CN115267937A - Braided river delta sand body configuration prediction method and system under offshore thin well pattern condition - Google Patents

Abstract

The invention provides a braided river delta sand body configuration prediction method and a braided river delta sand body configuration prediction system under an offshore thin well pattern condition, belonging to the technical field of oil and gas field exploitation geological research, wherein the method comprises the following steps: building a braided river delta prototype model; establishing a braided river delta configuration mode and a braided river delta configuration unit parameter library; calculating the thickness of the sand body of the target layer based on the well data; extracting seismic attributes of the target layer, and preferably selecting the seismic attributes capable of reflecting the sand body thickness; establishing a learning model between the thickness of the sand body explained by logging and the seismic attribute data preferably selected near the well point; carrying out intelligent well-to-seismic combined sand body prediction on various types of sand bodies with different stacking relations; and completing sand body configuration prediction. The system comprises a sand body configuration parameter base module, a data preparation module, a seismic attribute selection module, a machine learning regression model module, a colored inversion module and a sand body configuration prediction module. The invention improves the interpretation precision and reliability and overcomes the problem of colored inversion multi-solution.

Description

Braided river delta sand body configuration prediction method and system under offshore thin well pattern condition

Technical Field

The invention belongs to the technical field of oil and gas field exploitation geological research, and particularly relates to a braided river delta sand body configuration prediction method and a braided river delta sand body configuration prediction system under an offshore thin well pattern condition.

Background

The sand bodies in the delta phase of the braided river have extremely complex contact relationship in space, so that the reservoir has high heterogeneity, and the sand body reservoirs which are seemingly adjacently connected on the plane are actually blocked and disconnected by the deposition discontinuities. With the gradual and deep research on the sand body configuration, the prediction of the sand body configuration of the reservoir layer becomes more and more refined, and how to improve the prediction precision of the sand body configuration is always a hot point problem of geological interpretation work.

The prediction of the braided river delta sand body configuration in the current stage is mainly carried out in a land dense well pattern area, well logging data is taken as a core, and seismic attributes are taken as conditions of optimization constraints, so that the purpose of predicting the reservoir configuration is achieved. In the aspect of high-precision prediction of seismic attributes, on one hand, new more reasonable seismic attributes are searched for through a mathematical tool; on the other hand, different seismic attributes are adopted for fusion, some seismic attributes are selected according to personal experience, and some seismic attributes are selected through a specific algorithm, so that the purpose is to improve the correlation and simultaneously reserve effective seismic information as much as possible to obtain reliable, comprehensive and complete sand body configuration characteristic fusion seismic attributes.

Many researches show that the seismic response of the sand thickness of the target layer is related to seismic frequency, the resolution of high-frequency information is high, but the tuning thickness is small, and the method is suitable for predicting the thin sand body; the low-frequency information has large tuning thickness but low resolution and is suitable for predicting the thick sand body. When the sand thickness is greater than the tuning thickness, the amplitude and frequency information has obvious ambiguity. In addition, when the seismic data resolution is low, the seismic attribute distribution characteristics of the target layer are also related to the surrounding rock seismic attributes, for example, the high value of the amplitude attribute of the medium amplitude in the target layer can be the sand body of the medium thickness in the target layer, and can also be the response of the thick sand body of the adjacent stratum. Therefore, the sand body distribution prediction without considering frequency division and surrounding rock interference often causes a large error of a prediction result.

For seismic inversion, in practice of a dense well pattern area, model-based inversion is usually performed, and then, the distribution characteristics of sand bodies of a target stratum are finely described by combining rich well data. However, the conventional method is difficult to be applied to an offshore thin well pattern work area, on one hand, because the thin well pattern work area has less well data, and a large amount of researches show that the transverse heterogeneity of the sand body distribution of the braided river delta phase is strong, and the thickness difference of different single-layer sand bodies is also large, the thin-layer sand body is difficult to be finely carved by the model-based inversion and frequency division inversion in the prior art. The colored inversion has the characteristics of less logging data participation and faithful result to seismic data, responds to thin-layer sand bodies, has continuous inversion results, and can well complete the seismic inversion task under the condition of a thin well pattern. However, the result of the chromatic inversion also has multiple solutions, so that guidance needs to be performed according to the sand body distribution characteristic mode of the actual work area.

