Speaking at the ‘cracking the data challenge’ session of the virtual GLOBUC GO Digital conference, Michel Lutz, Chief Data Officer, TotalEnergies, sporting the de rigueur hoodie of the true data geek, described his excitement, on joining Total back in 2016, in finding so much data and such varied data types. Along with dozens of petabytes of seismics, millions of sensor data points, Total’s new energy business (Total recently rebranded as TotalEnergies) is ‘streaming data from wind turbines’. Lutz cited the HPC computing capacity of Total’s in-house Pangea 3 and the fact that ‘more and more data is moving to the cloud’.
Total has drunk deeply from the big data Kool-Aid, performing company-level actions re Total’s data culture. Total already has a community of data scientists across its businesses, with Al training and upskilling for people who want to learn and or/ become experts. Total has a data science boot camp, a digital academy and offers ‘AI for leaders’ to train top managers.
A digital transformation roadmap is set to make IT systems more data-efficient with, notably, a partnership with Microsoft on Azure ‘low code’ development. Lutz has signed-off on Total’s stated intent of deriving a ‘€1.5 billion value per year’ as of 2025 from its Digital Factory. This is to house some 300 developers at the heart of the Paris ‘digital district’. The factory is ‘already delivering 40-50 applications per year across all TE businesses’.
One upstream example is real time prediction of production and well
behavior analysis to improve reservoir performance. This replaces
today’s infrequent well tests with a real time, data driven virtual
test. This has proved challenging as ‘hundreds of ML models’ are
required. Other output from the factory includes power and gas
consumption analytics to control a facility’s energy footprint and
analysis of power in Total’s SAF battery plant to optimize production.
More from the Factory home page.
Peter van Den Heuvel revealed that Shell’s OSIsoft PI-based real time data infrastructure serves some 15,000 users around the world. 12 million events are captured every second from some 7 million connected instruments. Shell’s historical data stretches back over 25 years. Shell’s digitization effort builds on many PI tools (Super Collective, Asset Framework, Event Frames, Vision, Analytics) with extra help from Seeq. The PI AF SDK exposes data to other applications. These include Seeq, PowrBI, Petex, C3.ai and many others including SAP. Data from PI Collectives strategically located around the world flows into an extensive data platform running on Microsoft Azure. Some data comes back on site for consumption in C3.ai, Seeq, Power BI and an unidentified Digital Twin application (possibly from Askelos).
The system is set up to accommodate ad-how data access from other users and applications. But van Den Heuvel insists that while users can connect in many different ways, there are rules, making it clear what is allowed and what is not. Most connections pass through the OSIsoft API, with MuleSoft used for external connections. Originally Shell planned for one humungous global PI Super Collective. This proved ‘mission impossible’ because of technology and legal issues with data residency. Today, each regional server has its own PI Vision-based DIY application that lets users make their own trends and perform root cause analysis.
An encounter with (former) Shell visionary Johan Krebbers (of OSDU fame) convinced van Den Heuvel of the importance of the cloud. This has kicked off a long learning curve in the digital team in Shell as data moves into Azure. This was originally dubbed ‘PI in the Sky’ before Shell settled on SSIP, the Shell senor intelligence platform and ‘the next step in our digital transformation’. The plan is to be able to add-in imagery, drone data, etc. combine all of the above to ‘get more value from our data’.
Following a somewhat commercial presentation by Paula Doyle of Cognite’s ‘semantic’ data ops platform, Philipp Tippel (OMV) and Sofie Svartdal Berge (Cognite) teamed to present OMV E&P’s digital transformation, a.k.a. DigitUP. Tippel observed that earlier digital oilfield programs in mature fields seldom managed to go beyond the concept/pilot stage and see a global rollout. OMV, with help from Cognite has set out to change this. Enter the DigitUP program which is to run from 2020 to 2025, with an expected ‘tipping point’ circa 2024 (c.f. Total’s 2025 ‘jam tomorrow’). At which point OMV will ‘enter the world of autonomous operations’, with software driving first operations, then optimization, handling ‘all possible dynamic conditions’ with engineering support ‘on request’ (from the computer?).
Before then, OMV plans to operate on a ‘best day’ basis by 2022
… increasing production levels by repeating optimum performance
(‘best day’) every day. Best Day 1.0 involves ‘front running’ AI that
screens the entire production system to identify performance
deviations. After BD 1.0, more data sources will be added to the mix
with automated deviation detection and performance comparison. This
will send alarms to engineers and suggest action. Deferments are
classified and written back to the system of record. More from Cognite, and watch the video.
