A method, device and medium for predicting well logging curves

Abstract

The embodiments of the present specification disclose a method, device and medium for predicting a well logging curve, including: obtaining well logging input data from current well logging data, and geologically stratifying the well logging input data according to geological stratification characteristics to obtain each layer of well logging input data that retains the distribution characteristics of each layer of well logging data; multi-scale resampling of each layer of well logging input data to obtain a multi-sampling scale well logging feature map; determining a geological transformation probability feature map between each layer of well logging input data according to pre-acquired geological attribute data; merging the multi-sampling scale well logging feature map with the geological transformation probability feature map to obtain well logging input data; inputting the well logging input data into a pre-trained well logging curve prediction model to predict a well logging measurement curve.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method, device and medium for predicting a well logging curve. Background Art

[0002] Well logging curves are the most important data in well logging projects. They are the data basis for geological guidance and directional well operations. The quality of well logging curves directly affects the development output of the well. In practical applications, due to the complex underground conditions, the expansion of the well diameter, instrument failure, and improper operation may occur during the logging construction process, and the well logging data of some sections of the well are often distorted or missing; or due to cost considerations, the operator needs to abandon a certain well logging curve, resulting in incomplete well logging curve data for a certain well, which brings difficulties and challenges to subsequent geologists in geological interpretation and geological guidance.

[0003] At present, the commonly used method is rock physics modeling. The rock physics modeling method needs to use parameters such as formation pressure, temperature, and particle aspect ratio to establish a rock physics model. There are different modeling methods for different lithologies. Due to the complex underground reservoir conditions, the current rock physics modeling cannot accurately predict the well logging curve. Summary of the invention

[0004] One or more embodiments of the present specification provide a method, device and medium for predicting a well logging curve, which are used to solve the technical problems raised by the background technology.

[0005] One or more embodiments of this specification adopt the following technical solutions:

[0006] One or more embodiments of this specification provide a method for predicting a well logging curve, including:

[0007] Acquire well logging data to be input from the current well logging data, and perform geological stratification on the well logging data to be input according to geological stratification characteristics, so as to obtain well logging data to be input for each layer that retains the distribution characteristics of the well logging data for each layer;

[0008] Performing multi-scale resampling on the logging data to be input at each layer to obtain a multi-sampling scale logging characteristic map;

[0009] Determine the geological transformation probability characteristic map between the logging data to be input in each layer according to the geological attribute data acquired in advance;

[0010] The multi-sampling scale well logging characteristic map is combined with the geological transformation probability characteristic map to obtain well logging input data;

[0011] The logging input data is input into a pre-trained logging curve prediction model to predict a logging measurement curve.

[0012] One or more embodiments of this specification provide a well logging curve prediction device, including:

[0013] at least one processor; and,

[0014] a memory communicatively connected to the at least one processor; wherein,

[0015] The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0016] Acquire well logging data to be input from the current well logging data, and perform geological stratification on the well logging data to be input according to geological stratification characteristics, so as to obtain well logging data to be input for each layer that retains the distribution characteristics of the well logging data for each layer;

[0017] Performing multi-scale resampling on the logging data to be input at each layer to obtain a multi-sampling scale logging characteristic map;

[0018] Determine the geological transformation probability characteristic map between the logging data to be input in each layer according to the geological attribute data acquired in advance;

[0019] The multi-sampling scale well logging characteristic map is combined with the geological transformation probability characteristic map to obtain well logging input data;

[0020] The logging input data is input into a pre-trained logging curve prediction model to predict a logging measurement curve.

[0021] One or more embodiments of this specification provide a non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions are configured to:

[0022] Acquire well logging data to be input from the current well logging data, and perform geological stratification on the well logging data to be input according to geological stratification characteristics, so as to obtain well logging data to be input for each layer that retains the distribution characteristics of the well logging data for each layer;

[0023] Performing multi-scale resampling on the logging data to be input at each layer to obtain a multi-sampling scale logging characteristic map;

[0024] Determine the geological transformation probability characteristic map between the logging data to be input in each layer according to the geological attribute data acquired in advance;

[0025] The multi-sampling scale well logging characteristic map is combined with the geological transformation probability characteristic map to obtain well logging input data;

[0026] The logging input data is input into a pre-trained logging curve prediction model to predict a logging measurement curve.

