Injecting impure CO2 for enhanced gas recovery (CO2-EGR) offers a dual benefit by improving natural gas extraction while enabling CO2 sequestration. However, the interactions between CO2, N2, and CH4 under reservoir conditions require further investigation. This study employs Grand Canonical Monte Carlo (GCMC) and Molecular Dynamics (MD) simulations to quantify the adsorption and diffusion behaviors of CO2, N2, and CH4 in quartz nanopores over a pressure range of 1–24 MPa under varying water saturations and gas compositions. The results indicate that: (1) CO2 exhibits the broadest energy distribution and the strongest adsorption stability, occupying about 20 %–30 % more adsorption sites than CH4 or N2 and showing the least sensitivity to water saturation, with only a 30 % reduction at 50 % saturation, compared to 60 % for CH4, giving CO2 a clear competitive advantage. (2) The adsorption and desorption behaviors are strongly pressure dependent, as increasing pressure reduces the adsorption layer area and shifts gas distribution from adsorption dominated to free phase. Competitive adsorption analysis reveals that while CO2 dominates displacement at low pressures, mixtures that contain N2 achieve higher CH4 desorption efficiency above 13 MPa by mitigating diffusion resistance. (3) A higher N2 fraction improves CH4 diffusion coefficients, thereby facilitating gas mobility and ensuring superior recovery performance under high-pressure conditions. This study advances the fundamental knowledge of microscale gas behavior in tight sandstones and supports the feasibility of impure CO2 injection as a practical strategy for sustainable gas production.

Accurately predicting the phase equilibria of the CO2-oil-water system is vital for simulating CO2-enhanced oil recovery and CO2 geological storage in shale oil reservoirs. However, conventional three-phase flash models have limited accuracy in predicting the phase behavior of the CO2-oil-water system because they neglect the nonideality of water and the strong interactions between CO2 and H2O. These limitations become more pronounced in shale nanopores, where the compression of intermolecular distances significantly intensifies interactions between CO2 and H2O. In this study, we develop a novel three-phase equilibria calculation model based on the Peng-Robinson-Huron-Vidal equation of state (PR-HV EoS). To enhance predictive accuracy under the confinement effects, the model modifies the PR EoS to characterize the confinement effects through a fluid-wall interaction coefficient and an adsorption effect coefficient, and employs the non-random two-liquid (NRTL) activity coefficient model to characterize the interactions between CO2 and H2O. Additionally, we propose two new critical property shift correlations by fitting 44 sets of experimental data. The example calculations show that the model proposed in this study has higher accuracy than the conventional three-phase flash algorithm. Overall, this model can accurately calculate the phase behavior of the CO2-oil-water system in nanopores, providing an effective phase equilibria model for carbon capture, utilization, and storage (CCUS) simulation in shale oil reservoirs.

The development of natural gas reservoirs faces challenges such as low recovery rates and suboptimal economic performance. In water-drive gas reservoirs, water invasion is particularly prominent, leading to declining well productivity and low recovery efficiency. Conventional development techniques struggle to effectively control water invasion and enhance recovery. Therefore, exploring new enhanced recovery methods, particularly techniques targeting residual gas in water-flooded zones, is crucial for improving gas reservoir recovery and economic viability. This study focuses on the challenge of recovering residual gas in water-flooded zones during the late-stage development of water-drive gas reservoirs. It investigates the mechanisms of switching to nitrogen injection after water flooding and evaluates the effectiveness of nitrogen injection in enhancing gas recovery through core-scale numerical simulation. Experimental results show that at 30 MPa, the remaining gas saturation after water flooding was 28.1

The primary focus of this paper is to develop novel analytical solutions for estimating initial reserves, interblock connectivity and productivity indices (PIs) in heterogeneous gas reservoirs exhibiting gas/water boundary-dominated flow. To achieve this, a compartmented model is established to divide the heterogeneous reservoir into an ensemble of individual tank-type blocks, coupling each block to the neighboring blocks through crossflow of gas/water across the interfacial boundaries. The nonlinearity in the governing flow equations due to pressure-dependent fluid properties and saturation-dependent relative permeabilities is addressed by using pseudopressure-based definitions. Consequently, the block-based capacitance resistance model (CRMB) and the producer-based flowing material balance (FMBP) are proposed, where the flow fluctuations from interwell and interblock communications are handled by the assumption of time-stepwise variation and the definition of material-balance pseudotime, respectively. In addition to forward solutions, two inverse analysis approaches, including nonlinear regression and type-curve matching, are developed to reduce estimation uncertainty. The presented approaches are systematically compared and validated against numerical simulations of synthetic cases to assess the accuracy of CRMB and FMBP. Furthermore, a field case is analyzed to demonstrate the practical applicability of the proposed analysis approaches.