In the past research work, the reservoir sand body configuration is mostly predicted by depending on the logging data under the condition of a dense well pattern, and the configuration guidance of a deposition mode is rarely considered, so that the sand body configuration prediction is difficult to accurately perform under the condition of an offshore thin well pattern lacking logging data, the obvious multi-solution performance is realized, and the distribution of the reservoir sand body cannot be accurately predicted. Therefore, in addition to comprehensively utilizing various seismic attributes and logging curve data as much as possible, a machine learning method can be adopted, seismic data of different frequency bands and seismic attributes of surrounding rocks and a target layer are comprehensively considered, and a sand body configuration prototype parameter library is established to screen sand body prediction results, so that the multi-solution property can be effectively reduced, and the precision is improved.

Disclosure of Invention

The invention aims to provide a braided river delta sand body configuration prediction method and a braided river delta sand body configuration prediction system under the condition of an offshore open well pattern, which can solve the problems.

In order to achieve the purpose, the technical scheme of the invention is realized as follows: the braided river delta sand body configuration prediction method under the condition of the offshore open-pit pattern comprises the following steps:

s1: identifying sand bodies and mudstone interlayers in a prototype model area, carrying out omnibearing dissection on the sand body configuration in a reservoir stratum, analyzing the internal characteristics of a configuration unit, and measuring the geometrical parameter characteristics of the configuration unit so as to establish a braided river delta prototype model;

s2: taking a braided river delta prototype model as guidance, and establishing a braided river delta configuration mode and a braided river delta configuration unit parameter library in a research area by combining typical block dense well pattern anatomical data of the research area;

s3: establishing a high-precision stratum-structure framework in a research area according to the rock core, the well logging and the seismic data;

s4: calculating the thickness of a sand body of a target layer according to the core and the logging data;

s5: extracting seismic attributes of a target layer of a research area, and preferably selecting the seismic attributes capable of reflecting the sand thickness by analyzing the correlation between the seismic attributes and the well logging explained sand thickness;

s6: taking the well logging interpretation sand thickness of a target layer as a supervision data set, and carrying out machine learning between the well logging interpretation sand thickness and seismic attribute data preferably selected near a well point by adopting a supervision learning method so as to establish a learning model between the well logging interpretation sand thickness and the seismic attribute data;

s7: fusing the optimized seismic attributes by using the trained learning model, wherein the optimized frequency division seismic attributes are fused to form a frequency division multi-attribute intelligent fusion method, and the result is a frequency division intelligent fusion attribute; fusing the seismic attributes of the target layer and the surrounding rock layer, namely, an intelligent attribute fusion method for reducing surrounding rock interference, wherein the result is an intelligent fusion attribute for reducing surrounding rock interference;

s8: carrying out intelligent well-to-seismic sand body prediction on various types of sand bodies with different stacking relations based on a frequency division multi-attribute intelligent fusion method, an attribute intelligent fusion method for reducing surrounding rock interference and a colored inversion method;

s9: and on the basis of intelligent well earthquake combined sand body prediction, according to the established braided river delta configuration unit parameter library, constraining the prediction result, screening out the most reasonable prediction result, and completing sand body configuration prediction.

Further, step S1 includes:

s11: the morphological characteristics, lithofacies combination and quantitative parameters of the braided river delta configuration unit with high precision are obtained through instrument and field investigation;

s12: and summarizing the information in the S11, and extracting a braided river delta prototype model of the braided river delta according to types.

Further, the braided river delta configuration unit parameter library comprises a lithofacies library, a morphological structure library and a scale library.

Further, the braided river delta configuration mode and the braided river delta configuration unit parameter library comprise lithofacies combination, morphological structures and empirical formulas.

Further, step S3 includes:

s31: performing time depth calibration by using a synthetic seismic record method;

s32: and establishing an isochronous stratigraphic framework of the research area according to the horizon interpretation data in the seismic data and the hierarchical data provided by the logging data.

