Predictive analytics is attracting plenty of attention offshore

Offshore oil producers are increasingly using predictive analytics to make operational decisions, and this is having a knock-on effect on their relationships with services firms, industry sources have told Upstream Intelligence.

Predictive analytics is attracting attention in the oil industry (Image credit: inemaslow / iStock)

Related Articles

This trend has been particularly noticeable in the past six months, according to Jim Crompton, Director of Colorado-based Reflections Data Consulting. “After investing for decades and more into all of the data gathering, managers are now very interested in getting results out, to see where exactly they can reduce spend and be more efficient,” he said.

Among the areas where predictive analytics can create the biggest savings is equipment and process health. Based on various readings such as temperature and pressure, it is possible to monitor a specific piece of equipment and build a model which will predict when this piece will fail.

“All equipment fails, that is not an issue. But if your software can predict that it might fail in two months’ time then the company is going from scheduled maintenance to predictive maintenance. Being able to schedule when a company will do repairs and maintenance means avoiding incidents which are causing financial loss or environmental damage,” Crompton explained.

Shift toward pay-as-you-go

These developments in predictive analytics have led some to forecast swift changes to how oilfield-services companies tailor their offerings to operators.

“In the airline industry, companies are no longer buying turbines; but they are buying thrust propulsion by the hour. We see this change coming to the oil and gas industry,” said Phillipe Flichy, Senior Digital Oilfield Advisor at Baker Hughes.

In the Gulf of Mexico, electrical submersible pumps (ESPs) are commonly used to lift moderate to high volumes of fluids from wellbores. With predictive analytics coming into the picture, services firms could sell the lifting of a certain amount of hydrocarbons rather than the ESPs themselves, Flichy said.

Services companies selling a piece of equipment are typically penalized if the equipment stops working for a period of time during a contract period. If the equipment breaks down once that period is over, the financial burden is carried by the operator. If a critical piece of equipment malfunctions it can bring a well to a standstill.

The type of scenario envisaged by Flichy will see operators and services companies work more closely, with added benefit for both sides because they will have the ability to predict failure.

Procurement is another area where predictive analytics can create a significant difference for an operator’s bottom line, according to Crompton. He explained that if a firm can accurately gauge how it is spending money, not only on equipment and chemicals but also across services, this can create additional savings.

Market has room for more

Several companies are offering predictive-analytics software to the oil and gas industry, including GE Oil and Gas, Osprey and OAG Analytics.

One of the key components in the process of creating predictive analytics is machine learning, according to Luther Birdzell, chief executive of OAG Analytics. OAG provides big-data management and predictive analytics mainly to unconventional US onshore oil producers.

Programming for computers, the internet, or phone, usually consists of humans writing a set of explicit instructions for the computer to follow. In industrial use mathematicians and statisticians would work with data, find a correlation and then build models of “what if” scenarios by hand. But in the oil industry the amount of information available is so vast, complex and interrelated – for instance: the temperature and pressure in a well, well design, and the amounts of chemicals used – that it requires a different type of approach.

“Machine learning lets computers do more. You specify data sets to a computer and instead of a human writing that application the computer studies the data and writes its own, highly optimized application that predicts an outcome,” Birdzell said.

Machine learning has become so advanced that when used properly it is capable of building predictive models for the oil and gas industry that are virtually incomprehensible to humans, he added.

Despite what is currently available in terms of technology and software, only about 30% of oil companies are properly consuming their data to make the best operational decisions. Looking forward 12 months, Birdzell predicted the industry would further embrace the concept of the digital oilfield.

By Vanya Dragomanovich