際際滷

際際滷Share a Scribd company logo

AI in Energy: What's Next? Challenges and Opportunities

Y
Yohanes Nuwara

際際滷s that I presented at the International University Geophysics Competition (IUGC) 2022 in Institut Teknologi Bandung as a guest speaker. I presented about the challenge and opportunities of AI in oil and gas exploration, and some use cases in the energy sector.

1 of 24
Download to read offline
AI in Energy: Whats Next? Yohanes Nuwara Institut Teknologi Bandung, May 2022
What is AI? Supervised learning What song should I listen to after I wake up in the morning at 7 am? I want to hear hip hop genre. Suggest me a song to listen too now. The algo learns from your daily behavior (as teacher) e.g. what songs did you listen at 7 am and 9 pm this month The algo does clustering of songs based on genre and tempo of song Imagine we have smartphone that can recommend us song to listen Reinforcement learning Unsupervised learning I dont like this song, its too slow and mellow. Can you search a more energetic song? The algo learns from mistake and continue to optimize result What task to perform? You dont teach the machine Listening to music How machine does it? You teach the machine Mistake is the teacher AI is a computer algorithm that mimics human intelligence to do something beyond what human can do
Macroeconomic drivers of AI adoption in O&G Oil and gas companies invest in AI when oil price is low to reduce exploration costs and gain e鍖ciency COVID-19 causes oil price to drop signi鍖cantly to level of USD10/bbl O&G companies decided to cut exploration cost Digitalization helps company to save CAPEX and OPEX while gain e鍖ciency of exploration
AI is about Value Creation 3D seismic acquisition USD 2 Billion Deepwater drilling USD 30-60 Million Upstream conducts data acquisition just to collect more data Instead, AI can be used to 鍖nd new prospects based on data from nearby wells or 鍖elds Total exploration cost can be billions of USD This data for Gulf of Mexico (GoM) Prediction of volume of prospect (porosity, saturation) on one location based on geophysical data
Drivers of AI in O&G Space 2021 (2.034) 2027 (3.669) 2 3 4 USD Billion Market Value The value of AI in O&G market is currently USD 2.034 Bn, projected to grow by 10.81% per year Big Data Open Data Cloud Technology Source: Mordor Intelligence (Report)
Companies and Government are drivers of open data for public Equinor disclosed terrabytes of data open for public. The mission is to foster subsurface research. Volve Field, oil 鍖eld in Norwegian North Sea Hywind Tampen, wind farm in Scotland Sleipner 鍖eld CCS (CO2 storage) Northern Lights CCS (CO2 storage) UK government through Oil and Gas Authority owns a portal of open oil and gas 鍖eld records and datasets in the North Sea
Cloud Megaplayers Products like Cegal enables user to run reservoir static mapping in cloud Cloud technology enables more oil and gas companies to develop low-cost work鍖ows such as data processing and AI Reduced storage cost
Open Subsurface Data Universe (OSDU) OSDU is a universal platform for digital oil and gas companies Reduce silos of data between sectors Accelerate cloud computing of subsurface workflows Develop AI workflows
AI applications in Oil and Gas
Spatial reservoir prediction Predicting reservoir properties (porosity, water saturation, permeability) at other location using knowledge from nearby wells and multiple seismic attributes with neural network and machine learning algorithms Multipoint geostatistics