IBM host Dariusz Piotrowski kicked-off the Q&A asking ‘why has all this taken so long?’ Lutz responded that Total has been active in data and software for a long time although there was not much talk of data science when he joined, ‘no fancy terminology’. But there was expertise in a numerical approach to geoscience, notably with Total/Elf’s in-house developed SisMage seismic interpretation system. Today’s journey is more about a vision for a company-wide transformation, bringing talents together. Total didn’t wait until 2020 to be active in data. van Den Heuvel pointed out some of the problems that made digital hard to realize. OPC UA is still not embedded in many production systems. SSIP was conceived in a few weeks, but it took much longer to make fit for purpose and get the system endorsed. Data contextualization is all very well as a concept but stakeholders all have a different view of what things mean. It is hard to get all to speak the same language. Paula Doyle observed that while there is in general ‘lots of innovation’, companies are not so good at using data. This is in part due to a complex supplier network, contractors are not incentivized and may even be protective of acquired data. Engineers often have a ‘not invented here’ complex. On the other hand, lots of technologies are on their way to being ‘commodity’, companies don’t build their own ERP system. This should allow for a focus on where value will be created. The unfortunate reality today is that people take decisions based on a best guess. ‘A guy on a rig is making a decision based on a gut feel, that’s today’s industrial reality’. Are people the problem? In OMV the average age is 45-50*. ‘They don’t all talk IT’. ‘Young people have to explain it to them’.
* The first time we heard this kind of talk was 15-20 years ago, when today’s 40-50 year-olds were the smart young kids of the day. The ageist trope will be with us for as long as IT keeps up with its ‘fancy new terminology’.
A special session discussed OSDU and its potential for open collaboration, again moderated by IBM’s Dariusz Piotrowski, who asked what monetary benefits will accrue from the open subsurface data universe. Shell’s Johan Krebbers, father of OSDU, replied that it is too early to say but that once data is moved into single platform, ‘it will be there for many, many years’ (music to the ears of the cloud providers?). Schlumberger’s Jamie Cruise sees OSDU as an opportunity to consolidate data sources which should increase productivity and enable new workflows. Energistics’ Phil Neri commented that the move to the cloud is a ‘huge undertaking’. OSDU provides a ‘2-3 day data migration’ roadmap.
Piotrowski insisted, ‘how much time will it take and at what cost?’. Again, the response was less than direct! Cruise observed that OSDU should be seen as a part of the digital transformation, feeding AI/ML applications. Companies should ask ‘what do you want to curate in the platform’. All this will take time. ‘Be agile, talk to vendors about their data products’. Neri stressed the importance of metadata in OSDU. Today’s silos are ‘economical’ with metadata. Moving to a platform means this needs to change, making data ‘fully described’. This will be a ‘step change of huge value’.
The debate turned to the meaning of OSDU for smaller companies. Cruise anticipates ‘lots more innovation from vendors’. Krebbers added that OSDU gives small companies ‘equal access to data’. Cruise added from his own experience (with Target) that smaller providers needed to form multiple relationships with larger software houses. With OSDU, ‘you just write to the API’ adding ‘One API to rule them all*’. Neri agrees, OSDU is ‘opening the floodgates for small developers with innovative solutions’.
* There is many a true word spoken in jest. The OSDU API is supplied by Schlumberger!
And what of OSDU in the energy transition? Krebbers is confident
that new energy sources will fit into the new, and rebranded ‘open
energy data platform’. The session closed with a call out to the
community to get involved, sign up with OSDU.
Speaking in the midstream/downstream section, Asger Klindt introduced Maana’s
‘Fanar’ application that uses AI to optimize maritime fleet schedules,
matching cargoes with vessel availability. Fanar was proven at Aramco
Trading and in 2020 supported an average of 5 million barrels/day with
130 tankers and over 2000 voyages. Fanar is described as a low code
platform that runs in the Microsoft Azure cloud. The tool is claimed to
leverage information spanning vessels, terminals, bunkering locations,
chartering costs, weather, piracy and war risk, canal fees and more!
Fanar provides a single view of the business, automatically optimizing
schedules across the fleet and offering ‘what if’ scenarios for better
insights. ‘Cargo operations is a highly dynamic environment, planning
never stops, vessel statuses constantly change and veer from the
original plan which demands constant optimization, not just at the
beginning a voyage.’ More from the release.
Petronas’ Sharul Rashid presented on process safety and instrument analytics. Rashid enumerated some of the major process safety incidents of the past decades, from Longford (Australia, 1998), Pascagoula (Mississippi, 2002), to Buncefield (UK 2005) and Deepwater Horizon, (USA, 2010). History is repeating itself as process safety incidents continue to occur. Rashid offers five process safety questions that companies should ask themselves. 1. Do you understand what could go wrong? 2. Do you know what systems prevent this from happening? 3. Are they working? 4. What is your role? 5. How can we leverage digital technology? Rashid discussed the challenges of aligning old/legacy plant with the safety standard IEC61511. A Petronas refinery in Malaysia was commissioned in 1983 and had an ‘under-engineered’ safety instrumented function (SIF), including a single non-smart sensor that was safeguarding the furnace’s fuel gas burner throat against liquid carry-over from fuel gas header. This has now been upgraded with multiple smart sensors and a voting protocol. Rashid observed that while voting may ‘close the SIF gap’, it can increase the chances of a spurious trip.
In the upstream session, Marco Ferraz presented Lisbon, Portugal headquartered Galp’s
GeoScienceAdvisor, developed by Galp and IBM with financial support
from ANP. GSA provides AI/ML-based assistance to seismic interpreters,
performing seismic facies analysis and assessing geological risk and
probability of success. The screenshot looked a little like Eliis’
PaleoScan. More from GALP.
The sixth Global Business Club GO Digital Energy event will be held in Amsterdam on 7-8 June, 2022. More from GLOBUC.
© Oil IT Journal - all rights reserved.