Browsing by Subject "Drilling optimization"
Now showing 1 - 4 of 4
- Results Per Page
- Sort Options
Item Continuous learning of analytical and machine learning rate of penetration (ROP) models for real-time drilling optimization(2018-08-16) Soares, Cesar Mattos De Salles; Gray, Kenneth E., Ph. D.; Daigle, Hugh; Pyrcz, Michael; Millwater, Harry; Panchal, NeilkunalOil and gas operators strive to reach hydrocarbon reserves by drilling wells in the safest and fastest possible manner, providing indispensable energy to society at reduced costs while maintaining environmental sustainability. Real-time drilling optimization consists of selecting operational drilling parameters that maximize a desirable measure of drilling performance. Drilling optimization efforts often aspire to improve drilling speed, commonly referred to as rate of penetration (ROP). ROP is a function of the forces and moments applied to the bit, in addition to mud, formation, bit and hydraulic properties. Three operational drilling parameters may be constantly adjusted at surface to influence ROP towards a drilling objective: weight on bit (WOB), drillstring rotational speed (RPM), and drilling fluid (mud) flow rate. In the traditional, analytical approach to ROP modeling, inflexible equations relate WOB, RPM, flow rate and/or other measurable drilling parameters to ROP and empirical model coefficients are computed for each rock formation to best fit field data. Over the last decade, enhanced data acquisition technology and widespread cheap computational power have driven a surge in applications of machine learning (ML) techniques to ROP prediction. Machine learning algorithms leverage statistics to uncover relations between any prescribed inputs (features/predictors) and the quantity of interest (response). The biggest advantage of ML algorithms over analytical models is their flexibility in model form. With no set equation, ML models permit segmentation of the drilling operational parameter space. However, increased model complexity diminishes interpretability of how an adjustment to the inputs will affect the output. There is no single ROP model applicable in every situation. This study investigates all stages of the drilling optimization workflow, with emphasis on real-time continuous model learning. Sensors constantly record data as wells are drilled, and it is postulated that ROP models can be retrained in real-time to adapt to changing drilling conditions. Cross-validation is assessed as a methodology to select the best performing ROP model for each drilling optimization interval in real-time. Constrained to rig equipment and operational limitations, drilling parameters are optimized in intervals with the most accurate ROP model determined by cross-validation. Dynamic range and full range training data segmentation techniques contest the classical lithology-dependent approach to ROP modeling. Spatial proximity and parameter similarity sample weighting expand data partitioning capabilities during model training. The prescribed ROP modeling and drilling parameter optimization scenarios are evaluated according to model performance, ROP improvements and computational expenseItem End-to-end drilling optimization using machine learning(2018-08-07) Hegde, Chiranth Manjunath; Gray, Kenneth E., Ph. D.; Daigle, Hugh; Millwater, Harry; Pyrcz, Michael; Halal, AfifDrilling costs occupy a significant portion of oil and gas project’s budget. Optimization of drilling - increasing speed, reducing vibrations, and minimizing borehole instability - can lead to significant savings and hence have been extensively studied. Currently, most drilling optimization tools (or models) only tackle a single drilling metric: they seek to optimize either the rate of penetration (ROP), torque on bit (TOB), mechanical specific energy (MSE) or drilling vibrations. Models are often built independent of other metrics (without coupling) and do not accurately represent downhole conditions since drilling metrics are interrelated. This may lead to over or underestimation of the metric optimized which can severely reduce the effect of optimization. The objective of this dissertation is to introduce techniques, strategies, and algorithms that can be used to build a fully coupled drilling optimization model. Drilling optimization is studied by first optimizing ROP– where models for ROP prediction and inference are constructed using machine learning. Strategies and algorithms for determining optimal drilling parameters using ROP models are discussed. The unique problem posed by data-driven models are solved using meta-heuristic algorithms. A coupled model is constructed by building ROP, TOB, and MSE models conjointly using the random forests algorithm. Drilling vibrations – axial, lateral, and torsional – are modeled using a machine learning classification algorithm. This classification algorithm used to restrict the optimization space, ensuring that optimal parameters do not induce vibrations ahead of the bit. This model is used to investigate the effect of optimizing ROP and MSE on field data. A workflow is