CN117235628B - Well logging curve prediction method and system based on hybrid Bayesian deep network - Google Patents
Well logging curve prediction method and system based on hybrid Bayesian deep network Download PDFInfo
- Publication number
- CN117235628B CN117235628B CN202311493247.7A CN202311493247A CN117235628B CN 117235628 B CN117235628 B CN 117235628B CN 202311493247 A CN202311493247 A CN 202311493247A CN 117235628 B CN117235628 B CN 117235628B
- Authority
- CN
- China
- Prior art keywords
- data
- logging curve
- sample
- model
- classification labels
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 239000011435 rock Substances 0.000 claims abstract description 7
- 238000012549 training Methods 0.000 claims description 32
- 238000009826 distribution Methods 0.000 claims description 23
- 230000006870 function Effects 0.000 claims description 23
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 17
- 238000013528 artificial neural network Methods 0.000 claims description 11
- 238000005457 optimization Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 8
- 238000005070 sampling Methods 0.000 claims description 8
- 239000004927 clay Substances 0.000 claims description 7
- 238000004422 calculation algorithm Methods 0.000 claims description 6
- 230000003595 spectral effect Effects 0.000 claims description 4
- 238000013135 deep learning Methods 0.000 abstract description 9
- 230000004044 response Effects 0.000 abstract description 6
- 238000005516 engineering process Methods 0.000 abstract description 5
- 238000010586 diagram Methods 0.000 description 8
- 238000001228 spectrum Methods 0.000 description 4
- 239000012530 fluid Substances 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 239000011148 porous material Substances 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 239000004576 sand Substances 0.000 description 2
- YBJHBAHKTGYVGT-ZKWXMUAHSA-N (+)-Biotin Chemical compound N1C(=O)N[C@@H]2[C@H](CCCCC(=O)O)SC[C@@H]21 YBJHBAHKTGYVGT-ZKWXMUAHSA-N 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000001035 drying Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000002637 fluid replacement therapy Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 229920006395 saturated elastomer Polymers 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 230000035899 viability Effects 0.000 description 1
- FEPMHVLSLDOMQC-UHFFFAOYSA-N virginiamycin-S1 Natural products CC1OC(=O)C(C=2C=CC=CC=2)NC(=O)C2CC(=O)CCN2C(=O)C(CC=2C=CC=CC=2)N(C)C(=O)C2CCCN2C(=O)C(CC)NC(=O)C1NC(=O)C1=NC=CC=C1O FEPMHVLSLDOMQC-UHFFFAOYSA-N 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method and a system for predicting a logging curve based on a hybrid Bayesian deep network, wherein the method comprises the following steps: acquiring logging curve data of a plurality of wells in a complex lithology area; correlating the logging curve data with lithology types to form classification labels; based on the classification labels and the logging curve data, combining a petrophysical equation to obtain sample sets associated with different classification labels; constructing a hybrid Bayesian deep network model based on the sample set; optimizing the hybrid Bayesian deep network model; and predicting the newly input logging curve based on the optimized mixed Bayesian deep network model. According to the invention, an artificial intelligent Bayesian deep learning technology and a big data technology are fused to the logging curve prediction of rock physics, so that the classification or prediction problem is easier, and the system response is quicker.
Description
Technical Field
The invention belongs to the field of computer system engineering, and particularly relates to a hybrid Bayesian deep network-based logging curve prediction method and system.
Background
Among the oil and gas resource big data, the logging data is the most main oil reservoir description basic data in the oil and gas resource exploration and development process. However, in practical applications, due to problems such as wellbore expansion, mud or mud cake disturbance, some log data is often distorted or missing, and even some logs are abandoned for economic viability. And the key for realizing the high effect of logging big data to promote the effective application of oil and gas resource big data is the joint application of inter-well data and the effective mining of longitudinal logging information.
