CN112100930A - Formation pore pressure calculation method based on convolutional neural network and Eaton formula - Google Patents
Formation pore pressure calculation method based on convolutional neural network and Eaton formula Download PDFInfo
- Publication number
- CN112100930A CN112100930A CN202011249812.1A CN202011249812A CN112100930A CN 112100930 A CN112100930 A CN 112100930A CN 202011249812 A CN202011249812 A CN 202011249812A CN 112100930 A CN112100930 A CN 112100930A
- Authority
- CN
- China
- Prior art keywords
- pore pressure
- normal
- neural network
- well
- convolutional neural
- 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.)
- Granted
Links
- 239000011148 porous material Substances 0.000 title claims abstract description 58
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 39
- 230000015572 biosynthetic process Effects 0.000 title claims abstract description 38
- 238000004364 calculation method Methods 0.000 title abstract description 18
- 238000005056 compaction Methods 0.000 claims abstract description 70
- 238000005070 sampling Methods 0.000 claims abstract description 48
- 238000000034 method Methods 0.000 claims abstract description 27
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 238000012549 training Methods 0.000 claims description 15
- 238000005553 drilling Methods 0.000 claims description 11
- 230000006870 function Effects 0.000 claims description 10
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 4
- 208000019888 Circadian rhythm sleep disease Diseases 0.000 claims description 3
- 208000001456 Jet Lag Syndrome Diseases 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 3
- 230000002706 hydrostatic effect Effects 0.000 claims description 3
- 208000033915 jet lag type circadian rhythm sleep disease Diseases 0.000 claims description 3
- 238000001228 spectrum Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 5
- 238000012360 testing method Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 238000013519 translation Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000004836 empirical method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- 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
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V11/00—Prospecting or detecting by methods combining techniques covered by two or more of main groups G01V1/00 - G01V9/00
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
- G06F17/141—Discrete Fourier transforms
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Pure & Applied Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Discrete Mathematics (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Geophysics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
A method for calculating formation pore pressure based on a convolutional neural network and an Eaton formula comprises the following steps: overlapping and sampling the logging curves; sample 1D-2D transformation preprocessing based on short-time Fourier transformation, and converting a one-dimensional depth domain well logging curve sample into a two-dimensional deep frequency map; based on the intelligent recognition model of the normal compacted layer section of the convolutional neural network, extracting the segmented characteristics of the logging curve in a data driving mode, and recognizing the normal compacted layer section; constructing a normal compaction trend line fitting equation according to the identified logging curve of the normal compaction layer section; and (4) calculating the formation pore pressure profile by adopting an Eaton formula according to a normal compaction trend line equation. The invention adopts a stratum pore calculation method based on the combination of data driving and a physical model, avoids the artificial subjectivity in the process of constructing the normal compaction trend line, and improves the calculation precision of the stratum pore pressure.
Description
Technical Field
The invention relates to a stratum pore pressure calculation method based on a convolutional neural network and an Eaton formula, and belongs to the technical field of oil and gas drilling.
Background
The formation pore pressure is one of important parameters for researching the stability of a well wall, carrying out drilling engineering design and evaluating a reservoir, effectively and accurately carrying out the prediction of the formation pore pressure, and has important significance for fast and excellent drilling and protecting an oil gas reservoir. The Eaton method is the most common stratum pore pressure calculation method at present, and the method utilizes seismic or logging data to calculate the stratum pore pressure according to a normal compaction trend line, and has the characteristics of good continuity, high resolution, strong practicability and the like. The normal compaction trend line is constructed on the premise that the Eaton method is used for calculating the pore pressure of the stratum, according to the deposit compaction theory, the porosity of the stratum which is not compacted by the mudstone is larger than that of the normal compaction condition, and a parameter curve deviates from the normal compaction trend line in the logging process. In the prior art, a normal compaction interval is subjectively set and fitted to obtain a normal compaction trend line, and then the formation pore pressure in the whole well depth range in the longitudinal direction is calculated based on the normal compaction trend line. Due to the complexity of a drilling geological environment, the fuzziness of relevant interpretation data such as seismic logging and the subjectivity of manual judgment and the like, the constructed normal compaction trend line has uncertainty, and finally, errors of different degrees exist in a stratum pore pressure prediction result.
