Multi-source data comprehensive identification-based storm cause high-quality lithofacies identification method

Abstract

The invention belongs to the field of geological sedimentology, and particularly relates to a storm cause high-quality lithofacies identification method based on multi-source data comprehensive identification, which comprises the following steps of: A, data acquisition and preprocessing: S1, data acquisition and confirmation: automatically identifying the characteristics similar to a Baomah sequence in a rock core by using a convolutional neural network (CNN) according to the rock core, a field outcrop, a logging curve and logging data; outputting storm deposition sample information; s2, performing automatic lithofacies classification on the rock core image by adopting deep learning (U-Net) in combination with a sedimentary structure; and S3, automatically identifying curve characteristics of a'box-shaped '(storm dam main body) and'bell-shaped and finger-shaped' (storm dam side edge) by adopting time sequence analysis (LSTM or Transformer). According to the method, the high-quality lithofacies recognition model is established by integrating multivariate data, the recognition precision and efficiency are greatly improved, and a scientific basis is provided for oil and gas reservoir evaluation in a storm sedimentary environment.

Description

Technical Field

[0001] The present invention relates to the technical field of geological sedimentology, and in particular to a method for identifying high-quality lithofacies of storm genesis based on comprehensive identification of multi-source data. Background Art

[0002] Storm deposits are an important type of sedimentation in semi-deep lake environments. The high-quality lithofacies formed by them usually have high porosity and permeability and are an important component of oil and gas reservoirs.

[0003] Traditional identification methods rely primarily on core observations of oil-bearing levels and log characteristics. However, these methods fail to fully capture the high-quality lithofacies characteristics of storm deposits, which are characterized by complex features such as oscillating water flow and vertical "Bauma-like" sedimentary sequences. Multi-source data, such as sand-to-sand ratio, gamma value, porosity, compositional maturity, mercury removal efficiency, grain size sorting coefficient, and bioturbation index, can provide more comprehensive rock characterization information. However, a systematic approach to integrating these data for the identification of high-quality storm-derived lithofacies is currently lacking. Therefore, we propose a method for identifying high-quality storm-derived lithofacies based on comprehensive multi-source data to address this issue. Summary of the Invention

[0004] The purpose of the present invention is to solve the shortcomings of the existing technology and propose a storm-causing high-quality lithofacies identification method based on comprehensive identification of multi-source data.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] The method for identifying high-quality lithofacies of storm genesis based on comprehensive identification of multi-source data includes the following steps:

[0007] A: Data acquisition and preprocessing

[0008] S1: Data acquisition and verification, that is, using convolutional neural networks (CNN) to automatically identify the "Bauma-like sequence" features in the core based on cores, field outcrops, well logging curves, and mud logging data, and output storm sediment sample information;

[0009] S2: Automatic lithofacies classification of core images using deep learning (U-Net) combined with sedimentary structures;

[0010] S3: Use time series analysis (LSTM or Transformer) to automatically identify the “box-shaped” (main body of the storm dam) and “bell-shaped, finger-shaped” (side edge of the storm dam) curve features;

[0011] S4: Obtain rock porosity and permeability data through physical property testing;

[0012] S5: Obtain the pore distribution characteristics of the rock through mercury injection testing, including pore radius and mercury removal efficiency;

[0013] S6: Obtain the particle size distribution characteristics of the rock through particle size analysis, including average particle size, sorting coefficient and skewness;

[0014] S7: Obtaining reservoir quality differences among different storm lithofacies through thin section analysis parameters, including surface porosity;

[0015] B: Establishment of a high-quality lithofacies identification parameter system for storm genesis

[0016] S8: Feature importance analysis (random forest) was used to automatically select the optimal seven parameters and establish a high-quality lithofacies identification parameter system for storm deposits, namely, sand-to-ground ratio, gamma value, porosity, compositional maturity, mercury removal efficiency, particle size sorting coefficient, and bioturbation index;

[0017] C: Construction of comprehensive recognition model

[0018] S9: On the basis of assigning values ​​to each parameter, a pairwise comparison matrix of each element in the criterion layer is constructed to determine the relative weight of each criterion element;

[0019] S10: Calculation formula Q for the storm sediment high-quality lithofacies prediction coefficient based on sand-to-sand ratio, gamma value, porosity, compositional maturity, mercury removal efficiency, particle size sorting coefficient, and bioturbation index, i.e., comprehensive score;

[0020] S11: Calculate the comprehensive score and use unsupervised learning (K-means) to classify the high-quality lithofacies.

