Machine learning in the Energy Sector

The energy sector although old, is still fertile for innovation, innovative ideas are applied day in and day out in this critical industry. One is the utilization of big data and machine learning to increase efficiency and reduces risk via predicting equipment failures.

The energy sector, although old, is still fertile for innovation. Innovative ideas are applied frequently in this critical industry. An important example is a utilization of big data and machine learning to increase efficiency and reduces risk via predicting equipment failures.

“This cognitive revolution will overshadow anything we have seen thus far” said Amir Hussain, author of the book titled the Sentiment Machine in an interview with Fox News [1]


I asked myself this question a lot back when I was a process engineer at the world’s largest Energy company. How can the Energy sector make use of the advancements in Machine Learning? Are we doing enough? Is the sector too old to benefit from such advances? Or is the most fertile for such applications?

Upon researching this topic, I came across several useful applications of machine learning in this very crucial industry an industry that is responsible for meeting the world’s energy demand requiring 100 million barrels of oil per day.  [2]

Of the most interesting ones are:

  • Improving Exploration and Selection of Drilling Sites
  • Predicting when machines might fail
  • Increase safety, through evaluating risks
  • Avoiding decision biases


One aspect this article is going to focus on is predictive maintenance and pump/equipment selection as they help organizations like Saudi Aramco where many engineers are at work to improve the processes and increase efficiency, productivity and profitably.

Why does predictive maintenance matter?

Think about how much effort it takes to fix a failed pump at an oil rig or at a running plant, whether it’s a refinery, oil facility, petrochemical plant, etc. These plants cost billions of dollars to constructs and millions to operate. Time, efficiency, and productivity are the languages of the day. Of course, on top of all that comes safety. What happens when a pump fails? Let’s take the example of a submersible pump, deep underground. First, production stops because safety might be at risk. Highly specialized people and equipment need to be mobilized (usually a vendor or from HQ) in order to come and support the local operations team in fixing the issue. Imagine the cost associated with this massive operation resulting from this failure which can happen at any other plant. Now with machine learning companies can predict when such failures may occur and plan their maintenance plans accordingly. This is done via analyzing sensor data and former failure reports among other [3]

By monitoring equipment such as pumps for future failures, companies will have a more strategic schedule maintenance time (schedule as needed) and will be able to identify root causes of the issues more effectively. For machine learning to operate more effectively, more data is required (as tech savvies call it “Big Data”). One way we can have that is by channeling the equipment data gathering into one source in a corporate-wide cloud. This will enable engineers at headquarters to have enough data for their machines to analyze and have more accurate predictions of future failures.

Saudi Aramco is moving forward with implementing such technologies in its operations. The company recently invested $26M in Series B investment in Maana through its venture capital arm (Saudi Aramco Energy Ventures). Maana is an advanced analytics platform that incorporates big data into business applications. The Investment director of Saudi Aramco Energy Ventures commented on in the investment saying, “We decided to invest in Maana after seeing first-hand the business value that Maana’s platform delivers in optimizing assets and processes at some of the world’s largest industrial.” [4]

Saudi Aramco also recently announced a series A investment in FogHorn a leading developer of edge intelligence software for industrial and commercial IoT application. [5]

Despite these great efforts toward gaining the technology, I think the company needs to further its advancement in cloud computing capabilities in order to make use of the plethora of data it currently owns from the thousands of data collecting points in plants around the world. By investing more in the cloud computing solution, combining it with the data analytics software that is already in use, the company can take predictable maintenance to the next level. Enhanced predictable maintenance will lead to increased asset lifetimes and a significant reduction in equipment failures and downtime. In turn, this will result in significant cost savings from real-time actionable insights gained from sensors around all its equipment and facilities.

As Saudi Aramco and other players in the energy sector continue to advance in the machine learning space, the question around how these companies would deal with the cost associated with data storage, transfer, and maintaining of their IT infrastructure remains unanswered. There are also risks with cybersecurity and data protection to consider.

(775 words)


Works Cited

[1] Hussain, A., 2017. How the oil industry could benefit from artificial intelligence [Interview] (8 December 2017).

[2] Economics, B. P. E., 2018. BP Energy Outlook, United Kingdome: British Petroleum.

[3] Maana Media , 2017. [Online]
Available at:
[Accessed 12 November 2018].

[4] MAANA, 2018. [Online]
Available at:
[Accessed 12 November 2018].

[5] SAEV, 2017. SAEV. [Online]
Available at:
[Accessed 12 November 2018].

*Photo used is from <>


3D Printing Toyota’s Headlights: The Sourcing Decision


The AI Doctor will see you now – Machine learning transforming healthcare

Student comments on Machine learning in the Energy Sector

  1. Brilliant post! Many have the perception that the old oil business has reached saturation in terms of innovation which is far from the truth. The old mentality of “run to fail” which once thought to reduce maintenance cost has proven to be wrong as the cost of unplanned shutdown in terms of lost production and expedited repair far exceeds that savings of preventive and predictive maintenance. The infrastructure to collect data for expensive machines such as compressors, turbines and large pumps is already there in terms of instruments and sensors, what is left is processing this data in smart machines to make informed decisions about maintenance.

Leave a comment