The synergistic development of low-permeability reservoirs such as deep coalbed methane (CBM) and tight gas has emerged as a key technology to reduce development costs, enhance single-well productivity, and improve gas recovery. However, due to fundamental differences between coal seams and tight sandstones in their pore structure, permeability, water saturation, and pressure sensitivity, significant variations exist in their flow capacities and fluid production behaviors. To address the challenges of production allocation and main reservoir identification in the co-development of CBM and tight gas within deep gas-bearing basins, this study employs the transient multiphase flow simulation software OLGA to construct a representative dual-gas co-production well model. The regulatory mechanisms of the gas–liquid distribution, deliquification efficiency, and interlayer interference under two typical vertical stacking relationships—“coal over sand” and “sand over coal”—are systematically analyzed with respect to different tubing setting depths. A high-precision dynamic production allocation method is proposed, which couples the wellbore structure with real-time monitoring parameters. The results demonstrate that positioning the tubing near the bottom of both reservoirs significantly enhances the deliquification efficiency and bottomhole pressure differential, reduces the liquid holdup in the wellbore, and improves the synergistic productivity of the dual-reservoirs, achieving optimal drainage and production performance. Building upon this, a physically constrained model integrating real-time monitoring data—such as the gas and liquid production from tubing and casing, wellhead pressures, and other parameters—is established. Specifically, the model is built upon fundamental physical constraints, including mass conservation and the pressure equilibrium, to logically model the flow paths and phase distribution behaviors of the gas–liquid two-phase flow. This enables the accurate derivation of the respective contributions of each reservoir interval and dynamic production allocation without the need for downhole logging. Validation results show that the proposed method reliably reconstructs reservoir contribution rates under various operational conditions and wellbore configurations. Through a comparison of calculated and simulated results, the maximum relative error occurs during abrupt changes in the production capacity, approximately 6.37%, while for most time periods, the error remains within 1%, with an average error of 0.49% throughout the process. These results substantially improve the timeliness and accuracy of the reservoir identification. This study offers a novel approach for the co-optimization of complex multi-reservoir gas fields, enriching the theoretical framework of dual-gas co-production and providing technically adaptive solutions and engineering guidance for multilayer unconventional gas exploitation.

The estimation of space-varying geological parameters is often not computationally affordable for high-dimensional subsurface reservoir modeling systems. The adjoint method is generally regarded as an efficient approach for obtaining analytical gradient and, thus, proceeding with the gradient-based iteration algorithm; however, the infeasible memory requirement and computational demands strictly prohibit its generic implementation, especially for high-dimensional problems. The autoregressive neural network (aNN) model, as a nonlinear surrogate approximation, has gradually received increasing popularity due to significant reduction of computational cost, but one prominent limitation is that the generic application of aNN to large-scale reservoir models inevitably poses challenges in the training procedure, which remains unresolved. To address this issue, model-order reduction could be a promising strategy, which enables us to train the neural network in a very efficient manner. A very popular projection-based linear reduction method, i.e., propel orthogonal decomposition (POD), is adopted to achieve dimensionality reduction. This paper presents an architecture of a projection-based autoregressive neural network that efficiently derives an easy-to-use adjoint model by the use of an auto-differentiation module inside the popular deep learning frameworks. This hybrid neural network proxy, referred to as POD-aNN, is capable of speeding up derivation of reduced-order adjoint models. The performance of POD-aNN is validated through a synthetic 2D subsurface transport model. The use of POD-aNN significantly reduces the computation cost while the accuracy remains. In addition, our proposed POD-aNN can easily obtain multiple posterior realizations for uncertainty evaluation. The developed POD-aNN emulator is a data-driven approach for reduced-order modeling of nonlinear dynamic systems and, thus, should be a very efficient modeling tool to address many engineering applications related to intensive simulation-based optimization.

Introduction: For reservoirs characterized by vertically superimposed distributions of tight gas and deep coalbed methane (CBM), resource utilization efficiency remains low and economic returns are often poor. Dual-gas co-production offers a promising approach to enhance overall development efficiency. However, significant differences in reservoir pressure, gas–liquid ratio, and fluid properties between the upper and lower reservoirs lead to divergent pressure management strategies. Inappropriate coordination during co-production can result in pressure interference and adversely affect single-well productivity. The dual gas lift technique, with its multi-flow-path capability, is well-suited for accommodating varying reservoir pressures and fluid characteristics, making it particularly advantageous in dual-gas co-production applications.Method: In this study, the three-flow-path characteristics of the dual gas lift system are considered. A wellbore heat loss model is developed by incorporating heat transfer calculations through the tubing–tubing annulus gas injection channel into the thermal model. Using a “multiphase flow experimental setup with complex internal boundaries,” the flow behavior across different flow paths is analyzed, and a multiphase flow model tailored to dual gas lift operations is established through optimal selection. Based on the properties and development needs of vertically superimposed reservoirs, an optimized design methodology for pressure-divided and zone-specific production in dual gas lift wells is proposed. The full life cycle of a dual-gas co-production well is divided into four stages: initialization and pressure control, stable co-production, decline optimization, and late-stage low production. Corresponding production strategies are formulated for each phase.Result and Discussion: The results demonstrate that the dual gas lift system can effectively exploit the distinct characteristics of vertically superimposed reservoirs, enabling efficient and coordinated co-production while satisfying drainage and production requirements throughout the well's life cycle. Model predictions were validated against experimental pressure monitoring data, showing a pressure prediction accuracy exceeding 90% across various flow pattern.