CO₂ injection into gas reservoirs for enhanced gas recovery (CO₂-EGR) offers the dual benefits of geological carbon storage and increased natural gas production. However, the complex mixing of CO₂ and CH₄ in porous media limits accurate prediction of mixing-zone evolution and displacement efficiency. In this study, a high-temperature and high-pressure full-diameter core-flooding platform was developed, and a dynamic core breakthrough method was applied to tight sandstone and fractured core samples from the Sulige gas field. Longitudinal dispersion coefficients were determined, and the effects of temperature, pressure, water saturation, injection flow rate, and heterogeneity scale on CO₂–CH₄ mixing were systematically evaluated. The results demonstrate that the overall mixing process arises from the coupled contributions of molecular diffusion and convective dispersion. Under typical reservoir conditions, convective dispersion contributes approximately 90% of the total dispersion, whereas the diffusion contribution increases to about 30% under gaseous CO₂ conditions. The high viscosity of liquid CO₂ markedly suppresses mechanical dispersion. The dispersion coefficient increases exponentially with temperature and linearly with injection flow rate, while elevated pressure strongly enhances mixing in fractured cores. Water saturation exhibits a dual effect through changes in pore topology. Once water saturation exceeds 60%, improved channel connectivity caused by reduced tortuosity dominates, leading to a sharp increase in dispersion. Tortuosity was identified as the key geometric parameter governing dispersion-path complexity. Based on this finding, a three-region evaluation framework and a generalized dimensionless dispersion model coupling rock topology with fluid dynamic properties were established. The proposed model enables cross-scale prediction of dispersion intensity and provides a basis for estimating mixing-zone development and CO₂ breakthrough time in field-scale CO₂-EGR operations.

Starting from the first principle thinking,this study systematically reviews the development mechanisms of gas reservoirs and proposes the development concept of"full lifecycle enhanced gas recovery(EGR)".Following the principles of scientificity,practicality and comparability,a generational classification system for EGR technologies is established.The research indicates that the properties of natural gas dictate a development mechanism primarily driven by pressure depletion to release the elastic expansion energy of gas.This leads to a development model centered on primary depletion,supplemented by limited adjustments in late stages.Early development essentially lies in well pattern optimization and risk pre-control,while late development focuses on targeted local adjustments and integrated collaborative control.Primary gas recovery,relying on natural energy depletion,achieves a recovery factor of 25％-55％.Secondary gas recovery,through active regulation of the reservoir pressure field via techniques like blockage removal,and injection-production optimization,can enhance the recovery factor by 10-15 percentage points.Tertiary gas recovery,employing multiple mechanisms to alter the reservoir's physical and chemical fields synergistically,offers a potential further increase of 5-10 percentage points.Currently,primary recovery technologies are mature and well-established.Synergistic optimization of well patterns and fracture networks enables effective production from gas-drive reservoirs,while optimized development strategies facilitate orderly production from water-drive gas reservoirs.Secondary recovery technologies,in the field pilot stage currently,adopt active measures like enhanced water drainage,water shutoff,and gas injection to effectively control water influx and release trapped gas.Tertiary recovery remains largely in the laboratory or pilot test stage.Future efforts should focus on cross-generational technologies,such as"primary+secondary"and"primary+tertiary"combinations,to continuously improve recovery factors throughout the full lifecycle of gas reservoirs.

