Slide-Drilling Guidance System Optimizes Directional Drilling Path
The amount of uncertainty related to directional drilling makes accurate drilling challenging, leaving much to human know-how and interpretation. Additionally, few path-planning methods in the literature consider the directional steering tool being used.
The amount of uncertainty related to directional drilling makes accurate drilling challenging, leaving much to human know-how and interpretation. Additionally, few path-planning methods in the literature consider the directional steering tool being used. The formulation of the problem of optimization varies greatly between rotary steerable systems (RSS) and mud-motor configurations. Additional cost functions and constraints exist for mud-motor use, which significantly increases the complexity of the problem. A slide-drilling-guidance system is proposed to combat this issue and to help automate directional drilling.
Despite the numerous technological advancements that made directional drilling a reality, directional drilling is still viewed more as an art than as a science. The drillstring and penetrated rock formations are complex systems whose interaction makes prediction and planning highly uncertain.
Analytical, geometric approaches have sought to formulate the curvature of the well path in a way that eases the planning process and ensures smooth trajectories. The literature on directional drilling is rich in well-path-optimization methodologies based on operating constraints and well-path geometry. However, few of these methods seem to consider the tool used to generate the well path, and those that do typically use RSS because these yield better results from a technical perspective. Mud motors, however, are still used extensively because of cost effectiveness.
The complete paper proposes a directional-drilling guidance system targeted at downhole motors and slide drilling to generate optimal well paths and associated slide-drilling instructions. Instead of a simplistic shortest-path approach, various cost functions aimed at maximizing the value of the well are considered. For wellbore-propagation prediction, a computationally efficient and accurate model is used. The proposed methodology is built into a software package and is being developed and implemented currently for use in the real-time drilling (RTD) system of an operator.
The three main modules of the software are discussed in detail in the complete paper. The wellbore-propagation model predicts the evolution of the borehole and returns the predicted 3D path given the initial condition and a slide sheet. The fitness-evaluation module calculates the fitness of the 3D path using a cost function and returns a scalar value representing the overall value of the path. The genetic-algorithm solver generates candidate slide sheets and approaches an optimal slide sheet by iterating over generations.
Graphical User Interface (GUI). The GUI serves as a tool to visualize the well under consideration and to study the optimization results and the effectiveness of the approach. The GUI is written in MATLAB and uses the Runtime environment. Here, the data for the planned path and the actual survey data are loaded. The planned path and the surveyed path (up to the time for which such data are available) are displayed.
For model calibration, a previously drilled section of the well is selected and the respective slide sheet of the section is loaded. Afterward, the model is used to simulate a path within the given section of the slide sheet, starting from the first survey point of the selected section.
Once the model is calibrated, the system can be used to search for an optimal path and the associated drilling actions for the following sections of the well. In a field-use scenario, the most recent survey point would be selected to provide suggestions for the path being drilled in real time. For analysis purposes, previous survey points can be selected to compare what was drilled to what the optimizer might suggest. If desired, an initial guess for the slide sheet can be provided. The optimization parameters can be adjusted depending on the requirements. Finally, the optimizer is run given the initial condition, initial solution estimate, and appropriate optimization parameters. The fitness value that represents the objective function is displayed as well as the current iteration or epoch. These values are updated with each iteration of the optimization algorithm.
Once the optimization is complete, the resulting path is displayed. Other parameters such as drilling time are stored internally and can be accessed using the command window in MATLAB.
Integration With RTD Analytics System. The short-term plan for the proposed optimization framework is to make it production-ready so that it can be integrated into the RTD system as a decision-making tool for the directional driller or wellsite supervisor. The long-term plan is to make the framework as robust as possible so that it can be embedded into the rig’s closed-loop control system as a part of the algorithms for autonomous drilling.
The optimization framework is scheduled to be integrated into an operator’s RTD analytics system for testing purposes. This system is a flexible RTD platform built in-house by the operator. One of the important features of the RTD system is that it is designed for ease of plug-and-play testing. Thus, the integration of the proposed model is expected to be straightforward. The proposed optimization framework was set up as a Python library and can be easily installed into the Python environment of the RTD system. Within the RTD system, all inputs required for the optimization framework, such as the slide sheet, the actual survey, and the planned well path, are available in real time. When a real-time survey becomes available in the RTD system, the optimization calculation will be triggered. Once the calculation is complete, the resulting drilling instructions will be fed back to the RTD system and made available in the user interface of the RTD system for use at both the rig and the office.
Results and Discussion
To test the optimization framework, a case study using field data from a drilled well was completed. The algorithm was used to find an optimal path for the build section and the subsequent lateral section of the well. The model was calibrated using the original slide sheet and survey data. The surveyed path, the simulated path, and the positional error of the simulated path are presented in Fig. 1. The simulated path had approximately 1% positional error compared with the surveyed path. After model calibration, the algorithm was tasked to find a better path and a corresponding slide sheet.
The optimized path shows significant improvements over the surveyed (actual) path that was drilled. For the section that was considered [with a total measured depth (MD) of 1,192 ft], the optimized slide sheet has only eight slide actions compared with 13 in the original. The deviation of the optimized path from the planned path is also smaller compared with the actual deviation throughout the build section. The large tracking errors present in the original surveyed path were fixed in the optimal path, which resulted in an increase of over 500 ft MD in the lateral section, which translates into an increased pay zone. This did result in an increase in total slide MD by 22 ft (from 642 ft in the original slide sheet to 664 ft in the optimized slide sheet), but the economic benefits (increased MD in the pay zone would be expected to result in an increase in hydrocarbon production) justify this increase easily. Similarly, the reduction of the total number of slide actions would save on setup time and would lead to a reduction in drilling time. With these improvements, significant cost savings would be possible if the suggestions from the optimization framework were available during drilling.
The results of the proposed method are promising. Generally, global optimization algorithms are considered time-consuming for a problem of this scale and not suitable for real-time applications. To address this issue, heuristic optimization methods often are used to expedite the computational process. Among the many heuristic optimization methods, a genetic algorithm (GA) was selected and tested in this research effort. The results have shown that the GA approach has optimized the drilling process significantly, with limited computational expense. This makes the GA approach applicable for real-time drilling operations.
This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 194096, “Slide-Drilling Guidance System for Directional-Drilling Path Optimization,” by Can Pehlivantürk and John D’Angelo, The University of Texas at Austin, and Dingzhou Cao, Anadarko, et al., prepared for the 2019 SPE/IADC International Drilling Conference and Exhibition, The Hague, 5–7 March. The paper has not been peer approved.