CN-121995509-A - Method for intelligently dividing and identifying sedimentary microphase based on logging curve

CN121995509ACN 121995509 ACN121995509 ACN 121995509ACN-121995509-A

Abstract

The invention relates to the technical field of geophysical well logging and sedimentology intelligent interpretation, and particularly discloses a method for intelligently dividing and identifying sedimentary microphases based on a well logging curve. The method comprises the steps of obtaining gamma ray curves and cleaning rule configuration, generating a rough segmentation segment set through data cleaning, standardization and filtering, generating a segment sample set through interface alignment, boundary fine adjustment and segment judgment by combining logging interpretation conclusion, extracting characteristics based on operation period dependence, generating segment classification results through rule classification and model inference fusion treatment, and finally outputting a deposition microphase type result and forming closed loop feedback through sequence splicing, morphology rule mapping and configuration parameter self-adaptive updating. The invention realizes the automation and self-adaptive optimization of the deposition microphase identification process, and effectively improves the identification accuracy and the engineering application efficiency.

Inventors

YU BENZHI
CHEN LUFEI
WANG HAORAN
ZHANG YIAN

Assignees

武汉时代地智科技股份有限公司

Dates

Publication Date: 20260508
Application Date: 20251224

Claims (10)

1. A method for identifying a sedimentary microphase based on intelligent segmentation of a log, comprising: Acquiring GR curves and cleaning rule configuration, performing data cleaning and standardization processing, performing wavelet transformation denoising processing based on wavelet base orders and threshold strategy tables in the cleaning rule configuration, performing median filtering processing based on median filtering window widths in the cleaning rule configuration, and generating a coarse segmentation segment set; Acquiring a rough segmentation segment set and a logging interpretation conclusion, executing a process comprising searching half window width, matching priority, conflict judging strategy, interface alignment and boundary fine adjustment, including binding interpretation interfaces in candidate boundary neighborhood through searching half window width, carrying out small-range position correction according to local window statistics, executing segment segmentation and merging operation when the rough segmentation segment crosses a plurality of interpretation interfaces, carrying out segment boundary adjustment process, carrying out statistical analysis process based on a minimum segment length threshold and a long segment judgment threshold in cleaning rule configuration, carrying out index allocation process based on a sample number strategy in cleaning rule configuration, and generating a segment sample set; Performing extraction processing of target sampling interval, longest sequence length, short segment window width, long segment sliding stride, abnormal label mapping table, mask propagation strategy and characteristics, classifying slope trend rules for the short segment based on a monotonic continuous length threshold, a turning density threshold, a main turning strength threshold and a platform coverage proportion threshold in a rule version library, performing inference classification for the long segment based on a fusion weight in a model parameter by adopting an LSTM-FCN network, performing classification inference processing, and generating a segment classification result; And performing sequence splicing processing based on a minimum cavity threshold and a bridging strategy in cleaning rule configuration, performing rule mapping processing based on the form type, the segment length range, the stable segment coverage ratio, the adjacent class transition relation and the horizon definition in the microphase mapping table, and performing segment length threshold configuration updating processing based on updating the bit value snapshot and the abnormal duty ratio snapshot in the bin, so as to generate a cleaning rule configuration structure.
2. The method of claim 1, wherein generating the set of coarse segments further comprises: the GR curve and the cleaning rule configuration are obtained, and deletion correction and anomaly labeling processing are carried out on the basis of a deletion identification rule, an anomaly type dictionary, a depth unit and sampling interval standard, a shielding well Duan Minglu, a drift check baseline, a boundary extrapolation strategy, an interpolation mode priority and a log record level in the cleaning rule configuration, so that cleaning data are obtained; Extracting depth and curve values from cleaning data, performing depth alignment processing based on a target sampling interval given by cleaning rule configuration, and performing interval normalization processing based on a mapping policy table configured by the cleaning rule configuration to generate a standardized sequence; And carrying out wavelet denoising processing on the standardized sequence based on a wavelet base order and a threshold strategy table in the cleaning rule configuration, and carrying out median filtering processing based on a median filtering window width and a shielding section participation strategy in the cleaning rule configuration to generate a coarse segmentation section set structure.
3. The method of claim 2, wherein GR curve and cleansing rule configuration comprises: The GR curve is the corresponding relation between the well depth and the gamma response, and the recording unit, the sampling interval and the depth reference are given by an on-site recording system or a history calibration record; the purge rule configuration includes a miss-identification rule, an anomaly type dictionary, depth unit and sampling interval criteria, a mask well Duan Minglu, a drift check baseline, a boundary extrapolation policy, interpolation mode priority, and log record rating.
4. The method of claim 1, wherein performing the sequence stitching process based on the minimum hole threshold and bridging policy in the cleansing rule configuration comprises: The method comprises the steps of collecting and depositing microphase type results according to well sections, horizons and morphology types, extracting section length distribution, platform coverage proportion sections, turning density profile and adjacent transition relation count, counting the length distribution of short sections and long sections according to section classification results, stabilizing section coverage proportion profile and shielding adjacent mark duty ratio, and writing statistical products into an updating basis bin.
5. The method of claim 4, comprising updating a bin-dependent bitwise value snapshot, an exception duty cycle snapshot, and a cross-version lookup table for comparing a change magnitude with a previous version configuration.
6. The method of claim 1, wherein performing a segment length threshold configuration update process, the process of generating a cleansing rule configuration structure further comprises: Calculating boundary intervals of short section and long section distribution at section level, combining section length bandwidths corresponding to different forms in a deposited microphase type result, giving a suggested group of section length thresholds, and if bandwidths with obvious differences appear among different layers of the same well section, generating a layered threshold group with a layer limiting condition.
7. The method of claim 1, wherein generating the cleansing rule configuration structure further comprises: Checking the compatibility of candidate update and rule version library item by item, if there is conflict, recording rule conflict and making processing suggestion, and calculating the influence list of the affected links and items.
8. The method of claim 4, wherein the content of the cleansing rule configuration structure includes a threshold table, a segment length threshold, bridging and overlap arbitration parameters, a morphology candidate window scheme, a boundary-specific rule, a mask spreading policy, and versioning metadata, a structure header write configuration version number, a generation time, a source statistics digest, and a rollback policy.
9. The method of claim 1, wherein performing a segment length threshold configuration update process, the process of generating a cleansing rule configuration structure further comprises: And in the output preparation stage, performing difference alignment on the update and the previous configuration, writing a difference alignment result into a configuration difference abstract, and synchronizing the abstract to an operation and maintenance review channel.
10. The method of claim 1, wherein the log interpretation comprises: The logging interpretation conclusion pointer gives information such as horizon names, interface types, interface depth points or narrow ranges, credibility marks, interpretation version numbers and the like to the target well section, and the source can be interpretation personnel marks or automatic pushing of rule bases.