In the aspect of theoretical research methods, the current logging curve prediction technology mainly adopts three types of methods: (1) establishing a classical model based on geostatistics; (2) using a machine learning method to realize logging curve prediction; (3) And the logging curve prediction is realized by using an integrated learning and deep learning method. The statistical method depends on the completeness of the existing data, the geological rules of the parent curve and the child curve need to be counted, and the description and popularization capability of the corresponding complex reservoir are limited. Traditional neural network models such as an artificial neural network and a BP neural network often depend on larger training samples, the number of layers of the network is shallow, the generalization capability is weak, and overfitting is easy to cause, so that the problem of logging curve is solved by directly applying the neural network. The integrated learning and deep learning method can solve the problems of weak generalization capability, over-fitting and better prediction result of the conventional neural network to a certain extent, but the conventional deep learning logging data prediction method does not consider the complexity of data, and only the problems of data correlation analysis and classification are transmitted to a deeper network model for processing.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a logging curve classification based on spectrum classification, the combination and adjustment of lithology classification are carried out manually, after lithology curves are determined, the sample set is expanded by using a petrophysical equation according to sample sets of different classifications, the sample training is carried out by using a Bayesian deep learning network, and then the prediction of the curves is carried out by using a model.
The invention provides a method for predicting a logging curve based on a hybrid Bayesian deep network, which comprises the following steps:
acquiring logging curve data of a plurality of wells in a complex lithology area;
correlating the logging curve data with lithology types to form classification labels;
based on the classification labels and the logging curve data, combining a petrophysical equation to obtain sample sets associated with different classification labels;
constructing a hybrid Bayesian deep network model based on the sample set;
optimizing the hybrid Bayesian deep network model;
and predicting the newly input logging curve based on the optimized mixed Bayesian deep network model.
Wherein the log data comprises log longitudinal wave velocity, shear wave velocity, density, natural gamma, clay content, porosity and water saturation of the plurality of wells.
Wherein, the correlating the log data with lithology type to form a classification tag comprises:
lithology classification is performed on log data using spectral classification.
The method for obtaining the sample set associated with different classification labels based on classification labels and logging curve data and combining a petrophysical equation comprises the following steps:
and constructing longitudinal wave speed, transverse wave speed and density curves with different water saturation and different porosities according to the petrophysical model under different lithology.
Wherein, based on the sample set, constructing a hybrid bayesian deep network model includes:
fitting probability distribution of known data sample data by utilizing a multi-element Gaussian distribution, simulating and constructing a generator by utilizing a Monte Carlo random sampling method, and sending the generator into a discriminator to judge the authenticity of the distribution data.
Wherein the generator is for generating a quantity of analog samples.
Wherein the arbiter makes the determination by a loss function, wherein the loss function is expressed as follows:
wherein i is the number of nodes, j is the number of samples,is weight->Posterior probability of>As a result of the weight of the network,for learning sample maximum a posteriori estimation, +.>For the prior probability->Is the posterior probability of the label sample.
Wherein said optimizing said hybrid bayesian deep network model comprises:
iteratively training parameters of each layer of the discriminator through a model, and optimizing the discriminator model; then training a generator-discriminant model, and optimizing parameters of each layer of the generator under the condition that network parameters of each layer of the discriminant are unchanged; repeating the two optimization processes, and alternately training the generator and the discriminator to realize the training of the integral neural network.
Wherein, during the training process of the network, the weight is updated by using an optimization algorithm of gradient descent. The gradient descent algorithm adjusts model parameters w according to the gradient information of the loss function to the weight i,k The prediction result of the model on the training data is enabled to be closer to the real label.
Where k is the number of iterations, i is the number of nodes,for the kth iteration network weight, lr is the learning rate and E is the loss function.
The invention also provides a hybrid Bayesian deep network-based logging curve prediction system, which comprises:
the acquisition module is used for acquiring logging curve data of a plurality of wells in the complex lithology area;
the association module is used for associating the logging curve data with lithology types to form classification labels;
the sample expansion module is used for obtaining sample sets associated with different classification labels based on the classification labels and the logging curve data and combining a petrophysical equation;
a model building module for building a hybrid bayesian deep network model based on the sample set;
a model optimization module for optimizing the hybrid bayesian deep network model;
and the prediction module is used for predicting the newly input logging curve based on the optimized mixed Bayesian deep network model.