Further, step S4 includes:

s41: the sand body thickness calculating method is characterized in that the characteristics of a natural potential logging curve and a gamma logging curve are utilized, the lithofacies characteristics provided by rock core data are combined, single-well sand body interpretation is carried out, and the sand body thickness of a target layer is calculated.

Further, step S5 includes: and performing correlation analysis between the well logging explained sand body thickness and the seismic attribute data, and selecting the seismic attribute with the correlation coefficient higher than 0.5.

Further, if the preferred seismic attribute is a frequency division seismic attribute, a frequency division multi-attribute intelligent fusion method is adopted to fuse the trained supervised learning model with multiple frequency division attributes; and if the preferred seismic attributes comprise the seismic attributes of the surrounding rock stratum, adopting an attribute intelligent fusion method for reducing the surrounding rock interference to fuse the seismic attributes of the target layer and the surrounding rock stratum of the trained supervised learning model.

Furthermore, the colored inversion method comprises the steps of carrying out spectrum analysis on single-well wave impedance and seismic wave impedance, fitting a corresponding energy spectrum curve, setting a matching operator in a frequency domain, matching the single-well wave impedance spectrum curve with the seismic wave impedance spectrum curve, returning to a time domain, and applying the matching operator to seismic data for inversion.

The braided river delta sand body configuration prediction system under the condition of the offshore thin well pattern comprises a sand body configuration parameter library module, a data preparation module, a seismic attribute selection module, a machine learning regression model module, a colored inversion module and a sand body configuration prediction module;

the data preparation module is used for extracting sand thickness of logging interpretation of a target layer and seismic attribute data of the target layer, and the seismic attribute selection module is used for carrying out correlation analysis on the sand thickness of logging interpretation extracted by the data preparation module and the seismic attribute data of the target layer;

the machine learning regression model module is used for obtaining a trained learning model, the sand body configuration prediction module is used for fusing the seismic attributes selected by the seismic attribute selection module according to the learning model obtained by the machine learning regression model module, preliminarily predicting sand body distribution by combining the colored inversion results of the colored inversion module, and constraining the prediction result of the sand body configuration prediction module according to a braided river delta sand body configuration unit prototype parameter library established by the sand body configuration parameter library module to screen out the most reasonable prediction result.

Compared with the prior art, the braided river delta sand body configuration prediction method and the braided river delta sand body configuration prediction system under the offshore open-pit pattern condition have the following advantages:

(1) The invention establishes a braided river delta facies sand body configuration unit parameter library based on various characteristics of actual and typical braided river delta facies, and has guiding significance for actual configuration anatomy;

(2) A machine learning algorithm is introduced to intelligently fuse the seismic attributes, a nonlinear learning model is established, the complex relation between the seismic attributes and geological parameters can be represented in a refined manner, and the method is more objective than manual operation;

(3) The method has corresponding prediction methods for target layers under different conditions, comprehensively considers the interference of surrounding rocks on the seismic attributes of the target layers, improves the interpretation precision and reliability, and has universality in actual production;

(4) The scarcity of logging data under the condition of a thin well pattern is considered, a colored inversion method with strong objectivity is used for seismic data inversion, and in order to further reduce the multi-solution property of seismic interpretation, the method also restrains the prediction result through the established braided river delta configuration unit parameter library, so that the problem of the multi-solution property of colored inversion is solved;

(5) The configuration plan and the profile obtained by the method are more visual, and the configuration of the sand body can be quantitatively represented.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate an embodiment of the invention and, together with the description, serve to explain the invention and not to limit the invention. In the drawings:

FIG. 1 is a schematic overall flow diagram of the present invention;

FIG. 2 is a schematic illustration of a parameter bank of a braided river delta phase sand body configuration unit according to an embodiment of the invention;

fig. 3 is a schematic flow chart of a frequency division multiple attribute intelligent fusion method according to an embodiment of the present invention;

fig. 4 is a schematic flow chart of an attribute intelligent fusion method for reducing surrounding rock interference according to an embodiment of the present invention.

Detailed Description

It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict.

