Stephen Whitfield, Senior Editor
Finding the speed of penetration (ROP) sweet spot is key to optimizing rotary drilling and thus reducing costs. It’s not just about going as fast as possible. A high ROP can reduce the time required to reach the target depth, but if the ROP is too high it can compromise the drilling function and reduce the quality of the wellbore trajectory. As a result, the total depth that can be achieved may be reduced.
Finding the sweet spot is difficult due to the complex, nonlinear, and multiparameter nature of the drilling process and is influenced by multiple factors such as weight on the bit (WOB), mud properties, and bit type. Additionally, the industry is still primarily focused on traditional ROP prediction and optimization techniques, which often rely on human-analyzed empirical data and physical models that have limited ability to generalize the data across different formations and operating conditions.
Later this year, H&P will launch ROP Optimizer, a new technology to help address this challenge. It utilizes a cloud-based machine learning ROP optimization algorithm for sides, which helps accurately predict the optimal ROP in side sections while adhering to realistic and safe drilling constraints. Leverage historical and real-time data to train ROP prediction models based on input variables such as surface rotation speed, downhole motor rotation speed, and surface estimated WOB.
“If you think about the old way of developing purely physics-based analytical models for ROP, you couldn’t learn quickly and efficiently from well to well,” said Stefan Menin, technical fellow at H&P. “We always had to calibrate the model with new parameters. Sometimes the BHA is different, sometimes the rocks are different, sometimes the bits are different. But with this system, it’s a hybrid. We’re using physically-based modeling, but we’re also doing real-time machine learning.”
Dr. Menand said machine learning primarily addresses challenges in ROP prediction and optimization by capturing hidden patterns and complex relationships within large volumes of drilling data. However, using machine learning comes with its own challenges.
Most existing systems lack integration with physical principles and rely solely on data to optimize ROP, potentially overlooking important dynamics of the drilling system. Additionally, optimization algorithms often do not incorporate operational constraints on recommended drilling parameters, such as rotation speed and WOB. Therefore, it may not be feasible or even safe to implement these parameters in real-time field operations.
“Drillers know the minimum and maximum values they want to set,” Dr. Menand said. “They know, for example, that they want to measure the weight of a bit within a certain range. They know the limits of their equipment. We wanted to make sure that the software takes those limits into account. We want to make sure that the recommendations that this software makes are consistent with what the driller sees. If the software provides an output that exceeds the equipment’s specifications, he will immediately reject it.”
To integrate physical constraints into the software, the ROP optimization algorithm is enhanced with a predictive vibration model. The model calculates the resonant rotational speed required to avoid vibrations, i.e. the specific rotational speed at which the operating frequency of the drill bit matches its natural mechanical frequency. This threshold is embedded as a constraint within the model to ensure that the recommended rotation speed remains within safe operating limits.
“One of the goals of the ROP optimizer is to avoid bit malfunctions. You need proper bit engagement. Any drill bit needs to have a very targeted depth of cut, so you need to make sure the bit engagement is good. You also need to choose the right RPM to avoid resonant frequencies in that model to avoid vibrations,” said Dr. Menand.
The software workflow begins by collecting data from an offset well that is as close as possible to current well conditions in terms of formation, wellbore trajectory, BHA, bit and mud motor specifications. These wells serve as a baseline to compare the performance of the optimized models. Actual drilling data from the current well is collected through WITSML or directly from the rig. This data can also help you fine-tune your model’s performance.
A predictive vibration mapping model uses wellbore trajectory, drilling data, and BHA specifications to recommend surface rotation speeds that avoid vibrations. This output is fed into a ROP optimization algorithm to determine the optimal ROP within preset constraints on drilling parameters. The ROP optimizer sends optimal ROP recommendations to the rig at user-predetermined intervals (for example, every 10 feet) for a predicted distance beyond the current drilling depth.
The system runs on cloud servers. H&P has developed custom middleware (software programs that bridge different applications) for cloud servers and rigs. Recommendations are sent through that middleware to the rig control system, and the autodriller automatically executes the set points. However, Dr. Menand said the system has a manual override feature so drillers can take over commands from the software at any time.
“The driller is always in control,” he said. “Really, they’re the ones deciding what’s going on. If you want to build confidence in the system, you need that. If the driller sees what’s going on for some reason, they can abort the connection, take control, mitigate any potential problems, and then turn the optimizer back on if they want.”
Good results were obtained in virtual and field tests
Initial testing of the ROP optimization algorithm was conducted last year at H&P’s offices on a virtual rig, a software application that models a real drilling rig. The data used to train the algorithm was generated by a separate physical model that emulates the behavior of the target well. The bit-rock interaction model simulated the bit motion, and the torque and drag model calculated the surface torque. Separate physical models reproduced differential pressure, flow rate, ROP, and WOB.
In this test, the optimization agent collected data from the rig for each drilled foot. The internal predictive model is updated every time 100 feet worth of data is accumulated.
At the same time, the agent sends recommendations to the virtual rig every 10 feet. The rig executes this recommendation immediately and keeps all recommendation values constant until new recommendations arrive.
Virtual rig tests showed that the algorithm was able to drive the drilling system from selected initial conditions to optimal conditions by increasing both parameters step-by-step.
“Virtual testing was less about how the model behaved with the data. We wanted to make sure that the model could respond appropriately to changes in values and that those calculations remained robust,” Dr. Menand said.
H&P then conducted field tests on a rig in South Texas, running a ROP optimizer to drill a 7,000-foot lateral section. For this test, H&P’s Real-Time Operating Center (ROC) set up and monitored both the predictive vibration model and the ROP optimizer model. Real-time WITSML data is received from the rig and processed by the ROP Optimizer model to provide updated recommendations to the side sections.
ROC also connected the optimizer to the rig control system, allowing the optimizer to adjust the autodriller’s setpoints (WOB, ROP, surface rotation speed, differential pressure). With driller approval, updated recommendations were sent to the rig control system at regular drilling depth intervals. Updated recommendations were implemented regularly without any additional interaction from the driller. The frequency and limits of these recommendations were agreed with the operator.
Although the driller maintained the ability to resume manual control at any time if needed, the ROP optimizer managed operations in 97% of the drillings in this test, during which the driller adhered to 100% of the model’s recommendations. H&P noted a 26% increase in average ROP for the horizontal well’s side section compared to two offsets drilled on the same pad, same formation, and with the same BHA design.
The ROP Optimizer is currently undergoing additional field testing on a rig in West Texas and another in South Texas. The company plans to make the system available for commercial use on H&P rigs later this year. direct current
For more information, see IADC/SPE 230787, “Rotary Drilling Optimization Framework Using Machine Learning and Predictive Vibration Modeling for Real-Time Rig Control.”
