CN115373028A - Method for reconstructing logging curve of voiding lost circulation well section of fractured-vuggy carbonate reservoir - Google Patents
Method for reconstructing logging curve of voiding lost circulation well section of fractured-vuggy carbonate reservoir Download PDFInfo
- Publication number
- CN115373028A CN115373028A CN202211127777.5A CN202211127777A CN115373028A CN 115373028 A CN115373028 A CN 115373028A CN 202211127777 A CN202211127777 A CN 202211127777A CN 115373028 A CN115373028 A CN 115373028A
- Authority
- CN
- China
- Prior art keywords
- seismic
- attribute
- reservoir
- model
- seismic attribute
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/40—Seismology; Seismic or acoustic prospecting or detecting specially adapted for well-logging
- G01V1/44—Seismology; Seismic or acoustic prospecting or detecting specially adapted for well-logging using generators and receivers in the same well
- G01V1/46—Data acquisition
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/40—Seismology; Seismic or acoustic prospecting or detecting specially adapted for well-logging
- G01V1/44—Seismology; Seismic or acoustic prospecting or detecting specially adapted for well-logging using generators and receivers in the same well
- G01V1/48—Processing data
- G01V1/50—Analysing data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/60—Analysis
- G01V2210/61—Analysis by combining or comparing a seismic data set with other data
- G01V2210/616—Data from specific type of measurement
- G01V2210/6169—Data from specific type of measurement using well-logging
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/60—Analysis
- G01V2210/62—Physical property of subsurface
- G01V2210/624—Reservoir parameters
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/60—Analysis
- G01V2210/62—Physical property of subsurface
- G01V2210/624—Reservoir parameters
- G01V2210/6244—Porosity
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/60—Analysis
- G01V2210/62—Physical property of subsurface
- G01V2210/624—Reservoir parameters
- G01V2210/6246—Permeability
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/10—Numerical modelling
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Remote Sensing (AREA)
- Evolutionary Computation (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Geology (AREA)
- Geophysics (AREA)
- Environmental & Geological Engineering (AREA)
- General Engineering & Computer Science (AREA)
- Acoustics & Sound (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention relates to a method for reconstructing a logging curve of a voiding lost circulation well section of a fractured-vuggy carbonate reservoir, which comprises the following steps of: firstly, resampling a seismic inversion attribute body; establishing a seismic attribute parameter geological model, and then extracting the seismic attribute parameter curve from all the drilled wells in the target stratum along a shaft; then, taking key wells of all the acquired complete curves of the target layer as sample data, and constructing a plurality of deep neural network models of hidden layers; and training the neural network model through multiple iterations to establish a nonlinear mapping relation between the seismic attribute parameter curve and the logging curve, and finally, generating a missing logging curve of the target well by using the trained neural network model with the seismic attribute parameter of the logging curve-free well drilling as input. The method can effectively solve the problem of basic information loss of the leakage-loss air-defense section of the fracture-cave type oil reservoir.
Description
Technical Field
The invention relates to the technical field of production and development of fractured-vuggy carbonate reservoirs, in particular to a method for reconstructing a logging curve of a voided lost circulation section of a fractured-vuggy carbonate reservoir.
Background
The existing carbonate fracture-cave oil reservoir is in the forefront in the reserves and the yields of oil and gas in the world, and is a research object with high attention at home and abroad. According to statistics, more than 30% of the global carbonate reservoirs are fracture-cavity reservoirs. Large areas of carbonate fracture-cavity oil reservoirs are distributed in the west of China, have great production value and are important places for increasing, storing and increasing the petroleum yield in China. The fracture-cavity type oil reservoir forms a main reservoir space with erosion cavities and cracks under the combined action of multi-stage tectonic movement and ancient karst, and due to the characteristics of multi-reservoir space, multi-flow mode coexistence and the like of the fracture-cavity type oil reservoir, problems often occur in the oil field drilling process, wherein drilling to an emptying zone is a typical problem. When the well drilling is transferred to an emptying zone, open hole well completion is often adopted, the well logging data and geological data are usually absent, and the permeability of the producing hole near the well can not be mastered in the oil field development process, so that reasonable production data can not be determined. Therefore, the acquisition of the logging curve and geological data of the voided well of the fractured-vuggy reservoir is very important, and related geological factors can be accurately analyzed only by fully mastering the logging data, so that the production practice is well guided.
At present, the acquisition modes of the related well drilling logging curves are various, and conventional empirical prediction, analog logging technology, digital logging technology, imaging logging technology and the like exist, but the technology has a certain technical barrier for acquiring the logging curves of the fracture-cavity oil reservoir emptying well, so that the deviation of the prediction result is large.
