Surrogate and Dimensionality Reduction-Based Coupled Optimization of Well Pattern and Fracture Network in Tight Oil Reservoirs

Abstract Tight oil reservoirs have become a critical replacement domain for enhancing unconventional oil and gas reserves and production worldwide. The coupled optimization of well patterns and fracture networks is essential for achieving efficient development in such reservoirs. To overcome the limitations of existing methods, including insufficient coverage of optimization parameters, single-objective formulations, and low computational efficiency, this study proposes a surrogate-assisted coupled optimization method for well patterns and fracture networks with dimensionality reduction. Firstly, a dynamically evolving well pattern unit model is constructed based on spatial topology analysis and geometric transformation theory, while incorporating safety constraints, to quantitatively characterize the geometric configuration of well and fracture networks. On this basis, a multi-objective optimization model is established, aiming to maximize fracture-controlled reserves and achieve balanced and effective reservoir utilization, thereby simultaneously considering reservoir recovery and economic benefits. A surrogate-assisted evolutionary optimization algorithm, SADE–Sammon, is introduced. By combining dimensionality reduction mapping and surrogate-assisted iteration, this method efficiently achieves rapid convergence and global optimization in high-dimensional problems. The proposed approach is applied to a representative tight oil reservoir in Block G of the Bohai Bay Basin, China. The results demonstrate that, compared with conventional algorithms, computational efficiency is significantly improved, with the calculation time reduced by 43.14%. The optimized number of wells stabilizes between 8 and 10, with fracture-controlled reserves of approximately 2.0 × 104 m3. The economic net present value (NPV) increases from 60 million CNY to 120 million CNY. The coverage of high oil saturation zones improves from 37% to 81%, while the average oil saturation decreases by 17.3%. Field applications based on the optimization results were further conducted. In the pilot area, the average daily liquid production per well in the first year increased from 3.5 m3/d to 11.6 m3/d. The fracture-controlled reserves increased by 20.41%, and the predicted 10-year cumulative oil production per well increased by 9000 m3. This method effectively overcomes the limitations of traditional optimization techniques and achieves a global coupled optimization of development scale, reservoir utilization, and economic performance, providing robust technical support and theoretical guidance for the efficient development of tight oil reservoirs.