In recent years, the application of artificial intelligence and machine learning technology in the field of intelligent drilling is increasingly perfected, and an analysis method based on data driving brings a new idea for technical personnel in the analysis and research of engineering fields such as drilling design, geological characteristic parameter prediction and the like. The convolutional neural network is a supervised deep learning algorithm and is widely applied to the field of pattern recognition. The convolutional neural network is different from other machine learning algorithms, has the characteristics of local receptive field, weight sharing and pooling, and performs feature extraction on the image through the convolutional layer, so that the complexity of a network model is reduced through the weight sharing and pooling, the risk of overfitting is also reduced, and a deep learning framework for processing mass data can be constructed. The existing special patent mainly aims at the mode recognition of the convolutional neural network by time-frequency transformation of signals in a time domain, but the seismic and logging data information of a depth domain in the technical field of drilling cannot embody the advantage of the convolutional neural network in the aspect of processing big data.
Therefore, it is necessary to establish a formation pore pressure calculation method based on the combination of data driving and a physical model by considering a sampling rule and a signal processing mode and combining a convolutional neural network and an Eaton formula aiming at the calculation of the formation pore pressure in the technical field of drilling.
Disclosure of Invention
Aiming at the defects of the prior art, the invention discloses a stratum pore pressure calculation method based on a convolutional neural network and an Eaton formula.
Summary of the invention:
aiming at the uncertainty existing in the construction process of a normal compaction trend line in formation pore pressure calculation and the insufficient data quantity of logging information sampling in a depth domain, sampling is carried out in an overlapping sampling mode, short-time Fourier transform is carried out on a sampling sample, 1D-2D transformation preprocessing is carried out on a logging curve sampling sample, and the sampling sample is converted into a two-dimensional deep frequency map with depth domain and frequency domain information; then inputting the deep frequency image sampling sample into a convolutional neural network for carrying out feature adaptive learning of stratum compaction, establishing an intelligent identification model of a normal compaction layer section, and realizing end-to-end normal compaction layer section identification; and finally, applying the trained model to the target well to identify the normal compaction interval, fitting a normal compaction trend line equation, and substituting the normal compaction trend line equation into an Eaton formula to calculate the formation pore pressure in the whole well depth range in the longitudinal direction. By adopting the stratum pore calculation method based on the combination of data driving and a physical model, the artificial subjectivity in the normal compaction trend line construction process is avoided, and the calculation precision of the stratum pore pressure is improved.
The specific technical scheme of the invention is as follows:
a method for calculating the formation pore pressure based on a convolutional neural network and an Eaton formula is characterized by comprising the following steps:
1) selecting well logging data of the completed well to perform model training: carrying out overlapped sampling on logging curve parameters for calculating formation pore pressure, and respectively carrying out short-time Fourier transform to obtain a deep frequency map corresponding to each sampling sample; the logging curve parameters for calculating the formation pore pressure are logging information such as the layer velocity of the completed well and the like which can be used for calculating the formation pore pressure, and the logging information comprises but is not limited to: seismic interval velocity, acoustic moveout, resistivity, and apparent density;
2) inputting the deep frequency map obtained in the step 1) into a convolutional neural network for training, realizing the self-adaptive feature extraction of the normal compacted layer section, and obtaining an intelligent identification model of the normal compacted layer section;
3) performing overlapping sampling and short-time Fourier transform on the logging curve parameters of the target well pair applied to calculating the formation pore pressure according to the step 1) to obtain a deep frequency map corresponding to each sample; inputting the sample deep frequency map in the step into the intelligent identification model of the normal compaction layer section in the step 2) to identify the optimal normal compaction layer section;
4) fitting and constructing a normal compaction trend line equation and substituting the normal compaction trend line equation into an Eaton formula;
5) and calculating the stratum pore pressure profile of the target well to be measured.
Preferably according to the invention, in step 4), the normal compaction trend line equation is fit and constructed as follows:
wherein,
normal compacted layer speed, km/s;
normal compaction resistivity, Ω;
for normal compaction of apparent density, g/cm3；
Us/ft is the normal compaction sound wave time difference;
m, well depth;
is a constant;
the real surface speed, km/s;
the time surface resistivity, Ω;
apparent density of the earth's surface in g/cm3；
Time-surface acoustic moveout, us/ft.