Accurate production forecasting serves as a critical determinant for optimizing extraction strategies, guiding long-term field management in reservoir development. Both conventional methods and deep learning techniques face significant challenges in production forecasting due to the increasing complexities of reservoir extraction. Firstly, traditional production forecasting methods often fail to fully capture the complex reservoir behavior. Finally, these approaches demonstrate suboptimal performance in wells with limited data. These problems can lead to a decrease in prediction accuracy. To address these challenges, this paper introduces the Patching-iTransformer method and applies meta learning. The method improves prediction accuracy and overcomes the problem of few samples in production forecasting. Specifically, we implement a patching mechanism that segments the input time series, thereby converting the univariate time series into a two-dimensional representation. This architectural enhancement significantly strengthens the model's capability to capture latent interdependencies among variables. Currently, we develop a PiAM meta-learning algorithm with domain-specific adaptation for oil field applications by quantitatively assessing individual well contributions to reservoir exploitation. We use time series data from real wells to evaluate the accuracy of multiple wells under the PiAM model. The experimental results demonstrate that Patching-iTransformer achieved better performance improvements than the iTransformer method. R2 increased by 0.297, RMSE decreased by 11.64% and MAE decreased by 3.49%. PiAM meta-learning method demonstrated superior performance over the Patching-iTransformer model, showing a 0.535-point improvement in the R2 coefficient along with a reduction of 27.54% in RMSE and a decrease of 28.22% in MAE. (c) 2026 Publishing services by Elsevier B.V. on behalf of KeAi Communications Co. Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

The accelerating accumulation of anthropogenic CO2 emissions is a major driver of global climate change, emphasizing the urgent need for effective mitigation strategies. CO2-enhanced gas recovery (CO2-EGR) offers dual benefits of improving hydrocarbon recovery while enabling long-term geological storage. However, uncertainties in pore-scale transport mechanisms and rock-fluid interactions hinder reliable implementation. In this study, in-situ CT imaging combined with digital rock reconstruction was employed to investigate CO2-CH4 displacement in carbonate rocks. CT observations showed that prolonged exposure to CO2-saturated brine induced dissolution-dominated reactions with localized precipitation, which increased porosity but also elevated tortuosity and heterogeneity, as reflected in broadened pore size distributions and higher fractal dimensions. Numerical simulations based on a multicomponent convection-diffusion model revealed that mass transfer is governed by a diffusion-dominated mixing front and a convection-dominated trailing region. The mixing front, accounting for similar to 5.5 %-22.8 % of the total mixing zone, provides a clear boundary between diffusion- and convection-driven regimes, with pore connectivity exerting strong control on their persistence. Parametric analyses demonstrated that injection velocity is the primary factor reducing breakthrough and completion times, while pore pressure has a secondary influence via its impact on fluid properties. Temperature and injection composition exert relatively smaller effects at the pore scale, though their significance may increase under reservoir conditions. Rock-fluid reactions further prolonged displacement and enhanced the relative contribution of diffusion. These integrated insights advance the mechanistic understanding of coupled transport processes and offer practical guidance for predictive modeling and optimization of CO2-EGR strategies.

During the development of multi-layer tight sandstone gas reservoirs in Ordos Basin, China, it has not been easy to calculate accurately the production of each individual layer in gas wells. However, production allocation provides a vital basis for evaluating dynamic reserves and drainage areas of gas wells and remaining gas distributions of gas layers. To improve the accuracy and reliability of production allocation of gas wells, a new model was constructed based on the seepage equation, material balance equation, and pipe string pressure equation. In particular, this new model introduced the seepage equation with an elliptical boundary to accurately capture the fluid flow characteristics within a lenticular tight gas reservoir. The new model can accurately calculate the production and reservoir pressure of each individual layer in gas wells. In addition, the new model was validated and applied in the Sulige gas field, Ordos Basin. The following conclusions were drawn: First, The gas production contribution rates of pay zones based on the new model are fairly close to the measurements of the production profile logging, with errors less than 10%. Second, The overall drainage area of a gas well lies among those of each pay zone, and the total dynamic reserves of the well are close to the sum of the dynamic reserves of pay zones. Third, Higher permeability may lead to higher initial gas production of the pay zone, but the ultimate gas production contributions of pay zones are affected jointly by permeability and dynamic reserves. Finally, The new model has been successfully applied to the SZ block of the Sulige gas field, in which the fine evaluation of dynamic reserves, drainage areas, gas production, recovery factors, and remaining gas distributions of different layers was delivered, and the application results provide technical support for the future well placement and enhanced gas recovery of the block.