introduced linking all the aforementioned models into an end-to-end drilling optimization tool. The tool can be used as a recommendation system where field-measured data are used to determine and implement optimal drilling parameters ahead of the bit. The dissertation illustrates the use of statistical (or machine) learning techniques to address the problems encountered in drilling optimizationItem Improved torque and drag modeling using traditional and machine learning methods(2020-08-13) Oyedere, Mayowa Olugbenga; Gray, Kenneth E., Ph. D.; Foster, John T., Ph. D.; Daigle, Hugh; Millwater, Harry; Jones, JohnDuring the drilling process, the drillstring inadvertently comes in contact with the wellbore generating frictional losses in the rotating moment (torque) and axial force (drag). These losses reduce the rotational power available at the drill bit, thus making adequate torque and drag modeling a critical piece in the drilling puzzle. The simplifying assumptions of the widely used soft-string model for torque and drag modeling make it less accurate for new complex well designs, creating the need for the use of the more robust stiff-string model. This first part of this dissertation focuses on a new approach to developing a stiff-string model that can be easily implemented for well planning. The stiff-string model addresses the pitfalls of the soft-string model by using cubic splines for its well-path trajectory. It solves the three coupled, non-linear ordinary differential equations that describe the motion of the drillstring at each survey point to account for the shear forces and bending stiffness. The stiff-string model is then applied to design four horizontal wells. Drilling Optimization has consistently generated research interest over the years because of the cost-saving benefits of improving drilling efficiency. Rate of penetration (ROP) and torque-on-bit (TOB) predictions have become critical to the successful drilling optimization efforts. The second part of this dissertation focuses on the prediction and optimization of TOB using five regression-based machine learning algorithms. TOB was modeled as a function of rotary speed (RPM), weight-on-bit (WOB), flow rate, pump pressure, and unconfined compressive strength (UCS). Three direct search optimization algorithms—Nelder Mead, differential evolution, and particle swarm optimization (PSO)—were used to optimize TOB. The final part of this dissertation introduces a novel approach to ROP and TOB prediction by modeling it as a classification problem with two regions (low and high ROP and TOB respectively) based on a user-defined threshold. Five different classification algorithms were implemented and compared using the area under curve (AUC) classification metric. Finally, a probability gradient tool was developed to help inform the drilling engineer on the best combination of WOB and RPM to yield the desired drilling performanceItem Self-learning control of automated drilling operations(2018-06-22) Pournazari, Parham; Oort, Eric van; Fernandez, Benito R.; Chen, Dongmei; Barr, Ronald E.; Niekum, ScottIn recent years, drilling automation has sparked significant interest in both the upstream oil and gas industry and the drilling research community. Automation of various drilling tasks can potentially allow for higher operational efficiency, increased consistency, and reduced risk of trouble events. However, wide adoption of drilling automation has been slow. This can be primarily attributed to the complex nature of drilling, and the high variability in well types and rig specifications that prevent the deployment of off-the-shelf automation solutions. Such complexities justify the need for an automation system that can self-learn by interacting with the drilling environment to reduce uncertainty. The aim of this dissertation is to determine how a drilling automation system can learn from the environment and utilize this learning to control drilling tasks optimally. To provide an answer, the importance of learning, as well as its limitations in dealing with challenges such as insufficient training data, are explored. A self-learning control system is presented that addresses the aforementioned research question in the context of optimization, control, and event detection. By adopting an action-driven learning approach, the control system can learn the parameters that describe system dynamics. An action-driven approach is shown to also enable the learning of the relationship between control actions and user-defined performance metrics. The resulting knowledge of this learning process enables the system to make and execute optimal decisions without relying on simplifying assumptions that are often made in the drilling literature. Detection of trouble drilling events is explored, and methods for reduction of false/missed alarms are presented to minimize false interruptions of the drilling control system. The subcomponents of the self-learning control system are validated using simulated and actual field data from drilling operations to ascertain the effectiveness of the proposed methods.