[0021] Preferably, a pairwise comparison matrix of each factor is constructed to determine the relative weight of each criterion factor.

[0022] Preferably, the parameters are calculated automatically, and the product Ai of the elements in the i-th row of the matrix and its fourth root ai are output.

[0023] Preferably, the calculation formula of the weight corresponding to the output parameter is:

[0024]

[0025] Preferably, the weights corresponding to the output sand-to-land ratio are ε1=0.34; the weights corresponding to the output gamma value are ε2=0.24, and the weight corresponding to the output porosity is ε3=0.24; the weight corresponding to the output component maturity is ε4=0.072, the weight corresponding to the output mercury removal efficiency is ε5=0.072, the weight corresponding to the output particle size sorting coefficient is ε6=0.021, and the weight corresponding to the output bioturbation index is ε7=0.015.

[0026] Preferably, in S8, the analytic hierarchy process is used to first assign a scale to each parameter, and then perform weight calculation based on the assigned values ​​of each parameter, i.e., sand-to-ground ratio (SR)=4, gamma value (GR)=4, porosity —4, composition maturity (R)—2, mercury removal efficiency (Hg)—2, particle size sorting coefficient (Sc)—1, bioturbation index (Bio)—1 / 2.

[0027] Preferably, the formula for calculating the storm deposit high-quality lithofacies prediction coefficient based on the input seven parameters of sand-to-sand ratio, gamma value, porosity, composition maturity, mercury removal efficiency, particle size sorting coefficient, and bioturbation index is:

[0028]

[0029] Preferably, according to the size of the Q value, based on the comprehensive evaluation standard of fine-grained sedimentary rock reservoirs and combined with unsupervised learning automatic classification and rating, the storm deposit dominant lithofacies evaluation level is divided into three levels, that is, Q>50 is a dominant lithofacies, Q between 40-50 is a medium lithofacies, and Q<40 is a poor lithofacies.

[0030] Preferably, the construction of the comprehensive identification model also includes AI model verification and optimization, that is, inputting 8 sections of samples from the same study area to verify the accuracy of the high-quality lithofacies comprehensive scoring model.

[0031] Beneficial effects of the present invention:

[0032] (1) By integrating multi-source data such as sand-to-sand ratio, gamma value, porosity, compositional maturity, mercury removal efficiency, particle size sorting coefficient, and bioturbation index, the accuracy and reliability of storm-causing high-quality lithofacies identification were improved;

[0033] (2) A systematic high-quality lithofacies identification model was established, providing a scientific basis for the evaluation of oil and gas reservoirs in storm sedimentary environments.

[0034] (3) The method is simple and easy to use, and is suitable for high-quality lithofacies identification of different types of reservoirs. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flow chart of the storm-causing high-quality lithofacies identification method based on comprehensive identification of multi-source data proposed in the present invention;

[0036] Figure 2 This is a lithofacies core photograph of the high-quality lithofacies identification method for storm genesis based on comprehensive identification of multi-source data proposed in the present invention;

[0037] Figure 3 The bioturbation map of the storm-causing high-quality lithofacies identification method based on comprehensive identification of multi-source data proposed in this invention;

[0038] Figure 4 Schematic diagram of the ideal sedimentary profile and storm sedimentary sequence of the study area for the storm genesis high-quality lithofacies identification method based on comprehensive identification of multi-source data proposed in this invention. DETAILED DESCRIPTION

[0039] The technical solutions of the present invention will be described clearly and completely below with reference to specific embodiments. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are also within the scope of protection of the present invention.