There is little research on the strain characteristics of ultradeep dolomite rocks with deep burial, high stress, and developed fractures and macropores. This article comparatively studies the influence of stress and fracturing on the pore-throat structure of different types of ultradeep dolomite rocks by setting up stress experiments and fracturing experiments and using computed tomography (CT) scanning. In the stress experiment, the fluid-solid coupling effect on pore space was considered. The study shows that unlike shallow to medium carbonate reservoirs, when the stress is high but lower than the rock yield stress and in a flowing state, the studied ultradeep dolomite rock has a higher permeability after recovery from compression. When the stress is high, brittle deformation occurs inside the rock sample, and dolomite particles in the macropores are crushed and detached, blocking the middle pores with the gas flow; the two adjacent mesopores are merged into one macropore. In this way, the average pore size of the rock sample increases and the proportion of macropore volume increases. The increase in macropores plays a leading role in improving the reservoir's permeability. The edges of dolomite crystals are compressed open under high stress, resulting in more developed intercrystalline fractures, larger fracture widths, and strengthened communication between fractures. Besides the effect of stress, the fractures generated by fracturing greatly expand and connect the existing fractures, and then improve the connectivity of the reservoir, which is the main factor leading to the improvement of reservoir permeability after fracturing. In addition, rock mechanical parameters were obtained through triaxial mechanics experiments. Dolomite rock samples that have undergone multiple cycles of stress cycling have more developed microfractures and macropores and weakened axial compression resistance and are more prone to be fractured. This study can provide a theoretical basis and guidance for the efficient development of ultradeep dolomite.

Carbonate gas reservoirs are predominantly developed using acid fracturing techniques. The fluid flow area is mainly composed of a non-acid-fractured region, an acid-fractured region, and a near-wellbore region. At present, numerous studies have been conducted on productivity models considering various factors such as formation thickness, permeability, and skin factor. This study aims to overcome the inability of conventional models to holistically address permeability anisotropy, composite reservoir properties, acid fracturing fractures, gas-water two-phase flow, and stress sensitivity in carbonate gas reservoirs. Multiple conformal transformations were employed to establish an improved composite model for gas-water two-phase productivity calculation and analysis of horizontal wells in carbonate gas reservoirs. The traditional binomial deliverability equation and numerical simulation were utilized to verify the accuracy of the new productivity equation. Extensive factor analysis was conducted for gas well productivity, including horizontal wellbore length, effective formation thickness, fracture parameters, formation permeability, angle between the main permeability direction and the horizontal well, permeability anisotropy, fracture and matrix stress sensitivity, gas and water relative permeability, and water and gas production rate ratio. Furthermore, to the best of our knowledge, this paper is the first to propose the use of an X-shaped plate to describe the relationships between the well absolute open flow rate and the anisotropy of the formation permeability and the angle between the main permeability direction and the horizontal wellbore. The results of this study provide the theoretical guidance for gas well productivity evaluation, productivity classification, and rational production allocations for horizontal wells in carbonate gas reservoirs.

CCUS-EGR is becoming the most cost-effective method for energy saving and emission reduction globally. However, the acidic nature of CO2 can alter the reservoir permeability over time, affecting the stability and sustainability of gas injection. To explore this, long-term CO2 injection simulations were conducted on tight sandstone, carbonate rock, and volcanic rock. By applying gas and water permeability calculation methods, the reservoir permeability was monitored in real time throughout the long-term displacement process. Combined with NMR measurements, cross-lithology comparison analysis was conducted to investigate the evolution characteristics of pore structure and flow capacity, as well as the key influencing factors in various lithological reservoir samples. The results show that the influence of long-term CO2 replacement and shut-in well reinjection replacement on the seepage capacity of the rock samples mainly comes from the combined effects of clay expansion, mineral particle settling and plugging, and reaction dredging to increase infiltration, and the process of water-rock reaction involves mineral particles reacting, dissolving, dislodging, transporting, and plugging. The long-term replacement and shut-in well reinjection experimental process of dense sandstone seepage capacity slightly reduced or basically unchanged, due to the greater clay effect; with carbonate rock calcite and dolomite as the main components, CO2-water-rock reaction is dominated by mineral dissolution and clearing channels to increase seepage, with a significant increase in seepage capacity; feldspar is abundant in volcanic rocks, leading to mineral precipitation and pore-blocking during CO2-water-rock reaction. This results in mesopore enlargement and blockage of both micro- and macropores. The channel clearing and permeability enhancement are weaker compared with carbonate reservoirs, with only a slight increase in overall seepage capacity. These findings provide valuable guidance for the efficient implementation of the CCUS-EGR.

