In dipole acoustic logging,the dispersion of flexural waves complicates shear wave velocity extraction.Traditional dispersion correction algorithms rely on repeated forward modeling of the dispersion equation,which is extremely time-consuming.To address this issue,this paper proposes a dispersion correction method for dipole flexural waves based on a fully connected deep neural network.First,a large-scale dataset is generated by forward modeling using the dispersion equation,and a multi-layer fully connected deep neural network is constructed to learn the nonlinear mapping between formation parameters and dispersion curves,enabling fast forward simulation of dispersion curves.Second,a one-dimensional density clustering algorithm based on neighborhood consistency is introduced to automatically extract effective dispersion features from measured data.By combining this with the fully connected deep neural network bending flexural wave dispersion forward modeling method and the least-squares matching algorithm,an efficient workflow for dipole flexural wave dispersion correction is established to accurately extract shear wave velocity.The research results show that:①The fully connected deep neural network can generate one thousand predicted dispersion curves in only 0.277 s,achieving more than three orders of magnitude speedup compared to the dispersion equation(320.9 s)while maintaining a prediction error below 0.5%,thereby enabling real-time forward modeling at the millisecond level.②The density clustering algorithm effectively removes abrupt noise and identifies valid dispersion features,providing high signal-to-noise ratio inputs data for the correction workflow.③Application of the dipole flexural wave dispersion correction method to an actual well in the eastern South China Sea demonstrates that the corrected shear wave velocity is significantly more accurate than the original extraction result.Using this shear wave velocity for pay zone identification,the P-wave to S-wave velocity ratio and Lamé impedance sections accurately delineate sandstone reservoirs and gas-bearing zones,substantially enhancing the accuracy of seismic inversion and fluid identification.It is concluded that this method not only provides a new tool for accurate shear wave velocity extraction in dipole acoustic logging,but also serves as a valuable reference for constructing intelligent well-log data processing workflows.
Coalbed methane serves as a vital component of clean energy,and its efficient development encounters technical bottlenecks in the dynamic monitoring of production profiles.To tackle the difficulties in monitoring the production profiles during commingled drainage of low-pressure,thin-interbedded coalbed methane wells,as well as the insufficient accuracy of conventional methods and restricted instrument running,this paper takes well A1 in the Bijie pilot area of Guizhou province as the research object and proposes an evaluation method for coalbed methane well production profiles based on distributed optical fiber sensing.The study adopts a fixed-point optical fiber binding process outside the tubing to realize synchronous acquisition of full-wellbore temperature and acoustic signals under multiple working systems in well A1.A production profile interpretation framework is established based on the Ramey wellbore heat exchange model,Joule-Thomson effect and two-phase flow drift model.A coupled interpretation strategy is put forward,which uses the steady-state frequency band energy(FBE)analysis of distributed acoustic sensing(DAS)to constrain the temperature field inversion of distributed temperature sensing(DTS).Specifically,the steady-state frequency band energy of DAS in characteristic frequency bands is used to extract the water production rate information of each layer,so as to constrain the gas-liquid ratio parameters in the DTS temperature forward model.This strategy solves the strong multi-solution problem of traditional single DTS interpretation under gas-liquid two-phase flow conditions and realizes the fine quantitative allocation of gas and water production rates of each coal seam.Monitoring data clearly indicate the main gas-producing and water-producing layers,and the optical fiber interpretation results show good consistency with open-hole logging conclusions and fracturing operation parameters,which verifies the reliability of optical fiber monitoring in coalbed methane wells.This study provides a reliable dynamic monitoring method for the development of low-pressure,thin-interbedded coalbed methane and offers scientific guidance for the optimization and adjustment of production technologies in the process of coalbed methane development.
With the transition of well log evaluation towards data-driven paradigms, high-quality data has become the cornerstone for constructing high-confidence geological models. Consequently, data cleaning is a critical pre-processing step to ensure algorithm robustness and interpretation accuracy. Through a comprehensive literature review, this study traces the evolution of cleaning technologies since the 1970s, establishing a standardized framework from quality identification to performance evaluation. We analyze the formation mechanisms of noise, outliers, and missing values, identifying instrument acquisition, borehole environment, and operational factors as dominant controls. Synthesizing this development, we propose a five-stage evolutionary model comprising numerical analysis, signal processing, traditional machine learning, deep learning, and large language model (LLM). Based on this progression, we construct a technical classification system covering numerical analysis, signal processing, artificial intelligence, and multi-method fusion, comparing typical methods in terms of algorithmic principles and practical efficacy. Finally, forecasts future trends, highlighting intelligence, adaptive learning, real-time processing, cloud-edge collaboration, and knowledge-driven approaches. This research offers a theoretical foundation and practical reference for the iterative upgrading and standardization of well logging data processing technologies.
