CN111751878B - Method and device for predicting transverse wave speed - Google Patents
Method and device for predicting transverse wave speed Download PDFInfo
- Publication number
- CN111751878B CN111751878B CN202010434118.0A CN202010434118A CN111751878B CN 111751878 B CN111751878 B CN 111751878B CN 202010434118 A CN202010434118 A CN 202010434118A CN 111751878 B CN111751878 B CN 111751878B
- Authority
- CN
- China
- Prior art keywords
- curve
- transverse wave
- learning rate
- wave speed
- neural network
- 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 83
- 238000012549 training Methods 0.000 claims abstract description 92
- 238000003062 neural network model Methods 0.000 claims abstract description 71
- 238000012360 testing method Methods 0.000 claims description 38
- 230000006870 function Effects 0.000 claims description 22
- 239000011159 matrix material Substances 0.000 claims description 19
- 230000004913 activation Effects 0.000 claims description 17
- 238000004590 computer program Methods 0.000 claims description 16
- 238000010606 normalization Methods 0.000 claims description 12
- 238000011156 evaluation Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 11
- 238000003860 storage Methods 0.000 claims description 8
- 238000012544 monitoring process Methods 0.000 claims description 7
- 238000012935 Averaging Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000009826 distribution Methods 0.000 claims description 4
- 238000005315 distribution function Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 abstract description 14
- 238000004519 manufacturing process Methods 0.000 abstract description 9
- 238000010586 diagram Methods 0.000 description 14
- 238000004088 simulation Methods 0.000 description 6
- 238000013135 deep learning Methods 0.000 description 4
- 229910052500 inorganic mineral Inorganic materials 0.000 description 4
- 239000011707 mineral Substances 0.000 description 4
- 230000015572 biosynthetic process Effects 0.000 description 3
- 239000012530 fluid Substances 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000008021 deposition Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000002637 fluid replacement therapy Methods 0.000 description 2
- 238000009472 formulation Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 239000004215 Carbon black (E152) Substances 0.000 description 1
- 235000004257 Cordia myxa Nutrition 0.000 description 1
- 244000157795 Cordia myxa Species 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000004927 clay Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013213 extrapolation Methods 0.000 description 1
- 229930195733 hydrocarbon Natural products 0.000 description 1
- 150000002430 hydrocarbons Chemical class 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000010453 quartz Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000003079 shale oil Substances 0.000 description 1
- VYPSYNLAJGMNEJ-UHFFFAOYSA-N silicon dioxide Inorganic materials O=[Si]=O VYPSYNLAJGMNEJ-UHFFFAOYSA-N 0.000 description 1
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
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
- G01V1/30—Analysis
- G01V1/303—Analysis for determining velocity profiles or travel times
-
- 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
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Remote Sensing (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Acoustics & Sound (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- Health & Medical Sciences (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Geophysics (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention provides a method and a device for predicting transverse wave speed, wherein the method comprises the following steps: acquiring a conventional logging curve and a known transverse wave velocity curve; determining a preferred curve from the conventional log curve and the known cross-well velocity curve; normalizing the optimized curve to determine a normalized optimized curve; establishing a depth feedforward neural network model by combining a conventional logging curve, training by using the known well transverse wave speed, and determining a transverse wave speed prediction model; and inputting the normalized optimized curve into a transverse wave speed prediction model to determine the transverse wave speed. In the whole transverse wave speed prediction process, only the transverse wave speed curve of the well and the conventional logging curve of the predicted well are needed, the accurate prediction of the transverse wave speed is realized directly from data, the adjustment parameters are few, the requirements on personnel are low, and the method can be widely applied to actual production.
Description
Technical Field
The invention relates to the technical field of geophysics, in particular to a transverse wave speed prediction method and device.
Background
This section is intended to provide a background or context to the embodiments of the invention that are recited in the claims. The description herein is not admitted to be prior art by inclusion in this section.
The transverse wave speed is used as an important parameter in fluid replacement and prestack forward modeling and inversion, and in the seismic forward modeling process, the accuracy of the result directly influences the result after the subsequent fluid replacement, so that the understanding of the elastic parameters and seismic response change rules of geophysical workers under different fluid saturation conditions is influenced; in the seismic prestack inversion process, low-frequency information is lost in the seismic, a low-frequency model is often obtained through interpolation and extrapolation of a well logging, and the low-frequency model is often used for controlling the whole deposition background and deposition rule of a research area, so that the accuracy of the transverse wave speed on the well logging also influences the result of the whole prestack inversion, and the importance of the transverse wave speed prediction is reflected. In unconventional reservoirs such as shale gas, compact oil and the like, formation parameters such as formation stress, formation pressure and the like can be calculated only under the condition of accurate transverse wave speed. Therefore, the transverse wave velocity plays a vital role in the prediction and evaluation of conventional and unconventional hydrocarbon reservoir parameters.