Because the deposition process of the braided river delta is complex, and the contact relation between sand body configuration units is also very complex, the braided river delta sand body configuration unit parameter library is established for mode guidance through a laser scanner, a ground penetrating radar, a modern deposition satellite photo and typical block dense well network anatomical data, the stratum-structure framework under the sparse well network condition is accurately divided by combining a rock core, a logging and seismic data, and then a machine learning method is adopted to establish the nonlinear mapping relation between the sand body thickness and the target layer and the surrounding seismic attribute data according to the accurate division. The invention is described in detail below with reference to the figures and examples.

The invention provides a braided river delta sand body configuration prediction method and a braided river delta sand body configuration prediction system under an offshore thin well pattern condition. As shown in fig. 1, a research on a braided river delta prototype configuration mode is carried out by integrating ground penetrating radar data, laser scanning data and modern sedimentary satellite photos with a dense well network and anatomical data, and a braided river delta configuration unit parameter library is established, which includes a lithofacies library, a morphological structure library and a scale library. Establishing corresponding mode guidance according to a braided river delta sand body configuration parameter library, obtaining seismic response characteristic charts of different types of sand body superposition relations by using intelligent attribute fusion and colored inversion based on the thickness of a target layer well logging explained sand body, and completing sand body distribution and sand body superposition relation prediction under the mode guidance. The embodiment comprises the following steps:

s1: identifying sand bodies and mudstone interlayers in a prototype model area, carrying out omnibearing dissection on the sand body configuration in a reservoir stratum, analyzing the internal characteristics of a configuration unit, and measuring the geometrical parameter characteristics of the configuration unit so as to establish a braided river delta prototype model; specifically, information such as morphological characteristics, lithofacies combination, quantitative scale and the like of various configuration units of the braided river delta with high precision is acquired through various instruments and equipment and field investigation, and the lithofacies: the rock color, composition, structure, archaea contained in the structure, original occurrence of sediments and the like are limited by core data, and the lithofacies mainly refers to the rock color, composition and structure; quantitative scale: the length, width and height of the configuration unit and the correlation among the parameters are indicated; summarizing all the information, and extracting a braided river delta prototype model of the braided river delta according to types. As shown in fig. 2, the sand body and the mudstone interlayer in the prototype model area are identified through the acquisition, processing, interpretation, attribute analysis and the like of ground penetrating radar data, the response mark of the ground penetrating radar profile of the total structural boundary can complete the all-round anatomy of the sand body configuration in the reservoir by means of a laser scanner, the geometric parameter characteristics of the configuration unit can be visually measured through a satellite map, and then a modern braided river delta three-dimensional configuration model is established; the method comprises the steps of collecting three-dimensional laser data of a pigtail river delta outcrop section in the field, carrying out preprocessing such as splicing, denoising and data reduction on laser scanning data indoors, and establishing a field outcrop digital model through packaging, repairing and mapping; and measuring basic data of modern deposition of the braided river delta by using Google Earth software to perform the configuration research of the braided river delta prototype model.

S2: taking a braided river delta prototype model as guidance, and establishing a braided river delta configuration mode and a braided river delta configuration unit parameter library in a research area by combining typical block dense well pattern anatomical data of the research area (target area); the braided river delta configuration unit parameter library comprises a lithofacies library, a morphological structure library and a scale library. Wherein the lithofacies library comprises lithofacies types and lithofacies combinations; the morphological structure library comprises monomer morphologies and combination relations; the scale library includes absolute scale and relative scale. The braided river delta configuration mode and the braided river delta configuration unit parameter library comprise rock combinations, morphological structures and empirical formulas.

S3: establishing a high-precision stratum-structure framework of a work area according to the rock core, the well logging and the seismic data; specifically, firstly, a synthetic seismic record method (or cross-well VSP velocity data) is used for time-depth calibration, namely a spatial corresponding relation is established between seismic data of a time domain and well data of a depth domain; and secondly, establishing an isochronous stratigraphic framework (namely hierarchical data) of the research area according to the horizon interpretation data in the seismic data and the hierarchical data provided by the logging data.