[0040] Reference Figure 1-4 , a high-quality lithofacies identification method for storm genesis based on comprehensive identification of multi-source data, including the following steps:

[0041] A: Data acquisition and preprocessing

[0042] S1: Data acquisition and verification, that is, using convolutional neural networks (CNN) to automatically identify the "Bauma-like sequence" features in the core based on cores, field outcrops, well logging curves, and mud logging data, and output storm sediment sample information;

[0043] S2: Automatic lithofacies classification of core images using deep learning (U-Net) combined with sedimentary structures;

[0044] S3: Use time series analysis (LSTM or Transformer) to automatically identify the “box-shaped” (main body of the storm dam) and “bell-shaped, finger-shaped” (side edge of the storm dam) curve features;

[0045] S4: Obtain rock porosity and permeability data through physical property testing;

[0046] S5: Obtain the pore distribution characteristics of the rock through mercury injection testing, including pore radius and mercury removal efficiency;

[0047] S6: Obtain the particle size distribution characteristics of the rock through particle size analysis, including average particle size, sorting coefficient and skewness;

[0048] S7: Obtaining reservoir quality differences among different storm lithofacies through thin section analysis parameters, including surface porosity;

[0049] B: Establishment of a high-quality lithofacies identification parameter system for storm genesis

[0050] S8: Feature importance analysis (random forest) was used to automatically select the optimal seven parameters and establish a high-quality lithofacies identification parameter system for storm deposits, namely, sand-to-ground ratio, gamma value, porosity, compositional maturity, mercury removal efficiency, particle size sorting coefficient, and bioturbation index;

[0051] C: Construction of comprehensive recognition model

[0052] S9: On the basis of assigning values ​​to each parameter, a pairwise comparison matrix of each element in the criterion layer is constructed to determine the relative weight of each criterion element;

[0053] S10: Calculation formula Q for the storm sediment high-quality lithofacies prediction coefficient based on sand-to-sand ratio, gamma value, porosity, compositional maturity, mercury removal efficiency, particle size sorting coefficient, and bioturbation index, i.e., comprehensive score;

[0054] S11: Calculate the comprehensive score and use unsupervised learning (K-means) to classify the high-quality lithofacies.

[0055] In this embodiment, a pairwise comparison matrix of each factor is constructed to determine the relative weight of each criterion factor.

[0056] In this embodiment, the parameters are calculated fully automatically, and the product Ai of the elements in the i-th row of the matrix and its fourth root ai are output.

[0057] In this embodiment, the calculation formula of the weight corresponding to the output parameter is:

[0058]

[0059] In this embodiment, the weights corresponding to the output sand-to-land ratio are ε1=0.34; the weights corresponding to the output gamma value are ε2=0.24, and the weight corresponding to the output porosity is ε3=0.24; the weight corresponding to the output component maturity is ε4=0.072, the weight corresponding to the output mercury removal efficiency is ε5=0.072, the weight corresponding to the output particle size sorting coefficient is ε6=0.021, and the weight corresponding to the output bioturbation index is ε7=0.015.

[0060] In this embodiment, the hierarchical analysis method is used in S8 to first assign a scale to each parameter. On the basis of assigning values ​​to each parameter, weight calculation is performed, i.e., sand-to-ground ratio (SR)=4, gamma value (GR)=4, porosity —4, composition maturity (R)—2, mercury removal efficiency (Hg)—2, particle size sorting coefficient (Sc)—1, bioturbation index (Bio)—1 / 2.

[0061] In this embodiment, the formula for calculating the storm sediment high-quality lithofacies prediction coefficient based on the input of seven parameters, namely, sand-to-ground ratio, gamma value, porosity, compositional maturity, mercury removal efficiency, particle size sorting coefficient, and bioturbation index, is as follows:

[0062]

[0063] In this embodiment, based on the Q value and the comprehensive evaluation standard for fine-grained sedimentary rock reservoirs, combined with unsupervised learning automatic classification and rating, the storm deposit dominant lithofacies evaluation level is divided into three levels, that is, Q>50 is dominant lithofacies, Q between 40-50 is moderate lithofacies, and Q<40 is poor lithofacies.

