Summary A digitized workflow from predrill pore-pressure modeling with a Monte Carlo approach until update of the pressure prognosis while drilling from (for example) sonic or resistivity data is described. The approach has the potential to reduce the uncertainty in the predicted mud-weight window ahead of the bit. For the 3D pressure modeling, a basin modeling software is used, where the pressure compartments in the study area are defined by faults interpreted from seismic. Pressure generation and dissipation are calculated for the study area over millions of years, as the basin was subsiding and compaction was taking place. Key input parameters such as minimum horizontal stress, vertical stress, and frictional coefficients for failure criteria are varied. The output is pore-pressure profiles along the planned well path, with uncertainties. The work presented in this paper was carried out on a North Sea data set. The results show that the uncertainty in the pore pressures will highly influence the uncertainty span in both the fracture gradient and the collapse gradient. Representing the mud-weight window in terms of the most likely collapse and fracturing curve, with corresponding minimum and maximum pore-pressure-derived limits on each side, makes for a more realistic prediction. It presents the uncertainty in the result in a simple visual form, using a “traffic light” approach. While drilling, log data will automatically be used to update the pressure and mud-weight prognosis ahead of bit. The digital updated prognosis can help the drilling crew in decision making during drilling campaigns.
For automation of managed pressure drilling (MPD) to succeed, the automation system needs access to accurate measurement data and the ability to translate this into a correct representation of reality in the well. However, inaccuracies due to calibration problems and errors or omissions in manually entered data may combine with spurious behavior in the control model to make the automation system unreliable. This study presents a novel control scheme for automated MPD that addresses the problem of model reliability by using multiple control models to provide optimal control moves, even with the failure of one or two models. This work simulates the ability to optimize MPD operations with realistic measurement signals using Model Predictive Control (MPC) with a range of model types. Further, it provides a robust automated MPD system to reduce interruptions to drilling operations. This work makes use of a high fidelity dynamic well bore model in addition to low order and empirical control models. The three controllers feed into a switch that selects the best available controller recommendation and allows for a seamless transition between controllers. The switch control scheme enables automated switching between the controllers in the event of one or two models failing and also allows for the tuning and troubleshooting of one model while the others continue to run, all without any interruption to the drilling process. One of the key innovations in the ensemble switch is the seamless transition between controllers. This is accomplished by using the current process manipulated variable values as initial values for the optimization routines in the model predictive controllers that are not actively used to control the well. The strategy is tested in common drilling situations. Two typical drilling scenarios are simulated: normal drilling operations and a pipe connection procedure. The validity of the novel control structure in each scenario is verified through simulated outliers, drift, and noise, as well as simulated controller failure and lack of optimal solution convergence. The controller is able to maintain bit pressure within +/− 1 bar of the 400 bar set point during normal drilling operations despite temporary signal loss and poor data quality. Also, the bit pressure is held within +/− 5 bar of the 340 bar set point during a pipe connection procedure with no bit pressure measurements available to the controller. The techniques presented here can be used for more robust and stable automated MPD. Moreover, multiple models provide benefits that are typically associated with improved reliability due to hardware and safety systems redundancy allowing drilling to continue with fewer interruptions.
Physics-based hydraulic models are essential for proceeding to a high level of automation in drilling. Mathematical models can facilitate process understanding and problem detection, and determine appropriate actions in case of mismatch between model and data. Furthermore, calculations may replace measurements where and when the latter are not available, as normally occurs during connections or when instruments or signal transmissions fail. However, advanced hydraulic models rely on a large set of inputs, such as pipe and wellbore geometry, various tuning parameters and fluid properties. The models are therefore time-consuming and difficult to configure in the field, where third-party experts may be needed at each well, to properly initiate the automation system and adjust it during the drilling process. Although the methods described in this paper are relevant to any critical drilling operation, they are applied to Managed Pressure Drilling (MPD) as a widely deployed example of drilling automation. In MPD, hydraulic models predict downhole conditions and determine the requisite choke pressure for automatic adjustment. A new method for automatic configuration of key model parameters simplifies the tedious job of setting up the model and ensures that the automation system remains tuned to the well, even without onsite model tuning expertise. The proposed scheme is based on a simple method for separating inaccuracies due to co-linearity in frictional pressure losses and static mud weight. The search for optimal correction factors is based on a sequence of small oscillations of pump rate that can be applied during drilling without interrupting the operation. A massively parallel computing architecture improves the speed of the calibration algorithm proportional to the number of available CPU cores. A set of hydraulic model instances runs in parallel, allowing for efficient testing of changes in input signals within ranges of uncertainty. A method for selecting a subset of the best models that more accurately represent a given well is proposed. Computer simulations demonstrate how the novel calibration scheme allows automatic tuning of the friction factor and density correction factor, giving accurate prediction of the bottom hole pressure (BHP). The tuning scheme is run with a parallel architecture to demonstrate that correct values of unknown configuration parameters can be automatically determined sufficiently fast for real-time drilling control or as an advisory tool. The deployment of automation systems in drilling is hampered by the need for dedicated expert personnel to maintain systems that could have reduced the personnel needed on the rig. The proposed automated physics-based model tuning contributes to removing this roadblock, aiming at making automation systems a more cost-efficient option for drilling operations.
Objectives/Scope The objective of this work is to present a first step towards a hybrid approach between machine learning (ML) and physics-based modelling to provide decision support for drilling problems. The motivation for developing a hybrid approach is to obtain methods that are more reliable and easier to automate than physics-based models, while still have enough accuracy and predictivity. In this first step, we replicate the performance for predicting downhole pressure in a well of a high-fidelity simulator based upon physical principles by using ML methods. In addition, we also suggest a future roadmap. Methods, Procedures, Process A high-fidelity physics-based model for drilling and well control operations is used to generate vast amounts of data for two cases, drilling with not major events, and drilling into an over-pressured reservoir. Key simulation input parameters and assumptions are varied to create realistic scenarios. We replicate the high-fidelity simulator downhole pressure predictions by two supervised machine learning algorithms. Random forest (RF) and recurrent neural network (RNN). The hybrid approach is flexible and is also employed for kick detection and estimation of the mass of the influx. After using unaltered data from the high-fidelity simulator, we also demonstrate the ML methods on corrupted data with synthetic noise. Results, Observations, Conclusions RF and RNN obtained very high accuracy, predicting bottom hole pressure with small error margin. Good results were also obtained for the kick estimation and kick detection cases. Tested on corrupted data, RF trained with noise performed significantly better compared to RF trained without noise, at the cost of a slight reduction in accuracy in the error free scenario. Initials tests on real data are ongoing and further work is needed. Hybrid methods have the potential of performing well with noisy environments and are valuable tool to be used in drilling automation. Novel/Additive Information Combining highly advanced dynamic models for drilling and well control with modern ML methods has not been done earlier to the best knowledge of the authors. Demonstrating this on real data will be valuable because data-driven and physics-based approaches used separately are considered inadequate for future automated drilling concepts.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.