For operators that rely on electrical submersible pumps (ESPs) to meet their production goals, a failed pump at the wrong time is a costly pill to swallow. While great strides have been made in improving well monitoring systems, independent operator Apache Corp. and software developer Ayata are taking it one step further.

The companies are at work on a new project using big data analytics, namely prescriptive analytics, that will not only predict when an ESP will fail but also will recommend what necessary actions are needed to prevent the failure from occurring.

According to GE, there are more than 130,000 ESPs installed and operating around the world. A better understanding of when one or more of those pumps might fail would allow operators to better determine when pump repair or replacement should be scheduled without impacting production too severely.

“ESPs are critical for our industry as 60% of the world’s oil production runs through them,” Mike Bahorich, chief technology officer for Apache, said.

Apache has more than 1,200 pumps operating on its wells globally. With about one-third of its overall oil production flowing through ESPs, according to Bahorich, even the slightest uptick in production would garner significant returns.

“Just a 1% improvement in ESP performance around the world would provide over a half-million additional barrels of oil per day given the amount of oil pumped through ESPs,” he said. “Multiply that by [US] $100/bbl, and you’d have $53 million-plus per day just from improvements made to ESPs.”

So what is “prescriptive analytics,” exactly? According to Ayata CEO Atanu Basu, prescriptive analytics “not only helps you look ahead but also helps you make the right decisions to benefit from what is ahead. It does this is by combining different types of data such as text, image, audio, video, and numbers with complex mathematical algorithms.”

Currently, the companies’ technology teams are data mining a collaborative industry database known as the ESP Reliability Information and Failure Tracking System (RIFTS). The database contains information on more than 104,000 ESPs from approximately 750 fields worldwide representing 22 operating companies. The data were contributed by partners to the ESP RIFTS joint-industry project such as ConocoPhillips, Hess, Statoil, Chevron, Shell, and Apache. The data are used by the prescriptive analytics’ algorithms to forecast issues and develop recommendations.

“There are a number of variables in the database that we cannot change, for example reservoir temperatures,” Bahorich said. “There are, however, a number of things we can, like the particular brand of pump or cabling used. And while it is running, we can look at other variables like pressure, voltage, and sand intake.” Bahorich noted that the next step is to determine the best combinations for operations before moving into field testing. “There are many things about a reservoir that are unpredictable,” Bahorich said. “Oftentimes, the inability to predict is due to the Earth’s complexity. By using big data analytics, we’ll be able to get an early look at an issue and be able to adapt before it becomes a much larger issue.”