Compared with the prior art, the method and the device have the advantages that the study sample data are classified by utilizing a spectrum classification mode aiming at complex lithology areas, and the classification characteristics are more in accordance with the classification standard of rock physics by utilizing an artificial dry prediction mode, so that the problem that the complex lithology classification fitting is difficult in Bayesian deep study on the basis of small sample data is effectively solved. Meanwhile, an artificial intelligent Bayesian deep learning technology and a big data technology are fused into logging curve prediction of petrophysics, complex mathematical relations in geophysical subjects are gradually represented by forming more abstract high-level features from low-level features through learning and training of the samples, and feature representation of the samples in an original space is transformed into a new feature space, so that classification or prediction problems are easier, processing speed is faster, and system response is quicker.
Drawings
The above, as well as additional purposes, features, and advantages of exemplary embodiments of the present disclosure will become readily apparent from the following detailed description when read in conjunction with the accompanying drawings. Several embodiments of the present disclosure are illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings and in which like reference numerals refer to similar or corresponding parts and in which:
FIG. 1 is a flow chart illustrating a hybrid Bayesian deep network based log curve prediction method in accordance with an embodiment of the present invention;
FIG. 2 is a schematic diagram illustrating a spectral classification obtained log lithology classification according to an embodiment of the invention;
FIG. 3 is a schematic diagram illustrating a sample graph at different water saturation levels according to an embodiment of the present invention;
FIG. 4 is a schematic diagram showing a sample graph at different porosities according to an embodiment of the present invention;
FIG. 5 is a schematic diagram illustrating predicted transverse waves based on a hybrid Bayesian deep network log prediction model in accordance with an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail below with reference to the accompanying drawings, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The terminology used in the embodiments of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in this application and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise, the "plurality" generally includes at least two.
It should be understood that although the terms first, second, third, etc. may be used to describe … … in embodiments of the present invention, these … … should not be limited to these terms. These terms are only used to distinguish … …. For example, the first … … may also be referred to as the second … …, and similarly the second … … may also be referred to as the first … …, without departing from the scope of embodiments of the present invention.
It should be understood that the term "and/or" as used herein is merely one relationship describing the association of the associated objects, meaning that there may be three relationships, e.g., a and/or B, may represent: a exists alone, A and B exist together, and B exists alone. In addition, the character "/" herein generally indicates that the front and rear associated objects are an "or" relationship.
The words "if", as used herein, may be interpreted as "at … …" or "at … …" or "in response to a determination" or "in response to a detection", depending on the context. Similarly, the phrase "if determined" or "if detected (stated condition or event)" may be interpreted as "when determined" or "in response to determination" or "when detected (stated condition or event)" or "in response to detection (stated condition or event), depending on the context.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a product or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such product or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a commodity or device comprising such element.
The principle of the invention is as follows: the method for classifying the logging curves based on spectrum classification is provided, then lithology classification combination and adjustment are manually carried out, after lithology curves are determined, sample training is carried out on sample sets of different classifications by using a Bayesian deep learning network, and then curve prediction is carried out through a model.
Alternative embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
Embodiment 1,
As shown in fig. 1, the invention discloses a hybrid bayesian depth network-based logging curve prediction method, which comprises the following steps:
step 1: firstly, logging longitudinal wave speed, transverse wave speed, density, natural gamma, clay content, porosity and water saturation data of a plurality of wells are analyzed, lithology characteristics are analyzed, and logging curve data can be summarized into several lithology types.
Step 2: then, lithology classification is carried out on the logging curve data by utilizing a spectrum classification mode, if the classification is unreasonable, classification and combination can be carried out through manual judgment, and classification in the petrophysical sense is obtained, if the classification is reasonable, manual intervention is not needed.
Step 3: building a training sample: on the basis of the original logging curve, a sample set is expanded according to rock physical models under different lithology, and longitudinal wave speed, transverse wave speed and density curves with different water saturation and different porosities are constructed.
Step 4: training a neural network: fitting probability distribution of known data sample data by utilizing multiple Gaussian distribution based on a sample set, simulating and generating a certain amount of simulated samples (generators) by utilizing a Monte Carlo random sampling method, sending the simulated samples into a loss function to judge the authenticity of the distributed data (a discriminator), and training parameters of each layer of the discriminator through model iteration to optimize a discriminator model; then training a generator-discriminant model, and optimizing parameters of each layer of the generator under the condition that network parameters of each layer of the discriminant are unchanged; repeating the two optimization processes, and alternately training the generator and the discriminator to realize the training of the integral neural network.