The invention preferably samples the stratum speed curve of a certain stratum pore pressure target well to be calculated, selects the stratum above the well section for sampling in order to reduce the calculation amount, and according to the field experience, the newly tied upper strata always show normal compaction. And carrying out 1D-2D conversion pretreatment on each sample to obtain a deep frequency image of each sample to be identified, then bringing the deep frequency image into the established intelligent recognition model of the normal compacted interval to obtain the recognition result of each sample, and taking the sample with the highest normal compaction goodness of fit as the final normal compacted interval.
According to the invention, preferably, the step 5) of calculating the formation pore pressure profile of the target well to be measured comprises:
substituting the obtained formula (II) into an Eaton formula to predict the formation pore pressure profile of the whole well section, wherein the Eaton formula is as follows:
wherein,
is the pore pressure of the stratum in g/cm3；
G/cm of overburden pressure3；
Is hydrostatic column pressure, g/cm3；
The Eaton index is generally 1.6;
the stratum velocity is km/s, and can be obtained through seismic wave reflection time of seismic exploration before drilling and also through acoustic jet-lag logging.
According to a preferred embodiment of the present invention, the overlapped sampling in step 1) is: according to step length for logging curvekOverlapping sampling is performed. Wherein the sample length can be selected empirically or computationally assuming a length ofL，The well logging curves include a layer velocity curve, an acoustic time difference curve, a resistivity curve and an apparent density curve, and are all suitable for constructing a normal compaction trend line by adopting the method.
Preferably, the short-time fourier transform in step 1) is based on a 1D-2D transform preprocessing of the samples of the short-time fourier transform: intercepting a sampling sample by using a window function according to the well depth, carrying out Fourier transform on an intercepted logging curve, arranging frequency spectrums obtained by each section of intercepted sample according to a depth axis, and finally obtaining a two-dimensional deep frequency graph with time-varying characteristics, wherein a short-time Fourier transform formula:
in the formula,
in order to obtain a log of the well,
is centered at the well depth
The window function of (a) is determined,
the frequency domain parameters representing the fourier transform,
representing the depth domain parameters of the well log.
Preferably, the method for training the normal compacted interval intelligent recognition model in the step 2) comprises the following steps:
and calibrating the compaction types corresponding to the deep frequency maps of the completed wells, and combining the deep frequency maps of the calibrated compaction types into a training set to be input into the convolutional neural network for learning. Wherein, the calibrated deep frequency graph belongs to a completed well, and the stratum compaction type can be calibrated according to well history data; the convolutional neural network is preferably a classical convolutional neural network structure model such as LetNet-5, and the optimal convolutional neural network structure is assumed to be shown in FIG. 2.
Preferably, according to the present invention, the network parameters are continuously updated by using a back propagation algorithm for the convolutional neural network.
The technical advantages of the invention are as follows:
according to the method, the data-driven intelligent identification model of the normal compacted interval is established through short-time Fourier transform and the convolution deep well network, so that subjectivity and uncertainty of manual selection of the normal compacted interval in the traditional stratum pore pressure calculation process are avoided, and the stratum pore pressure calculation precision is improved.
Drawings
FIG. 1 is a flow chart of the technical solution of the present invention.
FIG. 2 is a diagram of the architecture of the present invention for setting the optimal convolutional neural network.
FIG. 3 is a graph of A-well zone velocity.
Fig. 4 is a sample plot for a well.
FIG. 5 is a time-frequency diagram of sample No. 1 of well A.
FIG. 6 is a pore pressure profile of a well A formation.
Detailed Description
The invention is described in detail below with reference to the following embodiments and drawings, but is not limited thereto.
The implementation case is as follows:
taking a certain block of well A as an example, taking the well A as a target well to be calculated for the formation pore pressure, taking 10 drilled adjacent wells of the well A as modeling training samples, selecting a recent upper strata of the well A with the depth of less than 1500m as a sampling interval of a normal compacted layer section to sample a test sample, wherein the layer speed curve is shown in figure 3, and the data of the figure is original data measured by a seismic or well logging tool.
A method for calculating formation pore pressure based on a convolutional neural network and an Eaton formula comprises the following steps:
1) selecting well logging data of the completed well to perform model training: carrying out overlapped sampling on logging curve parameters for calculating formation pore pressure, and respectively carrying out short-time Fourier transform to obtain a deep frequency map corresponding to each sampling sample;
the overlapped sampling is to perform overlapped sampling on the stratum speed curves of the well A and 10 drilled adjacent wells, the sampling sliding step length is 10m, the sample length is 200m, and the sampling frequency is 10 hz. The method comprises the following steps of firstly, sampling a target A well, sampling 10 adjacent wells, and calibrating the stratum compaction condition of the sampling samples of the 10 adjacent wells according to well history data, wherein the sampling samples of the target A well are a test set, the sampling samples of the 10 adjacent wells are a training set, and the stratum compaction condition of the sampling samples of the 10 adjacent wells is calibrated. The sampling condition of the A well is shown in FIG. 4, and FIG. 4 is a schematic diagram of sampling interval sampling of a normal compacted interval of the A well.