Seismic Fault Segmentation Automation in fault picking using computer vision based segmentation on 3D seismic data Output fault probability map Reduced time in expert interpretation
Missing Sonic Log Recovery Machine learning can be used to predict and generate synthetic sonic log on the interval where the log does not exist Extreme Gradient Boosting (XGBOOST) algorithm
Geosteering Classi鍖cation of subsurface lithology while drilling using machine learning algorithms Drilling data and cutting / mud log as training data Real-time prediction
Cognitive ML in Energy Exploration Research by Paolo DellAversana (Eni) Rock physics Electromagnetic Gravity Cognitive ML is a branch of AI that implements the science of brain and neural system New prospect Neural network Fuzzy system Joint inversion
Natural Language Processing Understanding technical reports Knowledge graph to represent summary of reports Text understanding and text mining can be used to process old technical reports into data and 鍖nd new opportunities, instead of drilling new well Linear Discriminant Analysis (LDA) to identify topics Named Entity Recognition (NER) to identify keywords
AI applications in Energy Transition
AI in Porous Medium Simulation Simulation of 鍖uid 鍖ow and permeability in porous medium is important not only in oil and gas, but also in energy storage and CCS Physics Informed Neural Network (PINN) to solve 鍖uid 鍖ow in porous media Micro CT Scan Imaging
AI as Rising Star in Geothermal Play Fairway Analysis (PFA) using ANN (Faulds et al, 2020) TU Delfts simulator called DARTS combine ML algorithms to predict production rate and pressure transient
Wind turbine placement optimization Selecting the best location of wind turbines to optimize power production from wind wakes Shell.ai Hackathon 2020 Optimization methods such as genetic algorithm, simulated annealing More advanced with deep learning and reinforcement learning Genetic algorithm Simulated annealing
How to start this 鍖eld?
21 AI for Students & Freshgrads (Roadmap) 1 5 4 3 2 Learn Python essentials (2-3 months) Basics of Python (NumPy, Pandas, Matplotlib, SciPy) Learn ML (1 month) Supervised and unsupervised learning with Scikit-Learn. Lots of online courses offered like Datacamp Learn Deep Learning (1 month) Neural network with Tensorflow Keras. Lots of online courses offered like Datacamp. Familiarize with project (1 month) Familiarize with small projects or notebooks that others have made e.g. in Kaggle, Medium, Hackathons Do self project Propose new project and initiate it
AI for Students & Freshgrads (Materials) Do a machine learning project I have a repository those who want to learn: http://github.com/yohanesnuwara/python-bootca mp-for-geoengineers Learn Python essentials I compiled some articles in Medium https://yohanesnuwara.medium.com/
References https://disruptivetechasia.com/big_news/ai-ml-and-dl-in-the-oil-and-gas-indus try-is-it-an-evolution-or-a-revolution/ https://www.ft.com/content/599af3d6-3f4e-11e9-b896-fe36ec32aece https://www.offshore-technology.com/comment/its-time-for-oil-gas-compani es-to-invest-in-ai/ https://www.cegal.com/en/resources/how-process-automation-can-increase -geoscience-e鍖ciency https://www.bnamericas.com/en/news/cost-of-offshore-studies-could-limit- mexican-og-investment
Thank you

