The oil and gas industries are some of the largest in the world, with a market size of over $7 trillion. Just over half of that is exploration and production, the rest being refining and distribution. It employs over 12 million people, with a further 60 million employed indirectly by its suppliers. While an oil rig might seem like a long way from a chatbot, the oil and gas industry has always been a heavy user of technology, and artificial intelligence (AI) is no exception.
The exploration sector, where oil and gas are discovered and extracted, has always been a high-stakes business, often with an element of luck involved. The largest oil field in the world, the Ghawar field in Saudi Arabia, was discovered in 1948 by three geologists who guessed that a bend in a dry river bed might be conducive to oil accumulation beneath. As technology developed, seismic processing was employed to map out the Earth’s interior structure. Interpreting seismic data is a complex task, and recently, neural networks have been one AI approach used in helping geologists interpret the vast amount of seismic data that is gathered. Machine learning techniques can be used to identify geological features, such as the faults where oil and gas can be trapped. When drilling commences, AI can be used to analyse data from the sensors attached to the drills, which measure things like rotary speed and mud flow. This data can be used to detect anomalies and identify potential hazards. Once an oil reservoir has been identified, AI models can be used to accurately model the reservoirs, suggesting the most efficient extraction strategies. A program designed by Chevron can detect methane leaks and shut them down without human intervention. This has been credited with reducing methane flaring, a major source of carbon emissions, by 60%.
Machine learning has long been used in predictive maintenance. AI models analyse sensor data from physical structures such as oil rigs and power lines, and detect changes in the patterns of the data to try and predict when a part is likely to fail. Since unexpected equipment failure can have huge consequences in oil production, anything that enables proactive maintenance and reduces the risk of part failure has a major economic impact. Some case studies have shown that AI predictive maintenance can reduce faults by 70%, significantly lowering maintenance costs and reducing the risk of potentially disastrous equipment failure. Robots that are driven by AI are already in use in equipment inspection for pipelines and offshore structures, able to operate in areas that are hazardous to humans. Robotic vehicles can operate at great depths on the seabed, some of which are beyond the capabilities of most human divers and submersibles.
Virtual replicas of physical assets, known as digital twins, are used to monitor and analyse the operations of equipment, using the sensors on equipment like valves, pipelines and drilling rigs. One example is that of gear track drilling rigs, with another real-life case being the monitoring of the integrity of mooring lines. It should be emphasised that digital twins are not especially new. They have been deployed in the oil and gas industry for over twenty years, using well-established techniques such as machine learning, well before the recent flood of interest in AI. One survey found that 44% of upstream organizations use AI in some form in their oil and gas exploration.
The use of AI is not restricted to the upstream sector of the oil industry. Oil refineries have huge amounts of physical equipment that can benefit from the same predictive maintenance and automated monitoring approaches. Real-time process optimisation, pipeline management and demand forecasting are where various AI techniques have been applied. Generative AI can be used to provide easy access to manuals and prefectures, as well as for personalised training courses for industry employees. This is quite apart from potential use in clerical and support tasks such as reviewing contracts. Naturally, AI also has applications in the retail and back-office areas of the energy industry, from the shops at service stations to procurement to support staff, just as in other industries. There are real benefits to be gained by the careful application of AI in the industry. One survey found that there was a 26% improvement in asset utilisation, and a 27% improvement in production uptime associated with the use of AI for predictive maintenance.
The deployment of AI in the energy industry is not without its challenges. Data quality is crucial to AI, especially to the neural networks used in generative AI, which rely on vast amounts of training data. Data in the oil industry is often in siloed specialist applications and is hard to integrate. Indeed, in one recent survey, integration of AI with existing systems was identified as the most important challenge, ahead of security and lack of skills. AI models are only as good as the data they are trained on, and if the data is incomplete or biased, then the output of the models reflects that. There is also a major concern about connecting AI to critical infrastructure, given the many well-documented cybersecurity risks associated with generative AI tools in particular. Oil and gas are highly regulated industries in most countries, so oil and gas companies need to be careful to ensure that any deployment of AI is within the regulatory framework of the various countries in which they operate. Specific regulation of AI is a fast-moving area, and companies need to keep a careful track of this in order to ensure compliance.
Despite these challenges and risks, the onward march of AI into the oil and gas industries seems set to continue. As AI becomes woven into the fabric of oil and gas operations, a balance needs to be struck between innovation and regulation. In 2017, The Economist magazine said that the world’s most valuable resource was no longer oil, but data. In the case of the oil industry, it is a blend of the two.
