Can AI decrease risk and boost efficiency for oil and gas companies?

Jon Lawson

In recent years, a new trend has enthusiastically grabbed the attention of the oil and gas industry, and that trend is digitisation. The promise is, through the use of new technologies – in particular those related to the collection and analysis of huge amounts of data – digitisation will bring oil and gas companies considerable business improvements. And for the most part, this is true. Yet the hype has left many companies puzzled. Some are confused about where to start and how to succeed in integrating these new technologies; and others are rightfully cautious about yet another innovation that typically translates into risky investment with returns not guaranteed.

However, when it comes to machine learning and artificial intelligence (AI) – major technologies within this digital transformation landscape – the promise for thw oil and gas industry looks much better than with a typical innovation. Despite “new industries” such as internet and e-commerce that are currently leading the way in this technology, it’s actually the older industries, including oil and gas, that possess the prerequisites needed to take full advantage of, and benefit from, these AI techniques.

Why the oil and gas industry? Why now?  

The oil and gas sector is very capital-intensive. Equipment, moving assets and associated logistics all generate huge costs and stifle change. Yet, it’s this same conservatism that makes AI such a great tool for oil and gas companies.

These companies have established routine processes, and consequently, an abundance of relevant data to learn from. At the same time, these processes contain a lot of fluctuations and uncertainties, which makes the use of AI worthwhile due to its ability to better deal with complexity. Ultimately, this is what is needed for AI-based optimisation: complex, stable operations with known KPIs and constraints, and past data on how a certain process went in the past. Algorithms can analyse all these historical datasets to learn from it in order to optimise and improve operational decision-making. Whereas associated production costs and high processing volumes ensure that even marginal improvement in one single process may result in millions of savings yearly.

Moreover, AI is one of the most affordable innovations in the industrial sector. Since it affects only operational decisions – making each more precise, efficient, or personalised – it does not require the change to the process itself. Integrating new process management software or purchasing equipment typically requires significant capital investment, and takes years to pay off. AI, instead, simply improves the processes that are already up and running, at almost no prior cost. With the recent oil price downturn that put great pressure on optimisation and operational efficiency, it definitely becomes a very attractive type of innovation.

How can AI be delivered?

The ultimate value of AI lies in its ability to deal with uncertainty – which both upstream and downstream oil and gas processes have plenty of. These are exactly the uncertainties that make companies unable to manage every process with precision and create inefficiencies resulting in equipment downtime, excessive spending of raw materials and energy, or throughput decrease.

Although there is extensive knowledge about, for example, petroleum geology or existing thermodynamic simulations of different refining processes, these theoretical models are only the best of the available assumptions. They describe and explain the processes with some precision, but can never fully account for all the fluctuations happening in the real-world.

This is where AI can add additional value. Through analysing the data that reflects how oil and gas processes run in the real world – on specific fields, or at a given processing unit installed at a certain plant, AI learns from this actual data and recognises hidden patterns or weak dependencies that are overlooked by the current tools. As a result, AI can ‘correct’ the output of the existing knowledge-based models and generate more precise predictions and recommendations for each and every case.

So, how exactly can it be best applied to existing use cases? And for what purposes is AI most suitably deployed?

AI methods can be appplied to numerous processes, starting from well drilling and completion to refining and petrochemical production. The goal is to identify the correct task and metric that AI is set to optimise, and supply it with enough historical examples.

For example, in drilling, AI models can be developed to recommend optimal drilling modes based on past history, so that the overall drilling speed is increased. Another model can be tasked with predicting potential anomalies in the drilling – such as stuck pipe, loss in circulation, or kicks – when their early signals are not yet noticeable to a human eye. Based on the model forecast, correct actions can be taken ahead of time to avoid equipment downtime or deviation from the target drilling interval. Similarly, choice of further well treatments during production phase can also be made with the aid of AI-based forecast, delivering extra precision on top of existing models.

In the downstream sector, almost every major processing step – from crude oil refining and gas fractionation to pyrolysis or production of polymers – can be tackled with AI to improve operational efficiencies. For example, in gas fractionation the challenge is to correctly adjust the unit operating modes in a timely manner because changes to the process – such as heating of the plates – are time sensitive. Using AI, it is possible to enhance existing process control by automatically recommending the choice of parameters, resulting in decrease of energy use and improved throughput.

Digital transformation has slowly taken its place in the agenda within the oil and gas industry. Yet, with demand and supply still stabilising and oil prices still low, many companies are desperately seeking a digital solution to their current problem. While considering investing heavily in IIoT (Industrial Internet of Things) to develop new capabilities, it is important not to overlook a technology that has been present in theory for quite some decades: AI. Instead of a ‘thing of the future’, AI should be treated as a new source of optimisation and improved efficiency – with no capital expenditure or changes to existing infrastructure and processes. It’s these AI applications that are most beneficial to traditional industries such as oil and gas, and can be applied now.

Recent Issues