For example, conventional empirical methods and imaging logging techniques are common techniques for obtaining well logs in actual mines, but the conventional methods have great inadaptability for predicting fracture-cavity-type emptying well logs. Firstly, the problem of large deviation of the accuracy of the conventional empirical method for predicting the logging curve is prominent, for the traditional sandstone compact reservoir, the deviation is relatively in a controllable range, but for a fracture-cavity type oil reservoir, due to the fact that the fracture-cavity type oil reservoir has multiple storage spaces and multiple flow modes, the defects displayed by the problem of large deviation of the conventional empirical method are multiplied. Secondly, acquiring logging curve data by adopting an imaging logging mode is one of the current mainstream modes, but the imaging logging technology needs to put a test probe into a frightening mode, so that the method has high test precision and wide adaptability, but the buried depth of the fractured-vuggy carbonate reservoir is often larger than 5000m, the probe is difficult to put into the fractured-vuggy carbonate reservoir by adopting the imaging logging mode, the implementation cost is high, the empty hole space of the fractured-vuggy carbonate reservoir is often m-level reservoir space, and the data acquired by imaging logging in the reservoir space is often deviated to a certain extent. Other known prior art does not mention a solution for reconstructing a logging curve of an emptying well section of a fractured-vuggy carbonate reservoir.
Disclosure of Invention
Aiming at the problems in the prior art, the technical problems to be solved by the invention are as follows: the lack of logging curve data of the released and lost well section of the fractured-vuggy carbonate reservoir leads to larger deviation of the prediction result of geology.
In order to solve the technical problem, the invention adopts the following technical scheme:
a method for reconstructing a logging curve of a voided lost circulation section of a fractured-vuggy carbonate reservoir comprises the following steps:
s100: determining that the tested oil reservoir is a carbonate fracture-cave oil reservoir, determining that a blow-down well section exists in the internal drilling well of the reservoir unit, performing seismic testing on the reservoir unit of the target layer with the blow-down well section, obtaining a seismic attribute body of a target reservoir unit, wherein the seismic attribute body of the target reservoir unit comprises reservoir geological information and logging information;
s200: establishing a carbonate fracture-cavity type reservoir geological model based on the reservoir geological information and the logging information, and constructing a seismic attribute parameter geological attribute model of the reservoir unit of the target stratum according to the carbonate fracture-cavity type reservoir geological model;
s300: taking seismic attribute parameters in a seismic attribute parameter geological attribute model as input, wherein the seismic attribute parameters comprise tensor, impedance and coherence, and acquiring seismic attribute parameter curves of all real drilling well emptying sections in a reservoir unit of a target stratum oil reservoir by applying a Make logs method through the seismic attribute parameter geological attribute model;
s400: establishing a nonlinear mapping relation between a seismic attribute parameter curve and a logging curve, and specifically comprising the following steps of:
s410: taking the seismic attribute parameter curves and the logging curves of all the real drilling well emptying sections as a sample set, and establishing a deep neural network model containing a plurality of hidden layers;
s420: taking a sample set as an input of a deep neural network model containing a plurality of hidden layers, outputting a result which is a nonlinear mapping relation between seismic attribute parameters and a logging curve, performing iterative training on the deep neural network model by using a Train evaluation model function in a BP (Back propagation) neural network module, and stopping training when the maximum iteration times are reached to obtain a trained deep neural network model;
s500: and (3) taking seismic attribute body data of a target layer without a logging curve as input, and obtaining a logging curve of the actual drilling and emptying well section of the predicted carbonate fracture-cave oil reservoir by using the trained deep neural network model, namely completing the reconstruction of the logging curve of the lost-circulation well section.
Preferably, the types of oil reservoir cells in S100 include: crack type, crack-hole type, and solution-hole type.
Preferably, the specific steps of constructing the seismic attribute parameter geological attribute model of the reservoir unit of the target reservoir in S200 are as follows:
s210: resampling the seismic inversion attribute shape of the geological model by using numerical simulation software, wherein the seismic inversion attribute shape comprises tensor, impedance and coherent body and other seismic attribute parameters;
s220: the seismic inversion attribute body is inverted and spread to the whole target stratum oil reservoir unit, and an inverted seismic attribute body is obtained;
s230: assigning the seismic attribute body obtained after inversion to an attribute model to obtain a seismic attribute parameter geological attribute model of the reservoir unit of the target reservoir;
constructing a seismic attribute parameter geological attribute model: the seismic body obtained by seismic testing can be fully utilized, and the geological property model finally generated by seismic attribute body reconstruction can accurately reflect the geological characteristics of the fracture-cavity type oil reservoir, and reference is made to the figure 2; the model is the basis for extracting seismic attribute parameters along a shaft, and the Make well logs method is based on the attribute model; in addition, the construction of the geological property model is one of the keys for carrying out numerical reservoir simulation, and after the geological model is constructed, the geological property model established by using the geological model method can fully realize the geological property coverage of the whole well area.
Preferably, the numerical software used for resampling the seismic inversion attribute shape of the geological model by using the numerical software in S210 is Petrel numerical simulation software.