S4: calculating the thickness of a sand body of a target layer according to the core and the logging data; and (3) performing single-well sand body interpretation by utilizing the characteristics of the natural potential logging curve and the gamma logging curve and combining the lithofacies characteristics provided by the rock core data, and calculating the sand body thickness of the target layer.

S5: extracting seismic attributes of a target layer of a research area, and preferably selecting one or more seismic attributes (the Pearson correlation coefficient is larger than 0.5) capable of reflecting the sand thickness by analyzing the correlation between the seismic attributes and the well logging explained sand thickness; namely, correlation analysis is carried out between the well logging explained sand body thickness and the seismic attribute data, and the seismic attribute with the correlation coefficient higher than 0.5 is selected.

S6: taking the well logging interpretation sand thickness of a target layer as a supervision data set, and carrying out machine learning between the well logging interpretation sand thickness data and seismic attribute data preferably selected near a well point by adopting a supervision learning method (such as a support vector machine, a neural network and the like) so as to establish a learning model (a nonlinear mapping relation) between the well logging interpretation sand thickness data and the seismic attribute data; the vicinity of the well point is a circular area with the well point position as the center, the radius of the circular area can be set according to the actual situation, the circular area is properly adjusted according to the seismic transverse sampling interval, the diameter is generally 25m and is approximately equal to the transverse sampling interval, and then the seismic attribute average value is obtained. And (3) taking the sand body thickness explained by the well data as the supervision data, adopting a supervision learning method, carrying out supervision learning between the well explained sand body thickness data and the optimized seismic attribute, and establishing a supervision learning model.

S7: fusing the optimized seismic attributes by using the trained learning model, wherein the fused optimized frequency division seismic attributes are a frequency division multi-attribute intelligent fusion method, and the result is a frequency division intelligent fusion attribute; and fusing the seismic attributes of the target layer and the surrounding rock layer to obtain the intelligent attribute fusion method for reducing the surrounding rock interference, wherein the result is the intelligent fusion attribute for reducing the surrounding rock interference. Specifically, if the preferred seismic attribute is a frequency division seismic attribute, a frequency division multi-attribute intelligent fusion method is adopted to fuse multiple frequency division attributes of the trained supervised learning model; and if the preferred seismic attributes comprise the seismic attributes of the surrounding rock stratum, adopting an attribute intelligent fusion method for reducing the surrounding rock interference to fuse the seismic attributes of the target layer and the surrounding rock stratum of the trained supervised learning model.

S8: and performing intelligent well-to-seismic sand body prediction on various types of sand bodies with different stacking relations based on a frequency division multi-attribute intelligent fusion method, an attribute intelligent fusion method for reducing surrounding rock interference and a colored inversion method. The intelligent fusion method comprises a frequency division multi-attribute intelligent fusion method and an attribute intelligent fusion method for reducing surrounding rock interference, and a proper intelligent fusion method is selected according to the difference of different horizon development patterns.

The frequency division multi-attribute intelligent fusion method is characterized in that seismic data volumes with different frequencies are selected for sand bodies with different thicknesses, and fusion is carried out by optimizing amplitude and frequency seismic attributes and using a machine learning algorithm, so that the method is suitable for target horizons with large sand body development scale difference.

The intelligent fusion method for reducing the surrounding rock interference extracts the seismic attributes of the amplitude class and the frequency class of the target layer and the upper and lower adjacent layers of the target layer, fuses the seismic attributes through a machine learning algorithm to reduce the surrounding rock interference and reduce the uncertainty of seismic attribute explanation, and is suitable for the target layer where the adjacent layer develops sand bodies and lacks a stable mudstone interlayer.

The colored inversion method mainly comprises the steps of performing spectral analysis on single-well wave impedance and seismic wave impedance, fitting corresponding energy spectrum curves, setting a matching operator in a frequency domain to match the single-well wave impedance spectrum curves with the seismic wave impedance spectrum curves, then converting the single-well wave impedance spectrum curves into a time domain, and applying the matching operator to seismic data to complete colored inversion.