The current method for predicting the transverse wave speed by using logging data mainly comprises an empirical formula method and a petrophysical modeling method. The empirical formula method is often a unit or multiple linear regression method to establish the functional relation between the transverse wave velocity curve and the conventional logging curves such as gamma, neutron and longitudinal wave time difference, and the fitting function is often too simple, so that the precision is difficult to meet the production requirement. In recent years, a plurality of domestic and foreign scholars develop a petrophysical modeling method to predict the transverse waves, the method needs to have accurate evaluation on stratum, has high requirements on earthquake reservoir prediction workers, needs to develop evaluation on key parameters such as mineral content, porosity and the like which meet the requirements of petrophysical modeling, and also needs to evaluate TOC, gas content and the like in shale oil. In addition, the input skeleton parameters are numerous, 8 parameters are involved under two mineral conditions, 4 parameters are added every time one mineral is added subsequently, and parameters such as stratum fluid property, temperature, pressure and the like are difficult to obtain, so that in the petrophysical simulation process of an unconventional reservoir such as compact oil, because of the numerous input parameters, the parameters which meet the conditions of an actual working area are difficult to obtain, and therefore, many difficulties are caused in the actual operation process, and the prediction accuracy is poor.
The Vp/Vs calculated by the linear fitting method through the empirical formula is basically a constant, and cannot meet the requirements of high-precision AVO forward modeling and prestack elastic parameter inversion.
The existing petrophysical modeling method has the defects that the adjustable parameters are numerous, some parameters are difficult to obtain, the flow is complex, operators are required to have professional backgrounds of logging interpretation and seismic petrophysical modeling, and the method is difficult to popularize and apply in actual production.
Therefore, how to provide a new solution to the above technical problem is a technical problem to be solved in the art.
Disclosure of Invention
The embodiment of the invention provides a method for predicting transverse wave speed, which directly starts from data and realizes accurate prediction of transverse wave speed, and comprises the following steps:
acquiring a conventional logging curve and a known transverse wave velocity curve;
determining a preferred curve from the conventional log curve and the known cross-well velocity curve;
normalizing the optimized curve to determine a normalized optimized curve;
establishing a depth feedforward neural network model by combining a conventional logging curve, training by using the known well transverse wave speed, and determining a transverse wave speed prediction model;
and inputting the normalized optimized curve into a transverse wave speed prediction model to determine the transverse wave speed.
The embodiment of the invention also provides a device for predicting the transverse wave speed, which comprises the following steps:
the data acquisition module is used for acquiring a conventional logging curve and a known well transverse wave speed curve;
the optimization curve determining module is used for determining an optimization curve according to a conventional logging curve and a known transverse wave speed curve;
the normalization module is used for carrying out normalization processing on the optimized curve and determining the normalized optimized curve;
the transverse wave speed prediction model determining module is used for establishing a depth feedforward neural network model by combining a conventional logging curve, training by utilizing the known well transverse wave speed, and determining a transverse wave speed prediction model;
and the transverse wave speed determining module is used for inputting the normalized optimal curve into the transverse wave speed prediction model to determine the transverse wave speed.
The embodiment of the invention also provides computer equipment, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor realizes the method for predicting the transverse wave speed when executing the computer program.
The embodiment of the invention also provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program for executing the method for predicting the transverse wave speed.
The embodiment of the invention provides a method and a device for predicting a transverse wave speed, which are characterized in that a conventional logging curve is optimized based on a known well transverse wave speed curve, an optimized curve is determined, normalization processing is carried out, a depth feedforward neural network model is established by combining the conventional logging curve on the basis, training is carried out by utilizing the known well transverse wave speed, a transverse wave speed prediction model is determined, and finally the normalized optimized curve is input into the transverse wave speed prediction model to determine the transverse wave speed; in the whole transverse wave speed prediction process, only the transverse wave speed curve of a well and the conventional logging curve of a predicted well are needed, the accurate prediction of the transverse wave speed is directly realized from data, the adjustment parameters are few, the requirements on personnel are low, and the method can be widely applied to actual production.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. In the drawings:
fig. 1 is a schematic diagram of a method for predicting a transverse wave velocity according to an embodiment of the present invention.
Fig. 2 is a flowchart of a method for predicting a transverse wave velocity according to an embodiment of the present invention.
FIG. 3 is a graph comparing the shear wave time difference obtained by the method for predicting the velocity of the shear wave, the shear wave time difference obtained by the petrophysical simulation and the shear wave time difference obtained by the empirical formula.
Fig. 4 is a graph of transverse wave moveout error using an empirical formula.
Fig. 5 is a graph of shear wave moveout error obtained using petrophysical simulation.
FIG. 6 is a schematic diagram of a shear wave time difference error obtained by using a method for predicting a shear wave velocity according to an embodiment of the present invention.
FIG. 7 is a schematic diagram of a computer device for performing a method for predicting shear wave velocity embodying the present invention.
Fig. 8 is a schematic diagram of a device for predicting a transverse wave velocity according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the embodiments of the present invention will be described in further detail with reference to the accompanying drawings. The exemplary embodiments of the present invention and their descriptions herein are for the purpose of explaining the present invention, but are not to be construed as limiting the invention.