The Petrel numerical simulation software provides a multi-disciplinary cooperation working process, and the biggest advantage of the Petrel numerical simulation software is that the Petrel numerical simulation software has an advanced innovation point and an advanced technology for seamlessly connecting the geophysical information, the geological model, the reservoir engineering and the drilling engineering information of the underground geologic body; the numerical simulation software has a better processing means for complex geology and complex fluid, particularly aims at the aspect of geological modeling, and the constructed geological model better reflects the real oil deposit condition and provides a better geological model foundation for the numerical simulation of the oil deposit in the later period; the numerical simulation software is more accurate in identification and description of the fracture-cavity type oil reservoir emptying lost well section, and particularly can fully disperse attribute parameters to each node of a test work area when a seismic body is inverted and a seismic parameter geological attribute model is constructed; various modules are inserted into the numerical simulation software, a plurality of tools are needed for mutual cooperation in well logging curve reconstruction, the numerical simulation software meets the requirement, in addition, a BP neural network system and a genetic inversion algorithm are built in the software, the function is strong, and the tool requirement of the invention can be met.
Preferably, the inversion method used in resampling the geologic model seismic inversion attribute shape in S210 is a genetic inversion method.
The genetic inversion method is realized by an inversion module containing a part of multilayer neural network, and is usually used for deducing seismic attributes such as impedance, tensor and the like in the numerical simulation process of an oil field. The data required to be input by the module is seismic attribute data, but the method does not need an initial attribute model as input data, so that the acquisition of data volume is reduced, and the calculation amount is reduced; and the probability of acquiring the global minimum error by the module is usually greater than that of other methods based on neural network inversion, so that the convergence of seismic attribute inversion is better restrained.
Compared with the prior art, the invention has at least the following advantages:
the carbonate fracture-cavity oil reservoir has complex and various reservoir spaces, wherein the reservoir spaces contain various reservoir spaces, when a well is drilled to an emptying zone, particularly to a karst cave zone and a large fracture zone, open hole completion is often adopted, and related logging data cannot be directly acquired, so that the stratum attribute of the well section is rarely known in the drilling process. The logging curve reconstruction is carried out on the carbonate rock fracture-cave type emptying lost circulation section by adopting the technology of the invention:
1. the geological characteristics of the well section can be fully known by acquiring the logging curve of the emptying lost section, the reservoir type of the well section can be fully identified, the crude oil storage condition of the emptying well section can be acquired, and a certain guide is provided for later-stage production and development;
2. the reconstruction of the logging curve of the well without the logging curve only generates the important parameters of the logging curve, such as porosity, permeability and the like, and the parameters have important significance for geological reserve recalculation, rock attribute model construction and the like in the later period;
3. according to the well logging curve reconstruction technology, the occurrence condition of the crude oil of the carbonate fracture-cavity oil reservoir can be more accurately analyzed, and a basis is provided for optimizing a reasonable injection-production well pattern;
4. the technology is convenient to operate, does not relate to complex equation derivation, consumes less time for the engineering personnel to learn, and is more favorable for knowing the field production practice.
Drawings
FIG. 1 is a schematic diagram of log data generation for a blow-down interval.
Fig. 2 is a schematic diagram of BP neural network training.
FIG. 3 is a flowchart of a BP neural network generating a well log.
FIG. 4 is a schematic diagram of TP7 well reservoir geological information screening and well location distribution of actual drilling wells.
FIG. 5 is a diagram of a TP7 well seismic attribute parameter attribute model.
FIG. 6 is a schematic diagram of a reconstructed logging curve of a void section of a real well of a TP7 well.
FIG. 7 is a schematic view of the permeability curve of a solid well of a TP7 well along a bore of a well.
Detailed Description
The present invention is described in further detail below.
The method comprises the steps of firstly resampling a seismic inversion attribute body, establishing a seismic attribute parameter geological model covering the depth range of the whole target stratum, assigning the model characteristics to the attribute model by using a geological modeling module, then extracting seismic attribute parameter curves from all real drilling wells in a work area along a shaft by using a Make logs method, then establishing a neural network model of a plurality of hidden layers by using the real drilling wells with complete logging curves and seismic attribute parameters in the work area as sample data, establishing a nonlinear mapping relation between the seismic attribute parameter curves and the logging curves by training the model through multiple iterations, and finally generating the missing logging curves of the target well emptying missing well section by using the seismic attribute parameters of the real drilling wells without the logging curves as input and using the trained neural network model, thereby solving the problem of missing information of the emptying and missing well sections of the carbonate fracture-type oil reservoir.
