2 Preview of the content
 1.3 Oil and gas industry overview
 1.4 Brief history of oil exploration
 1.5 Oil and gas as limited resources
 1.6 Challenges of oil and gas
 References

2. Data Science, Statistics, and Time-Series

 2.1 Measurement, uncertainty, and record keeping
    2.1.1 Uncertainty
    2.1.2 Record keeping
 2.2 Correlation and timescales
 2.3 The idea of a model
 2.4 First principles models
 2.5 The straight line
 2.6 Representation and significance
 2.7 Outlier detection
 2.8 Residuals and statistical distributions
 2.9 Feature engineering
 2.10 Principal component analysis
 2.11 Practical advice
 References

3. Machine Learning
 
 3.1 Basic ideas of machine learning
 3.2 Bias-variance complexity trade-off
 3.3 Model types
    3.3.1 Deep neural network
    3.3.2 Recurrent neural network or long short-term memory network
    3.3.3 Support vector machines
    3.3.4 Random forest or decision trees
    3.3.5 Self-organizing maps (SOM)
    3.3.6 Bayesian network and ontology
 3.4 Training and assessing a model
 3.5 How good is my model?
 3.6 Role of domain knowledge
 3.7 Optimization using a model
 3.8 Practical advice
 References

4. Introduction to Machine Learning in the Oil and Gas Industry

 4.1 Forecasting
 4.2 Predictive maintenance
 4.3 Production
 4.4 Modeling physical relationships
 4.5 Optimization and advanced process control
 4.6 Other applications
 References

5. Data Management from the DCS to the Historian

 5.1 Introduction
    5.1.1 Convergence of OT and IT
    5.1.2 A maturity model for OT/IT convergence
    5.1.3 Digital Oilfield 2.0 headed to the edge
 5.2 Sensor data
    5.2.1. There are problems with data from sensors:
 data quality challenges
    5.2.2. Validation, estimation, and editing (VEE)
 5.3. Time series data
 5.4. How sensor data is transmitted by field networks
    5.4.1. From Plant to Field: Communications Protocols (HART, Fieldbus, OPC, OPC-UA and Wireless Hart)
    5.4.2. Wireless SCADA radio
    5.4.3. Which protocol is best?
 5.5. How control systems manage data
    5.5.1. Cloud-based SCADA and web-based SCADA
 5.6. Historians and information servers as a data source
    5.6.1. What can you record in a data historian?
 5.7. Data visualization of time series data—HMI (human machine interface)
    5.7.1. Asset performance management systems (APM)
    5.7.2. Key elements of data management for asset performance management
 5.8. Data management for equipment and facilities
    5.8.1. What is a document management system?
 5.9. Simulators, process modeling, and operating training systems
 5.10. How to get data out of the field/plant and to your analytics platform
    5.10.1. Data visualization
    5.10.2. Data analytics
    5.10.3. Three historical stages of industrial analytics
 5.11. Conclusion: do you know if your data is correct?
 References

6. Getting the Most Across the Value Chain

 6.1. Thinking outside the box
 6.2. Costing a project
 6.3. Valuing a project
    6.3.1. How to measure the benefit
    6.3.2. Measuring the benefit
 6.4. The business case
 6.5. Growing markets, optimizing networks
 6.6. Integrated strategy and alignment
 6.7. Case studies: capturing market opportunities
 6.8. Digital platform: partner, acquire, or build?
 6.9. What success looks like

7. Project Management for a Machine Learning Project

 7.1. Classical project management in oil & gas—a (short) primer
 7.2. Agile—the mindset
 7.3. Scrum—the framework
    7.3.1. Roles of scrum
    7.3.2. Events
    7.3.3. Artifacts
    7.3.4. Values
    7.3.5. How it works
 7.4. Project execution—from pilot to product
    7.4.1. Pilot setup
    7.4.2. Product owner
    7.4.3. Development team
    7.4.4. Scrum master
    7.4.5. Stakeholders
 7.5. Management of change and culture
 7.6. Scaling—from pilot to product
    7.6.1. Take advantage of a platform
    7.6.2. Establish a team and involve the assets
    7.6.3. Keep developing
    7.6.4. Involve UX expertise
 References