[0064] In this embodiment, the construction of the comprehensive identification model also includes AI model verification and optimization, that is, inputting 8 sections of samples from the same study area to verify the accuracy of the high-quality lithofacies comprehensive scoring model.

[0065] Example 2

[0066] Reference Figure 1-4 , a high-quality lithofacies identification method for storm genesis based on comprehensive identification of multi-source data, including the following steps:

[0067] Step S1: Data input, i.e., based on well logging curves, mud logging data, core scanning images, and high-definition photos of field outcrops (UAV aerial photography + LiDAR 3D modeling), a convolutional neural network (CNN) is used to automatically identify the "Bauma-like sequence" features in the core and output storm sediment sample information;

[0068] Step S2: Automatically classify the core images into lithofacies using deep learning (U-Net) combined with sedimentary structures;

[0069] Step S3: Time series analysis (LSTM) is used to automatically identify the “box-shaped” (main body of the storm dam) and “bell-shaped, finger-shaped” (side edge of the storm dam) curve features;

[0070] Step S4: Data input, i.e., high-quality lithofacies identification parameters such as sand-to-sand ratio, porosity, permeability, pore size, mercury removal efficiency, average particle size, sorting coefficient, composition maturity, surface porosity, and bioturbation density;

[0071] Step S5: Using feature importance analysis (random forest) to automatically screen the optimal seven parameters and establish a high-quality lithofacies identification parameter system for storm deposits;

[0072] Step 6: Using the hierarchical analysis method, first assign a corresponding scale to each parameter. On the basis of assigning values ​​to each parameter, a hierarchical relationship matrix is ​​established to calculate the weights.

[0073] Step 7: Automatically generate the calculation formula Q for the storm sedimentary lithofacies comprehensive rating prediction coefficient;

[0074] Step 8: Use batch data processing (Python automated script) to calculate the Q value of all 6-segment samples, and combine unsupervised learning (K-means) to automatically divide the ratings;

[0075] Step 9: AI model verification and optimization, that is, inputting 8 sections of samples from the same study area to verify the accuracy of the high-quality lithofacies comprehensive scoring model.

[0076] The specific process of the hierarchical analysis method:

[0077] Process 1: First, input seven evaluation indicators, namely sand-to-sand ratio, gamma value, porosity, compositional maturity, mercury removal efficiency, particle size sorting coefficient, and bioturbation index. Automatic scaling optimization (Bayesian optimization) adjusts the initial scale of each parameter and generates the scale corresponding to each indicator, namely sand-to-sand ratio (SR) -4, gamma value (GR) -4, porosity —4, composition maturity (R)—2, mercury removal efficiency (Hg)—2, particle size sorting coefficient (Sc)—1, bioturbation index (Bio)—1 / 2;

[0078] Process 2: Construct a pairwise comparison matrix of each factor and determine the relative weight of each criterion factor;

[0079]

[0080]

[0081] in, Indicates the importance of the parameter to the parameter

[0082] Process 3: Fully automatic parameter calculation, outputting the product Ai of the elements in the i-th row of the matrix and its fourth root ai:

[0083] A1=512;

[0084] A2=128;

[0085] A3=128;

[0086] A4=1;

[0087] A5=1;

[0088] A6=1 / 128;

[0089] A7=1 / 16384;

[0090] Process 4: Output the weights corresponding to the parameters according to formula 1:

[0091]

[0092] The weights corresponding to the output sand-to-soil ratio are ε1 = 0.34; the weights corresponding to the output gamma value are ε2 = 0.24, the weight corresponding to the output porosity is ε3 = 0.24; the weight corresponding to the output composition maturity is ε4 = 0.072, the weight corresponding to the output mercury removal efficiency is ε5 = 0.072, the weight corresponding to the output particle size sorting coefficient is ε6 = 0.021, and the weight corresponding to the output bioturbation index is ε7 = 0.015;

