Search

CN-121069476-B - Method for identifying layer boundary in geological envelope based on artificial intelligence

CN121069476BCN 121069476 BCN121069476 BCN 121069476BCN-121069476-B

Abstract

The application provides a method for identifying a layer boundary in a geological envelope based on artificial intelligence, which comprises the steps of determining seismic exploration data and logging data of a target work area, extracting seismic reflection characteristics from the seismic exploration data based on a trained first neural network model, simultaneously extracting logging sequence characteristics from the logging data by using a trained second neural network model, generating geological characteristic characterization of the target work area according to the seismic reflection characteristics and the logging sequence characteristics, determining implicit topological structure characterization representing the geological envelope in the target work area according to the geological characteristic characterization based on formation contact relation constraint and lithology sequence constraint aiming at the target work area, and processing the implicit topological structure characterization based on a trained third neural network model so as to predict the boundary of the geological envelope in the target work area. The method ensures that the geological envelope horizon boundary prediction result more accords with the actual geological law.

Inventors

  • SU JIARUI

Assignees

  • 长江大学

Dates

Publication Date
20260512
Application Date
20250915

Claims (7)

  1. 1. A method for predicting geologic envelope horizon boundaries based on artificial intelligence, comprising: Step1, determining seismic exploration data and logging data of a target work area; Step 2, extracting seismic reflection characteristics from seismic exploration data based on the trained first neural network model, and extracting logging sequence characteristics from logging data by adopting the trained second neural network model; Step 3, generating geological characteristic characterization of the target work area according to the seismic reflection characteristics and the logging sequence characteristics; step 4, determining an implicit topological structure representation of a representation geological envelope in a target work area according to geological characteristic representation based on formation contact relation constraint and lithology sequence constraint of the target work area; Step 5, processing the implicit topological structure characterization based on the trained third neural network model to predict the horizon boundary of the geological envelope in the target work area; Step 2, extracting seismic reflection characteristics from seismic exploration data based on a first neural network model after training, and simultaneously extracting logging sequence characteristics from logging data by adopting a second neural network model after training, wherein the method specifically comprises the following steps of: step 2.1, performing multi-scale feature extraction on the data after seismic exploration denoising based on a first neural network model after training so as to obtain initial seismic reflection features; Step 2.2, based on the trained second neural network model, vertical sequence feature extraction is carried out on the logging data so as to generate initial logging sequence features; Step 2.3, eliminating redundant features which are not related to stratum boundaries in the initial seismic reflection features and the initial logging sequence features to generate screened seismic reflection features and screened logging sequence features; Step 3, generating a geological characteristic representation of the target work area according to the seismic reflection characteristic and the logging sequence characteristic, wherein the geological characteristic representation specifically comprises the following steps: Step 3.1, based on a space coordinate system of a target work area, performing space coordinate anchoring on the seismic reflection characteristic and the logging sequence characteristic, and calculating association weights according to association degrees of the seismic reflection characteristic and the logging sequence characteristic after the space coordinate anchoring and the stratum boundary respectively to generate a characteristic association mapping table; step 3.2, carrying out weighted fusion on the seismic reflection characteristics and the logging sequence characteristics based on the characteristic association mapping table, and verifying and eliminating the fused contradictory characteristics through stratum topological relation so as to generate a fused geological characteristic set; Step 3.3, carrying out space continuous processing on the fusion geological feature set based on the geological background of the target work area region, and marking geological significance corresponding to each feature so as to generate geological feature characterization of the target work area; Determining an implicit topological structure representation of a representation geological envelope in a target work area based on formation contact relation constraint and lithology sequence constraint of the target work area, wherein the implicit topological structure representation of the representation geological envelope in the target work area is determined according to geological characteristic representation; step 5, processing the implicit topological structure characterization based on the trained third neural network model to predict the horizon boundary of the geological envelope in the target work area, specifically including: Step 5.1, carrying out boundary sensitive characteristic enhancement on the implicit topological structure characterization based on an implicit structure characteristic enhancement layer in the trained third neural network model so as to generate enhanced implicit structure characteristics; Step 5.2, based on a multi-scale boundary feature capturing layer in the third neural network model after training, carrying out multi-scale boundary feature extraction on the reinforced implicit construction features so as to generate a multi-scale boundary feature set; Step 5.3, performing horizon boundary preliminary prediction on the multi-scale boundary feature set based on a horizon boundary preliminary prediction layer in the trained third neural network model to generate initial horizon boundary data; and 5.4, performing topology verification and optimization on the initial horizon boundary data based on a boundary topology verification optimization layer in the trained third neural network model to generate final horizon boundary data of the geological envelope in the target work area.