To address the problems of insufficient sensitivity analysis on the throttling pressure drop of gas nozzles,great difficulties in downhole gas nozzle selection and layered flow rate allocation in layered gas injection wells with CO2-bearing natural gas,and to meet the engineering demands of complex deep-sea well conditions,this paper takes the layered gas injection gas nozzles for CO2-bearing natural gas as the research object.Based on the Fluent numerical simulation platform,a numerical calculation system including the energy equation,mass equation,momentum equation,multi-component transport model and standard k-ε turbulence model is established.Using the control variable method,this paper investigates the effects of CO2 molar ratio,gas nozzle diameter,flow rate,temperature,outlet pressure and multi-factor coupling on the throttling pressure drop of gas nozzles,reveals the pressure distribution law inside the gas nozzle,and conducts field tests on an oil production platform in the South China Sea.The results show that:①The throttling pressure drop of the gas nozzle is mainly concentrated in the contraction zone of the nozzle.When the CO2 molar ratio is 20%～80%,the throttling pressure drop decreases in a quadratic function with the increase of CO2 content,and the pressure drop changes more significantly when the CO2 molar ratio is 20%～50%.②Under the action of a single factor,the throttling pressure drop increases in a quadratic function with the increase of flow rate,increases linearly with the rise of temperature,and decreases gradually with the increase of gas nozzle diameter and outlet pressure.③Under the flow rate-temperature coupling effect,the flow rate dominates the increase of pressure drop,and the decrease of temperature can slightly offset the increase of pressure drop.Under the outlet pressure-temperature coupling effect,the two factors synergistically enhance the inhibition effect on pressure drop.④In the field test,a gas nozzle with a diameter of 5.2 mm is adopted,the actual wellhead gas injection pressure is 0.6 MPa higher than the target pressure,and the error between the actual injection flow rate(7.2×104 m3/d)and the target injection flow rate(6.1×104 m3/d)is less than 20%,which meets the requirements of field layered gas injection allocation.The research on the variation law of the throttling pressure drop of CO2-bearing natural gas nozzles under multi-factor and coupling effects can provide reliable theoretical and technical support for the accurate selection of downhole nozzles and the dynamic regulation of layered gas injection flow rate in layered gas injection wells with high CO2 content.
To address the issue that casing scale interferes with the calculation of formation macroscopic capture cross-section(Sigma)and reduces the accuracy of residual oil saturation evaluation in pulsed neutron-neutron(PNN)logging through cased holes,this paper establishes a three-dimensional numerical model of PNN instrument,wellbore and formation using the Monte Carlo simulation software.The simulation reproduces the slowing-down and capture processes of neutrons in the scaled borehole and formation.The effects of scale type,thickness,and composition on thermal neutron time spectra,neutron distribution,and measured Sigma are analyzed.A self-compensation correction method for the formation Sigma is developed and validated with field log data.The results show that:①Casing scale replaces borehole fluid,reducing the fast neutron slowing-down capability.Corrosion scale containing elements with high thermal neutron capture cross-section(e.g.,Cl)shortens the thermal neutron lifetime,leading to overestimation of the formation Sigma,whereas sedimentary scale weakens the thermal neutron capture capability,resulting in underestimation of Sigma.②The formation Sigma and the near-to-far detector thermal neutron count ratio increase linearly with the thickness of corrosion scale,while they decrease linearly with the thickness of sedimentary scale.The magnitude of this effect varies with formation porosity.For each 1 cm increase in corrosion scale thickness,formation Sigma increases by 3.37 c.u.in a formation with 8%porosity and by 3.69 c.u.in a formation with 15%porosity.③A higher NaCl content in the corrosion scale leads to a more pronounced influence on the formation Sigma.④The correction amount of the formation Sigma exhibits a linear relationship with the difference between the cased-hole neutron porosity and the compensated neutron porosity.The proposed self-compensation correction method achieves an average absolute error of only 0.29 c.u.after correction.It is concluded that casing scale significantly disturbs the logging response by altering neutron slowing-down and capture processes.The self-compensation correction method based on the thermal neutron count ratio and compensated neutron porosity effectively eliminates the scale effect,substantially improving the accuracy of formation Sigma calculation and the reliability of residual oil saturation evaluation,thus providing key technical support for dynamic monitoring of residual oil in cased holes.
