There are few industries where the failure of a single piece of equipment can lead to such catastrophic events, in terms of life and property, as in the oil and gas industry. Events such as Deepwater Horizon are still very much part of living memory. At the very least, unexpected downtime in any facet of operations is a huge cost. Consequently, adoption of new technology in this industry is always assessed for its impact on workplace safety and reliability, and the industry has traditionally been regarded as conservative in respect to uptake of new technologies.
That is, with the exception of sensor and data measurement tools! Their widespread adoption, partly in response to safety concerns, means that all points along the entire oil and gas value chain, from exploration, development, drilling, production, transportation, processing, distribution and downstream refining, are the subject of monitoring and measurement. Today, the oil and gas industry is one of the largest generators of big data (exabytes!) on a daily basis.
The volume of collected data can itself be a problem, as processing huge amounts of generated data can result in significant delays and uncertainty. In effect, it’s sometimes difficult to see the wood for the trees.
The advent of artificial intelligence (AI), which needs massive volumes of information to work effectively, is opportune. Digital transformation of the oil and gas industry is now rapidly underway, marked by several recent strategic collaborations between oil and gas companies and AI technology providers.
Currently, the two main applications for AI appear to reside in automation, which has the inherent benefit of removing personnel from hazardous environments, and trending analysis, to predict problems before they occur.
One such example of the latter application is SparkPredict™, a cognitive analytics platform which monitors topside and subsea installations. It uses automated development of neural networks to discern patterns in collected sensor data, enabling identification of suboptimal operations and impending failures before they occur. Aker BP is deploying these AI solutions across their offshore production platforms to increase production and efficiency.
In November 2019, Abu Dhabi National Oil Company (ADNOC) announced that it had entered into a 10-year partnership with Honeywell. ADNOC will deploy Honeywell’s asset monitoring and predictive analytics platform to maximise asset efficiency and integrity across ADNOC’s upstream and downstream operations. The AI platform also provides for a digital twin. Digital twins have the unique ability to provide visualisation in 4D, effectively enabling analysts to ‘go back in time’ to identify signal anomalies from sensors that could be used to predict future breakdowns or malfunctions – back to the future indeed!
Other early movers in the AI field are BP and Shell.
BP has invested in Houston start-up, Belmont Technology Inc., to develop a cloud-based geoscience platform using AI, nicknamed “Sandy”. BP feeds the platform geology, geophysics, reservoir and historic project information and Sandy intuitively links that information together, providing a more accessible graphics format for interrogating BP’s subsurface assets. It has the capacity to provide responses to specific questions posed in natural language, such as “What factors control production in the Chirag field?” Sandy can then map out several scenarios, some unforeseen, enabling rapid informed decision-making. BP hopes that the platform will accelerate project lifecycles, from exploration to reservoir modelling, reducing time to collect, interpret and simulate models by up to 90%.
Shell is collaborating with Microsoft on the Azure C3 Internet of Things software platform. While it has invested heavily in AI, it’s reported that it has already saved over a billion in cost reductions, production increases and additional customer margins. Shell’s ambitious AI program involves nearly 300 projects, including providing machine vision for robots and drones, supporting use cases such as inspection. Another project in development involves AI-enabled onshore drilling, in which real-time data coming from the drill bit helps geologists chart a more accurate course for the well and reduce drill bit wear-and-tear.
Because AI inherently depends on algorithms embedded in software, and algorithms in their own right are not patentable, one might assume that this valuable technology is not patentable. In fact, the patent record tends to suggest that practical applications of AI are patentable, and companies from the USA and China are the major filers of patents in this area.
Oil field service providers such as Halliburton and Schlumberger are particularly active in filing patents in AI. Titles such “Learning-based Bayesian Optimization to Optimize controlled drilling parameters”; “Drill bit repair type prediction using machine learning”; “Automated mutual improvement of oilfield models”; and “Disentanglement for inference on seismic data and generation of seismic data” give one a sense of the variety of AI patent applications in the oil and gas space.
The advent of AI also poses philosophical questions for patent law, particularly as the field approaches the holy grail of AI, the moment of singularity. Currently, the question of whether an artificial intelligence system can be identified as an inventor has been posed by the University of Surrey, which lodged two patent applications created autonomously by AI. The patent offices of the USA, United Kingdom and Europe are grappling with this question of inventorship, which also goes hand in hand with patent ownership.
A decision that an AI system is a ‘person’ who could own a patent would have far-reaching consequences. In theory, as the owner of a patent, the AI system would be able to assert and enforce its rights. Or, on the other hand, would an AI system potentially be able to be held liable for patent infringement and damages (payable in bitcoin perhaps)?
It seems as if we are entering into the world of Philip K. Dick – could we reach a point where AI would utilise machine learning to avoid patent infringement, or better yet, invent in a manner which doesn’t infringe prior patent rights in the first place?
At Wrays, we will continue to follow these fascinating updates to technology and their legal and practical implications for our clients.