Front-line workers arguably play the most important role in the global energy supply chain. The decisions they make day-to-day\u2013 whether opening a valve or shutting down a pump \u2013 has the potential to impact billions of dollars\u2019 worth of trade flows and prevent fatal accidents.<\/p>\n
Oil and gas companies spare no expense in monitoring key parameters to help their front-line staff make better decisions; a typical facility could have thousands of sensors monitoring several variables every second. [1] Hundreds of them are variables that can be manipulated to control its processes.<\/p>\n
Put yourself in the shoes of a panel operator in a control room: how many of these parameters would you monitor before taking action?<\/p>\n
According to McKinsey, no more than ten. [1]<\/p>\n
The wealth of data available and the human limitations of processing it is only one example of how AI can augment human capabilities to further optimize processes. Royal Dutch Shell – the Anglo-Dutch energy giant \u2013 recognizes this trend, which is why it\u2019s investing heavily in its machine learning (ML) capabilities to improve its operations.<\/p>\n
These trends are important to Shell for three main reasons.<\/p>\n
Unlike innovation in other manufacturing industries, innovation in oil and gas was historically driven by the need to expand resource bases, not reduce costs. Deepwater well development, horizontal drilling, and LNG shipping are prime examples of the herculean feats achieved by oil companies. However, as in any boom-bust market, the bull oil market up until 2014 made companies complacent and their costs ballooned accordingly. Secular headwinds – beginning with the 2014 price crash led by the boom in shale oil production \u2013 have eroded the high margins that buoyed the industry. Cost discipline became an existential priority, and ML solutions have the potential to significantly reduce costs. [2]<\/p>\n
In addition, national oil companies are beginning to exert more control over the available resource base, leaving nothing for international oil companies (IOCs). IOCs can no longer compete on reserve size, but on their technological capabilities that add value to their commercial partners.<\/p>\n
Finally, the predictive capabilities offered by AI applications have the potential to exponentially improve safety in operating sites and retail locations in a cost-effective way. [3]<\/p>\n
Shell has identified three potential applications for AI in the short term: predictive maintenance, real time decision-making for horizontal drilling, and image recognition to detect safety risks in retail operations. [4]<\/p>\n
Horizontal drilling is a time sensitive activity where decisions need to be made instantaneously based on many dynamic variables. By using the data already available from previous wells, AI can chart the course for an underground drill bit by optimizing its speed and direction, thereby boosting well productivity and reducing the need to replace drill bits. The data collected can further improve productivity in other wells, highlighting the network effects inherent to AI. The ultimate goal would be to have an automated horizontal drilling program without the need for human intervention, improving the productivity of geologists who can oversee more wells. [5]<\/p>\n
Unplanned shutdowns due to equipment failure is an acute problem in downstream industries, contributing to low productivity and high maintenance costs. ML can be used to predict the likelihood of failures by processing the massive streams of real time data. By pre-emptively performing maintenance, brownfield assets could improve production efficiency by 10% and bottom line impact by up to $200mm. [6] Shell is rolling out two programs; one in Australia to monitor equipment for coal seam gas production, and another to detect anomalies in downstream valves. [5]<\/p>\n
Finally, Shell is partnering with Microsoft and C3 IoT to create image recognition software that uses AI to detect safety threats at its retail gas station locations, such as smoking and theft. [3]<\/p>\n