At present, it is necessary to change the development method to improve the development of gas reservoirs entering the middle to late stage of development. CCUS-EGR can drive out residual gas while burying carbon, which is an economically feasible method. For most reservoirs, water intrusion is severe or there is a large amount of pore water, so it is necessary to study the impact of CO2-water–rock reaction on reservoir physical properties and pore structure. Rock sample displacement experiments were implemented, CO2 were injected into carbonate rock samples and tight sandstones saturated with bound water for displacement, and the changes in the permeability of the rock samples during the displacement process were monitored. The pore structure of rock samples before and after CO2 water rock reaction was measured using nuclear magnetic resonance equipment, and the different water rock reaction occurring in different rock types were studied combined with XRD testing. This study provides a reference for on-site CO2 injection testing. For carbonate rock samples mainly com-posed of dolomite and calcite, supercritical CO2 has a negative effect on the permeability of the rock samples. Under the temperature and pressure conditions of the reservoir, CO2 dissolves in water to form carbonic acid. Carbonic acid reacts with CaCO3 to form HCO3−/CO32− ions, which dissolve in the solution. When the HCO3− ions in the solution become supersaturated, calcium carbonate precipitates. The micropores in carbonate rocks are expanded by acid dissolution, while the macropores are filled by precipitated calcium carbonate. After injecting CO2 for 1175 min, the porosity of the rock sample decreased by 10.44

Summary Based on the nonlinear relationship between the cumulative gas production and the total pressure difference, a segmental material balance equation was applied, and an improved flow material balance (FMB) equation was proposed to calculate the dynamic reserves of shale gas reservoirs with a variable gas drainage radius. In the early stage, the shale gas well drainage radius gradually increased. The spread range of the formation pressure increased, but fractures gradually closed because of the enhancement of the effective stress. This resulted in stress sensitivity. In the middle to late stages, the gas drainage radius can be regarded as unchanged. The rate of increase in the pressure spreading range decreased, and the rate of decrease in the fracture closure decreased. The stress sensitivity can be ignored. To explain these phenomena, a segmental material balance equation was established. Furthermore, an improved FMB equation was obtained based on the productivity equation using the potential superposition theorem, and the drainage radius of horizontal wells was regarded as a variable for the last dynamic reserve calculation. Finally, the dynamic reserves of four shale gas wells were calculated. The comparison indicated that the proposed improved equation predictions agreed more closely with actual development experience than the conventional models based on the dynamic recovery rate calculation and the correlation coefficient obtained by data fitting. The proposed method improves the dynamic reserve calculations and contributes to well productivity evaluation.

In view of the characteristics of low permeability, strong heterogeneity, small effective sand scale and poor connectivity of tight gas reservoir, well pattern infilling optimization is the main method to improve gas recovery. Taking Sulige gas field as an example, based on the study of sedimentary facies, effective sand-body distribution, combined with production performance and engineering parameters, the well control drainage was obtained. Meanwhile, the interwell connectivity and interference probability was qualitatively analyzed by interference well test, and the critical well spacing density was determined. In order to quantitatively reveal the gas grabbing degree at different well spacing density and determine whether the final cumulative gas production of gas wells with interference is economical and feasible, the evaluation index of "gas well production interference ratio (GWPIR)" was defined specifically, and the intersection plot of GWPIR relating to gas reserve abundance and well spacing density was draw. Different from the traditional method of well number interference probability, through the combination of GWPIR plot and economic evaluation, well pattern optimization from both two perspectives of pursuing higher economic benefits and recovery degree can be realized, providing technical support for improving gas recovery and long-term stable production of gas field.

Summary In unconventional shale and tight reservoirs, the concept of stimulated reservoir volume (SRV) is used to correlate the volume of total injected proppant with well performance. The SRV configuration consists of primary fractures connected to the wellbore and secondary fractures intersecting primary fractures. SRV productivity is determined by fracture conductivity, fracture dimensions, and network complexity, which also vary with time. This work presents an extension of the unified-fracture-design (UFD) approach to account for not only the pseudosteady state (PSS) but also transient flow regimes and ultimately optimize SRV for maximizing well performance. A generalized productivity index (PI) for both the transient and PSS regimes is presented to improve well performance by searching for the maximum PI over time. In addition, a surrogate model is developed to accelerate the optimization. This study demonstrates that the UFD enables the determination of the optimal fracture network conductivity and complexity that contribute to the maximum PI with a given proppant volume. The optimal SRV design is time-dependent until the PSS is reached. The surrogate model not only improves the computational efficiency but also delivers high precision, which means far less computational burden than the traditional parametric-sensitivity analysis.