Aiming at a target layer with larger difference of sand body development scale, fusing the optimized frequency division seismic attributes by adopting a frequency division multi-attribute intelligent fusion method, as shown in figure 3, firstly, decomposing an original data body into frequency division data bodies with different central frequency bands by adopting a wavelet frequency division technology, and optimizing low-frequency, medium-frequency and high-frequency data bodies according to a target layer sand body thickness distribution interval, wherein the high frequency is suitable for a thin layer, and the low frequency is suitable for a thick layer; extracting multiple seismic attributes such as amplitude, frequency and phase common to a target interval from an original seismic data volume, performing correlation analysis on the seismic attributes and the sand thickness, and preferably selecting one or more attributes with high correlation (correlation coefficient is more than 0.5) with the sand thickness; setting the average value of each attribute around each well as a training data set, setting the sand thickness of well logging interpretation on a target layer well as a monitoring data set, selecting a Support Vector Machine (SVM) algorithm to perform machine learning between the training data set and the monitoring data set, and establishing a nonlinear regression model between the well logging interpretation sand thickness and the fusion attribute; and finally, obtaining a fusion attribute graph capable of quantitatively representing the thickness plane distribution of the sand body of the target layer by applying the trained regression model.

Aiming at a target horizon where a sand body develops at an adjacent horizon and a stable mudstone interlayer is lacked, adopting an attribute intelligent fusion method for reducing the interference of surrounding rocks to fuse the seismic attributes of the target horizon and a surrounding rock stratum, as shown in figure 4, firstly, extracting a plurality of seismic attributes such as amplitude class, frequency class, phase class and the like common to a target interval from an original seismic data body, carrying out correlation analysis on the seismic attributes and the sand body thickness, and preferably selecting one or more attributes with high correlation (correlation coefficient is more than 0.5) with the sand body thickness; extracting 2-3 kinds of preferable seismic attributes of upper and lower surrounding rocks and a target layer in a stratum slicing mode respectively along a top interface and a bottom interface of the target layer upwards and downwards by adopting a time window with 1/4 wavelength, setting the extracted seismic attributes as a training data set, and setting the thickness of a sand body interpreted by well logging on the target layer as a monitoring data set; selecting a Support Vector Machine (SVM) algorithm to perform machine learning between a training data set and a monitoring data set, and establishing a nonlinear regression model between well logging interpretation sand thickness and fusion attributes; and finally, fusing the seismic attributes of the upper and lower surrounding rocks and the target layer into a new attribute capable of quantitatively reflecting the planar distribution of the thickness of the sand body of the target layer by applying the trained regression model.

In the example, a machine learning algorithm is adopted as a Support Vector Machine (SVM), training data input by a kernel function through a radial basis function can be expressed as (xi, yi), wherein i =1,2,3 \8230, n and n are the number of training samples, xi belongs to input data (seismic attributes), and xi belongs to R; yi is the target data (sand thickness) and is recorded as yi ∈ R.

The support vector machine regression model expression is as follows:

f(x)＝<w，x>+b

wherein w, x belongs to R, b belongs to R, f (x) is the output result after operation, w is the weight value which can change along with gradient descending, w, x represents the dot product of w and x, and b is a constant adjusting factor.

In addition, a genetic neural network, a deep learning algorithm, or the like may be employed as the machine learning algorithm.

And (3) carrying out reliability evaluation on the obtained regression model, wherein the specific method comprises the following steps: if the accuracy of the test set is greater than or equal to 80%, accepting the model; otherwise, the steps are repeated to adjust the initial parameters until the accuracy of the test set is greater than or equal to 80%, and the model is output.

The method adopts a colored inversion method, can overcome the defect of sparse well pattern, and more truly reflects the distribution characteristics of profile sand bodies.

S9: and on the basis of intelligent well earthquake combined sand body prediction, according to the established braided river delta configuration unit parameter library, constraining the prediction result, screening out the most reasonable prediction result, and completing sand body configuration prediction. The braided river delta facies sand body configuration parameter library constraint is that an intelligent attribute fusion result and a colored inversion result are screened through an established configuration prototype parameter library, and a prediction result which meets the requirement of the parameter library and also meets the actual condition is selected.