Step 5: testing and generating a model: inputting the test data set data into the trained variation to generate an countermeasure network, and outputting the generated simulated sample curve.
Embodiment II,
The invention provides a method for predicting a logging curve based on a hybrid Bayesian deep network, which comprises the following steps:
acquiring logging curve data of a plurality of wells in a complex lithology area;
correlating the logging curve data with lithology types to form classification labels;
based on the classification labels and the logging curve data, combining a petrophysical equation to obtain sample sets associated with different classification labels;
constructing a hybrid Bayesian deep network model based on the sample set;
optimizing the hybrid Bayesian deep network model;
and predicting the newly input logging curve based on the optimized mixed Bayesian deep network model.
Wherein the log data comprises log longitudinal wave velocity, shear wave velocity, density, natural gamma, clay content, porosity and water saturation of the plurality of wells.
Wherein, the correlating the log data with lithology type to form a classification tag comprises:
lithology classification is performed on log data using spectral classification.
The method for obtaining the sample set associated with different classification labels based on classification labels and logging curve data and combining a petrophysical equation comprises the following steps:
and constructing longitudinal wave speed, transverse wave speed and density curves with different water saturation and different porosities according to the petrophysical model under different lithology.
Wherein, based on the sample set, constructing a hybrid bayesian deep network model includes:
fitting probability distribution of known data sample data by utilizing a multi-element Gaussian distribution, simulating and constructing a generator by utilizing a Monte Carlo random sampling method, and sending the generator into a discriminator to judge the authenticity of the distribution data.
Wherein the generator is for generating a quantity of analog samples.
Wherein the arbiter makes the determination by a loss function, wherein the loss function is expressed as follows:
wherein i is the number of nodes, j is the number of samples,is weight->Posterior probability of>As a result of the weight of the network,for learning sample maximum a posteriori estimation, +.>For the prior probability->Is the posterior probability of the label sample.
Wherein said optimizing said hybrid bayesian deep network model comprises:
iteratively training parameters of each layer of the discriminator through a model, and optimizing the discriminator model; then training a generator-discriminant model, and optimizing parameters of each layer of the generator under the condition that network parameters of each layer of the discriminant are unchanged; repeating the two optimization processes, and alternately training the generator and the discriminator to realize the training of the integral neural network.
Wherein, during the training process of the network, the weight is updated by using an optimization algorithm of gradient descent. The gradient descent algorithm adjusts model parameters w according to the gradient information of the loss function to the weight i,k The prediction result of the model on the training data is enabled to be closer to the real label.
Third embodiment,
As shown in fig. 2-5, the method of the present embodiment, taking a certain actual data as an example, specifically includes the following steps:
step 1: and analyzing according to a well curve in a target work area, wherein reservoir characteristics in the work area are divided into high-impedance characteristics, low-impedance characteristics, upper-high-low-impedance characteristics and lower-high-impedance characteristics, and non-reservoir is of mudstone and mudstone sandstone types. Therefore, the curves of a plurality of wells are divided into five types of mudstones, argillaceous sandstones, sandstones with high resistance, sandstones with low resistance, and sandstones with high resistance from top to bottom, as a cluster center selection criterion in step 2.
Step 2: the longitudinal wave speed, transverse wave speed, density, clay content, porosity and water saturation curves of a plurality of wells are connected end to end and combined into a plurality of attribute samples x ij （i=1,2,3,…nJ=1, 2,3,4,5, 6), constructing a smoothing factor S versus x ij Filtering; obtaining partial mudstone, argillaceous sandstone, sandstone with high impedance, sandstone with low impedance and sandstone with high impedance, and obtaining depth domain attribute samples u corresponding to the classification values by interpretation in the step 1 ijk (i=1, 2,3, … n0; j=1, 2,3,4,5,6, k=1, 2,3,4, 5) is set as an initial clustering center, attribute samples are intercepted by utilizing a sliding time window, the Euclidean distance between the attribute sample data and the clustering center is calculated, and when the integral square sum is minimum through iterative calculation, a clustering result L is output i Such as the lithology curve classification of fig. 2.
