2017 saw a feeding frenzy on several fronts. Digital twin, internet of things, big data, artificial intelligence and predictive analytics have featured large in our (and everyone else’s) reporting. But it’s sometimes hard to tell them apart. The digital twin concept from product lifecycle management was enthusiastically reworked in big data/AI offerings from GE, IBM, Siemens, AMEC/FW, ABB, Emerson and many others. The feeding frenzy continues unabated in 2018, witness our current lead.
My Obama Peak prediction did not stand for long, it was bested in November as US crude oil production blasted through the 2015 high and on to new records. I offer no defense other than the observation that forecasts are a bit like armies of monkeys trying to write Shakespeare. Someone is bound to get it right. My resolution for 2018. Stick with reporting, no more forecasts.
Mergers and acquisitions in the upstream software space included Palantir and PetroVR/Caesar Systems, Paradigm and Emerson, Pason and Verdazo, Quorum and WellEZ to name but a few. 2018 starts with a bang on the A&M front too – see our Done Deals section on page 8.
On the standards front, the big thing in 2017 was progress on Cfihos, the capital facilities handover standard. Cfihos though, reflects a long-term degradation in upstream data knowhow. In the 1990s it was the hard but smart Express language of DLIS and Epicentre. This was replaced by various XML-based ‘utility’ standards and (for ISO 15926 but not much else) the simple but ultimately inadequate RDF in the 2000s. Now there are calls to replace XML with the latest tech du jour, JSON. But Cfihos is going ‘back to basics’ with an Excel spreadsheet-based ‘standard.’ Basics are all very well, but Excel? Really? Other significant standardization efforts last year include the alignment of Energistics’ standards with the new common ETP protocol and ExxonMobil’s initiative along with The Open Group for a new process control standard.
2017 saw the rise of a plethora of competing ‘platforms,’ all designed to capture your big data and store it in a vendor’s proprietary/open cloud (cognitive dissonance intended). Such offerings came from GE, DNV GL, SAP, Siemens and Schneider. Others moving cloud-ward included Schlumberger with its ‘Delfi’ ‘cognitive’ E&P environment in the Google cloud and Halliburton’s announcement of a deal with Microsoft on the Azure’s platform.
Along with the rise of the IoT and the cloud came the realization that it is impossible to have every sensor plugged into the cloud. Enter ‘edge’ computing, again with a plethora of offerings, not least the Linux Foundation’s EdgeX. Other IT novelties included FME’s use of a Docker ‘swarm’ in geoprocessing and on INT’s introduction of its Ivaap microservices-based back end. We also reported (less enthusiastically) on an avalanche of blockchain offerings in oil and gas.
Much of the big data/AI hype that has come across our desk is ‘forward-looking’ stuff. As I said, it’s better to report than predict. So we went forth and formed our own picture of what is happening in this space. Agile Scientific’s machine learning hackathon provocatively set out to ‘cut out the science’ in geoscience. In some cases, the application of AI appears to have merit (machine learning for mineral microscopy). Other applications are more contentious, such as replacing seismic modeling or reservoir simulation with ‘black box mapping’ from ‘labelled’ cross sections or sketches. But it seems likely that the AI pioneers may contribute to speeding awkward parts the workflow, such as handling large volumes of dirty data (see page 4).
At the EAGE Workshop on data science in geoscience (N° 7 2017) Total’s Michel Lutz summarized the AI phenomenon as the ‘democratization’ of neural nets and decision trees thanks to open source software. Lutz reported successful production forecasts in shale wells, augmenting decline curve analysis with data-driven analytics. Shell reported on its GeoDNN-deep neural net-based seismic feature extraction, co-developed with MIT. Shell is also using ML in reservoir engineering with ‘AutoSum,’ a prototype tool for summarizing large ensembles of reservoir models to help understand key sensitivities. Agile Data Decisions reported work on the seminal CDA unstructured data challenge, a ‘fantastic dataset’ of logs and reports from decades of North Sea exploration.
One possibly lasting spin off of the big data/AI movement is the acceptance of open source software in the upstream. Only a few years ago this was considered anathema. One important piece of open source kit that is getting traction is the venerable Lucene/Solr search engine. This is now baked into commercial offerings from Fuse, EnergyIQ, Voyager Search, and this month’s novelty, BCT’s Geodatafy (page 12). Fuse has leveraged Lucene for a decade, so maybe 2017 is more of a coming out for open source software in the upstream.
Looking ahead we have some great big data/ML reporting in store for you in 2018, see page 4 and our upcoming report from the IFPen DataScience in Energy event.
Probably the biggest news of 2017 actually slipped into January 2018 with Schlumberger’s announcement that its seismic acquisition business ‘does not meet our return expectations going forward, even factoring in an eventual market recovery.’ Schlumberger has ‘exited’ land and marine acquisition. The venerable WesternGeco, which historically embeds GSI, Prakla, SSL and others, is going ‘asset light.’ Wow! If acquisition is going ‘asset light’ where will our big data be coming from in 2018?
© Oil IT Journal - all rights reserved.