S100-S500 are all steps realized by the technology, wherein the seismic attribute body, the geological information and the logging information related to S100 are the basis for constructing a carbonate fracture-cave reservoir geological model in S200; s200, analyzing the information in S100 and constructing a fracture-cave type oil reservoir geological model, wherein the model has a large number of karst caves, and a well is often positioned at the karst caves, so that a large number of air-defense leakage-loss well sections exist, and the well sections lack well logging curves, so that oil reservoir workers have certain problems in the work of geological understanding, reserve recalculation, oil reservoir exploitation and the like; extracting seismic attribute parameter curves (all wells) along a shaft by using a Make logs method in S300, and providing conditions for constructing a non-mapping relation related to the seismic attribute parameters and the logging curves in S400; the nonlinear correlation mapping relation constructed in the S400 is an important ring for generating the logging curve of the emptying lost well section, namely, the model constructed in the previous step is connected with the model constructed in the previous step to generate a seismic parameter curve, and the seismic parameter curve is led out in the next step to generate the logging curve by utilizing the seismic parameter curve; s500 is the final step, and a logging curve of the carbonate fracture-cave type oil reservoir emptying lost circulation section is generated by utilizing the seismic parameter curve along the shaft generated in the S300 and the related nonlinear mapping relation established in the S400, so that curve reconstruction is realized.
A method for reconstructing a logging curve of a voided lost circulation section of a fractured-vuggy carbonate reservoir comprises the following steps:
s100: determining that the tested oil reservoir is a carbonate fracture-cave oil reservoir, determining that an emptying well section exists in the internal drilling well of an oil reservoir unit, and performing seismic testing on a target-stratum oil reservoir unit with the emptying well section to obtain a seismic attribute body of the target-stratum oil reservoir unit, wherein the seismic attribute body of the target-stratum oil reservoir unit comprises oil reservoir geological information and well logging and logging information;
the reservoir unit types in S100 comprise: crack type, crack-hole type, and solution-hole type.
S200: establishing a carbonate fracture-cavity type reservoir geological model based on the reservoir geological information and the logging information, and constructing a seismic attribute parameter geological attribute model of the reservoir unit of the target stratum according to the carbonate fracture-cavity type reservoir geological model;
the specific steps of constructing the seismic attribute parameter geological attribute model of the reservoir unit of the target stratum in the S200 are as follows:
s210: resampling the seismic inversion attribute shape of the geological model by using numerical simulation software, wherein the seismic inversion attribute shape comprises tensor, impedance and coherent body and other seismic attribute parameters;
in the step S210, the numerical software used for resampling the seismic inversion attribute shape of the geological model by using the numerical software is Petrel numerical simulation software;
the inversion method used in resampling the seismic inversion attribute shape of the geological model in S210 is a genetic inversion method.
S220: the seismic inversion attribute body is inverted and spread to the whole target reservoir stratum unit to obtain an inverted seismic attribute body;
s230: assigning the seismic attribute body obtained after inversion to an attribute model to obtain a seismic attribute parameter geological attribute model of the reservoir unit of the target reservoir; wherein the attribute model is a prior art model.
S300: taking seismic attribute parameters in a seismic attribute parameter geological attribute model as input, wherein the seismic attribute parameters comprise tensor, impedance and coherence, and acquiring seismic attribute parameter curves of all real drilling well emptying sections in a reservoir unit of a target stratum by applying a Make logs method through the seismic attribute parameter geological attribute model;
the Make logs method is the prior art, and all seismic attribute parameter curves of all drilling wells (including well logging curve wells and non-well logging curve wells) in a test work area can be extracted along a shaft by applying the Make logs method to generate complete seismic attribute parameter curves of real drilling wells along the shaft; the implementation process is a Make logs method, and seismic attribute information along the shaft can be generated by utilizing a grid attribute model. The method enters a Make logs module by double-clicking a Wells tab, extracts and attaches seismic data on a grid attribute model to a shaft.
S400: establishing a nonlinear mapping relation between a seismic attribute parameter curve and a logging curve, and specifically comprising the following steps:
s410: taking the seismic attribute parameter curves and the logging curves of all the real drilling well emptying sections as a sample set, and establishing a deep neural network model containing a plurality of hidden layers;
s420: taking a sample set as an input of a deep neural network model containing a plurality of hidden layers, outputting a result which is a nonlinear mapping relation between seismic attribute parameters and a logging curve, performing iterative training on the deep neural network model by using a Train evaluation model function in a BP (Back propagation) neural network module, and stopping training when the maximum iteration times are reached to obtain a trained deep neural network model;
the method comprises the following steps that a Train estimatization model function in a BP neural network module is the prior art, automatic iterative training can be carried out on a deep neural network, iterative training is carried out on the process based on the Train estimatization model function in the BP neural network module in Petrel numerical simulation software, key real well drilling complete seismic attribute parameters and well logging curves serve as input parameters, iterative learning training is carried out for multiple times, a deep neural network training model with a plurality of hidden layers is built, the model takes the seismic attribute parameters and the well logging curves as blueprints, and a nonlinear mapping relation of the seismic attribute parameters and the well logging curves is built;
the BP neural network system is composed of a large number of neurons in an interactive mode, and is a system for performing parallel processing and nonlinear conversion on input seismic attribute parameters and logging curve information. The system is a kind of unidirectional transmission multilayer forward mesh, and solves the problem of the connection right of the hidden unit. The system adopts BP algorithm, carry on two times of transmission calculation to earthquake attribute parameter and logging curve information input, transmit calculation forward first; and secondly, the calculation is transmitted backwards, and the error is gradually reduced by the two calculations.
