Dan Jeavons provided an update on his 2016 MathWorks talk on Matlab and advanced analytics in Shell. Digital innovation and AI are now a key components of Shell’s ‘accelerating’ energy transition with its bold ambition for net zero 2050. There will be massive shifts in energy in the next decade. Shell is to leverage AI and data science to transform and manage the energy system. There are already some 5,000 members of Shell’s AI network and 10,000 equipment items are ‘monitored by AI’. Other digital transformations involve Shell EV charging. Shell deploys some 100 AI apps running against 1.9 trillion rows of data. All is now in a common data frame, accessed from remote digital centers with embedded AI. Citizen data scientists are using Matlab and Azure ML. Amja Chaudry drilled-down into Shell’s use of MathWorks tools. With help from MathWorks execs, Shell embarked on experiments and proofs of concepts to see if they could be deployed cloud native. This led Shell to acquire the Matlab Production Server to deploy its in-house developed algorithms. The first product to be rolled (in 2016) was the Quest Solution for daily CO2 monitoring and alerting from laser measurement across the Canadian CCS facility.
Matlab has been used for over 25 years across Shell and usage is now consolidated into a single license agreement and center of excellence. This now involves a DevOps approach to ‘accelerate and operationalize’ deployment. In 2019 Shell rolled-out CatCheck Connect on Azure, an app to evaluate catalyzer health. In 2020 the Matlab WebApps Server was deployed to support Shell’s ‘MADA’ (modern data analysis) tool, and also to develop a enterprise app for subsurface geologic feature prediction. In 2021 Shell formed the OpenAI initiative, a collaboration with C3ai Baker Hughes and Microsoft to commercialize its know-how. The latest MathWorks tool is the Matlab Online Server that is allowing solutions developed with Simulink and Simscape to be cloud-hosted.
Jeavons summed up observing that there is now the potential for digital to become ‘the way we do business’, rather than an optional extra. ‘There is a confluence of science and data-driven modeling, everything Shell does is physical’. Applications like CatcheckConnect combine science and data, leveraging Matlab’s strength. Simulink is used to combine data-driven chemometrics with process monitoring and high resolution mass spectroscopy to optimize GTL technology. Jeavons concluded by welcoming MathWorks* into the OpenAI initiative.
Another noteworthy presentation at the MathWorks Expo looked into
different ways of combining Matlab with open source machine learning
environments including PyTorch and TensorFlow. The Matlab Deep Learning Model Hub offers some 50 pretrained models and a connection to TensorFlow and PyTorch repositories.
Watch this and other presentations from the show here.
* MathWorks and Kongsberg Digital joined the OpenAI initiative in 2021.
© Oil IT Journal - all rights reserved.