The short-time Fourier transform in the step 1) is based on sample 1D-2D transform preprocessing of the short-time Fourier transform: intercepting a sampling sample by using a window function according to the well depth, carrying out Fourier transform on an intercepted logging curve, arranging frequency spectrums obtained by each section of intercepted sample according to a depth axis, and finally obtaining a two-dimensional deep frequency graph with time-varying characteristics, wherein a short-time Fourier transform formula:
in the formula,
in order to obtain a log of the well,
as a window function centered at the depth of the well,
the frequency domain parameters representing the fourier transform,
representing the depth domain parameters of the well log.
And carrying out short-time Fourier transform on the sampling samples of the well A and the adjacent well, wherein the window function is a Hamming window, the window length is set to be 512, the translation step length of the window each time is 1, and the frequency scale is set to be 0.5:0.5: 5.5. And finally, obtaining a time-frequency diagram of each sampling sample, wherein the time-frequency diagram of the No. 1 sample of the well A is shown in figure 5. The selection of the parameters needs to be determined through multiple tests, and experiments show that the depth resolution is poorer as the window length is longer, and the frequency resolution is better; the shorter the window length, the better the depth resolution and the poorer the frequency resolution. The smaller the window translation step size is, the better the depth precision of the deep frequency image is, but the calculation amount is increased. The frequency scale is determined according to the sampling frequency, and the maximum scale is generally 1/2 of the sampling frequency.
2) Inputting the deep frequency map obtained in the step 1) into a convolutional neural network for training, realizing the self-adaptive feature extraction of the normal compacted layer section, and obtaining an intelligent identification model of the normal compacted layer section;
the method for training to obtain the intelligent recognition model of the normal compacted layer section comprises the following steps:
and calibrating the compaction types corresponding to the deep frequency maps of the completed wells, and combining the deep frequency maps of the calibrated compaction types into a training set to be input into the convolutional neural network for learning. Wherein, the calibrated deep frequency graph belongs to a completed well, and the stratum compaction type can be calibrated according to well history data; the convolutional neural network preferably selects classic convolutional neural network structure models such as LetNet-5, the optimal convolutional neural network structure is selected in the embodiment as shown in FIG. 2, and parameters of each layer of the optimal neural network structure are selected as shown in Table 1 and are a 7-layer CNN model: In-C1-P1-C2-P2
-F-Out. Taking the adjacent well sampling sample deep frequency map of the calibration compaction type obtained in the step as a training set, and selecting a loss function as a logarithmic loss function:
wherein y is an output variable, x is an input variable,Lin order to be a function of the loss,Nin order to input the amount of samples,Mas to the number of possible categories, the number of categories,
is a binary index and represents the categoryjWhether it is an input instance x i The true category of (a) of (b),
predicting input x for a model i Belong to the categoryjThe probability of (c).
And updating and learning network parameters by taking a random gradient descent algorithm in the back propagation algorithm as an optimization algorithm, and finally establishing a normal compaction layer section intelligent identification model based on the convolutional neural network.
TABLE 1 optimal convolutional neural network architecture
3) Performing overlapping sampling and short-time Fourier transform on the logging curve parameters of the target well pair applied to calculating the formation pore pressure according to the step 1) to obtain a deep frequency map corresponding to each sample; inputting the sample deep frequency map in the step into the intelligent identification model of the normal compaction layer section in the step 2) to identify the optimal normal compaction layer section;
4) fitting and constructing a normal compaction trend line equation and substituting the normal compaction trend line equation into an Eaton formula;
in step 4), fitting and constructing a normal compaction trend line equation as follows:
wherein,
normal compacted layer speed, km/s;
normal compaction resistivity, Ω;
for normal compaction of apparent density, g/cm3；
Us/ft is the normal compaction sound wave time difference;
m, well depth;
is a constant;
is composed of
The real surface speed, km/s;
is composed of
The time surface resistivity, Ω;
is composed of
Apparent density of the earth's surface in g/cm3；
Time-surface acoustic moveout, us/ft.
Specifically, the deep frequency map of the well A sampling sample obtained in the sampling mode in the step 1) is taken as a test set to be brought into the normal compaction layer section intelligent identification model established in the step 2), and the identification result of each sample is obtained, as shown in table 2. The method can know that the normal compaction goodness of fit of the sample corresponding to the well section is the highest, and the sample well section 935-1135 m is selected as the normal compaction well section of the well A. And (3) carrying out normal compaction trend line equation fitting according to the well section layer speed curve to obtain a fitting equation as follows:
and K1 and K2 are calculated by a sampling polynomial fitting method according to the layer speed and depth data of the normal compaction layer section.
TABLE 2A identification of Normal compacted intervals for a well
| Sample number | ① | ② | ③ | ④ | ⑤ |
| Normal compaction | 0.852 | 0.733 | 0.531 | 0.422 | 0.574 |
| Abnormal compaction | 0.148 | 0.267 | 0.469 | 0.578 | 0.426 |
5) Calculating a stratum pore pressure profile of a target well to be measured, comprising:
substituting the obtained formula (II) into an Eaton formula to predict the formation pore pressure profile of the whole well section, wherein the Eaton formula is as follows:
wherein,
is the pore pressure of the stratum in g/cm3；
G/cm of overburden pressure3；
Is hydrostatic column pressure, g/cm3；
The Eaton index is generally 1.6;
the stratum velocity is km/s, and can be obtained through seismic wave reflection time of seismic exploration before drilling and also through acoustic jet-lag logging.
Specifically, according to the normal compaction trend line equation of the well A obtained in the step 4), calculating the normal compaction layer velocity of the well depth corresponding to the well layer velocity measuring points of the well A, and then substituting into an Eaton formula to calculate the formation pore pressure of each measuring point, wherein the Eaton formula is as follows:
and finally obtaining a pore pressure profile of the A well stratum, as shown in figure 6. By comparing the formation pore pressure actual measurement points, the formation pore pressure profile obtained by calculation of the method has good fitting degree with the actual measurement points, and compared with the conventional empirical method in which a normal compaction interval is subjectively set, the calculation precision of the formation pore pressure is improved.
In addition, the network parameters are continuously updated by adopting a back propagation algorithm for the convolutional neural network.
Claims (7)
1. A method for calculating the formation pore pressure based on a convolutional neural network and an Eaton formula is characterized by comprising the following steps:
1) selecting well logging data of the completed well to perform model training: carrying out overlapped sampling on logging curve parameters for calculating formation pore pressure, and respectively carrying out short-time Fourier transform to obtain a deep frequency map corresponding to each sampling sample;
2) inputting the deep frequency map obtained in the step 1) into a convolutional neural network for training, realizing the self-adaptive feature extraction of the normal compacted layer section, and obtaining an intelligent identification model of the normal compacted layer section;
3) performing overlapping sampling and short-time Fourier transform on the logging curve parameters of the target well pair applied to calculating the formation pore pressure according to the step 1) to obtain a deep frequency map corresponding to each sample; inputting the sample deep frequency map in the step into the intelligent identification model of the normal compaction layer section in the step 2) to identify the optimal normal compaction layer section;
4) fitting and constructing a normal compaction trend line equation and substituting the normal compaction trend line equation into an Eaton formula;
and calculating the stratum pore pressure profile of the target well to be measured.
2. The method for calculating the formation pore pressure based on the convolutional neural network and the Eaton formula as claimed in claim 1, wherein in step 4), the fitting and construction of the normal compaction trend line equation is as follows:
wherein,
normal compacted layer speed, km/s;
normal compaction resistivity, Ω;
for normal compaction of apparent density, g/cm3；
Us/ft is the normal compaction sound wave time difference;
m, well depth;
is a constant;
is composed of
The real surface speed, km/s;
is composed of
The time surface resistivity, Ω;
is composed of
Apparent density of the earth's surface in g/cm3；
Time-surface acoustic moveout, us/ft.
3. The method for calculating the formation pore pressure based on the convolutional neural network and the Eaton formula as claimed in claim 1, wherein the step 5) of calculating the formation pore pressure profile of the target well to be measured comprises:
substituting the obtained formula (II) into an Eaton formula to predict the formation pore pressure profile of the whole well section, wherein the Eaton formula is as follows:
wherein,
is the pore pressure of the stratum in g/cm3；
G/cm of overburden pressure3；
Is hydrostatic column pressure, g/cm3；
The Eaton index is generally 1.6;
the stratum velocity is km/s, and can be obtained through seismic wave reflection time of seismic exploration before drilling and also through acoustic jet-lag logging.
4. The method for calculating the formation pore pressure based on the convolutional neural network and the Eaton formula as claimed in claim 1, wherein the overlapped sampling in step 1) is: according to step length for logging curvekOverlapping sampling is performed.
5. The method for calculating the formation pore pressure based on the convolutional neural network and the Eaton formula as claimed in claim 1, wherein the short-time fourier transform in step 1) is based on sample 1D-2D transform preprocessing of the short-time fourier transform: intercepting a sampling sample by using a window function according to the well depth, carrying out Fourier transform on an intercepted logging curve, arranging frequency spectrums obtained by each section of intercepted sample according to a depth axis, and finally obtaining a two-dimensional deep frequency graph with time-varying characteristics, wherein a short-time Fourier transform formula:
in the formula,
in order to obtain a log of the well,
is centered at the well depth
The window function of (a) is determined,
the frequency domain parameters representing the fourier transform,
representing the depth domain parameters of the well log.
6. The method for calculating the formation pore pressure based on the convolutional neural network and the Eaton formula as claimed in claim 1, wherein the method for training to obtain the intelligent recognition model of the normal compacted interval in step 2) comprises:
and calibrating the compaction types corresponding to the deep frequency graphs of the completed wells, and inputting the time-frequency graph sets of the calibrated compaction types into the convolutional neural network for learning by combining as a training set.
7. The method of claim 1, wherein the network parameters are continuously updated using a back propagation algorithm for the convolutional neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011249812.1A CN112100930B (en) | 2020-11-11 | 2020-11-11 | Formation pore pressure calculation method based on convolutional neural network and Eaton formula |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011249812.1A CN112100930B (en) | 2020-11-11 | 2020-11-11 | Formation pore pressure calculation method based on convolutional neural network and Eaton formula |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112100930A true CN112100930A (en) | 2020-12-18 |
| CN112100930B CN112100930B (en) | 2021-02-02 |
Family
ID=73785090
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011249812.1A Active CN112100930B (en) | 2020-11-11 | 2020-11-11 | Formation pore pressure calculation method based on convolutional neural network and Eaton formula |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112100930B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113283182A (en) * | 2021-07-06 | 2021-08-20 | 中海石油(中国)有限公司 | Method, device, medium and equipment for predicting and analyzing formation pressure |
| CN113486539A (en) * | 2021-08-09 | 2021-10-08 | 中国石油大学(华东) | Method for analyzing uncertainty of formation pressure in whole deep water drilling process |
| CN114693005A (en) * | 2022-05-31 | 2022-07-01 | 中国石油大学(华东) | Three-dimensional underground oil reservoir dynamic prediction method based on convolution Fourier neural network |
| CN115059448A (en) * | 2022-06-01 | 2022-09-16 | 中国石油大学(华东) | Stratum pressure monitoring method based on deep learning algorithm |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107843927A (en) * | 2016-09-20 | 2018-03-27 | 中国石油化工股份有限公司 | Shale formation pressure prediction method and device based on well shake joint speed |
| CN109509111A (en) * | 2017-09-15 | 2019-03-22 | 中国石油化工股份有限公司 | The prediction technique and system of prospect pit strata pressure |
| CN109736784A (en) * | 2018-04-27 | 2019-05-10 | 长江大学 | Sedimentary rock formations pore pressure prediction calculation method |
| 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 |
-
2020
- 2020-11-11 CN CN202011249812.1A patent/CN112100930B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107843927A (en) * | 2016-09-20 | 2018-03-27 | 中国石油化工股份有限公司 | Shale formation pressure prediction method and device based on well shake joint speed |
| CN109509111A (en) * | 2017-09-15 | 2019-03-22 | 中国石油化工股份有限公司 | The prediction technique and system of prospect pit strata pressure |
| CN109736784A (en) * | 2018-04-27 | 2019-05-10 | 长江大学 | Sedimentary rock formations pore pressure prediction calculation method |
| 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 |
Non-Patent Citations (5)
| Title |
|---|
| LEILA ALIOUANE 等: "Pore Pressure prediction in shale gas reservoirs using neural network and fuzzy logic with an application to Barnett Shale.", 《GEOPHYSICAL RESEARCH ABSTRACTS》 * |
| MORTEZA AZADPOUR 等: "Pore pressure prediction and modeling using well-logging data in oneofthegas fields in south of Iran", 《JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING》 * |
| 李昌盛 等: "基于遗传算法优化BP神经网络的地层破裂压力预测方法", 《西安石油大学学报（自然科学版）》 * |
| 李萌: "用阵列感应测井评价地层压力", 《中国优秀硕士学位论文全文数据库 基础科学辑》 * |
| 陈子剑: "东海西湖凹陷低孔低渗气藏地层孔隙压力研究", 《中国博士学位论文全文数据库 工程科技Ⅰ辑》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113283182A (en) * | 2021-07-06 | 2021-08-20 | 中海石油(中国)有限公司 | Method, device, medium and equipment for predicting and analyzing formation pressure |
| CN113283182B (en) * | 2021-07-06 | 2023-09-05 | 中海石油(中国)有限公司 | Formation pressure prediction analysis method, device, medium and equipment |
| CN113486539A (en) * | 2021-08-09 | 2021-10-08 | 中国石油大学(华东) | Method for analyzing uncertainty of formation pressure in whole deep water drilling process |
| CN113486539B (en) * | 2021-08-09 | 2024-08-09 | 中国石油大学(华东) | Stratum pressure uncertainty analysis method in whole deep water drilling process |
| CN114693005A (en) * | 2022-05-31 | 2022-07-01 | 中国石油大学(华东) | Three-dimensional underground oil reservoir dynamic prediction method based on convolution Fourier neural network |
| CN115059448A (en) * | 2022-06-01 | 2022-09-16 | 中国石油大学(华东) | Stratum pressure monitoring method based on deep learning algorithm |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112100930B (en) | 2021-02-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112100930B (en) | Formation pore pressure calculation method based on convolutional neural network and Eaton formula | |
| CN110954948B (en) | Physical property parameter inversion method and system for rock physical constraint reservoir | |
| CN112162316B (en) | High-resolution well-seismic fusion prestack inversion method driven by AVO waveform data | |
| CN114861563B (en) | Method, device, medium and equipment for predicting formation pressure in physical embedding deep learning | |
| CN104533400A (en) | Method for reconstructing logging curve | |
| CN116305850B (en) | Stratum thermal conductivity prediction method based on random forest model | |
| CN111722283B (en) | Stratum velocity model building method | |
| CN115049173B (en) | Deep learning and Eaton method coupling driving stratum pore pressure prediction method | |
| CN114529110A (en) | Lithofacies inversion method and system based on deep neural network model | |
| CN114488311A (en) | Transverse wave time difference prediction method based on SSA-ELM algorithm | |
| CN114065909A (en) | Well logging curve completion method based on CNN _ AB _ Bi-LSTM | |
| CN114114414A (en) | Artificial intelligence prediction method for 'dessert' information of shale reservoir | |
| CN118153420A (en) | Deep learning and digital twin-based wellbore trajectory prediction method, system, equipment and medium | |
| CN110118994B (en) | Continental facies hydrocarbon source rock quantitative prediction method based on seismic inversion and machine learning | |
| Sarkheil et al. | The fracture network modeling in naturally fractured reservoirs using artificial neural network based on image loges and core measurements | |
| CN111751878A (en) | Method and device for predicting transverse wave velocity | |
| CN116413831A (en) | Multi-well synthetic seismic record automatic calibration and construction interpretation and reservoir prediction method | |
| CN115809587A (en) | Prestack gas-containing prediction method based on combination of deep network and AVO analysis | |
| CN112859172A (en) | Longitudinal and transverse wave micro-logging data processing method and device | |
| CN108875109B (en) | Method and system for predicting abnormal formation pressure | |
| CN117310805B (en) | Evolution characterization method for formation of fracture porosity related to faults | |
| CN115327643B (en) | Machine learning training sample expansion and evaluation method for intelligent oil gas detection | |
| CN116047598B (en) | Oil exploration method and system based on artificial intelligence | |
| CN118210059A (en) | Reservoir fracture prediction method and system | |
| CN114594523A (en) | Physical constraint and data-driven nuclear magnetic resonance T construction2Spectrum method and device |
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 |