  2. 2. The method of claim 1, wherein step 1, determining the seismic survey data and the well log data for the target work area, specifically comprises: step 1.1, performing self-adaptive noise suppression processing on original seismic exploration data of a target work area to generate data after seismic exploration denoising; Step 1.2, performing depth correction and outlier rejection processing on original logging data of a target work area to generate a logging data correction result; And 1.3, carrying out data space coordinate matching on the data after seismic exploration denoising and the logging data correction result so as to generate seismic exploration data and logging data of a target work area.
  3. 3. The method of claim 2, wherein step 1.1, performing adaptive noise suppression processing on raw seismic survey data of the target work area to generate seismic survey denoised data, specifically comprises: Step 1.1.1, dividing original seismic exploration data of a target work area according to space continuous windows, calculating amplitude standard deviation, frequency distribution characteristics and amplitude change rate between adjacent windows of each divided window data, and determining window dividing statistical characteristics of the seismic data according to the amplitude standard deviation and the frequency distribution characteristics; Step 1.1.2, according to the window statistics characteristics of the seismic data, combining the pre-marked effective reflection signal samples and noise samples, judging the type of the window-by-window signal of the original seismic exploration data of the target work area, applying dynamic weight filtering to the window marked as noise, and reserving the original characteristics of the window marked as the effective signal to generate the denoised seismic exploration data.
  4. 4. The method according to claim 1, wherein step 2.1, based on the trained first neural network model, performs multi-scale feature extraction on the denoised data of the seismic survey to obtain the initial seismic reflection feature, specifically includes: step 2.1.1, performing feature extraction of different spatial scales on the data after seismic exploration denoising based on a multi-scale convolution feature extraction layer in a trained first neural network model so as to generate a preliminary multi-scale seismic feature map; step 2.1.2, performing attention weight calculation and feature fusion on the preliminary multi-scale seismic feature map based on a cross-scale feature attention fusion layer pair in the trained first neural network model so as to generate a fused seismic feature map; step 2.1.3, carrying out residual operation on the fused seismic feature map based on a residual noise suppression layer in the trained first neural network model so as to generate a denoised seismic feature map; And 2.1.4, performing dimension compression on the denoised seismic feature map based on a feature dimension compression and screening layer in the trained first neural network model to remove redundant feature dimensions, and performing effective feature screening to retain features strongly related to horizon boundaries so as to generate initial seismic reflection features.
  5. 5. The method of claim 1, wherein step 2.2 of performing vertical sequence feature extraction on the log data based on the trained second neural network model to generate the initial log sequence feature comprises: 2.2.1, performing vertical window division and sequence conversion on the well logging data based on a vertical window serialization layer in the trained second neural network model so as to generate a vertical well logging sequence fragment set; 2.2.2, carrying out residual operation on the vertical logging sequence fragment set based on a sequence residual denoising layer in the trained second neural network model so as to generate a denoised vertical logging sequence fragment set; 2.2.3, performing attention weight calculation on the denoised vertical logging sequence fragment set based on the multi-feature attention enhancement layer pairs in the trained second neural network model to generate an attention enhanced logging sequence feature set; And 2.2.4, carrying out feature correlation analysis and screening on the logging sequence feature set after attention enhancement based on a sequence feature screening layer in the trained second neural network model so as to generate initial logging sequence features.
  6. 6. The method of claim 1, wherein step 3 of generating a geologic characterization of the target work area from the seismic reflection signature and the log sequence signature further comprises: And 3.0, generating a geological background of the target work area according to the geological outcrop data and the drilling core description data of the target work area and the periphery.
  7. 7. The method according to claim 6, wherein the generating step 3.0, according to the geological outcrop data and the drilling core description data existing in the target work area and the periphery, the geological background of the target work area specifically includes: Step 3.0.1, classifying and labeling lithology types, sedimentary structures and fossil combinations on the geological outcrop data and the drilling core description data existing in the target work area and the periphery to generate a classified geological basic data set; Step 3.0.2, performing vertical sequence association analysis on lithology combinations and sedimentary structures in the geological basic data set after classification, and extracting key features reflecting a sedimentary environment to generate a sedimentary environment feature set; Step 3.0.3, based on the deposition environment feature set, combining the existing construction motion records around the target work area, and establishing time sequence association between the deposition environment feature and the construction motion stage so as to generate a construction evolution time sequence table; And 3.0.4, comprehensively checking the deposition environment characteristics and the construction evolution stage based on the deposition environment characteristic set and the construction evolution time sequence table, and supplementing stratum development rule description to generate a geological background of the target work area region.

