Chevron and Halliburton proved a fully autonomous closed-loop fracturing program in Colorado and are incorporating additional diagnostics into the system.
The closed-loop system, which includes sensing, decision logic, and execution layers, autonomously and in real time dynamically adjusts completion parameters based on data from the subsurface. It was developed as part of a continuous improvement effort, not a science project, Awais Navaiz, technical advisor at Halliburton, said while presenting SPE 230613 at SPE’s Hydraulic Fracturing Technical Conference and Exhibition in February.
Jason Bell, Rockies completions engineer at Chevron, said during the presentation the focus of the deployment was to prove the technology.
“We were trying to rapidly deploy the system, prove up the technology, and we needed to be able to work with some bounds. We’re trying to convince our companies that this is something that’s useful and important, and we needed to be able to do it quickly, and we couldn’t disrupt the schedule, and we couldn’t disrupt the asset,” he said.
Specifically, the paper’s authors wrote, the program jointly created by Chevron and Halliburton aimed to create and deploy a fully autonomous, sensor-informed, closed-loop fracturing system in an unconventional asset. During the deployment in Colorado, it needed to demonstrate continuous and reliable autonomous execution, gather and use noninvasive diagnostics to obtain actionable subsurface insights without disrupting operations, and show how automated workflows can help transform subsurface feedback into dynamic changes on the pad. The workflow condenses decision-making from minutes or hours by a human to a few seconds by the autonomous system.
The first phase of the deployment involved diagnostics to enable an understanding of current fracturing performance, followed by efficiency optimization during which the fracturing treatment was automated. The final phase was integrating subsurface feedback to enable the closed-loop capability to execute treatment changes autonomously.
Navaiz said the team obtained diagnostics via disposable fibers, which are low-cost and noninvasive, to collect data for monitoring baseline key performance indicators around how the fracture system is behaving and growing.
Surface automation, a key foundation of the program, reduced human decisions on location by about 90% and increased computer-made decisions by 14 times, he said. “The computer can execute a stage the same exact way day in and day out, thousands of stages, given the guardrails that are defined. (With) humans, there's always variability.”
Navaiz said the program defined the fully closed-loop operation as a self-regulating, self-contained process that used feedback to control its performance. “Basically, the process is considered fully closed-loop when the feedback is continuously integrated into the system without any manual intervention,” he said.
For the deployment by Chevron and Halliburton, subsurface data was fed continuously into the system, which used the information to make decisions based on a predefined logic. “This decision is sent out to the field and executed by the frac pumps, and no humans touch this process,” he said.
Energy-Engineering Workflow
The energy-engineering workflow connected the diagnostics feedback with the surface automation. “That, for us, was the first flag in the ground in terms of closing the loop,” Navaiz said. “The first workflow that we developed was primarily targeting redistributing energy in the reservoir, dynamically, using fracture-growth feedback. And the premise of this was uniformity.”
The energy-engineering workflow redistributes the energy in the reservoir dynamically based on feedback about fracture growth to encourage more uniformity in the fractures. “When we design our frac stages, we assume everything to be equidistant. Equidistant clusters, equidistant stages, equidistant wells and pads. Everything is planned to be that way. And our assumption in that is we want everything to be exactly uniform. We want uniform stages, uniform wells, uniform DSU (drill spacing unit) fracture growth. But in reality, it looks drastically different,” he said.
An alternative to sending slurry into fast-propagating stages is to limit volumes until a better-behaving stage is encountered and then reinject the “banked” volume into that, he said. The closed-loop energy-engineering workflow does that (Fig. 1).
Bell said determining which stages were performing well or poorly required lots of data, which was collected via disposable fiber optics. “We collected data over 1,500 stages, just in the background to the whole frac program as the frac program just moved along. Just collecting data, sending it to the cloud, getting it analyzed, cataloging it, defining it for us. During that whole period, we didn’t change anything with our frac program. Same stage design, same everything. We made no changes, just collecting background data, cataloging it, characterizing it,” he said.
Characterizing the data enabled the team to determine which regions had “fast” and which had “slow” propagating fracturing systems, he said.
Navaiz said when the computer ascertained a slow frac was occurring, it would pull slurry volume from the piggy bank and apply it to the slow-propagating stage (Fig. 2).
The system is modular, he added, so if necessary, the observe, decide, or act portion could be swapped. “You could replace whatever you want in all of this. You want to change your observation from offset fibers to another diagnostic, be our guest. You want to change the action or you want to change the decision logic, go ahead,” he said.
Bell said the system’s modularity gives it a plug-and-play nature that can easily adapt to different scenarios.
“Maybe fiber’s not the answer, and you don’t have offset DUCs (drilled but uncompleted wells), but there’s another diagnostic that you can plug in. That’s the power of the system,” he said. In the future, he added, other sensing tools may be used.
Deployment in Colorado
Bell said the initial deployment ran in an open-loop setting before the closed-loop deployment. In the first closed-loop application, he said, the team pumped more than 90% of the stages with full end-to-end automation. He said the approach allowed the team to de-risk and stress-test the system, and it was deployed on simul-fracture pairings.
“We completed and demonstrated a fully autonomous, subsurface-driven frac application with closed-loop feedback. No human intervention,” he said. The system “took sensory data, processed that data in the cloud, made a decision up with it, sent that decision—whatever the change was going to be, whether there was going to be a change or not—back to the frac fleet, and the frac equipment automatically made that change. No human intervention.”
Looking to the Future
The authors concluded an intelligent surface fracturing operation that is guided by downhole measurements can trigger dynamic completion decisions to optimize a completions program. Future work is likely to involve more complex decision-logic architecture, and additional diagnostics are being considered for inclusion in the workflows, they wrote.
Navaiz said the overarching agenda the team is working on now revolves around maintaining operational efficiency.
“The limitation that we’re now trying to tackle is making our operations so versatile that we can actually make changes on independent wells with simul-frac, trimul-frac, quad-frac if we need to and not impact this completions factory,” he said.
For Further Reading
SPE 230613 Transforming Hydraulic Fracturing: The First-Ever Closed-Loop Completions Program by A. Navaiz and P.F. Stark, Halliburton; M. Paradeis, formerly Chevron USA Inc., now with Subterra Energy Consulting; J. Bell, D. Beasley, E. White, and H. Lynch, Chevron; and F. Adil, J.B. Tran, C. Cox, and J. Doucette, Halliburton.