[0093] The formula for calculating the high-quality lithofacies prediction coefficient of storm deposits based on the input of seven parameters, including sand-to-sand ratio, gamma value, porosity, compositional maturity, mercury removal efficiency, particle size sorting coefficient, and bioturbation index, is:

[0094]

[0095] Process 5: Input sandstone samples of various lithofacies from the six sections of the study area, namely massive bedding sandstone, massive bedding sandstone with mudstone tear clasts, hummocky cross-bedding sandstone, wavy cross-bedding sandstone, wave-formed ripple-bedding sandstone, deformed bedding sandstone, and parallel-bedding sandstone. Take multiple samples for each lithofacies and input the sand-to-ground ratio, gamma value, porosity, compositional maturity, mercury removal efficiency, grain size sorting coefficient, and bioturbation index.

[0096] Process 6: Input the identification index into the prediction coefficient formula to obtain different calculation results, average the prediction coefficients of each lithofacies to obtain Q 块状 =57.73;Q 撕裂屑块状 =54.95;Q 丘状 =51.2;Q 液状 =46.1;Q 浪成 =44.3;Q 平形 =28.9;Q 变形 =37.3.

[0097] Process 7: Based on the Q value, the comprehensive evaluation standard for fine-grained sedimentary rock reservoirs is combined with unsupervised learning (K-means) to automatically classify and rank the storm sedimentary dominant lithofacies into three levels: Q > 50 for dominant lithofacies, Q between 40 and 50 for moderate lithofacies, and Q < 40 for poor lithofacies.

[0098] Process 8: Sandstone samples from the Chang 8 section of the same study area were used for validation (Table 2). The accuracy of the high-quality lithofacies comprehensive scoring model was verified, and the model generalization ability was evaluated (if the accuracy rate was >80%, it passed the validation).

[0099] Process 9: Average the prediction coefficients of each lithofacies to obtain Q 块状 =57.29;Q 撕裂屑块状 =53.71;Q 丘状 =49.82;Q 液状=45.96;Q 浪成 =44.37;Q 平形 =26.82;Q 变形 =38.23, and the accuracy rate after generalization ability evaluation is 85.7%, which passed the verification.

[0100] The present invention improves the accuracy and reliability of high-quality lithofacies identification of storm genesis by integrating multi-source data such as sand-to-sand ratio, gamma value, porosity, composition maturity, mercury removal efficiency, particle size sorting coefficient, and bioturbation index. A systematic high-quality lithofacies identification model is established, providing a scientific basis for the evaluation of oil and gas reservoirs in storm sedimentary environments. The method is simple and easy to implement and is applicable to the high-quality lithofacies identification of different types of reservoirs.

[0101] The above is a detailed introduction to the storm-causing high-quality lithofacies identification method based on comprehensive identification of multi-source data provided by the present invention. This article uses specific examples to illustrate the principles and implementation methods of the present invention. The description of the above examples is only intended to help understand the method and core concept of the present invention. It should be pointed out that for those skilled in the art, without departing from the principles of the present invention, various improvements and modifications can be made to the present invention, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. A high-quality lithofacies identification method for storm genesis based on comprehensive identification of multi-source data, characterized by: The following steps are involved: A: Data acquisition and preprocessing S1: Data acquisition and verification, that is, using convolutional neural networks to automatically identify "Bauma-like sequence" features in cores based on cores, field outcrops, well logging curves, and mud logging data, and output storm sediment sample information; S2: Automatic lithofacies classification of core images using deep learning combined with sedimentary structures; S3: Automatically identify "box-shaped" and "bell-shaped, finger-shaped" curve features using time series analysis; S4: Obtain rock porosity and permeability data through physical property testing; S5: Obtain the pore distribution characteristics of the rock through mercury injection testing, including pore radius and mercury removal efficiency; S6: Obtain the particle size distribution characteristics of the rock through particle size analysis, including average particle size, sorting coefficient and skewness; S7: Obtaining reservoir quality differences among different storm lithofacies through thin section analysis parameters, including surface porosity; B: Establishment of a high-quality lithofacies identification parameter system for storm genesis S8: Using feature importance analysis to automatically select the optimal seven parameters, a high-quality lithofacies identification parameter system for storm deposits was established, namely, sand-to-sand ratio, gamma value, porosity, compositional maturity, mercury removal efficiency, particle size sorting coefficient, and bioturbation index; C: Construction of comprehensive recognition model S9: On the basis of assigning values ​​to each parameter, a pairwise comparison matrix of each element in the criterion layer is constructed to determine the relative weight of each criterion element; S10: Calculation formula Q for the storm sediment high-quality lithofacies prediction coefficient based on sand-to-sand ratio, gamma value, porosity, compositional maturity, mercury removal efficiency, particle size sorting coefficient, and bioturbation index, i.e., comprehensive score; S11: Calculate the comprehensive score and use unsupervised learning to classify high-quality lithofacies.