Description

Method for intelligently dividing and identifying sedimentary microphase based on logging curve Technical Field The invention relates to the technical field of geophysical well logging and sedimentology intelligent interpretation, in particular to a method for intelligently dividing and identifying sedimentary microphases based on a well logging curve. Background Lithology, granularity, sortability, clay content, vertical sequence, sand morphology and distribution, etc. are important causative markers in identifying and recognizing sedimentary phases. These causative markers are the result of hydrodynamic factors in various deposition environments, while hydrodynamic conditions control the changes in petrophysical properties, such as formation natural potential, natural gamma, etc. The logging curve is the physical response of various physical properties along the depth change of the well hole, so that the accurate rock-electricity relation of the coring well is established, the logging curve is further popularized to the non-coring well, and the reservoir characteristics of the non-coring well are reversely deduced. The change of the cause mark in the longitudinal and transverse directions is effectively fed back by using the logging curve shape, so that valuable data is provided for identifying the sedimentary facies, and the method is an effective way for identifying the sedimentary facies. The natural gamma well logging curve is in the form of a box, a bell, a funnel and a finger, and as shown in fig. 1, the natural gamma well logging curve reflects the energy change or the relative stability of the sediment during the deposition, for example, the box reflects the stable rapid accumulation of hydrodynamic conditions or the stable deposition of environment, the bell reflects the deposition environment with gradually weakened energy of water flow or gradually reduced supply of material source, and the funnel is opposite to the bell and reflects the deposition environment with strengthened hydrodynamic force or gradually increased supply of material source. They have obvious anomalies in different sedimentary microphase zones and in different reservoirs. In particular, the method for measuring the rock stratum in the well naturally exists, and the method for measuring and acquiring the well logging curve is simple in technology and low in cost. Thus, it is widely used in dividing and studying deposition bands. At present, the deposition microphase division based on logging data is mainly interpreted manually according to the morphology of a logging curve (such as a box type, a bell type and a funnel type in a logging gamma curve represent different deposition environments and deposition loops). This approach is affected by the knowledge of human experience and proficiency, and is not only time-consuming and labor-consuming, but also has some subjectivity and uncertainty in interpretation results. The advanced technologies such as big data analysis, deep learning and the like are applied to oil-gas geology research, so that the problem of idle exploration and attempt of big data analysis resources in the current petroleum industry is solved. The premise of identifying the logging phase is to accurately divide the identification unit. The underground sedimentary structure is complex and changeable, so that the logging phases of the same sedimentary microphase are different, the identification difficulty for experienced interpreters is low, but the difficult effect display for computer automatic identification is severe. The neural network which has been developed in recent years can autonomously learn and extract the characteristics between data, and provides a new means for identifying the sedimentary microphase of the logging. The network autonomously learns the curve characteristics instead of manually extracting, so that the characteristics of the data can be maintained to the greatest extent. The existing method for automatically dividing and identifying the sedimentary microphase by utilizing the logging curve based on the neural network mainly comprises the steps of converting the logging curve into pictures, and applying the pictures to the field of target detection or image division for automatically dividing and identifying the sedimentary microphase. Then, the neural network is utilized to automatically divide the labels, a large amount of manpower is needed to manufacture the labels, the workload is large, and a large amount of manpower and material resources are consumed. In order to accurately and conveniently automatically divide and identify the deposition microphase, the invention provides a method for intelligently dividing and identifying the deposition microphase based on a logging curve, which improves the speed of automatic division and the accuracy of classification, and reduces the cost of certain manual label manufacturing. In addition, in the field of geophysical well logging and se