The braided river delta sand body configuration prediction system under the offshore thin well pattern condition comprises a sand body configuration parameter base module, a data preparation module, a seismic attribute selection module, a machine learning regression model module, a colored inversion module and a sand body configuration prediction module;

the sand body configuration parameter library module is a braided river delta facies characteristic quantitative parameter knowledge base which is built in a plurality of modes and is used for establishing mode guidance;

the data preparation module is used for extracting sand thickness of logging interpretation of a target layer and seismic attribute data of the target layer, and the seismic attribute selection module is used for performing correlation analysis on the sand thickness of logging interpretation extracted by the data preparation module and the seismic attribute data of the target layer and selecting seismic attributes with correlation coefficients higher than 0.5;

the machine learning regression model module takes the sand thickness as a target data set, takes the seismic attribute values of a target stratum and upper and lower adjacent strata as a training sample data set, takes all the target data and the training data as input data, and establishes nonlinear mapping through a machine learning algorithm to obtain a trained learning model;

the inversion module is used for performing spectral analysis on single-well wave impedance and seismic wave impedance, fitting a corresponding energy spectrum curve, setting a matching operator in a frequency domain, matching the single-well wave impedance spectrum curve with the seismic wave impedance spectrum curve, then returning to a time domain, applying the matching operator to seismic data, and completing colored inversion;

the sand body configuration prediction module takes a learning model obtained by the machine learning regression model module as a basis, integrates the seismic attributes preferably selected by the seismic attribute selection module, preliminarily predicts sand body distribution by combining the colored inversion result of the colored inversion module, restrains the prediction result of the sand body configuration prediction module according to a braided river delta sand body configuration unit prototype parameter library established by the sand body configuration parameter library module, analyzes the section distribution characteristics and the superposition relation of the sand body of a target layer, and screens out the prediction result which is most reasonable and can represent the actual sand body distribution characteristics, thereby completing sand body configuration prediction.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and should not be taken as limiting the scope of the present invention, which is intended to cover any modifications, equivalents, improvements, etc. within the spirit and scope of the present invention.

Claims (10)

1. The braided river delta sand body configuration prediction method under the condition of the offshore thin well pattern is characterized by comprising the following steps of:

s1: identifying sand bodies and mudstone interlayers in a prototype model area, carrying out omnibearing dissection on the sand body configuration in a reservoir stratum, analyzing the internal characteristics of a configuration unit, and measuring the geometrical parameter characteristics of the configuration unit so as to establish a braided river delta prototype model;

s2: taking a braided river delta prototype model as guidance, and establishing a braided river delta configuration mode and a braided river delta configuration unit parameter library in a research area by combining typical block dense well pattern anatomical data of the research area;

s3: establishing a high-precision stratum-structure framework in a research area according to the rock core, the well logging and the seismic data;

s4: calculating the thickness of a sand body of a target layer according to the core and the logging data;

s5: extracting seismic attributes of a target layer of a research area, and preferably selecting the seismic attributes capable of reflecting the sand thickness by analyzing the correlation between the seismic attributes and the well logging explained sand thickness;

s6: taking the well logging interpretation sand thickness of a target layer as a supervision data set, and carrying out machine learning between the well logging interpretation sand thickness and seismic attribute data preferably selected near a well point by adopting a supervision learning method so as to establish a learning model between the well logging interpretation sand thickness and the seismic attribute data;

s7: fusing the optimized seismic attributes by using the trained learning model, wherein the fused optimized frequency division seismic attributes are a frequency division multi-attribute intelligent fusion method, and the result is a frequency division intelligent fusion attribute; fusing seismic attributes of the target layer and the surrounding rock layer to obtain an attribute intelligent fusion method for reducing surrounding rock interference, wherein the result is an intelligent fusion attribute for reducing surrounding rock interference;

s8: carrying out intelligent well-to-seismic sand body prediction on various types of sand bodies with different stacking relations based on a frequency division multi-attribute intelligent fusion method, an attribute intelligent fusion method for reducing surrounding rock interference and a colored inversion method;

s9: and on the basis of intelligent well earthquake combined sand body prediction, according to the established braided river delta configuration unit parameter library, constraining the prediction result, screening out the most reasonable prediction result, and completing sand body configuration prediction.

2. The method for predicting the braided river delta sand body configuration under the offshore open-work condition according to claim 1, wherein the step S1 comprises:

s11: the morphological characteristics, lithofacies combination and quantitative parameters of the braided river delta configuration unit with high precision are obtained through instrument and field investigation;

s12: and summarizing the information in the S11, and extracting a braided river delta prototype model of the braided river delta according to types.

3. The method for predicting the braided river delta sand body configuration under the offshore open-work condition according to claim 1, wherein the method comprises the following steps: the braided river delta configuration unit parameter library comprises a lithofacies library, a morphological structure library and a scale library.

4. The method for predicting the braided river delta sand body configuration under the offshore open-work condition according to claim 1, wherein the method comprises the following steps: the braided river delta configuration mode and the braided river delta configuration unit parameter library comprise lithofacies combination, morphological structures and empirical formulas.

5. The method for predicting the braided river delta sand body configuration under the offshore open-work condition according to claim 1, wherein the step S3 comprises:

s31: time depth calibration is carried out by using a synthetic seismic record method;

s32: and establishing an isochronous stratigraphic framework of the research area according to the horizon interpretation data in the seismic data and the hierarchical data provided by the logging data.

6. The method for predicting the braided river delta sand body configuration under the offshore open-work condition according to claim 1, wherein the step S4 comprises:

s41: the sand body thickness calculating method is to utilize the characteristics of the natural potential well logging curve and the gamma well logging curve and combine the lithofacies characteristics provided by the core data to carry out single-well sand body interpretation and calculate the sand body thickness of the target layer.

7. The method for predicting the braided river delta sand body configuration under the offshore open-work condition according to claim 1, wherein the step S5 comprises: and performing correlation analysis between the well logging explained sand body thickness and the seismic attribute data, and selecting the seismic attribute with the correlation coefficient higher than 0.5.

8. The method for predicting the braided river delta sand body configuration under the offshore open-work condition according to claim 1, wherein the method comprises the following steps: if the preferred seismic attribute is a frequency division seismic attribute, performing multiple frequency division attribute fusion on the trained supervised learning model by adopting a frequency division multiple attribute intelligent fusion method; and if the preferred seismic attributes comprise the seismic attributes of the surrounding rock stratum, adopting an attribute intelligent fusion method for reducing the surrounding rock interference to fuse the seismic attributes of the target layer and the surrounding rock stratum of the trained supervised learning model.

9. The method for predicting the braided river delta sand body configuration under the offshore open-work condition according to claim 1, wherein the method comprises the following steps: the colored inversion method comprises the steps of performing spectral analysis on single-well wave impedance and seismic wave impedance, fitting corresponding energy spectrum curves, setting a matching operator in a frequency domain to match the single-well wave impedance spectrum curves with the seismic wave impedance spectrum curves, then converting the single-well wave impedance spectrum curves into a time domain, and applying the matching operator to seismic data for inversion.

10. Braided river delta sand body configuration prediction system under marine thin-well pattern condition its characterized in that: the system comprises a sand body configuration parameter library module, a data preparation module, a seismic attribute selection module, a machine learning regression model module, a colored inversion module and a sand body configuration prediction module;

the data preparation module is used for extracting sand thickness of logging interpretation of a target layer and seismic attribute data of the target layer, and the seismic attribute selection module is used for carrying out correlation analysis on the sand thickness of logging interpretation extracted by the data preparation module and the seismic attribute data of the target layer; the machine learning regression model module is used for obtaining a trained learning model, the sand body configuration prediction module is used for fusing the seismic attributes selected by the seismic attribute selection module according to the learning model obtained by the machine learning regression model module, preliminarily predicting sand body distribution by combining the colored inversion results of the colored inversion module, and constraining the prediction result of the sand body configuration prediction module according to a braided river delta sand body configuration unit prototype parameter library established by the sand body configuration parameter library module to screen out the most reasonable prediction result.

Images