Description

Method for identifying layer boundary in geological envelope based on artificial intelligence Technical Field The invention relates to the field of mineral resource exploration and data interpretation, in particular to a method for identifying a layer boundary in a geological envelope based on artificial intelligence. Background In the fields of oil and gas resource exploration, mineral resource development, engineering geological investigation and the like, accurate determination of horizon boundaries of geological envelopes in a target work area is one of core technical requirements. The accurate positioning of the geological envelope horizon boundary directly influences the formation division, resource reserve evaluation and scientificity of development scheme design, and especially in complex geological structure areas, boundary prediction is realized through comprehensive analysis of multi-source geological data so as to meet the requirement of actual engineering on geological information refinement. The conventional scheme for predicting the horizon boundary of the geological envelope takes the seismic exploration data as main input and a small amount of logging data as auxiliary verification, adopts a threshold segmentation method to extract a seismic reflection event as a preliminary indication of the horizon boundary after denoising the seismic data through a filtering algorithm with fixed parameters, and finally relies on a traditional statistical regression model to fit the preliminarily indicated boundary by combining manually screened logging lithology parameters to obtain a horizon boundary prediction result. The prior art scheme has obvious technical limitations that as the single type of seismic characteristics are processed only through a fixed algorithm and an adaptation mechanism of the actual geological rules (such as stratum contact relation and lithology sequence distribution) of a target work area is not established, complementary information of multi-source data (earthquake and logging) cannot be effectively integrated, and the problems that in a geological condition complex area (such as stratum contact relation is various and lithology vertical transition is complex), the predicted horizon boundary is prone to space continuity fracture or has larger deviation from the actual geological envelope boundary are difficult to meet the requirement of high-precision geological exploration. Disclosure of Invention Object of the invention The main object of the present patent is to provide a method for identifying a layer boundary in a geological envelope based on artificial intelligence, so as to overcome or alleviate the above-mentioned problems in the prior art. (II) technical scheme The invention provides a method for identifying a layer boundary in a geological envelope based on artificial intelligence, which comprises the following steps: Step1, determining seismic exploration data and logging data of a target work area; step 2, extracting seismic reflection characteristics from seismic exploration data based on the first neural network model after training, and simultaneously extracting logging sequence characteristics from logging data by adopting the second neural network model after training; Step 3, generating geological characteristic characterization of the target work area according to the seismic reflection characteristics and the logging sequence characteristics; step 4, determining an implicit topological structure representation of a representation geological envelope in a target work area according to geological characteristic representation based on formation contact relation constraint and lithology sequence constraint of the target work area; And 5, processing the implicit topological structure characterization based on the trained third neural network model to predict the horizon boundary of the geological envelope in the target work area. (III) beneficial effects 1. Aiming at the limitation of processing single type seismic characteristics and manually screening logging lithology parameters by a fixed algorithm, the method adopts a first neural network model which is completed by training to extract seismic reflection characteristics from seismic exploration data, and simultaneously adopts a second neural network model which is completed by training to extract logging sequence characteristics from logging data, which is different from the mode that the traditional fixed parameter filtering can only process single seismic characteristics and the manual screening logging parameters are easy to be subjectively influenced, the two neural network models can dynamically learn the extraction rule of the seismic reflection characteristics and the logging sequence characteristics through training without depending on fixed algorithm or manual intervention, can more comprehensively capture multidimensional reflection information in the seismic data and vertical sequence informa