2. The method for identifying high-quality lithofacies of storm genesis based on comprehensive identification of multi-source data according to claim 1 is characterized in that: Construct a pairwise comparison matrix of each factor and determine the relative weight of each criterion factor.

3. The storm-causing high-quality lithofacies identification method based on comprehensive identification of multi-source data according to claim 1 is characterized in that: Fully automatic parameter calculation, outputting the product Ai of the elements in the i-th row of the matrix and its fourth root ai.

4. The method for identifying high-quality lithofacies of storm genesis based on comprehensive identification of multi-source data according to claim 1 is characterized in that: The calculation formula for the weight corresponding to the output parameter is:

5. The method for identifying high-quality lithofacies of storm genesis based on comprehensive identification of multi-source data according to claim 4 is characterized in that: The weights corresponding to the output sand-to-soil ratio are ε1 = 0.34; the weights corresponding to the output gamma value are ε2 = 0.24, and the weight corresponding to the output porosity is ε3 = 0.24; the weight corresponding to the output compositional maturity is ε4 = 0.072, the weight corresponding to the output mercury removal efficiency is ε5 = 0.072, the weight corresponding to the output particle size sorting coefficient is ε6 = 0.021, and the weight corresponding to the output bioturbation index is ε7 = 0.

015.

6. The method for identifying high-quality lithofacies of storm genesis based on comprehensive identification of multi-source data according to claim 5, characterized in that: In S8, the analytic hierarchy process is used to assign a scale to each parameter. On the basis of assigning values ​​to each parameter, weight calculation is performed, i.e., sand-to-ground ratio (SR) = 4, gamma value (GR) = 4, porosity —4, composition maturity (R)—2, mercury removal efficiency (Hg)—2, particle size sorting coefficient (Sc)—1, bioturbation index (Bio)—1 / 2.

7. The method for identifying high-quality lithofacies of storm genesis based on comprehensive identification of multi-source data according to claim 1 is characterized in that: The formula for calculating the high-quality lithofacies prediction coefficient of storm deposits based on the input of seven parameters, including sand-to-sand ratio, gamma value, porosity, compositional maturity, mercury removal efficiency, particle size sorting coefficient, and bioturbation index, is:

8. The method for identifying high-quality lithofacies of storm genesis based on comprehensive identification of multi-source data according to claim 1 is characterized in that: According to the size of the Q value, based on the comprehensive evaluation standard of fine-grained sedimentary rock reservoirs and combined with unsupervised learning automatic classification and rating, the evaluation level of storm deposit dominant lithofacies is divided into three levels, that is, Q>50 is dominant lithofacies, Q between 40-50 is moderate lithofacies, and Q<40 is poor lithofacies.

9. The method for identifying high-quality lithofacies of storm genesis based on comprehensive identification of multi-source data according to claim 1, characterized in that: The construction of the comprehensive identification model also includes AI model verification and optimization, that is, inputting 8 sections of samples from the same study area to verify the accuracy of the high-quality lithofacies comprehensive scoring model.

Images