8. The Business of AI Adoption

 8.1. Defining artificial intelligence
 8.2. AI impacts on oil and gas
    8.2.1. Upstream impacts
    8.2.2. Downstream impacts
    8.2.3. Production and midstream impacts
    8.2.4. New business models
 8.3. The adoption challenge
    8.3.1. The uncertainties of new technology
    8.3.2. AI in the field
 8.4. The problem of trust
    8.4.1 Work is evolving
    8.4.2. Driverless transportation
    8.4.3. Trust and the machine
    8.4.4. The human-smart machine trust gap
    8.4.5. Trusting a smart machine
    8.4.6. Trusting the smart machine developer
    8.4.7. Making it real
    8.4.8. Getting to trust
 8.5. Digital leaders lead
    8.5.1. Finding the digital leader
    8.5.2. Moving beyond trials and pilots
    8.5.3. The role of trials and pilots
    8.5.4. The economics of pilot projects
    8.5.5. Moving to enterprise deployment
 8.6. Overcoming barriers to scaling up
    8.6.1. The scale mismatch
    8.6.2. Supplier consolidation
    8.6.3. The corporate accelerator
    8.6.4. The oil company investor
 8.7. Confronting front line change
    8.7.1. The corporate parallels
    8.7.2. Early warning signs
 8.8. Doing digital change
    8.8.1. A typical change champion
    8.8.2. Organizational reaction to change

9. Global Practice of AI and Big Data in Oil and Gas Industry

 9.1. Introduction
 9.2. Integrate digital rock physics with AI to optimize oil recovery
    9.2.1. The upstream business
    9.2.2. Digital core technology
    9.2.3. Modeling wettability at the pore-scale
 9.3. The molecular level advance planning system for refining
    9.3.1. Prediction of crude oil mixing and molecular properties
    9.3.2. Scheduling optimization at the molecular leve
    9.3.3. Collaborative optimization of the entire industry chain
 9.4. The application of big data in the oil refining process
    9.4.1. Principle and methodology
    9.4.2. A case study of CCR process unit
 9.5. Equipment management based on AI
    9.5.1. Equipment hazard monitoring and warning
    9.5.2. Equipment fault recognition and diagnosis
    9.5.3. Equipment health status, residual life prediction and other management
 References

10. Soft Sensors for NOx Emissions

 10.1. Introduction to soft sensing
 10.2. NOx and SOx emissions
 10.3. Combined heat and power (CHP)
 10.4. Soft sensing and machine learning
 10.5. Setting up a soft sensor
 10.6. Assessing the model
 10.7. Conclusion
 References

11. Detecting Electric Submersible Pump Failures
 
 11.1. Introduction
 11.2. ESP data analytic
 11.3. Principal Component Analysis
 11.4. PCA diagnostic model
 11.5. Case study: diagnosis of the ESP broken shaft
    11.5.1. Selection of the ESP broken shaft variables
    11.5.2. Score of principle components
    11.5.3. Pump broken shaft identification
 11.6. Conclusions
 References

12. Predictive and Diagnostic Maintenance for Rod Pumps

 12.1. Introduction
    12.1.1. Beam pumps
    12.1.2. Beam pump problems
    12.1.3. Problem statement
 12.2. Feature engineering
    12.2.1. Library-based methods
    12.2.2. Model-based methods
    12.2.3. Segment-based methods
    12.2.4. Other methods
    12.2.5. Selection of features
 12.3. Project method to validate our model
    12.3.1. Data collection
    12.3.2. Generation of training data
    12.3.3. Feature engineering
    12.3.4. Machine learning
    12.3.5. Summary of methodology
 12.4. Results
    12.4.1. Summary and review
    12.4.2.Conclusion
 References

13. Forecasting Slugging in Gas Lift Wells

 13.1. Introduction
 13.2. Methodology
 13.3. Focus projects
    13.3.1. Dashboarding landscape/architecture
 13.3.2. Slugging
 13.4. Data structure
 13.5. Outlook
 13.6. Conclusion
 Further reading

Index