；
；
；
；
Wherein i is a depth domain index, j is an attribute dimension index, k is a cluster center number index, m is the number of sample attributes in a cluster, n is the number of sample points in each sample depth domain, and D is the number of classifications. The formula CSS is called intra-cluster sum of squares (cluster Sum of Square), also called ineertia. And the sum of squares in the clusters in one dataset is added to obtain the overall sum of squares TotalCSS (Total Cluster Sum of Square), the smaller the value is, the more similar the sample in each cluster is, and the better the clustering effect is.
Step 3: aiming at the characteristics of the sand shale in the local area, a petrophysical equation of the sand shale is constructed, and fluid replacement with different saturation and pore replacement with different porosities are carried out based on a Gassmann equation and an Xu-White equation.
In the middle ofAnd->Rock bulk modulus and shear modulus for saturated fluids, +.>And->For drying the bulk modulus and shear modulus of the rock framework, we can calculate by Xu-White equation,/I>Bulk modulus for matrix can be calculated by VRH mean equation, < >>For matrix porosity, ++>Is the bulk modulus of the fluid in the pores, +.>For the bulk modulus of water,for bulk modulus of qi, < >>Is water saturation, ++>For longitudinal wave velocity +.>Is transverse wave velocity. Fluid and pore replacement can be accomplished by varying the water saturation and porosity.
Calculate the water saturation sw= [0%,20%,40%,60%,80%,100%]Lower longitudinal, transverse wave velocity and density curves, as shown in fig. 3; calculation of porosityAccording to ratio= [0.6 0.8 1 1.1 1.2 1.3 ]]The scaled up and down longitudinal and transverse wave velocity and density curves are expanded into a sample set x as shown in fig. 4.
Step 4: setting the transverse wave speed as a label sample y, setting the longitudinal wave speed, density, water saturation, clay content, porosity and classification curve L as a learning sample x, and constructing a Bayes deep learning network, wherein a Bayes formula can be written as follows:
where we want to find the posterior probability p (w|x, y) of the network weight w, the prior probability p (w) is that we can be empirically and blindly guessed, e.g. we initially set p (w) to a standard normal distribution, and the likelihood p (y|x, w) is a function of the network weight w. However, p (w|x, y) is often difficult to calculate, so using variance estimation, p (w|x, y) is approximated with one distribution q (w|θ), and the similarity between the two distributions p (w|x, y) is measured with KL divergence q (w|θ). The loss function of the variational inferred BNN (Bayesian deep learning) is as follows:
where i is the number of nodes and j is the number of samples. Based on the chain rule of calculus, we can effectively calculate the gradient of log-likelihood with respect to model parameters by using an automatic differentiating tool, and iteratively train the model parameters according to the calculated gradient to find the local optimum of the log-likelihood. The network model comprises a seven-layer neural network, an input layer, an output layer, four middle layers and an hidden variable layer, wherein the middle layers comprise a generator and a discriminator model, and the hidden variable layer uses a Gaussian distribution probability density function. CollectingFitting probability distribution of known data sample data with multiple Gaussian distribution, setting weight function as normal distribution, simulating Li Yongmeng terCarlo random sampling method to generate a certain quantity of simulated samples log q (w|theta), log p (w), log p (y|w, x) (model generator), sending into loss function to judge true and false data, and updating distribution(wherein θ is a distribution parameter, α is an update weight, and E is a loss function), training the weight parameters w of each layer of the arbiter by using a gradient descent algorithm, and optimizing the arbiter model; then training a generator-discriminant model, and optimizing parameters of each layer of the generator under the condition that network parameters of each layer of the discriminant are unchanged; repeating the two optimization processes, and alternately training the generator and the discriminator to realize the training of the integral neural network.
Step 5: and selecting a test well, and optimizing and predicting the transverse wave speed y by using the input longitudinal wave speed, density, clay content, porosity and water saturation x through a model. As shown in fig. 5, the correlation value between the predicted transverse wave curve and the actual transverse wave curve reaches 0.991155.