Fig. 1 is a schematic diagram of a method for predicting a transverse wave speed according to an embodiment of the present invention, and as shown in fig. 1, the embodiment of the present invention provides a method for predicting a transverse wave speed, so as to implement accurate prediction of a transverse wave speed, where the method includes:
step 101: acquiring a conventional logging curve and a known transverse wave velocity curve;
step 102: determining a preferred curve from the conventional log curve and the known cross-well velocity curve;
step 103: normalizing the optimized curve to determine a normalized optimized curve;
step 104: establishing a depth feedforward neural network model by combining a conventional logging curve, training by using the known well transverse wave speed, and determining a transverse wave speed prediction model;
step 105: and inputting the normalized optimized curve into a transverse wave speed prediction model to determine the transverse wave speed.
The method for predicting the transverse wave speed provided by the embodiment of the invention is characterized in that a conventional logging curve is optimized based on a known well transverse wave speed curve, an optimized curve is determined, normalization processing is carried out, a depth feedforward neural network model is established by combining the conventional logging curve on the basis, training is carried out by utilizing the known well transverse wave speed, a transverse wave speed prediction model is determined, and finally the normalized optimized curve is input into the transverse wave speed prediction model to determine the transverse wave speed; in the whole transverse wave speed prediction process, only the known transverse wave speed curve of the well and the conventional logging curve of the predicted well are needed, the data is directly transmitted, the adjustment parameters are few, the requirements on personnel are low, and the method can be widely applied to actual production.
Aiming at the problems that the conventional petrophysical modeling at present needs to adjust a plurality of skeleton parameters, a high-precision transverse wave speed prediction result is needed, the estimation of mineral content, porosity and fluid content in the early stage is needed, the corresponding needed parameter adjustment work is complicated, high requirements are put forward on personnel in the parameter adjustment process, the threshold is high, and the method is difficult to popularize and apply in actual production and application. In view of the above problems, an embodiment of the present invention provides a method for predicting a transverse wave velocity, which may include:
acquiring a conventional logging curve and a known transverse wave velocity curve; determining a preferred curve from the conventional log curve and the known cross-well velocity curve; normalizing the optimized curve to determine a normalized optimized curve; establishing a depth feedforward neural network model by combining a conventional logging curve, training by using the known well transverse wave speed, and determining a transverse wave speed prediction model; and inputting the normalized optimized curve into a transverse wave speed prediction model to determine the transverse wave speed.
In an embodiment, the conventional logging curve is obtained from well data about to be predicted for transverse wave speed, and the conventional logging data belongs to basic data in the exploration and development process and is easy to obtain; the known well shear wave velocity profile includes a plurality of measured shear wave velocity profiles for known wells, which are obtained from a small number of existing array acoustic logging data.
In an embodiment of the present invention, when the method for predicting a shear wave velocity provided by the embodiment of the present invention is implemented, determining a preferred curve according to a conventional log curve and a known well shear wave velocity curve includes:
according to the vector correlation calculation method, calculating the correlation coefficient between the conventional well logging curve and the known well transverse wave velocity curve, and carrying out correlation coefficient matrix evaluation on the conventional well logging curve according to the correlation coefficient to determine the optimal curve.
In the embodiment, the conventional well logging curve (serving as an input curve) is optimized, the correlation coefficient between the conventional well logging curve and the known transverse wave velocity curve is calculated according to a vector correlation calculation method, the conventional well logging curve is automatically sequenced according to the correlation coefficient, and the correlation coefficient matrix evaluation is carried out on the conventional well logging curve according to the correlation coefficient to determine the optimized curve.
The foregoing pair of conventional log curves includes: at least one or a combination of a gamma curve, a neutron curve, a density curve and a sonic curve.
In an embodiment of the present invention, when the method for predicting a transverse wave speed provided by the embodiment of the present invention is implemented, in one embodiment, normalization processing is performed on a preferred curve, and a normalized preferred curve is determined, including:
fitting the statistical distribution characteristics of the preferred curve by adopting a Gaussian distribution function, and determining the mean value and the variance of the preferred curve;
and normalizing the preferred curve according to the mean value and the variance of the preferred curve, and determining the normalized preferred curve.
In a specific implementation of the method for predicting the transverse wave speed provided by the embodiment of the present invention, in one embodiment, a normalized preferred curve is determined according to the following manner:
wherein x' is a normalized preferred curve; x is a preferred curve; μ is the mean of the preferred curve; σ is the variance of the preferred curve.
The foregoing description of the determination of the normalized preferred curve is illustrative, and those skilled in the art will appreciate that the above-described formulation may be modified and added with other parameters or data in a manner as desired, or that other specific formulations may be provided, and that such modifications are within the scope of the present invention.
In an embodiment of the present invention, a depth feedforward neural network model is built in combination with a conventional logging curve, training is performed by using a known well shear wave velocity, and a shear wave velocity prediction model is determined, including:
establishing a depth feedforward neural network model by combining a conventional logging curve, dividing the known well transverse wave speed into a training set and a testing set, optimizing an activation function, and selecting the activation function with smaller training error and testing error;
selecting a learning rate by adopting a dichotomy method, and determining a global optimal value of the deep feed-forward neural network model;
randomly initializing a weight matrix, inputting a training set and a testing set into a deep feedforward neural network model, training the deep feedforward neural network model according to an activation function and a global optimal value, and monitoring errors of the training set and the testing set;
when the errors of the training set and the testing set do not reach the set value, randomly initializing a weight matrix again;
when the errors of the training set and the testing set reach a set value, the model is stored;
and carrying out parallel averaging on the stored models, and determining a transverse wave speed prediction model.