S500: and (3) taking seismic attribute body data of a target layer without a logging curve as input, and obtaining a logging curve of the actual drilling and emptying well section of the predicted carbonate fracture-cave oil reservoir by using the trained deep neural network model, namely completing the reconstruction of the logging curve of the lost-circulation well section.
The process is based on a Neural net module, a logging curve of an emptying well section is generated by using a Make well log function in the module, a deep Neural network training model is used as a template, seismic attribute parameters of an emptying lost well without the logging curve are used as input variables, and the nonlinear mapping relation between the seismic attribute parameters and the logging curve is analyzed by using the function, so that the logging curve mapped with the seismic attribute parameters of the emptying well section is simulated and generated.
The process of training the seismic attribute parameters and the logging curve information by the BP neural network system comprises the following steps: initializing grid weight; the training set takes out a pair of data inputs; calculating an output vector; judging the relation between the output vector and the constraint; and satisfying the constraint output, thereby obtaining the nonlinear mapping relation between the seismic attribute parameters and the logging curve.
Examples
Referring to fig. 1-3, a method for reconstructing a logging curve of a relief well of a fractured-vuggy carbonate reservoir specifically comprises the following steps:
the method comprises the following steps: determining that the tested oil reservoir is a carbonate fracture-cavity oil reservoir, determining that an emptying well section exists in an internal drilling well of an oil reservoir unit, performing seismic testing on the oil reservoir to obtain a seismic attribute body of the unit, and establishing a carbonate fracture-cavity oil reservoir geological model based on existing data;
step two: establishing a Seismic Attribute parameter geological Attribute model covering the whole target interval based on the geological model, resampling a Seismic inversion Attribute body (tensor, impedance and coherence) of the geological model by using a Volume Attribute module in Petrel numerical simulation software, inverting the Seismic inversion Attribute body to be distributed in the whole target interval by using a Genetic inversion method, and then assigning the Seismic Attribute body (tensor, impedance and AFE) obtained by inversion to the Attribute model by using a Geometrical modeling module through a Seismic re-acquisition method, thereby establishing the Seismic Attribute parameter geological Attribute model covering the depth range of the whole target interval (including an emptying well interval);
step three: and acquiring a seismic attribute parameter curve of the emptying section of the actual drilling well by applying a Make logs method based on the seismic attribute parameter attribute geological model. By applying the method, all seismic attribute parameter curves of all wells (with logging curve wells and without logging curve wells) in a test work area can be extracted along a shaft, and a complete seismic attribute parameter curve of a real well along the shaft is generated;
step four: based on the key real well with the complete seismic attribute parameters and the logging curve as a sample set, a Neural net (BP Neural network system) module is used for deep learning training to establish a nonlinear mapping relation between the seismic attribute parameter curve and the logging curve. The process is based on a Train evaluation model function in a Neural net module in Petrel numerical simulation software to carry out iterative training, a key real well drilling complete seismic attribute parameter and a well logging curve are used as input parameters to serve as input data, multiple times of iterative learning training are carried out, and therefore a deep Neural network training model containing a plurality of hidden layers is established, the model takes the seismic attribute parameter and the well logging curve as a blueprint, and a nonlinear mapping relation of the seismic attribute parameter and the well logging curve is established;
step five: and predicting and generating a logging curve of the actual drilling and emptying well section of the carbonate fracture-cave type oil reservoir by utilizing a BP neural network system based on the neural network training model. The process is also based on a Neural net module, a logging curve of the emptying well section is generated by using a Make well log function in the module, a Neural network training model in step four is used as a template, seismic attribute parameters of the emptying lost well without the logging curve are used as input variables, and the nonlinear mapping relation between the seismic attribute parameters and the logging curve is analyzed by using the function, so that the logging curve mapped with the seismic attribute parameters of the emptying well section is simulated and generated. The well logging curve is based on the interpretation results of porosity, permeability, reservoir type and the like, the problem of basic well logging information loss of the carbonate fracture-cave type oil reservoir leakage emptying well section is effectively solved, and theoretical guidance is provided for oil field production practice.
In the second step, the Genetic inversion module is an inversion module containing a part of multilayer neural network, and is usually used for deriving seismic attributes such as impedance, tensor and the like in the oil field numerical simulation process. The module needs to input data seismic attribute data, but the method does not need an initial attribute model as input, and the probability of acquiring the global minimum error is often greater than that of the rest methods based on neural network inversion, so that the convergence of seismic attribute inversion is better restrained.
The Genetic inversion (Genetic inversion) module contains an implicit multilayer neural network, and the network work flow is as follows: establishing initial community (seismic attribute data) input; coding; fitness function (reflecting the superiority and inferiority of individual seismic attribute data in the attribute community); screening (preferentially screening seismic attribute data); intersecting (preferably, a plurality of intersections of the post-seismic attribute data are interchanged to generate a plurality of seismic attribute data); mutation (single seismic attribute data is partially mutated, and newly generated seismic attribute data is used for replacing existing data); and (6) outputting.
The technology belongs to the prior art, and the construction of a seismic attribute parameter geological attribute model based on a genetic inversion method comprises four parts: 1, performing parameter re-acquisition based on the seismic attribute body; 2, learning the input seismic data by utilizing the genetic hidden layer; 3, establishing a corresponding relation between the seismic attribute body and the space position of the well region; and 4, generating a geological attribute model of the seismic attribute parameters of the whole well area by utilizing the corresponding relation based on a neural network system.
Wherein the module activates:
input layer/hidden layer relationship:
the above equations are the activation function, input-neural network hidden layer relationships, respectively. Wherein y is hidden_lavyer For hiding data, w input,,n 、y input,i 、w input, The hidden layer property is constructed by using the input variables and n as the number of the input variables. Based on the hidden layer and earthquake extractionGenerating an address attribute model of the seismic attribute parameters by utilizing genetic inversion; the method has the function of embodying the constructed seismic attribute parameter geological attribute model including the seismic attribute parameters of all parts of the whole well zone, and creates a precondition for extracting a seismic parameter curve along a shaft by a Make logs method in the next step.
Output layer relationship:
under the module, a gradient descent method (conjugate gradient) is adopted for workflow weight updating, errors are propagated reversely, and convergence to a global minimum value is facilitated through multiple iterations.
In step three, the Make logs method may have generated the along-the-wellbore seismic attribute information using a mesh attribute model. The method enters a Make logs module by double-clicking a Wells tab, extracts and attaches seismic data on a grid attribute model to a shaft.
In the fourth step, the BP neural network system is composed of a large number of neurons in an interactive mode, and the system conducts parallel processing and nonlinear conversion on input seismic attribute parameters and logging curve information. The system is a kind of unidirectional transmission multilayer forward mesh, and solves the problem of the connection right of the hidden unit. The system adopts BP algorithm, carry on the transmission calculation twice to earthquake attribute parameter and logging curve information input, first, transmit the calculation forward; secondly, the calculation is transmitted backwards, and the error is gradually reduced by the two calculations.
The process of training the seismic attribute parameters and the logging curve information by the BP neural network system comprises the following steps: initializing grid weight; a training set takes out a pair of data input; calculating an output vector; judging the relation between the output vector and the constraint; and satisfying the constraint output, thereby obtaining the nonlinear mapping relation between the seismic attribute parameters and the logging curve.
The method comprises the steps of firstly resampling a seismic inversion attribute body, establishing a seismic attribute parameter geological model covering the depth range of the whole target stratum, assigning the model characteristics to the attribute model by using a geological modeling module, then extracting seismic attribute parameter curves from all real drilling wells in a work area along a shaft by using a Make logs method, then establishing a neural network model of a plurality of hidden layers by using the real drilling wells with complete logging curves and seismic attribute parameters in the work area as sample data, establishing a nonlinear mapping relation between the seismic attribute parameter curves and the logging curves by training the model through multiple iterations, and finally generating the missing logging curves of the target well emptying missing well section by using the seismic attribute parameters of the real drilling wells without the logging curves as input and using the trained neural network model, thereby solving the problem of missing information of the emptying and missing well sections of the carbonate fracture-type oil reservoir.
Data experiments
The invention establishes a carbonate fracture-cavity type reservoir geological model by using Petrel numerical simulation software based on carbonate fracture-cavity type reservoir geological information and well logging and logging information, wherein the model contains a plurality of drilling wells and seismic attribute data, and the actual drilling well comprises a plurality of emptying wells. The TP7 well region of the fracture-cavity oil reservoir is taken as an application case for test analysis, and the geological information of the region is shown in figure 2 and comprises multistage reservoir spaces such as fracture cavities, karst caves, fractures, cracks and the like.
As shown in FIG. 4, the well zone comprises 20 actual wells, wherein wells such as TP257H, TP271H, TP201CH, TP267X, etc. have lost circulation zones; the wells such as TP15X, TP271H, TP240 and TP202 have complete logging curves, so that the lost circulation prevention well section is marked as a well group I for the convenience of expression, and the rest of the wells with complete logging curve information are marked as a well group II (the TP7 well position relation is shown in detail in figure 4).
Firstly, resampling a TP7 well geological model seismic inversion Attribute body by using a Volume Attribute module in Petrel numerical simulation software, wherein sampling information comprises the following steps: tensor, wave impedance, AFE. And (3) sampling, and simultaneously spreading the TP7 well zone seismic inversion attribute body to the whole target interval by using a Genetic inversion method, wherein the seismic attribute body has complete target interval attribute information. And then, assigning the well attribute model with the TP7 well Seismic attribute body (tensor, impedance and AFE) obtained by inversion by using a Geometrical modeling module in Petrel numerical simulation software through a Seismic re-acquisition method, so that the attribute model has Seismic inversion attributes (the attribute model is shown in detail in figure 5).
And then, the obtained attribute model containing the seismic attribute parameters is used for assigning the geological model to all real drilled wells. And selecting seismic attribute parameter data in the attribute model by using a Make logs function in a Well module in Petrel numerical simulation software, putting the Make logs in all real drilled wells of the TP7 Well zone, and extracting seismic attribute parameters in the attribute model along the shaft. So that all wells of the TP7 well zone have the seismic attribute parameter curve.
Taking a well group two in a TP7 well zone as a key well (a well in a well injection group two has a complete well logging curve and an earthquake attribute parameter curve), taking the well logging curve and the address attribute parameter curve of the well group as input samples, applying a Train estimation model function in a Neural net (BP Neural network system) module to carry out deep learning training on input data, and establishing a deep Neural network training model containing a plurality of hidden layers through a plurality of iterative learning trainings. And finally establishing a nonlinear mapping relation between the TP7 well region key well seismic attribute parameter curve and the logging curve. The training mode takes the seismic attribute parameter data and the well logging curve as the blueprint, and establishes the nonlinear mapping relation of the seismic attribute parameter and the well logging curve (the process is shown in detail in fig. 2-3).
And finally, predicting and generating a logging curve of the solid drilling and emptying well section of the TP7 well region carbonate fracture-cave type oil reservoir by utilizing a BP neural network system based on the neural network training model (see figure 6 in detail). In the process, the seismic attribute parameter data of a TP7 well group II are used as input variables, a Neural net module is also applied, and a logging curve of an emptying well section is generated by using a Make well log function in the module. The deep neural network training model is used as a training blueprint, input variables (seismic attribute parameter data) are input into a BP neural network system, the deep neural network training model is used for analysis, a logging curve corresponding to the seismic attribute parameter data is obtained by utilizing the training model analysis rule, and finally a logging curve of the carbonate fracture-cave type oil reservoir leakage emptying well section of the TP7 well region is established (the seismic attribute parameter data and the logging data of the emptying well section after reconstruction are respectively shown in the figure 7).
Finally, the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting, although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions may be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all of them should be covered in the claims of the present invention.
Claims (5)
1. A carbonate fracture-cave type oil reservoir emptying lost circulation section well logging curve reconstruction method is characterized by comprising the following steps: the method comprises the following steps:
s100: determining that the tested oil reservoir is a carbonate fracture-cave oil reservoir, determining that an emptying well section exists in the reservoir unit of the oil reservoir during drilling, performing seismic testing on the reservoir unit of the target layer oil reservoir with the emptying well section, obtaining a seismic attribute body of a target stratum reservoir unit, wherein the seismic attribute body of the target stratum reservoir unit comprises reservoir geological information and logging information;
s200: establishing a carbonate fracture-cave type reservoir geological model based on the reservoir geological information and the logging information, and constructing a seismic attribute parameter geological attribute model of the reservoir unit of the target layer according to the carbonate fracture-cave type reservoir geological model;
s300: taking seismic attribute parameters in a seismic attribute parameter geological attribute model as input, wherein the seismic attribute parameters comprise tensor, impedance and coherence, and acquiring seismic attribute parameter curves of all real drilling well emptying sections in a reservoir unit of a target stratum oil reservoir by applying a Make logs method through the seismic attribute parameter geological attribute model;
s400: establishing a nonlinear mapping relation between a seismic attribute parameter curve and a logging curve, and specifically comprising the following steps:
s410: taking the seismic attribute parameter curves and the logging curves of all the real drilling well emptying sections as a sample set, and establishing a deep neural network model containing a plurality of hidden layers;
s420: taking a sample set as an input of a deep neural network model containing a plurality of hidden layers, outputting a result which is a nonlinear mapping relation between seismic attribute parameters and a logging curve, performing iterative training on the deep neural network model by using a Train evaluation model function in a BP (Back propagation) neural network module, and stopping training when the maximum iteration times are reached to obtain a trained deep neural network model;
s500: and (3) taking seismic attribute body data of a target layer without a logging curve as input, and obtaining a logging curve of the actual drilling and emptying well section of the predicted carbonate fracture-cave type oil reservoir by using the trained deep neural network model, namely completing the reconstruction of the logging curve of the lost and emptying well section.
2. The method for reconstructing the logging curve of the open-hole and lost circulation section of the fractured-vuggy carbonate reservoir as claimed in claim 1, wherein the method comprises the following steps: the reservoir unit types in S100 comprise: crack type, crack-hole type, and solution-hole type.
