Summary Shale-oil horizontal wells are increasingly characterized by large inclination depth, long horizontal sections, and densely distributed multiple targets, making efficient trajectory design critical to drilling safety and development performance. Although dual-2D trajectory configurations have shown operational advantages, current designs still rely heavily on empirical templates and lack a unified, quantitative optimization framework under narrow target windows and stringent engineering constraints. To address the reliance of existing dual-2D trajectory designs on empirical templates and the lack of a unified, quantitative biobjective optimization model under multiple targets and stringent engineering constraints, we propose a biobjective dual-2D trajectory-optimization framework for shale-oil horizontal wells. A segmented parametric trajectory model couples an upper 3D azimuth-turning section with a lower 2D landing and buildup section while explicitly enforcing curvature, dogleg, tool capability, and target-window constraints. Building on this model, an enhanced distributed differential evolution (DE) algorithm (ECMADE), is developed to solve the high-dimensional, strongly constrained, multimodal optimization problem and to generate Pareto-optimal trajectory schemes that jointly balance target-entry accuracy and designed well depth. Research has shown that (1) the proposed biobjective dual-2D trajectory optimization framework couples the upper 3D azimuth-turning section with the lower 2D landing and buildup section. Under strict curvature, dogleg severity, and engineering constraints, it achieves high-precision target entry while reducing the designed measured depth (MD), thereby substantially improving trajectory planning performance for shale-oil horizontal wells. (2) In the Gulong GY3 field case, the optimized trajectory yields a total MD of 5,285.78 m, which is 36.22 m shorter than the original design, while the target-entry error is only 0.000111781 m2. The resulting trajectory is continuous and smooth, indicating good drillability and compliance with all constraints. (3) The ECMADE-based optimization produces a relatively uniformly distributed Pareto front, enabling practical trade-offs between designed well depth and target-entry error. In terms of convergence behavior and solution-set quality, the proposed method consistently outperforms particle swarm optimization (PSO), classical DE, and nondominated sorting genetic algorithm II (NSGA-II), as evidenced by a more compact and better-structured solution set, greater concentration around the principal knee regions, and overall lower errors within comparable depth ranges. (4) Ablation experiments based on the evolutionary curves of the best fitness across generations show that the full algorithm maintains higher fitness levels and faster improvement rates throughout the iterations. These results confirm pronounced synergies among cooperative multisubpopulation evolution, multioperator parallel search, state-triggered information exchange and elite migration, and success-history-based adaptive parameter control; stable convergence and the best overall solution quality are achieved only when all modules are integrated.
Zhu et al. (2026) studied this question.