[0027] At least one of the above technical solutions adopted in the embodiments of this specification can achieve the following beneficial effects:

[0028] The embodiment of this specification performs geological stratification on the acquired well logging input data according to the geological stratification characteristics, and obtains each layer of well logging input data that retains the distribution characteristics of each layer of well logging data, which can better predict lithology changes at depths where lithology changes are more drastic. At the same time, the well logging input data of each layer is multi-scale resampled to obtain a multi-sampling scale well logging characteristic map to extract the well logging curve response at different scales, so that the results output by the well logging curve prediction model are more accurate. In addition, based on the geological attribute data, the geological transformation probability characteristic map between each layer of well logging input data is determined, and the formation change information is displayed from a probability perspective, so that the results output by the well logging curve prediction model are more accurate. The embodiment of this specification combines the above-mentioned multi-sampling scale well logging characteristic map with the geological transformation probability characteristic map to obtain the well logging input data. After the well logging input data is input to the well logging curve prediction model, the output well logging measurement curve is more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art description. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. In the drawings:

[0030] Figure 1 A schematic flow chart of a method for predicting a well logging curve provided in one or more embodiments of this specification;

[0031] Figure 2 A data processing flow chart provided for one or more embodiments of this specification;

[0032] Figure 3 A schematic diagram of a natural gamma logging curve provided for one or more embodiments of this specification;

[0033] Figure 4 A multi-sampling scale logging curve characteristic diagram (left side) and a geological transformation probability characteristic diagram (right side) provided for one or more embodiments of this specification;

[0034] Figure 5 A schematic diagram comparing the predicted results and actual measurement results provided in one or more embodiments of this specification;

[0035] Figure 6 A schematic diagram of the structure of a logging curve prediction device provided in one or more embodiments of this specification. DETAILED DESCRIPTION

[0036] The embodiments of this specification provide a method, an apparatus, a device, and a medium.

[0037] At present, the commonly used methods are rock physics modeling and multivariate fitting methods. The rock physics modeling method needs to use parameters such as formation pressure, temperature, and grain aspect ratio to establish a rock physics model. There are different modeling methods for different lithologies. The multivariate linear fitting method needs to use the correlation between the correction curve and multiple curves in the undistorted layer to establish a multivariate fitting formula, thereby correcting the distorted logging curve.

[0038] Due to the complexity of underground reservoir conditions, there is currently no physical model or empirical model that can accurately describe the mapping relationship between these logging curves, and it is also impossible to generate other logging curves based on some logging curves.

[0039] The embodiment of this specification proposes a logging curve reconstruction method based on logging curve transformation and convolutional neural network CNN, which is used to solve the problem of missing a certain logging curve in the logging project.

[0040] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0041] Figure 1 A flow chart of a method for predicting a well logging curve provided in one or more embodiments of this specification is provided. The process can be executed by a well logging curve prediction system. For complex underground conditions, well logging curves of different rock formations can be accurately predicted. Certain input parameters or intermediate results in the process allow manual intervention and adjustment to help improve accuracy.

[0042] The method steps of the embodiment of this specification are as follows:

[0043] S102, obtaining well logging data to be input from the current well logging data, and geologically stratifying the well logging data to be input according to geological stratification characteristics to obtain well logging data to be input for each layer that retains the distribution characteristics of the well logging data for each layer.

[0044] In an embodiment of the present specification, the well logging data to be input may be a known well logging curve. For example, the well logging data to be input may be one or more of an average drilling speed well logging curve, a gamma well logging curve, a resistivity well logging curve, and a porosity well logging curve. Since these well logging curves are relatively easy to obtain, the above-mentioned well logging curves may be obtained in advance, so that other well logging curves that are not easy to obtain or have data missing may be calculated through these well logging curves. The well logging curves that are not easy to obtain or have data missing are the well logging measurement curves predicted subsequently, and the well logging measurement curves may be lithology density well logging curves.