Example IV
The invention also provides a hybrid Bayesian deep network-based logging curve prediction system, which comprises:
the acquisition module is used for acquiring logging curve data of a plurality of wells in the complex lithology area;
the association module is used for associating the logging curve data with lithology types to form classification labels;
the sample expansion module is used for obtaining sample sets associated with different classification labels based on the classification labels and the logging curve data and combining a petrophysical equation;
a model building module for building a hybrid bayesian deep network model based on the sample set;
a model optimization module for optimizing the hybrid bayesian deep network model;
and the prediction module is used for predicting the newly input logging curve based on the optimized mixed Bayesian deep network model.
Fifth embodiment (V),
The disclosed embodiments provide a non-transitory computer storage medium storing computer executable instructions that perform the method steps described in the embodiments above.
It should be noted that the computer readable medium described in the present disclosure may be a computer readable signal medium or a computer readable storage medium, or any combination of the two. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples of the computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this disclosure, a computer-readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In the present disclosure, however, the computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, with the computer-readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: electrical wires, fiber optic cables, RF (radio frequency), and the like, or any suitable combination of the foregoing.
The computer readable medium may be contained in the electronic device; or may exist alone without being incorporated into the electronic device.
Computer program code for carrying out operations of the present disclosure may be written in one or more programming languages, including an object oriented programming language such as Java, smalltalk, C ++ and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any kind of network, including a local Area Network (AN) or a Wide Area Network (WAN), or can be connected to AN external computer (for example, through the Internet using AN Internet service provider).
The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The units involved in the embodiments of the present disclosure may be implemented by means of software, or may be implemented by means of hardware. Wherein the names of the units do not constitute a limitation of the units themselves in some cases.
The foregoing description of the preferred embodiments of the present invention has been presented for purposes of clarity and understanding, and is not intended to limit the invention to the particular embodiments disclosed, but is intended to cover all modifications, alternatives, and improvements within the spirit and scope of the invention as outlined by the appended claims.
Claims (6)
1. A method for predicting a logging curve based on a hybrid Bayesian deep network comprises the following steps:
acquiring logging curve data of a plurality of wells in a complex lithology area;
correlating the logging curve data with lithology types to form classification labels;
based on the classification labels and the logging curve data, combining a petrophysical equation to obtain sample sets associated with different classification labels;
constructing a hybrid Bayesian deep network model based on the sample set;
optimizing the hybrid Bayesian deep network model;
predicting a newly input logging curve based on the optimized mixed Bayesian depth network model;
the method for obtaining the sample set associated with different classification labels based on classification labels and logging curve data and combining a petrophysical equation comprises the following steps:
constructing longitudinal wave speed, transverse wave speed and density curves with different water saturation and different porosities according to rock physical models with different lithologies;
wherein said constructing a hybrid bayesian deep network model based on said sample set comprises:
fitting probability distribution of known data sample data by utilizing a multi-element Gaussian distribution, simulating and constructing a generator by utilizing a Monte Carlo random sampling method, and sending the generator into a discriminator to judge whether the distribution data is true or false;
wherein the generator is for generating a quantity of analog samples; wherein the arbiter makes the determination by a loss function, wherein the loss function is expressed as follows:
wherein i is the number of nodes, j is the number of sampling points,is weight->Posterior probability of>For network weight, ++>For learning sample maximum a posteriori estimation, +.>For the prior probability->Is the posterior probability of the label sample.
2. The method of claim 1, wherein the log data comprises log longitudinal wave velocity, shear wave velocity, density, natural gamma, clay content, porosity, and water saturation of the plurality of wells.
3. The method of claim 1, wherein said correlating the log data with lithology types to form classification labels comprises:
lithology classification is performed on log data using spectral classification.
4. The method of claim 1, wherein the optimizing the hybrid bayesian depth network model comprises:
iteratively training parameters of each layer of the discriminator through a model, and optimizing the discriminator model; then training a generator-discriminant model, and optimizing parameters of each layer of the generator under the condition that network parameters of each layer of the discriminant are unchanged; repeating the two optimization processes, and alternately training the generator and the discriminator to realize the training of the integral neural network.