In an embodiment of the present invention, a method for predicting a transverse wave velocity is provided, where selecting a learning rate by using a dichotomy, determining a global optimal value of a deep feedforward neural network model includes:
selecting a first learning rate to train the deep feed-forward neural network model, and observing a training error of the first learning rate; simultaneously, a second learning rate is selected to train the deep feed-forward neural network model, and training errors of the second learning rate are observed; wherein the first learning rate is greater than the second learning rate
Taking the midpoint of the first learning rate and the second learning rate as the midpoint learning rate, training the deep feed-forward neural network model, and determining a midpoint learning rate training error;
judging whether the deep feed-forward neural network model converges or not according to the training error of the first learning rate, the training error of the second learning rate and the training error of the middle learning rate;
if the deep feedforward neural network model is not converged, taking the midpoint again from the midpoint learning rate to the second learning rate until the deep feedforward neural network model is converged, and determining the global optimal value of the deep feedforward neural network model.
In an embodiment, the aforementioned deep feedforward neural network model established in combination with the conventional logging curve may be a multi-layer learning network composed of fully connected layers; in order to train the deep feedforward neural network model, dividing the known transverse wave speed of the well into a training set and a testing set, and respectively inputting the training set and the testing set into the deep feedforward neural network model for training; during training, different activation functions may be suitable for fitting the transverse wave speeds of different work areas, and the present activation functions such as relu, elu, leaky-relu, selu, gelu and the like need to be optimized in the process, so that the activation function with smaller training errors and smaller testing errors is optimized. And then selecting a learning rate by adopting a dichotomy, and determining a global optimal value of the deep feed-forward neural network model, wherein the method comprises the following steps: firstly, training a deep feed-forward neural network model by selecting a first learning rate, and observing a training error of the first learning rate; simultaneously, a second learning rate is selected to train the deep feed-forward neural network model, and training errors of the second learning rate are observed; after the training errors of the first learning rate and the second learning rate are obtained, taking the midpoint between the first learning rate and the second learning rate as a midpoint learning rate, training the deep feed-forward neural network model, and determining the midpoint learning rate training errors; if the deep feedforward neural network model is not converged, the midpoint is fetched again from the direction from the midpoint learning rate to the second learning rate until the deep feedforward neural network model is converged, and the global optimal value of the deep feedforward neural network model is determined, so that the efficiency in calculation can be ensured, and the quick search of the global optimal value is realized. The first learning rate may be a larger learning rate when implemented, and the second learning rate may be a smaller learning rate. In one example, the first learning rate may be 110% -500%, and the second learning rate may be 5% -95%, where the first learning rate may be reduced according to actual requirements, for example, 110%, 120%, 130%, 140%, 150% >, 495%, 500% may optionally be selected from one or other values; the second learning rate may also be increased according to actual needs, e.g., 5%, 10%, 15%, 20%, 25% >. 90%, 95%, optionally with one or other values therebetween. Then randomly initializing a weight matrix, training a deep feed-forward neural network model according to an activation function and a global optimal value, and monitoring errors of a training set and a testing set; when the errors of the training set and the testing set do not reach the set value, randomly initializing a weight matrix again; when the errors of the training set and the testing set reach a set value, the model is stored; at this time, a plurality of models are stored; and carrying out parallel averaging on the stored multiple models to determine a transverse wave speed prediction model.
Fig. 2 is a flowchart of a method for predicting a shear wave velocity according to an embodiment of the present invention, where, as shown in fig. 2, a flow for predicting a shear wave velocity includes:
inputting a curve; the input curve comprises an input conventional logging curve;
processing the input curve through the related correlation coefficient matrix evaluation to determine a preferred curve;
normalizing the optimized curve to determine a normalized optimized curve;
constructing a deep learning network by combining a conventional logging curve; the deep learning network comprises a deep feedforward neural network model;
randomly initializing a weight matrix, inputting a training set and a testing set into a deep feedforward neural network model, training the deep feedforward neural network model according to an activation function and a global optimal value, and monitoring errors of the training set and the testing set;
judging errors of the training set and the testing set;
when the errors of the training set and the testing set do not reach the set value, randomly initializing a weight matrix again;
when the errors of the training set and the testing set reach the set value, the models are stored, and finally a plurality of models can be obtained: model 1, model 2 … … model n;
averaging the stored n models in parallel to determine a transverse wave speed prediction model;
and inputting the normalized optimized curve into a transverse wave speed prediction model to predict the transverse wave speed.
According to the embodiment of the invention, a conventional logging curve is utilized, different learning models are constructed by adopting a deep learning algorithm, and different models are connected in parallel to predict the transverse wave speed, so that the transverse wave speed prediction precision is improved. On the basis of similarity matrix evaluation of an input conventional logging curve, a logging curve most relevant to the transverse wave speed is selected as a preferable curve, a deep feedforward neural network is constructed by combining the conventional logging curve on the basis, different prediction models are obtained by randomly initializing different weight coefficient matrixes, in the model training and testing process, a model with high prediction precision is reserved by monitoring a curve (error curve) formed by errors of a training set and a testing set, and finally the different models with high precision are connected in parallel to further improve the prediction precision. On the basis of the result of inaccurate logging stratum evaluation, the conventional logging curve is directly utilized to predict the transverse wave velocity, the model is optimized only by utilizing the error curve, and the method has few intermediate steps, few adjustable parameters and easy development of large-scale application and operation.
According to the method for predicting the transverse wave speed, provided by the embodiment of the invention, the actual prediction results of a plurality of gas fields of Sichuan basin and Erdos basin show that the method has the advantages of high precision of the predicted transverse wave speed and convenience in operation.
FIG. 3 is a graph comparing the shear wave time difference obtained by the method for predicting the velocity of the shear wave, the shear wave time difference obtained by the petrophysical simulation and the shear wave time difference obtained by the empirical formula. In fig. 3, CAL is a borehole diameter curve, GR is a natural gamma curve, RT is a deep resistivity curve, RXO is a shallow resistivity curve, vdcl is a clay content, vqua is a quartz content, port is a total porosity, cnl is a neutron curve, den is a density curve, dtc is a longitudinal wave time difference curve, dts_emp is a transverse wave time difference curve (inverse of transverse wave velocity) calculated by an empirical formula, dts_rm is a transverse wave time difference curve obtained by petrophysical modeling, keras_dts is a transverse wave time difference curve obtained by the present invention, and DTS is a measured transverse wave time difference curve.
FIG. 4 is a graph of transverse wave moveout error using an empirical formula; FIG. 5 is a graph of shear wave time difference error using petrophysical simulation; FIG. 6 is a schematic diagram of a shear wave time difference error obtained by using a method for predicting a shear wave velocity according to an embodiment of the present invention; by combining fig. 3 to fig. 6, it can be obtained that the transverse wave time difference error obtained by adopting the transverse wave speed prediction method of the embodiment of the invention is far smaller than the transverse wave time difference error obtained by adopting an empirical formula and the transverse wave time difference error obtained by adopting petrophysical simulation; further, the longitudinal and transverse wave velocity ratio (vpvs_dl) is calculated by using the transverse wave velocity obtained by the transverse wave velocity prediction method according to the embodiment of the invention, the measured longitudinal and transverse wave velocity ratio (VpVs), the longitudinal and transverse wave velocity ratio (vpvs_emp) predicted by an empirical formula, the longitudinal and transverse wave velocity ratio (vpvs_rm) obtained by petrophysical modeling are smaller in error, the error is less than 3%, and the errors of the other methods are all close to 10%.
Fig. 7 is a schematic diagram of a computer device for running a method for predicting a transverse wave speed according to the present invention, and as shown in fig. 7, an embodiment of the present invention further provides a computer device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, where the processor implements the method for predicting a transverse wave speed when executing the computer program.
The embodiment of the invention also provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program for executing the method for predicting the transverse wave speed.
The embodiment of the invention also provides a device for predicting the transverse wave speed, which is described in the following embodiment. Because the principle of solving the problem of the device is similar to that of a transverse wave speed prediction method, the implementation of the device can refer to the implementation of the transverse wave speed prediction method, and the repetition is omitted.
Fig. 8 is a schematic diagram of a device for predicting a transverse wave speed according to an embodiment of the present invention, and as shown in fig. 8, the embodiment of the present invention further provides a device for predicting a transverse wave speed, which may include:
a data acquisition module 801 for acquiring a conventional log and a known cross-well velocity profile;
a preferred curve determination module 802 for determining a preferred curve from the conventional log curve and the known cross-well velocity curve;
the normalization module 803 is configured to normalize the preference curve, and determine a normalized preference curve;
the shear wave speed prediction model determining module 804 is configured to establish a depth feedforward neural network model in combination with a conventional logging curve, perform training by using a known well shear wave speed, and determine a shear wave speed prediction model;
the transverse wave speed determining module 805 is configured to input the normalized preferred curve into a transverse wave speed prediction model to determine a transverse wave speed.
In an embodiment of the present invention, when the apparatus for predicting a transverse wave speed is implemented, a curve determining module is preferably configured to:
according to the vector correlation calculation method, calculating the correlation coefficient between the conventional well logging curve and the known well transverse wave velocity curve, and carrying out correlation coefficient matrix evaluation on the conventional well logging curve according to the correlation coefficient to determine the optimal curve.
In an embodiment, when the device for predicting the transverse wave speed provided by the embodiment of the present invention is implemented, the normalization module is specifically configured to:
fitting the statistical distribution characteristics of the preferred curve by adopting a Gaussian distribution function, and determining the mean value and the variance of the preferred curve;
and normalizing the preferred curve according to the mean value and the variance of the preferred curve, and determining the normalized preferred curve.
In an embodiment of the present invention, the method further includes determining a normalized preferred curve according to the following manner:
wherein x' is a normalized preferred curve; x is a preferred curve; μ is the mean of the preferred curve; σ is the variance of the preferred curve.
In an embodiment of the present invention, when the device for predicting a transverse wave speed provided by the embodiment of the present invention is implemented, in one embodiment, a module for determining a model for predicting a transverse wave speed is specifically configured to:
establishing a depth feedforward neural network model by combining a conventional logging curve, dividing the known well transverse wave speed into a training set and a testing set, optimizing an activation function, and selecting the activation function with smaller training error and testing error;
selecting a learning rate by adopting a dichotomy method, and determining a global optimal value of the deep feed-forward neural network model;
randomly initializing a weight matrix, training a deep feed-forward neural network model according to an activation function and a global optimal value, and monitoring errors of a training set and a testing set;
when the errors of the training set and the testing set do not reach the set value, randomly initializing a weight matrix again;
when the errors of the training set and the testing set reach a set value, the model is stored;
and carrying out parallel averaging on the stored models, and determining a transverse wave speed prediction model.
In an embodiment of the present invention, when the device for predicting a transverse wave speed provided by the embodiment of the present invention is implemented, the module for determining a transverse wave speed prediction model is further configured to:
selecting a first learning rate to train the deep feed-forward neural network model, and observing a training error of the first learning rate; simultaneously, a second learning rate is selected to train the deep feed-forward neural network model, and training errors of the second learning rate are observed; wherein the first learning rate is greater than the second learning rate
Taking the midpoint of the first learning rate and the second learning rate as the midpoint learning rate, training the deep feed-forward neural network model, and determining a midpoint learning rate training error;
judging whether the deep feed-forward neural network model converges or not according to the training error of the first learning rate, the training error of the second learning rate and the training error of the middle learning rate;
if the deep feedforward neural network model is not converged, taking the midpoint again from the midpoint learning rate to the second learning rate until the deep feedforward neural network model is converged, and determining the global optimal value of the deep feedforward neural network model.
In summary, the method and the device for predicting the shear wave speed provided by the embodiment of the invention are characterized in that a conventional logging curve is optimized based on a known well shear wave speed curve, an optimized curve is determined, normalization processing is carried out, a depth feedforward neural network model is established by combining the conventional logging curve on the basis, training is carried out by utilizing the known well shear wave speed, a shear wave speed prediction model is determined, and finally the normalized optimized curve is input into the shear wave speed prediction model to determine the shear wave speed; in the whole transverse wave speed prediction process, only the transverse wave speed curve of a well and the conventional logging curve of a predicted well are needed, the accurate prediction of the transverse wave speed is directly realized from data, the adjustment parameters are few, the requirements on personnel are low, and the method can be widely applied to actual production. According to the invention, on the basis of the transverse wave speed of the known well, a correlation algorithm in deep learning and machine learning is introduced, known data is split into a training set and a testing set, an error curve in the testing set is monitored in the training process, a reliable model is obtained, and the models obtained through multiple learning are connected in parallel to obtain a final high-precision transverse wave speed prediction model. The method directly starts from data, has few adjustment parameters and low requirements on personnel, and can be widely applied in actual production.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The foregoing description of the embodiments has been provided for the purpose of illustrating the general principles of the invention, and is not meant to limit the scope of the invention, but to limit the invention to the particular embodiments, and any modifications, equivalents, improvements, etc. that fall within the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (12)
1. A method for predicting shear wave velocity, comprising:
acquiring a conventional logging curve and a known transverse wave velocity curve; the known well shear wave velocity profile comprises a plurality of measured shear wave velocity profiles for the known wells;
determining a preferred curve from the conventional log curve and the known cross-well velocity curve;
normalizing the optimized curve to determine a normalized optimized curve;
establishing a depth feedforward neural network model by combining a conventional logging curve, training by using the known well transverse wave speed, and determining a transverse wave speed prediction model;
inputting the normalized optimized curve into a transverse wave speed prediction model to determine transverse wave speed;
the method for determining the shear wave velocity prediction model comprises the following steps of:
establishing a depth feedforward neural network model by combining a conventional logging curve, dividing the known well transverse wave speed into a training set and a testing set, optimizing an activation function, and selecting the activation function with smaller training error and testing error;
selecting a learning rate by adopting a dichotomy method, and determining a global optimal value of the deep feed-forward neural network model;
randomly initializing a weight matrix, inputting a training set and a testing set into a deep feedforward neural network model, training the deep feedforward neural network model according to an activation function and a global optimal value, and monitoring errors of the training set and the testing set;
when the errors of the training set and the testing set do not reach the set value, randomly initializing a weight matrix again;
when the errors of the training set and the testing set reach a set value, the model is stored;
and carrying out parallel averaging on the stored models, and determining a transverse wave speed prediction model.
2. The method of claim 1, wherein determining a preferred profile from a conventional log profile and a known shear well velocity profile comprises:
according to the vector correlation calculation method, calculating the correlation coefficient between the conventional well logging curve and the known well transverse wave velocity curve, and carrying out correlation coefficient matrix evaluation on the conventional well logging curve according to the correlation coefficient to determine the optimal curve.
3. The method of claim 1, wherein normalizing the preference profile to determine a normalized preference profile comprises:
fitting the statistical distribution characteristics of the preferred curve by adopting a Gaussian distribution function, and determining the mean value and the variance of the preferred curve;
and normalizing the preferred curve according to the mean value and the variance of the preferred curve, and determining the normalized preferred curve.
4. A method according to claim 3, characterized in that the normalized preference profile is determined in the following way:
wherein x' is a normalized preferred curve; x is a preferred curve; μ is the mean of the preferred curve; σ is the variance of the preferred curve.
5. The method of claim 1, wherein selecting the learning rate using a dichotomy, determining a global optimum for the deep feed forward neural network model, comprises:
selecting a first learning rate to train the deep feed-forward neural network model, and observing a training error of the first learning rate; simultaneously, a second learning rate is selected to train the deep feed-forward neural network model, and training errors of the second learning rate are observed; wherein the first learning rate is greater than the second learning rate;
taking the midpoint of the first learning rate and the second learning rate as the midpoint learning rate, training the deep feed-forward neural network model, and determining a midpoint learning rate training error;
judging whether the deep feed-forward neural network model converges or not according to the training error of the first learning rate, the training error of the second learning rate and the training error of the middle learning rate;
if the deep feedforward neural network model is not converged, taking the midpoint again from the midpoint learning rate to the second learning rate until the deep feedforward neural network model is converged, and determining the global optimal value of the deep feedforward neural network model.
6. A transverse wave speed prediction apparatus, comprising:
the data acquisition module is used for acquiring a conventional logging curve and a known well transverse wave speed curve; the known well shear wave velocity profile comprises a plurality of measured shear wave velocity profiles for the known wells;
the optimization curve determining module is used for determining an optimization curve according to a conventional logging curve and a known transverse wave speed curve;
the normalization module is used for carrying out normalization processing on the optimized curve and determining the normalized optimized curve;
the transverse wave speed prediction model determining module is used for establishing a depth feedforward neural network model by combining a conventional logging curve, training by utilizing the known well transverse wave speed, and determining a transverse wave speed prediction model;
the transverse wave speed determining module is used for inputting the normalized optimized curve into the transverse wave speed prediction model to determine the transverse wave speed;
the transverse wave speed prediction model determining module is specifically used for:
establishing a depth feedforward neural network model by combining a conventional logging curve, dividing the known well transverse wave speed into a training set and a testing set, optimizing an activation function, and selecting the activation function with smaller training error and testing error;
selecting a learning rate by adopting a dichotomy method, and determining a global optimal value of the deep feed-forward neural network model;
randomly initializing a weight matrix, inputting a training set and a testing set into a deep feedforward neural network model, training the deep feedforward neural network model according to an activation function and a global optimal value, and monitoring errors of the training set and the testing set;
when the errors of the training set and the testing set do not reach the set value, randomly initializing a weight matrix again;
when the errors of the training set and the testing set reach a set value, the model is stored;
and carrying out parallel averaging on the stored models, and determining a transverse wave speed prediction model.
7. The apparatus of claim 6, wherein the preference profile determination module is configured to:
according to the vector correlation calculation method, calculating the correlation coefficient between the conventional well logging curve and the known well transverse wave velocity curve, and carrying out correlation coefficient matrix evaluation on the conventional well logging curve according to the correlation coefficient to determine the optimal curve.
8. The apparatus of claim 6, wherein the normalization module is configured to:
fitting the statistical distribution characteristics of the preferred curve by adopting a Gaussian distribution function, and determining the mean value and the variance of the preferred curve;
and normalizing the preferred curve according to the mean value and the variance of the preferred curve, and determining the normalized preferred curve.
9. The apparatus of claim 8, wherein the normalization module is further configured to determine the normalized preference profile as follows:
wherein x' is a normalized preferred curve; x is a preferred curve; μ is the mean of the preferred curve; σ is the variance of the preferred curve.
10. The apparatus of claim 6, wherein the shear wave velocity prediction model determination module is further configured to:
selecting a first learning rate to train the deep feed-forward neural network model, and observing a training error of the first learning rate; simultaneously, a second learning rate is selected to train the deep feed-forward neural network model, and training errors of the second learning rate are observed;
taking the midpoint of the first learning rate and the second learning rate as the midpoint learning rate, training the deep feed-forward neural network model, and determining a midpoint learning rate training error;
judging whether the deep feed-forward neural network model converges or not according to the training error of the first learning rate, the training error of the second learning rate and the training error of the middle learning rate;
if the deep feedforward neural network model is not converged, taking the midpoint again from the midpoint learning rate to the second learning rate until the deep feedforward neural network model is converged, and determining the global optimal value of the deep feedforward neural network model.
11. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method of predicting transverse wave speed according to any one of claims 1 to 5 when the computer program is executed.
12. A computer-readable storage medium storing a computer program for executing a method of predicting a shear wave velocity according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010434118.0A CN111751878B (en) | 2020-05-21 | 2020-05-21 | Method and device for predicting transverse wave speed |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010434118.0A CN111751878B (en) | 2020-05-21 | 2020-05-21 | Method and device for predicting transverse wave speed |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111751878A CN111751878A (en) | 2020-10-09 |
| CN111751878B true CN111751878B (en) | 2023-05-30 |
Family
ID=72673433
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010434118.0A Active CN111751878B (en) | 2020-05-21 | 2020-05-21 | Method and device for predicting transverse wave speed |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111751878B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113466962B (en) * | 2021-07-06 | 2023-03-28 | 西安交通大学 | Transverse wave prediction method and system based on Gaussian process regression |
| CN117669785A (en) * | 2022-08-26 | 2024-03-08 | 中国石油天然气股份有限公司 | Transverse wave time difference prediction method and device |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110703330A (en) * | 2018-07-10 | 2020-01-17 | 中国石油化工股份有限公司 | Carbonate rock logging correction and transverse wave velocity prediction method based on rock physics |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10802171B2 (en) * | 2017-04-28 | 2020-10-13 | Pioneer Natural Resources Usa, Inc. | High resolution seismic data derived from pre-stack inversion and machine learning |
| CN107526117B (en) * | 2017-07-06 | 2019-08-13 | 天津科技大学 | Based on autocoding and transfinites and learn the acoustic speed prediction technique of joint network |
| CN107678059B (en) * | 2017-09-05 | 2019-06-28 | 中国石油大学(北京) | A kind of method, apparatus and system of reservoir gas-bearing identification |
| CN107817535B (en) * | 2017-09-27 | 2019-07-09 | 中国石油天然气股份有限公司 | The determination method and apparatus of short lap |
| CN109212617B (en) * | 2018-08-24 | 2020-09-08 | 中国石油天然气股份有限公司 | Automatic identification method and device for electric imaging logging phase |
| CN112703429B (en) * | 2018-09-14 | 2024-02-13 | Bp北美公司 | Machine learning-based seismic attribute analysis |
| CN111144001A (en) * | 2019-12-26 | 2020-05-12 | 湖南科技大学 | Mine shaft engineering TBM control method based on BP neural network |
-
2020
- 2020-05-21 CN CN202010434118.0A patent/CN111751878B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110703330A (en) * | 2018-07-10 | 2020-01-17 | 中国石油化工股份有限公司 | Carbonate rock logging correction and transverse wave velocity prediction method based on rock physics |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111751878A (en) | 2020-10-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11704579B2 (en) | Earth modeling methods using machine learning | |
| US8504341B2 (en) | Methods, systems, and computer readable media for fast updating of oil and gas field production models with physical and proxy simulators | |
| AU2007211291B2 (en) | Methods, systems, and computer-readable media for fast updating of oil and gas field production models with physical and proxy simulators | |
| US11699099B2 (en) | Confidence volumes for earth modeling using machine learning | |
| CN111751878B (en) | Method and device for predicting transverse wave speed | |
| CN104533400A (en) | Method for reconstructing logging curve | |
| CN114723095A (en) | Missing well logging curve prediction method and device | |
| CN111561313B (en) | Compact sandstone reservoir parameter prediction method based on physical model and machine learning | |
| CN113610945A (en) | Ground stress curve prediction method based on hybrid neural network | |
| CN113947198A (en) | Logging curve reconstruction method based on nonlinear autoregressive neural network model | |
| CN114488311A (en) | Transverse wave time difference prediction method based on SSA-ELM algorithm | |
| Shi et al. | Prediction of shear wave velocity using machine learning technique, multiple regression and well logs | |
| CN116840906A (en) | Uncertainty evaluation method for reservoir physical property parameter prediction | |
| Eftekharifar et al. | 3D Petrophysical modeling using complex seismic attributes and limited well log data | |
| Echeverria et al. | A robust scheme for spatio-temporal inverse modeling of oil reservoirs | |
| CN108875109B (en) | Method and system for predicting abnormal formation pressure | |
| Echeverría et al. | Robust scheme for inversion of seismic and production data for reservoir facies modeling | |
| CN112946754B (en) | Reservoir porosity prediction method and device | |
| Sun | Data-driven rational function neural networks: a new method for generating analytical models of rock physics | |
| Hamada et al. | Parallel Self Organizing Neural Network (PSONN) Prediction of Water Saturation in Carbonate Reservoirs | |
| CN114428313A (en) | Rock physical elastic parameter forward modeling method, forward modeling device and electronic equipment | |
| CN117094428A (en) | Prediction method for deep carbonate rock hydrocarbon source rock localization parameters based on machine learning | |
| CN114594523A (en) | Physical constraint and data-driven nuclear magnetic resonance T construction2Spectrum method and device | |
| CN116227335A (en) | Three-dimensional ground stress prediction method of parallel neural network based on physical constraint | |
| CN113064216A (en) | Reservoir porosity prediction method, system, equipment and readable storage medium |
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 |