From traditional fossil fuels to future clean energy, natural gas is a very important bridge. Ultra-deep carbonate water-drive gas reservoir is an important part of natural gas development. Using a numerical simulator, a single well mechanism model considering the physical properties and seepage characteristics of the target ultra-deep carbonate water drive gas reservoir is established to study the influence of water body parameters, seepage parameters and fracturing measures on its development. The reservoir characteristic parameters and flow characteristic parameters suitable for the target ultra-deep gas reservoir are extracted from the existing experiments. Based on the experimental data, a single well mechanism model of the target ultra-deep carbonate gas reservoir is established using a numerical simulator. Use single well production data for production history matching, and find out the main control factors affecting the development of water drive gas reservoir in the process of history matching. Based on the historical production data of a single well, the production index of a single well in the next 10 years is predicted. The results show that for the target ultra-deep carbonate rock water flooding gas reservoir, reducing the volume of water body, reducing the velocity of water invasion, and lowering the gas-water interface are beneficial to reduce the water production rate, weaken the gas-water Jamin effect caused by water invasion, and reduce the depletion of near-wellbore pressure. After considering the high-velocity non-Darcy effect of gas, the gas production rate of gas wells decreases, while the water production rate increases, and a lot of pressure is consumed at the near-wellbore area of the gas reservoir. In this condition, it is a must to greatly reduce the bottomhole pressure of gas wells to achieve the target gas production rate. The gas-water two-phase permeability curve considering high pressure and high temperature increases both gas and water production rate, and making the historical matching results closer to the actual single well production data. Fracturing improves the gas-phase seepage capacity and alleviate the gas-water Jiamin effect in the near-wellbore area. Under the same gas production rate, the bottomhole pressure will increase significantly. But fracturing has little effect on the water production rate of the target gas well. The novelty of this study is that the mechanism model fully simulates the reserve characteristics and seepage characteristics of the target ultra-deep carbonate rock water drive gas reservoir by using experimental data considering reservoir temperature and pressure conditions. Studying the main controlling factors of the ultra-deep carbonate water-drive gas reservoir are conducive to formulating development technology policies for the efficient development of similar gas reservoirs.

Traditional methods for forecasting production rate, such as Arps, analytical techniques, and recurrent neural network (RNN)-based deep learning, are mainly point prediction techniques developed within the framework of single- well forecasting. These methods often face limitations stemming from single- well historical production data and model assumptions, hindering their ability to consider the influence of development patterns of other production wells within the block on the target well. In addition, they struggle to predict the multiple production rate time series simultaneously and often fail to quantify uncertainty in predictions or adequately exploit extensive relevant historical production data. To tackle these challenges, we propose a model based on the deep autoregressive recurrent neural network (DeepAR), leveraging related multiwell production rate data to enable global modeling and probabilistic forecasting. This model incorporates dynamic covariate and static categorical variable data, integrating Bayesian inference and using gradient descent algorithms and maximum likelihood estimation methods to derive a comprehensive historical- future production probability evolution pattern shared across multiple wells. Leveraging data from 943 tight gas wells, a comprehensive evaluation of the DeepAR model's performance was undertaken, encompassing the comparison of prediction accuracy with long short- term memory (LSTM), classification prediction, coldstart prediction, and single- well multitarget prediction scenarios, summarizing the applicability conditions for each. The research findings highlight that DeepAR integrates the acquired comprehensive production probability evolution pattern with specific production historical data of the target well to formulate a "comprehensive + specific" production probability prediction approach, resulting in improved stability and accuracy. On average, DeepAR demonstrates a 58.79% reduction in normalized deviation (ND) compared to the LSTM model, showcasing enhanced stability, particularly in scenarios involving frequent well shut- ins and openings. Moreover, DeepAR can learn static categorical features, with the classification model resulting in a 27.15% reduction in the ND compared to the unclassified model. Furthermore, DeepAR adeptly addresses the challenge of limited data availability, achieving cold- start prediction and facilitating multitarget single- well training and prediction while considering the interdependency among multiple variables over time and effectively mitigating the issue of missing auxiliary variables during the prediction phase. This study contributes to a broader understanding of production forecasting and analysis of production dynamics methods from a big data perspective.