5. The method of claim 4, wherein the weights are updated during training of the network using a gradient descent optimization algorithm that adjusts the model parameters w based on the gradient information of the loss function versus the weights i,k Making the prediction result of the model on the training data more similar to the real label, wherein
Where k is the number of iterations, i is the number of nodes,for the kth iteration network weight, lr is the learning rate and E is the loss function.
6. A hybrid bayesian-based deep network log prediction system, comprising:
the acquisition module is used for acquiring logging curve data of a plurality of wells in the complex lithology area;
the association module is used for associating the logging curve data with lithology types to form classification labels;
the sample expansion module is used for obtaining sample sets associated with different classification labels based on the classification labels and the logging curve data and combining a petrophysical equation;
a model building module for building a hybrid bayesian deep network model based on the sample set;
a model optimization module for optimizing the hybrid bayesian deep network model;
the prediction module is used for predicting a newly input logging curve based on the optimized mixed Bayesian deep network model;
the method for obtaining the sample set associated with different classification labels based on classification labels and logging curve data and combining a petrophysical equation comprises the following steps:
constructing longitudinal wave speed, transverse wave speed and density curves with different water saturation and different porosities according to rock physical models with different lithologies;
wherein said constructing a hybrid bayesian deep network model based on said sample set comprises:
fitting probability distribution of known data sample data by utilizing a multi-element Gaussian distribution, simulating and constructing a generator by utilizing a Monte Carlo random sampling method, and sending the generator into a discriminator to judge whether the distribution data is true or false;
wherein the generator is for generating a quantity of analog samples; wherein the arbiter makes the determination by a loss function, wherein the loss function is expressed as follows:
wherein i is the number of nodes, j is the number of sampling points,is weight->Posterior probability of>For network weight, ++>For learning sample maximum a posteriori estimation, +.>For the prior probability->Is the posterior probability of the label sample.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311493247.7A CN117235628B (en) | 2023-11-10 | 2023-11-10 | Well logging curve prediction method and system based on hybrid Bayesian deep network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311493247.7A CN117235628B (en) | 2023-11-10 | 2023-11-10 | Well logging curve prediction method and system based on hybrid Bayesian deep network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117235628A CN117235628A (en) | 2023-12-15 |
| CN117235628B true CN117235628B (en) | 2024-01-26 |
Family
ID=89088330
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311493247.7A Active CN117235628B (en) | 2023-11-10 | 2023-11-10 | Well logging curve prediction method and system based on hybrid Bayesian deep network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117235628B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106556867A (en) * | 2015-09-29 | 2017-04-05 | 中国石油天然气股份有限公司 | Phase-controlled porosity inversion method based on Bayesian classification |
| CN109577972A (en) * | 2018-12-21 | 2019-04-05 | 西南石油大学 | Sandy gravel materials rock mechanics parameters Logging Evaluation Method based on lithology breakdown |
| CN109919184A (en) * | 2019-01-28 | 2019-06-21 | 中国石油大学(北京) | A kind of more well complex lithology intelligent identification Methods and system based on log data |
| CN111783825A (en) * | 2020-05-26 | 2020-10-16 | 中国石油天然气集团有限公司 | Well logging lithology identification method based on convolutional neural network learning |
| CN116244657A (en) * | 2023-04-13 | 2023-06-09 | 南京理工大学 | Train axle temperature abnormality identification method based on generation of countermeasure network and ensemble learning |
| CN116451859A (en) * | 2023-04-19 | 2023-07-18 | 淮阴工学院 | Bayesian optimization-based stock prediction method for generating countermeasure network |
| CN116821786A (en) * | 2023-05-26 | 2023-09-29 | 中海石油(中国)有限公司 | Hypotonic reservoir classification method, device and medium |
-
2023