3. The method for reconstructing the logging curve of the open-hole and lost circulation section of the fractured-vuggy carbonate reservoir as claimed in claim 2, wherein the method comprises the following steps: the specific steps of constructing the seismic attribute parameter geological attribute model of the reservoir unit of the target stratum in the S200 are as follows:
s210: resampling the seismic inversion attribute shape of the geological model by using numerical simulation software, wherein the seismic inversion attribute shape comprises tensor, impedance and coherent body and other seismic attribute parameters;
s220: the seismic inversion attribute body is inverted and spread to the whole target stratum oil reservoir unit, and an inverted seismic attribute body is obtained;
s230: and assigning the seismic attribute body obtained after inversion to an attribute model to obtain a seismic attribute parameter geological attribute model of the reservoir unit of the target reservoir.
4. The method for reconstructing the logging curve of the open-hole and lost circulation section of the fractured-vuggy carbonate reservoir as claimed in claim 3, wherein the method comprises the following steps: and in the step S210, numerical software used for resampling the seismic inversion attribute shape of the geological model by using the numerical software is Petrel numerical simulation software.
5. The method for reconstructing the logging curve of the open-hole and lost circulation section of the fractured-vuggy carbonate reservoir as claimed in claim 3, wherein the method comprises the following steps: the inversion method used in resampling the geologic model seismic inversion attribute shape in S210 is a genetic inversion method.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2022103829406 | 2022-04-12 | ||
| CN202210382940.6A CN114740531A (en) | 2022-04-12 | 2022-04-12 | Method for reconstructing logging curve of voiding lost circulation well section of fractured-vuggy carbonate reservoir |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115373028A true CN115373028A (en) | 2022-11-22 |
Family
ID=82280806
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210382940.6A Withdrawn CN114740531A (en) | 2022-04-12 | 2022-04-12 | Method for reconstructing logging curve of voiding lost circulation well section of fractured-vuggy carbonate reservoir |
| CN202211127777.5A Pending CN115373028A (en) | 2022-04-12 | 2022-09-16 | Method for reconstructing logging curve of voiding lost circulation well section of fractured-vuggy carbonate reservoir |
Family Applications Before (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210382940.6A Withdrawn CN114740531A (en) | 2022-04-12 | 2022-04-12 | Method for reconstructing logging curve of voiding lost circulation well section of fractured-vuggy carbonate reservoir |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN114740531A (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115421181B (en) * | 2022-07-27 | 2023-10-20 | 北京超维创想信息技术有限公司 | Three-dimensional geological model phase control attribute modeling method based on deep learning |
-
2022
- 2022-04-12 CN CN202210382940.6A patent/CN114740531A/en not_active Withdrawn
- 2022-09-16 CN CN202211127777.5A patent/CN115373028A/en active Pending
Also Published As
| Publication number | Publication date |
|---|---|
| CN114740531A (en) | 2022-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| RU2496972C2 (en) | Device, method and system of stochastic investigation of formation at oil-field operations | |
| US10767448B2 (en) | Multistage oilfield design optimization under uncertainty | |
| US7933750B2 (en) | Method for defining regions in reservoir simulation | |
| RU2486336C2 (en) | Method of formation breakdown simulation and its estimation, and computer-read carrier | |
| CN104755960B (en) | Improvement to the rate pattern for handling seismic data is modeled based on basin | |
| CN109478208A (en) | The iteration of integrated data and process integration for oil exploration and production assessment and repeatable workflow | |
| US20090319243A1 (en) | Heterogeneous earth models for a reservoir field | |
| US20070016389A1 (en) | Method and system for accelerating and improving the history matching of a reservoir simulation model | |
| US20060184329A1 (en) | Method system and program storage device for optimization of valve settings in instrumented wells using adjoint gradient technology and reservoir simulation | |
| CN103403768B (en) | Method and system about the model of subterranean strata | |
| US10961826B2 (en) | Subsurface modeler workflow and tool | |
| RU2002122397A (en) | Comprehensive reservoir optimization | |
| CN111596978A (en) | Web page display method, module and system for lithofacies classification by artificial intelligence | |
| CN112394404B (en) | Progressive reservoir fine characterization method | |
| CN113534261A (en) | Reservoir gas content detection method and device based on intelligent optimization integrated network | |
| CN115373028A (en) | Method for reconstructing logging curve of voiding lost circulation well section of fractured-vuggy carbonate reservoir | |
| CN107633556A (en) | It is a kind of quantitatively to obtain the probabilistic method of three dimensional ore deposit geological model | |
| CN110988997A (en) | Hydrocarbon source rock three-dimensional space distribution quantitative prediction technology based on machine learning | |
| Hou et al. | Data-driven optimization of brittleness index for hydraulic fracturing | |
| RU2670801C1 (en) | System of integrated conceptual design of hydrocarbon fields | |
| CN115877447A (en) | Reservoir prediction method for seismic restraint three-dimensional geological modeling under straight-flat combined well pattern condition | |
| CN114862609A (en) | Oil reservoir exploitation method for ultra-high water-cut oil field | |
| WO2016187238A1 (en) | Auto-validating earth interpretation and modeling system | |
| van Wees et al. | Accelerating geothermal development with a play-based portfolio approach | |
| CN111815769B (en) | Modeling method, computing device and storage medium for thrust covered zone construction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |