Optimizing the trajectory of directional wellbores is essential to minimize drilling costs and the impacts of potential drilling problems. It poses multi-objective optimization challenges. Well-design optimization models initially focus on wellbore-length minimization, but ideally also need to consider minimizing the surface torque during drilling and address, among other constraints, collision avoidance with offset wells. A novel trajectory-optimization model is described that computes the separation factor along the wellbore. It employs a genetic optimization algorithm with an objective function that maximizes the minimum separation factor along the entire length of a wellbore. Plausible well trajectories are identified within a feasible solution space defined by user-identified constraints. The simplicity and effectiveness of the proposed model are demonstrated using a case study involving real well data from the Reshadat oil field offshore southern Iran. In the case considered, a proposed well trajectory is identified as unsafe in terms of its minimum separation factor with an offset well and is re-planned with the proposed model to achieve a safer trajectory.

This study presents a novel approach for optimizing well paths in extended reach drilling (ERD) wells. Different trajectories can be used for ERD wells, each with its pros and cons. Previous research overlooked certain objective functions in single-objective optimization and lacked an autonomous method for selecting the best solution from Pareto optimal solutions in multi-objective optimizations. Furthermore, they lacked comparing different profiles in well design. Risk assessment and operational factors, which greatly influence optimization and drilling success, were insufficiently considered. This study utilized the multi-objective genetic algorithm (MOGA) and the technique for order preference by similarity to an ideal solution (TOPSIS) method to select the optimal well path based on torque, wellbore length, risk (e.g., keyseat), and required tools. First, all possible trajectories were determined, and MOGA identified the optimal path with minimal torque and length. The fuzzy decision-making method automatically selected the best solution from the Pareto optimal solution set. The associated risks and required tools are evaluated for each trajectory. Finally, the TOPSIS method selected the optimal trajectory based on torque, length, risks, and required tools. The case study demonstrated that the undersection path was the most advantageous trajectory for ERD wells, with a 60% closeness to the ideal state. The multiple build trajectory achieved 57% closeness, while the build and hold and double build paths had lower closeness values (43 and 28%, respectively). Consequently, it can be inferred that in the context of ERD wells, it is preferable to carry out the deviation process at deeper depths.

In recent years, rate of penetration (ROP) prediction using machine learning has attracted considerable interest. However, few studies have addressed ROP prediction uncertainty and its relation to training data and model inputs. This paper presents the application of a quantile regression deep neural network (QRDNN) for ROP prediction on multi-lateral wells drilled in the Alvheim field of the North Sea. The quantile regression framework allows the characterization of the prediction uncertainty, which can inform the end-user on whether the model predictions are reliable. Three different training strategies for the QRDNN model are investigated. The first strategy uses individual hole sections of the multi-lateral well to train the model, which is then tested on sections of similar hole size. In the second strategy, the models are trained for specific formations encountered in the well, assuming the formation tops are known for both the training and test sections. The third strategy uses training data from offset wells from the same field as the multi-lateral well, exploring different offset–well combinations and input features. The resulting QRDNN models are tested on several complete well sections excluded from the training data, each several kilometers long. The second and third strategies give the lowest mean absolute percentage errors of their median predictions of 27.3% and 28.7% respectively—all without recalibration for the unknown test well sections. Furthermore, the third model based on offset training gives a robust prediction of uncertainty with over 99.6% of actual values within the predicted P10 and P90 percentiles.