Rise of the Machines: How to Employ Machine Learning and Artificial Intelligence to Prevent Oil and Gas Well Failures10/03/2016
Machine learning can be used to monitor wells around the clock and alert engineers to a possible sign of failure
This is the last in a three-part series about machine learning’s applications in upstream oil and gas. This blog details how machine learning can be used to help teams predict and prevent well failures.
Eliminating failures and nonproductive time is a top goal for management at every upstream oil and gas company. Drilling and production engineers are well-acquainted with the grind that can result from constantly putting out fires. Many of these engineers find themselves saying, “There has to be a better way.” Those seven words have been the staple of this series. Similar to blogs 1 and 2, the push for a better way can lead your organization into a new world of efficiency.
Fighting fires is all too common in production engineering, and many production engineers view their ability to solve problems on the fly as their greatest skill. While efficient problem-solving is something to be proud of, the goal of production engineering is to diagnose problems before a failure occurs.
With a plethora of wells under management, it can be hard for staffs both in the field and in the office to constantly monitor every well. We’ve discussed in previous blogs how tedious and time-consuming tasks are perfect candidates for a machine learning or artificial intelligence solution. With the advent of SCADA systems, computers have the ability to monitor wells around the clock and alert engineers to a possible sign of failure.
Setting up a well-monitoring system is a practical solution for any size company. The first step is to determine the average operating conditions for a set of wells. Examples of operating parameters include wellhead pressure for wells not on artificial lift, polished rod load for rod pumps, and internal motor temperature for electric submersible pumps. These ranges can be determined by setting an operating range within a set of standard deviations of the mean. If the parameter exceeds the operating range, an alert can be sent to personnel who can determine the best course of action. This alert can include snapshots of vital information to help personnel quickly assess the situation and make decisions. This system can be set in place for any parameter being monitored by the SCADA system.
Another common task for production engineers is determining the artificial lift needs for the entire lifecycle of the well. Over time engineers can review the pressure profile of wells to determine the best time to switch artificial lift methods. Once this profile is studied and understood, artificial intelligence can be used to monitor wells and update personnel on the timeframe for artificial lift changes. Paired with best engineering judgement, this tool can ensure that the production needs of every well are properly met in a timely manner.
There are a limited number of hours in the day for technical staffs to fight fires, yet properly designed systems can alert personnel and give them the vital information to quickly and efficiently solve problems.
So how will you use predictive analytics to maximize your workday?
Optimize Your Production Operations
Check out the resources on our production microsite to learn more about production-optimization strategies. This site includes customer-success videos, case-study webinars, and white papers containing lots of valuable production best practices.