- 2023-11-10 CN CN202311493247.7A patent/CN117235628B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106556867A (en) * | 2015-09-29 | 2017-04-05 | 中国石油天然气股份有限公司 | Phase-controlled porosity inversion method based on Bayesian classification |
| CN109577972A (en) * | 2018-12-21 | 2019-04-05 | 西南石油大学 | Sandy gravel materials rock mechanics parameters Logging Evaluation Method based on lithology breakdown |
| CN109919184A (en) * | 2019-01-28 | 2019-06-21 | 中国石油大学(北京) | A kind of more well complex lithology intelligent identification Methods and system based on log data |
| CN111783825A (en) * | 2020-05-26 | 2020-10-16 | 中国石油天然气集团有限公司 | Well logging lithology identification method based on convolutional neural network learning |
| CN116244657A (en) * | 2023-04-13 | 2023-06-09 | 南京理工大学 | Train axle temperature abnormality identification method based on generation of countermeasure network and ensemble learning |
| CN116451859A (en) * | 2023-04-19 | 2023-07-18 | 淮阴工学院 | Bayesian optimization-based stock prediction method for generating countermeasure network |
| CN116821786A (en) * | 2023-05-26 | 2023-09-29 | 中海石油(中国)有限公司 | Hypotonic reservoir classification method, device and medium |
Non-Patent Citations (1)
| Title |
|---|
| "基于贝叶斯深度学习的高维小样本数据分类研究";魏波涛;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》;第C042-3183页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117235628A (en) | 2023-12-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tahmasebi et al. | Data mining and machine learning for identifying sweet spots in shale reservoirs | |
| Gholami et al. | Applications of artificial intelligence methods in prediction of permeability in hydrocarbon reservoirs | |
| Gupta et al. | Looking ahead of the bit using surface drilling and petrophysical data: Machine-learning-based real-time geosteering in volve field | |
| Wang et al. | Marcellus shale lithofacies prediction by multiclass neural network classification in the Appalachian Basin | |
| Tadeusiewicz | Neural networks in mining sciences–general overview and some representative examples | |
| Rostamian et al. | Evaluation of different machine learning frameworks to predict CNL-FDC-PEF logs via hyperparameters optimization and feature selection | |
| Shi et al. | Brittleness index prediction in shale gas reservoirs based on efficient network models | |
| Zhu et al. | Rapid identification of high-quality marine shale gas reservoirs based on the oversampling method and random forest algorithm | |
| J Al-Mudhafar | Integrating lithofacies and well logging data into smooth generalized additive model for improved permeability estimation: Zubair formation, South Rumaila oil field | |
| Tan et al. | Evaluation of complex petroleum reservoirs based on data mining methods | |
| CN114638255A (en) | Tight sandstone fracture logging comprehensive identification method combined with deep learning | |
| Xue et al. | An automated data-driven pressure transient analysis of water-drive gas reservoir through the coupled machine learning and ensemble Kalman filter method | |
| CN114741944A (en) | Method for predicting logging curve by using machine learning and deep learning algorithm | |
| CN114114414A (en) | Artificial intelligence prediction method for 'dessert' information of shale reservoir | |
| Verma et al. | Quantification of sand fraction from seismic attributes using Neuro-Fuzzy approach | |
| Du et al. | A deep learning framework using graph convolutional networks for adaptive correction of interwell connectivity and gated recurrent unit for performance prediction | |
| CN117473305A (en) | Method and system for predicting reservoir parameters enhanced by neighbor information | |
| CN110118994B (en) | Continental facies hydrocarbon source rock quantitative prediction method based on seismic inversion and machine learning | |
| Djebbas et al. | Integrating hydraulic flow unit concept and adaptive neuro-fuzzy inference system to accurately estimate permeability in heterogeneous reservoirs: Case study Sif Fatima oilfield, southern Algeria | |
| CN117235628B (en) | Well logging curve prediction method and system based on hybrid Bayesian deep network | |
| Verga et al. | Improved application of assisted history matching techniques | |
| Anifowose et al. | A functional networks-type-2 fuzzy logic hybrid model for the prediction of porosity and permeability of oil and gas reservoirs | |
| Li et al. | Lithology classification based on set-valued identification method | |
| Huang et al. | Quantitative analysis of the main controlling factors of oil saturation variation | |
| Temizel et al. | Hybrid modeling in unconventional reservoirs to forecast estimated ultimate recovery |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |