An integrated physics–machine learning framework for time-dependent injectivity decline and fracture growth in horizontal water-injection wells

A system for compressed flue gas energy storage in aquifers is proposed, combining compressed gas energy storage with carbon dioxide (CO2) sequestration. Transient simulations of the cyclic injection-production process in aquifers with anticlinal symmetric structures were conducted, incorporating the flow, phase, saturation, and chemical reaction equations. The dynamic characteristics of bottom-hole pressure, bottom-hole temperature, and the gas-phase saturation field under designed operating conditions were analysed. The total carbon dioxide storage capacity throughout the dynamic process was calculated. Energy and exergy efficiencies of the compressed flue gas in a single injection-production cycle were also evaluated. The results show that over time, bottom-hole pressure, temperature, and gas-phase saturation fields gradually stabilise into periodic patterns during the cyclic injection-production process. Carbon dioxide storage capacity increases continuously, while energy efficiency declines before stabilising, and exergy efficiency rises before stabilising. The effects of gas storage construction duration and injection-production rates on the dynamic characteristics of the cyclic process were also examined. Longer construction durations result in smaller bottom-hole pressure and temperature differences during injection-production, higher carbon dioxide sequestration, and improved exergy efficiency in the Huff-n-Puff operation. Conversely, higher injection-production rates lead to greater bottom-hole pressure and temperature differences and increased carbon dioxide sequestration, but lower exergy efficiency in the Huff-n-Puff operation.

Purpose: This study presents an integrated data-driven framework for optimising gas injection strategies in enhanced oil recovery processes. Design/Methodology/Approach: The proposed framework uses Long Short-Term Memory (LSTM) to model nonlinear temporal dependencies. The key operational and reservoir variables, such as gas injection rate (GIR), bottom-hole pressure, separator pressure, reservoir temperature, tubing inner diameter, gas-oil ratio (GOR), and gas composition, are considered. In the proposed method, after preprocessing, feature selection is done using the Sequential Forward Selection (SFS) method. Then, the Whale Optimisation Algorithm (WOA) was employed to optimise injection strategies and tune LSTM hyperparameters. The objective is to maximise the net present value (NPV) subject to operational constraints. Research Limitation: Uncertainties and changing reservoir conditions may limit the generalisability of the proposed framework without further real-time adaptation. Findings: The results demonstrate that the hybrid WOA–LSTM framework outperforms LSTM and GRU models in both prediction accuracy and economic evaluation. In the multivariate scenario, the model's RMSE is 2.22, MAE is 1.09, accuracy is 97.20%, and NPV is $27.42 million. The results confirm the effectiveness of integrating metaheuristic optimisation and deep learning to enhance production forecasting and decision-making. Practical Implication: It can enable oil field operators to improve the production efficiency and maximise economic returns while respecting the operational constraints. Social Implication: Optimising gas injection resources reduces waste and enhances energy efficiency. In air pollution, the proposed model reduces hydrocarbon production and improves air quality. Originality/Value: This study introduces a novel hybrid framework that combines LSTM-based forecasting with WOA for prediction. The model offers a powerful solution for complex reservoir management problems.

Managed pressure drilling (MPD) is the core technology for developing formations with high pressure and narrow density windows. It precisely maintains the bottomhole pressure (BHP) within the safe operating window defined by formation pore pressure and fracture pressure by actively regulating the wellbore pressure profile. If pressure control becomes unstable, it can easily trigger gas kicks or lost circulation, posing a severe threat to operational safety. However, existing model predictive control (MPC) schemes have significant limitations: pure data-driven models exhibit poor generalization under complex conditions, while control algorithms based on traditional mechanistic models struggle to meet the stringent real-time requirements of field control cycles due to high-complexity numerical iteration processes. To balance control precision and real-time performance, this paper proposes a physics-informed model predictive control framework (PINC-MPC). During the training phase, physical prior knowledge such as the law of mass conservation is embedded into the neural network as constraints to construct a physically consistent deep surrogate model, enabling it to characterize complex wellbore characteristics. In the control phase, this surrogate model replaces the time-consuming numerical solving process of the mechanistic model within the MPC loop, achieving near-real-time state prediction and rolling optimization while ensuring physical fidelity. Experimental results indicate that PINC-MPC demonstrates superior control performance. Its median single-step solving time is only 16.81 ms, achieving an 11.1-fold acceleration compared to the mechanistic model-based scheme (187.3 ms). In a 5000 s full-cycle closed-loop control experiment, the total time required for the former is only 1.68 s, while the latter reaches 18.73 s, representing an efficiency improvement of approximately 91%. In terms of control accuracy, the integrated absolute error (IAE), reflecting the total deviation of the control process, significantly decreased from 63.40 MPa·s for the industrial successive linearization MPC (SLMPC) to 12.90 MPa·s, an improvement of 79.7%. Especially in extreme dynamic conditions such as simulated pump shutdowns for pipe connections and sudden gas kicks, the framework demonstrates excellent predictive ability and response efficiency. It can proactively trigger compensation actions to keep BHP fluctuations within 0.30 MPa, significantly outperforming the traditional SLMPC method. The research results prove that PINC-MPC provides an efficient, precise, and robust nonlinear control strategy for MPD systems, offering important engineering reference value for enhancing the automation level of intelligent drilling systems.

Abstract During production adjustment in ultra-deep gas wells, changes in throttle opening cause conversion between fluid kinetic and pressure potential energy, leading to significant transient pressure fluctuations. Focusing on this process, a multi-field coupled transient flow model was developed based on transient flow theory, incorporating dynamic gas compressibility variations and real-time fluid friction response. A sensitivity analysis conducted using Well X as an engineering case revealed that bottom-hole pressure, adjustment amplitude, and adjustment duration significantly affect wellhead pressure fluctuations. Higher bottom-hole pressure and larger adjustment amplitudes increase peak fluctuation values, while shorter adjustment durations intensify transient effects despite accelerating production state transition, necessitating operational optimization. This study provides technical support for safe and efficient production adjustment in ultra-deep gas wells.

ABSTRACT At present, the oil pipe perforation gas lift drainage technology has been widely applied and reported, but there are few reports on its application in condensate gas reservoirs developed by deep circulation gas injection. In order to broaden the application scope of oil pipe perforation gas lift drainage technology, further tap the technical potential, and propose a simple, effective, and low-cost drainage oil and gas production process technology for condensate gas reservoirs developed by deep circulation gas injection, this paper establishes gas lift start-up pressure calculation model, start-up gas injection volume calculation model, and optimal hole diameter optimization model based on the principles of mass conservation and energy conservation, and forms a calculation method for oil pipe perforation gas lift drainage characteristics. Due to the process design concept of using perforated holes instead of gas lift valves in this article, the perforation depth and hole diameter are permanently fixed and cannot be adjusted in the later stage. When designing parameters, it is necessary to focus on controlling the perforation depth and hole diameter, taking into account both gas injection efficiency and column safety. Therefore, this article establishes a parameter calculation model for oil pipe perforation gas lift holes by studying the calculation method of local resistance loss caused by hole contraction and expansion in the actual flow field of the oil pipe annulus, combined with the pressure gradient equations of single-phase flow in the annulus and multiphase flow in the oil pipe. Based on the node analysis method and the above model, a process parameter system suitable for condensate gas wells was formed, and the technical feasibility was verified through field experiments. Key technical issues such as simulation of gas-liquid three-phase flow in condensate gas wells, calculation of borehole sudden shrinkage resistance, and evaluation of tubing strength after perforation have been addressed. Based on geological parameters and production dynamics, a parameter design criterion is proposed with core parameters including tubing drilling depth, tubing drilling aperture, wellhead gas injection volume, bottom hole pressure, injection pressure, start-up pressure, daily gas production at the wellhead, and daily liquid production at the wellhead. Through on-site experiments, it was found that the production of the test well using oil pipe perforation gas lift method successfully solved the problem of abnormal production due to well bore fluid accumulation. The daily oil production level increased by 1.86 times and the daily gas production level increased by 3.95 times. The daily profit can reach 37,000 yuan, and the annual revenue can be about 12.21 million yuan. The use of oil pipe perforation gas lift can effectively reduce the operating cost of gas lift measures, which can be significantly reduced by 70% compared to the operating cost of gas lift column. The process adaptability has been effectively verified. This research achievement breaks through the limitations of traditional gas lift technology in high condensate oil and complex phase wells, providing important technical support for efficient development of deep condensate gas reservoirs, and has significant economic benefits and engineering promotion value.

Natural gas, as a clean and efficient energy source, occupies an increasingly important position under the background of the "dual-carbon" strategy and energy structure transformation. However, affected by various complex factors such as reservoir heterogeneity, gas-liquid-solid multiphase flow, inter-well interference, and lack of bottom-hole pressure, traditional gas well productivity prediction and production allocation methods have problems such as insufficient accuracy, poor adaptability, and high dependence on manual work, making it difficult to meet the needs of efficient development of offshore and unconventional gas reservoirs. Therefore, to meet the requirements of accuracy, adaptability, and intelligence, this paper studies the intelligence of gas well productivity prediction and its production allocation model, analyzes and compares current machine learning algorithms such as XGBoost, Random Forest, Support Vector Machine (SVM), and Long Short-Term Memory Neural Network (LSTM) in gas well dynamic analysis, points out the existing problems in current research, and looks forward to the development trends of gas well productivity prediction and production allocation in the directions of multi-factor coupling, intelligent integration, and full-life-cycle dynamic optimization.

The depletion of reservoir energy in mature oil fields poses a significant challenge to sustainable hydrocarbon production. In the prolific Niger Delta basin, many reservoirs have progressed beyond their natural flow potential, necessitating the deployment of artificial lift systems to maintain economic production rates and maximize recovery. This study presents a performance evaluation of natural flow and Electrical Submersible Pump (ESP)-assisted wells in a depleting Niger Delta oil reservoir using an integrated Material Balance (MBAL) and PROSPER modeling approach. The workflow commenced with a rigorous reservoir-level analysis using MBAL to characterize the reservoir's drive mechanism and estimate the Original Oil in Place (OOIP). The analysis confirmed a strong water drive mechanism with an OOIP of approximately 2989.9 MMSTB. Subsequent well-level modeling in PROSPER was used to simulate the well's performance under natural flow conditions, identifying the point of flow cessation at a reservoir pressure of approximately 2000 psig and a water cut of 60%. To address this production gap, an ESP system was designed and simulated. The results demonstrated that the ESP intervention successfully lowered the flowing bottomhole pressure, increased drawdown, and restored production to a stable rate of 1,204 STB/day of oil (3,010.5 STB/day total liquid). A comparative forecast revealed that without the ESP, the well would have ceased production, leaving significant recoverable reserves untapped. The study conclusively demonstrates that the integrated MBAL-PROSPER workflow provides a robust technical framework for diagnosing production constraints, optimizing artificial lift timing, and extending field life.

Carbonate reservoirs, characterized by extensive fractures and cavities, are prone to gravity displacement during drilling when the bottom-hole pressure approaches equilibrium. This phenomenon, driven by density differences between drilling and formation fluids, can result in simultaneous overflow and leakage, posing significant well control risks such as kicks or blowouts. The occurrence of gravity displacement downhole makes its timely detection through conventional annular flow monitoring techniques challenging. This study investigates the triggering conditions and safe density window for gravity displacement in fractured and cavernous formations. Through theoretical analysis and experimental simulation, we examined the displacement mechanisms in both fractured and cavernous conditions. Computational fluid dynamics (CFDs) simulations were used to validate critical fluid column heights for fractured formations and the proposed safe density window. Based on these findings, practical methods to mitigate the hazards associated with gravity displacement overflow are proposed. The results offer valuable guidance for the field identification and mitigation of such incidents, contributing to managed pressure drilling and enhancing drilling safety in complex carbonate reservoirs.

During multistage fracturing, cyclic loading can easilylead tocement sheath failure and annular pressure leakage. To clarify theevolution of stress states and their impact on annular sealing integrity,this study applies shakedown theory. The cement sheath is treatedas an ideal elastoplastic material, and the Mohr–Coulomb criterionis used to derive an analytical solution for the shakedown limit load.The effects of in situ stress, wellbore geometry, material parameters,and fracturing operations on the cement sheath are systematicallyanalyzed. Results show that the shakedown limit load of the cementsheath is influenced by in situ stress, wellbore geometry, and cementproperties. The limit load increases with higher in situ stress butdecreases as the diameter ratio increases. Both the cohesion and internalfriction angle of the cement significantly enhance its load-bearingcapacity. In Well L119, the bottom-hole pressure during fracturingexceeded the elastic limit. When the inner wall pressure reached 123.10MPa, the cement sheath yielded plastically. The plastic zone nonlinearlypropagated from the inner wall to a radius of 242.5 mm. Therefore,cement sheaths in deeper sections exhibit better structural stability,while the sealing integrity of shallow sections should be carefullymonitored during multistage fracturing. Optimizing the well designwith a diameter ratio of 0.64 and using cement with a higher cohesionand internal friction angle can effectively improve the resistanceof the cement to damage and its load-bearing limit. This study offersa quantitative basis for optimizing fracturing parameters and providespractical guidance for wellbore integrity management.

Investigating the impact of relative permeability characteristics and pressure management on oil recovery during low-salinity water injection

Heterogeneous stacking strategy for modeling flowing bottom-hole pressure of oil wells

Unconventional shale gas is vital for clean energy and energy security. However, its ultralow permeability leads to low primary recovery. This study investigates CO2-enhanced shale gas recovery (ESGR) in the Yanchang shale reservoir by using a CMG-GEM compositional simulator with a dual porosity/permeability model, evaluating well placements and both continuous and huff-and-puff injection methods. Findings show that continuous CO2 injection increases CH4 recovery by 7.59% over no injection. For huff-and-puff, a shorter 1 year injection period yielded the highest CH4 recovery; extending injection to 2 and 3 years caused declines of 1.93 and 3.89%, respectively. Conversely, using five injection cycles resulted in the highest cumulative CH4 recovery. Starting CO2 injection later in the production lifecycle also optimized recovery, with a 1.15% increase observed after 10 years of initial CH4 production, as it utilizes more favorable reservoir pressure conditions. Moreover, for CO2 storage, the reservoir exhibited 99.565% efficiency during continuous injection. In huff-and-puff, longer injection durations improved storage, with a 3 year period achieving 98.35% efficiency. Similarly, five injection cycles yielded the highest storage efficiency, at 99.12%. Delaying the injection start time also significantly enhanced CO2 storage, with efficiency improving from 96.38% after 1 year to 98.35% after 10 years, leveraging improved pressure dynamics over time. In addition, a sensitivity analysis confirmed that key reservoir parameters, such as matrix porosity, permeability, and pressure, significantly influence both gas recovery and storage capacity. Critical hydraulic fracture parameters, including half-length, spacing, conductivity, and bottom-hole pressure, are essential for optimizing gas flow and CO2 injection efficiency. This study applies to tight shale gas formations worldwide, offering insights into optimizing hydraulic design and injection strategies to enhance shale gas production and CO2 sequestration, supporting global carbon management and climate change mitigation.

This study investigates the dynamic evolution and efficacyof oxygen-reducingair huff-and-puff (HnP) for enhanced oil recovery in tight reservoirs,employing a novel large-scale two-dimensional plate model (30 ×30 × 3.5 cm) and a single-dimensional elongated core to simulateoxygen-depletion processes across 13 HnP cycles. Through systematicexperimental trials, we examined key operational factorswellshut-in duration and production pressure differentialsto quantifyrecovery efficiency. Results reveal that oil displacement is primarilydriven by pressure-difference-induced fluid dynamics (63.04% of totalrecovery), with diffusion-mass transfer as a secondary mechanism (36.96%),highlighting minimal light hydrocarbon extraction and limited crudeoil fractionation capacity. Crucially, HnP performance is governedby shut-in duration and pressure differentials, with a single cyclecategorized into three distinct stages: single-phase gas flowback,free gas drive, and declining productivity. The free gas drive stagedominates oil production, evidenced by the gas-to-oil ratio, oil rate,and bottom-hole pressure variations, underscoring its critical rolein optimizing recovery strategies for tight reservoirs.

Frequent drilling fluid lost circulation in the Kuqa foreland area of the Tarim Oilfield severely constrains drilling efficiency and safety. The complex formation structures and diverse lost circulation types in this region are compounded by a lack of systematic classification in existing studies and weak correlation between mechanism analysis and field plugging measures, leading to a deficiency in quantitative decision-making for lost circulation prevention and control. Based on lithology analysis, loss zone pressure differential calculation, well log interpretation, and core observations, this study establishes an integrated “formation–lithology–pressure” diagnostic and classification method for lost circulation. A systematic classification framework comprising five types of lost circulation channels and mechanisms was developed. Based on this, the dominant lost circulation types and characteristics of three typical vertical formations in the Kuqa foreland were clarified: ① The supra-salt sandy conglomerate formations (e.g., Q1x, N2k) are dominated by permeability loss, where the loss rate (V) and bottomhole pressure differential (ΔP) exhibit a strong positive correlation (V ∝ ΔP). On-site application of graded bridging plugging formulations achieved a first-attempt success rate of ≥90%. ② The salt–gypsum formations (E1-2km) are primarily characterized by induced fracture loss, with a weak correlation between V and ΔP and dynamic fracture opening/closing behavior. Conventional rigid plugging materials showed limited effectiveness, resulting in a first-attempt success rate of <50%. ③ The K1bs formation is dominated by vertically developed natural fracture loss, where V and ΔP also demonstrate a strong positive correlation. In a specific Keshen block, a power-law relationship between the fracture aperture (W) and loss rate was established (W = 0.26·V0.62, R2 = 0.98), providing a basis for predicting fracture aperture and optimizing plugging formulations, with a plugging success rate of ≥80%. The classification system and quantitative criteria developed in this study effectively link lost circulation mechanisms, dynamic characteristics, and engineering countermeasures, offering theoretical support and a decision-making framework for optimizing lost circulation prevention and control measures and improving success rates in the Kuqa foreland area.

ABSTRACT The mechanisms of CO2 Miscible Front Migration is a key factor for enhanced oil recovery in CO2 flooding, it is difficult to be described due to complex interactions between fluids, therefore, the CO2 Miscible Front Migration characteristics are unclear. In this paper, we established an injection-production pair model for coupled Gas-Oil Two-Phase Flow in Miscible Flooding based on the theory of Fluid Interactions in Miscible Flooding and gas-oil two-phase fluid flow in porous media, considering the heterogeneity, reservoir dip angle and miscibility degree. The results show that with the increase in gas viscosity, dip angle and decrease of intralayer heterogeneity, the gas breakthrough time is later and the distance of miscible front migration is longer; With the increase of interlayer heterogeneity and reservoir pressure, the gas breakthrough time is earlier and the distance of miscible front migration is shorter. At the same time, due to the negligence effect between CO2 and oil, and well pattern in the analytical model, an ideal model based on reservoir parameters from Block G66 × 1 fault-block reservoir in Jidong Oilfield is established. The model identified the CO2-crude oil miscible zone by using oil saturation and CO2 molar fraction in oil and characterized the migration patterns of the CO2 miscible front through bottom-hole pressure, gas-oil ratio, and dynamic front-position monitoring. Simulation results show when areal heterogeneity increases from 1 to 9, CO2 breakthrough time is earlier, the contribution of the pure oil flow zone to recovery factor decreased from 72.62% to 26.31%, the contribution of CO2 mass transfer zone to recovery factor increased from 14.06% to 47.11%. When the formation dip angle increased from 0° to 35°, CO2 breakthrough time is earlier, the contribution of the pure oil flow zone to recovery factor increased from 59.32% to 65.62%, the contribution of CO2 mass transfer zone to recovery factor decreased from 17.54% to 16.32%. With initial water saturation increasing from 50% to 90%, displacement efficiency continued decreasing, CO2 flooding breakthrough time is later. The contribution of the pure oil flow zone to recovery factor increased from 55.68% to 82.58%, and the contribution of CO2 mass transfer zone to recovery factor decreased from 27.42% to 10.20%. The results provide the basis for front prediction and control in CO2 miscible flooding development.

Controlling bottom hole pressure (BHP) during tripping-out is a key challenge in ultra-deep well drilling. Under high-temperature and high-pressure (HTHP) conditions, ultra-deep wells feature long tripping-out cycles, where thermal effects are prone to causing BHP reduction and increasing kick risk. However, existing pressure control technologies struggle to adapt to the requirements of narrow safe density windows in deep formations. This study establishes a transient tripping-out temperature field model, taking the PS6 ultra-deep vertical well as a case study to calculate the variations in temperature, equivalent static density (ESD), and BHP during tripping-out at 2910 m and 9026 m. A weighted drilling fluid supplementation method is presented, with supplementary parameters designed and its feasibility verified. The results indicate that during the entire tripping-out process, the bottom hole temperature at 2910 m increases by 17.5 °C and BHP rises by 0.016 MPa; at 9026 m, the temperature increases by 72.6 °C and BHP decreases by 2.410 MPa. Compared with the traditional “heavy mud cap” technology, the presented method can control BHP within a smaller fluctuation range (within 0.339 MPa) during tripping-out, better adapting to the safe tripping requirements of narrow safe density windows in deep formations and effectively mitigating kick risk.

Lithuania covers the deepest parts of the Baltic basin and contains many reservoirs that have been explored for Hydrocarbon production and gas storage projects, including CO2 and hydrocarbon gas storage. Studies have also been conducted to assess the storage potential of these reservoirs for gases like CO2 and Hydrogen. In the studies, four saline aquifers, including Syderiai, Vaskai, and D11, and depleted hydrocarbon reservoirs in the Gargzdai structure were evaluated for potential CO2 storage. However, the long-term fate of these gases’ migration at the field scale has not been reported previously. In response to the existing gap, this study aims to evaluate the risks and challenges associated with subsurface CO2 and Hydrogen storage by conducting numerical simulations at two injection rates, of fluid migration, pH variations, and geomechanical responses using the tNavigator platform, complemented by laboratory experiments on outcrops representative of Syderiai formation, to achieve a detailed understanding of geochemical interactions between rocks and fluids. The results reveal distinct gas-specific behaviors: CO2 exhibits enhanced solubility trapping, density-driven convective mixing, and pronounced pH reduction, whereas Hydrogen demonstrates rapid buoyant migration, higher pressure buildup, and negligible geochemical reactivity. Both gases demonstrate short-term storage viability in the Syderiai aquifer under the modeled conditions, with pressure and total vertical stress remaining below the bottom-hole pressure limit of 450 bars. This integrated simulation and experimental study enhances our understanding of Lithuanian reservoirs for the safe, long-term storage of both CO2 and Hydrogen.

The aim of the study was to identify key technological indicators affecting productivity and the risk of sand production in the operation of sand-bearing wells at an offshore field. The methodology included field and laboratory studies of 32 production wells of various geometries, conducted from January 2024 to June 2025. Parameters such as flow rate, temperature gradient, bottomhole and formation pressure, and vibration frequency were monitored using digital sensors and processed using dimensionality reduction and machine learning methods. The results showed significant differences between vertical and horizontal wells: with an average flow rate of 74.71 m 3 /day, vertical wells had a productivity coefficient of 11.01 m 3 /day·MPa, while horizontal wells had a productivity coefficient of 22.56 m 3 /day·MPa at a flow rate of 66.10 m 3 /day. The principal component method revealed the greatest significance of the temperature gradient and flow rate (load coefficients of 0.667), as well as the decisive role of vibration activity in the formation of unstable modes (coefficient of 0.851), defined in this study as operational regimes exhibiting rapid changes in flow rate and pressure variance exceeding 15% within a 24-hour period. The calculated Spearman’s coefficient ( ρ = 0.88, p < 0.0001) between temperature fluctuations and productivity changes confirmed the direct influence of thermodynamics on filtration processes. Among the predictive models, XGBoost demonstrated the best regression accuracy (RMSE = 3.45; MAPE = 8.23%; R 2 = 0.91). However, to assess the risk of sand production as a classification task, additional metrics were calculated: F1-score = 0.91, AUC = 0.94, Precision = 0.88, Recall = 0.93, confirming the model’s suitability for this purpose. The practical significance of the results obtained lies in the possibility of using the developed approaches by technological monitoring services, design organisations, and field operators to build intelligent control systems aimed at reducing accidents, increasing production stability, and optimising the operating modes of sand-bearing reservoirs

Summary Real-time monitoring and accurate control of bottomhole pressure (BHP) play a decisive role in maintaining the mechanical balance of the formation-wellbore system during drilling operations. Due to the difficulty in real-time transmission of downhole parameters, surface measurements often fail to reflect the true downhole conditions, limiting the adaptability of existing intelligent models to dynamically changing complex drilling environments and constraining their field applications. In this paper, we propose a surface-parameter-driven BHP prediction approach with real-time calibration. By analyzing the multiphase flow mechanics in the wellbore and the coupling correlation between surface and downhole data, key surface parameters are selected as auxiliary prediction tasks to construct a multitask temporal neural network model that outputs both surface features and BHP jointly. During model inference, an online error feedback mechanism dynamically updates the shared hidden layer weights using real-time surface data and prediction errors, significantly enhancing the model’s adaptability to current complex downhole conditions. Results demonstrate that compared with traditional single-task models, the proposed method reduces prediction error by 6%, achieving an accuracy of 93% in BHP monitoring. This study provides an innovative framework for real-time and precise monitoring of BHP under complex downhole environments.