More Related Content

AI in Energy: What's Next? Challenges and Opportunities

  • 1. AI in Energy: Whats Next? Yohanes Nuwara Institut Teknologi Bandung, May 2022
  • 2. What is AI? Supervised learning What song should I listen to after I wake up in the morning at 7 am? I want to hear hip hop genre. Suggest me a song to listen too now. The algo learns from your daily behavior (as teacher) e.g. what songs did you listen at 7 am and 9 pm this month The algo does clustering of songs based on genre and tempo of song Imagine we have smartphone that can recommend us song to listen Reinforcement learning Unsupervised learning I dont like this song, its too slow and mellow. Can you search a more energetic song? The algo learns from mistake and continue to optimize result What task to perform? You dont teach the machine Listening to music How machine does it? You teach the machine Mistake is the teacher AI is a computer algorithm that mimics human intelligence to do something beyond what human can do
  • 3. Macroeconomic drivers of AI adoption in O&G Oil and gas companies invest in AI when oil price is low to reduce exploration costs and gain e鍖ciency COVID-19 causes oil price to drop signi鍖cantly to level of USD10/bbl O&G companies decided to cut exploration cost Digitalization helps company to save CAPEX and OPEX while gain e鍖ciency of exploration
  • 4. AI is about Value Creation 3D seismic acquisition USD 2 Billion Deepwater drilling USD 30-60 Million Upstream conducts data acquisition just to collect more data Instead, AI can be used to 鍖nd new prospects based on data from nearby wells or 鍖elds Total exploration cost can be billions of USD This data for Gulf of Mexico (GoM) Prediction of volume of prospect (porosity, saturation) on one location based on geophysical data
  • 5. Drivers of AI in O&G Space 2021 (2.034) 2027 (3.669) 2 3 4 USD Billion Market Value The value of AI in O&G market is currently USD 2.034 Bn, projected to grow by 10.81% per year Big Data Open Data Cloud Technology Source: Mordor Intelligence (Report)
  • 6. Companies and Government are drivers of open data for public Equinor disclosed terrabytes of data open for public. The mission is to foster subsurface research. Volve Field, oil 鍖eld in Norwegian North Sea Hywind Tampen, wind farm in Scotland Sleipner 鍖eld CCS (CO2 storage) Northern Lights CCS (CO2 storage) UK government through Oil and Gas Authority owns a portal of open oil and gas 鍖eld records and datasets in the North Sea
  • 7. Cloud Megaplayers Products like Cegal enables user to run reservoir static mapping in cloud Cloud technology enables more oil and gas companies to develop low-cost work鍖ows such as data processing and AI Reduced storage cost
  • 8. Open Subsurface Data Universe (OSDU) OSDU is a universal platform for digital oil and gas companies Reduce silos of data between sectors Accelerate cloud computing of subsurface workflows Develop AI workflows
  • 10. Spatial reservoir prediction Predicting reservoir properties (porosity, water saturation, permeability) at other location using knowledge from nearby wells and multiple seismic attributes with neural network and machine learning algorithms Multipoint geostatistics
  • 11. Seismic Fault Segmentation Automation in fault picking using computer vision based segmentation on 3D seismic data Output fault probability map Reduced time in expert interpretation
  • 12. Missing Sonic Log Recovery Machine learning can be used to predict and generate synthetic sonic log on the interval where the log does not exist Extreme Gradient Boosting (XGBOOST) algorithm
  • 13. Geosteering Classi鍖cation of subsurface lithology while drilling using machine learning algorithms Drilling data and cutting / mud log as training data Real-time prediction
  • 14. Cognitive ML in Energy Exploration Research by Paolo DellAversana (Eni) Rock physics Electromagnetic Gravity Cognitive ML is a branch of AI that implements the science of brain and neural system New prospect Neural network Fuzzy system Joint inversion
  • 15. Natural Language Processing Understanding technical reports Knowledge graph to represent summary of reports Text understanding and text mining can be used to process old technical reports into data and 鍖nd new opportunities, instead of drilling new well Linear Discriminant Analysis (LDA) to identify topics Named Entity Recognition (NER) to identify keywords
  • 17. AI in Porous Medium Simulation Simulation of 鍖uid 鍖ow and permeability in porous medium is important not only in oil and gas, but also in energy storage and CCS Physics Informed Neural Network (PINN) to solve 鍖uid 鍖ow in porous media Micro CT Scan Imaging
  • 18. AI as Rising Star in Geothermal Play Fairway Analysis (PFA) using ANN (Faulds et al, 2020) TU Delfts simulator called DARTS combine ML algorithms to predict production rate and pressure transient
  • 19. Wind turbine placement optimization Selecting the best location of wind turbines to optimize power production from wind wakes Shell.ai Hackathon 2020 Optimization methods such as genetic algorithm, simulated annealing More advanced with deep learning and reinforcement learning Genetic algorithm Simulated annealing
  • 20. How to start this 鍖eld?
  • 21. 21 AI for Students & Freshgrads (Roadmap) 1 5 4 3 2 Learn Python essentials (2-3 months) Basics of Python (NumPy, Pandas, Matplotlib, SciPy) Learn ML (1 month) Supervised and unsupervised learning with Scikit-Learn. Lots of online courses offered like Datacamp Learn Deep Learning (1 month) Neural network with Tensorflow Keras. Lots of online courses offered like Datacamp. Familiarize with project (1 month) Familiarize with small projects or notebooks that others have made e.g. in Kaggle, Medium, Hackathons Do self project Propose new project and initiate it
  • 22. AI for Students & Freshgrads (Materials) Do a machine learning project I have a repository those who want to learn: http://github.com/yohanesnuwara/python-bootca mp-for-geoengineers Learn Python essentials I compiled some articles in Medium https://yohanesnuwara.medium.com/
  • 23. References https://disruptivetechasia.com/big_news/ai-ml-and-dl-in-the-oil-and-gas-indus try-is-it-an-evolution-or-a-revolution/ https://www.ft.com/content/599af3d6-3f4e-11e9-b896-fe36ec32aece https://www.offshore-technology.com/comment/its-time-for-oil-gas-compani es-to-invest-in-ai/ https://www.cegal.com/en/resources/how-process-automation-can-increase -geoscience-e鍖ciency https://www.bnamericas.com/en/news/cost-of-offshore-studies-could-limit- mexican-og-investment