Automation Anxiety: Balancing Big Data and
By Stephen Rassenfoss | 14 March 2014
While the future of drilling is expected to include automated systems driven by big data gathered in real time, the reality is that computer-controlled components that use limited data—delivered as needed and offering measurable benefits—are what’s likely to sell now.
Using real-time data to improve automated drilling operations was the focus of a recent panel discussion hosted by the SPE Drilling Systems Automation Technical Section in Fort Worth, Texas. The dialogue between the panel of experts and the knowledgeable audience focused on what innovations are likely to be adopted when it comes to automation and data gathering. The short answer: there is a demand for data innovations offering benefits that cover the cost and trouble of change. Additionally, innovators need to find measure the value they can add.
One of the most direct justifications for embracing innovation has been drilling wells faster. But automated drilling control systems that promise faster speeds are competing with current technology that has already been used to slash the number of days needed to drills onshore.
“Selling based on drilling faster will not necessarily get you enough” added value to sell a system wired to deliver high-speed data for computerized control systems, said Brett Borland, manage for drilling engineering for global wells at ConocoPhillips. “Going from drilling 90 ft/hr to 150 ft/hr may not be enough meat on the bone to sell it.”
So what will it take to convince oil companies to buy into data-driven automation? Borland said that ConocoPhillips has gained valuable insights from an ongoing test of a system using wired drilling pipe to provide real-time drilling data, which it will use to determine how it drills a variety of wells.
ConocoPhillips is currently running a five-well test in south Texas using drill pipe wired with coaxial cable and digital controls from National Oilwell Varco designed to reduce what Borland calls “downhole dysfunction.” That includes periods of excessive vibration that can slow drilling and damage equipment.
The reward for quickly detecting and correcting those problems would be reduced spending on replacing damaged downhole hardware, as well as the lost cost of repairing damaged equipment. Another plus is that reducing the number of times drilling pipe needs to be tripped in and out of a wellbore reduces the risk of well incidents and injuries associated with pulling a drillstring.
“Why do we need to automate rigs?” said Mike Loudermilk, operations manager for drilling and completions for Lime Rock Resources. “We have got to protect life and property.”
The cost of one serious injury claim can cover the cost of a mechanical pipe handling system, he said. But it is hard to put a price on incidents avoided, and it can be difficult to convince drilling crews to embrace new machinery, even if the technology is proven, Loudermilk said.
The 18 months spent planning the ConocoPhillips test included creating a system to motivate workers on site to use it while drilling. “All those on the drilling crew are incentivized to make it work,” Borland said. “We are not looking to drill automated wells from the office. We are looking to automate drilling programs while we are out there.”
Gathering real-time data has proved to be easier than making sure it is used effectively. “We can get tons of data,” Loudermilk said, including the ability to view that data and monitor drilling progress on a smartphone. “Here is the problem: If I call that driller,I hired him to drill that well. If I tell him how to do it, I know something bad will happen.”
The organizational issues related to effectively using data go beyond the rig site. During the discussion, a member of the audience said that he used to work for a company that provided operators with rapid analytical updates based on drilling data. The service never caught on. He then asked if drilling data analysis is more likely to be used now by operators and drillers.
Borland responded that the value of data gathering depends on whether an organization is set up to direct it to employees with the tools to analyze it, and whether there are leaders in place who are committed to making sure that what has been learned is applied in the field. Any manager involved with data gathering needs to make sure it “fits into the processes of the company,” he said.
While “real-time” has become a buzzword as more sensors are installed in wells, the delivery speed required varies widely. It can range from multiple updates every second for a system that automatically adjusts the weight on the drill bit and its rotation speed, to shipping a disc in the mail for later analysis.
A lot of drilling data is also being discarded after the well is completed because, with more wells being drilled faster, engineers supervising multiple projects do not have time to look back over past jobs.
Data from multiple sensors provide many more points of view, but work is needed to provide engineers a clearer picture of what is going on, said Robello Samuel, a Halliburton technology fellow in drilling. The data needs to be synthesized based on an engineering point of view and then clearly displayed because “the easiest way to communicate data is to the eye.”
Addressing the question of whether the industry is doing a good job using the data it gathers now, the answer was: “Absolutely we do not. We have not scratched the surface of it,” said Shahab Mohaghegh, a petroleum engineering professor at West Virginia University. He is known for using methods, such as artificial intelligence or neural networks, to seek patterns in data sets and learn from them, and has started a company to commercialize the work.
Recently, Mohaghegh has been working on a predictive drilling system. So far it has been tested using drilling data gathered from wells drilled by a large oil company. Mohaghegh said those results indicate it can predict when there is a high probability of trouble ahead, such as a stuck pipe.
Another data analysis company, Ayasdi Government Services, which developed its methods seeking hard-to-detect patterns in huge volumes of data collected by the US military, is now seeking data analysis work in oil and gas. The amount of data does not need to be big. What is needed is a data-based solution sufficiently complex to contain connections that are not obvious, said Benjamin Mann, vice president for energy at Ayasdi.
“Big data is a big cliché,” he said. As for the data speed required: “It is not real time. It is in the time we need to use it.”
Stephen Rassenfoss is the Emerging Technology Senior Editor for the Journal of Petroleum Technology.