[0045] The well logging data to be input is geologically stratified according to the geological stratification characteristics to obtain the well logging data to be input for each layer retaining the distribution characteristics of the well logging data of each layer. The well logging data to be input can be geologically stratified according to the geological stratification characteristics. During the geological stratification, the well logging curves in the well logging data to be input can be used to divide the layers, such as the intra-layer difference method. For example, when the well logging data is at a depth of 900 meters, there are 90 possible strata during the period, and there may be multiple types of lithological strata that appear alternately in these 90 layers; then, the Gaussian distribution parameters corresponding to the distribution of the well logging data of each layer are determined respectively, and the Gaussian distribution parameters can be the expected value and variance value of the well logging data of the same layer; finally, the well logging data to be input is normalized according to the Gaussian distribution parameters to obtain the well logging data to be input for each layer retaining the distribution characteristics of the well logging data of each layer, so that the lithology changes can be better predicted at the depth where the lithology changes dramatically.

[0046] It should be noted that lithology refers to some attributes of rock characteristics, such as natural radioactivity, resistivity, porosity, density, color, composition, structure, cement, cement type, special minerals, etc. Lithology can also refer to some attributes of soil characteristics.

[0047] S104, performing multi-scale resampling on the logging data to be input in each layer to obtain a multi-sampling scale logging characteristic map.

[0048] In the embodiments of the present specification, the logging data to be input for each layer can be sampled separately according to a plurality of pre-set sampling intervals to obtain sampling data corresponding to each sampling interval. This process is called resampling. The logging responses of different formation thicknesses can be extracted through resampling. When the sampling interval is small, the extracted information reflects the response of the thick layer and the thin layer coupled together. When the sampling interval is large and low-frequency, the extracted information mainly reflects the information of the thick layer. Then, the sampling data corresponding to each sampling interval are merged to obtain a multi-sampling scale logging characteristic map.

[0049] Furthermore, in the embodiments of the present specification, when merging the sampling data corresponding to the sampling intervals to obtain a multi-sampling scale logging characteristic diagram, the sampling data corresponding to the sampling intervals can be processed separately by linear interpolation to unify the sampling data corresponding to the sampling intervals into the same number of sampling points, and determine the sequence of the sampling data corresponding to the sampling intervals; the sequences of the sampling data corresponding to the sampling intervals are merged to obtain a multi-sampling scale logging characteristic diagram.

[0050] S106, determining a geological transformation probability characteristic map between the logging data to be input in each layer according to the geological attribute data acquired in advance.

[0051] In the embodiments of this specification, the above contents are implemented through the following specific schemes:

[0052] According to the geological attribute data, the attribute value distribution interval corresponding to each type of rock can be determined. For example, the geological attribute data is the natural gamma value of the rock. In the geological attribute data, the natural gamma value distribution interval of rock A is 0-20, the natural gamma value distribution interval of rock B is 20-40, and the natural gamma value distribution interval of rock C is 40-60; the rock type of each layer of well logging data to be input can be determined according to the attribute value distribution interval corresponding to each type of rock. For example, the rock types include A, B and C, and the geological stratification can include 6 layers. The rock types of each layer can be A, C, B, C, A, B from low to high; according to the rock type of each layer of well logging data to be input, the transformation probability of any two types of rocks is determined; according to the transformation probability of the any two types of rocks and the geological stratification in which each well logging data to be input in each layer of well logging data to be input, the transformation probability between any two well logging data to be input is determined; according to the transformation probability between any two well logging data to be input, the geological transformation probability characteristic map between the well logging data to be input in each layer is determined.

[0053] Furthermore, before determining the geological transformation probability characteristic map between the layers of well logging data to be input, in order to mine more geological features, the transformation probability between any two layers of well logging data to be input can be determined based on the transformation probability of the any two types of rocks and the rock types of the layers of well logging data to be input; finally, based on the transformation probability between the any two layers of well logging data to be input, the geological transformation probability characteristic map between the layers of well logging data to be input can be determined. Subsequently, either the geological transformation probability characteristic map between the layers of well logging data to be input can be used as input data, or the transformation probability characteristic map between any two layers of well logging data to be input can be used as input data.

[0054] S108, merging the multi-sampling scale well logging characteristic map and the geological transformation probability characteristic map to obtain well logging input data.

[0055] A 5-dimensional tensor is pre-constructed in the logging curve prediction model, the first dimension represents the number of samples, the second dimension represents the time step, the third dimension represents the image width, the fourth dimension represents the image height, and the fifth dimension represents the number of image channels. In the embodiment of this specification, the number of samples is the number of well logging data to be input, the time step can be set to 4, the image width and height are both preset values, and the number of image channels is 2. The multi-sampling scale logging feature map is used as the first channel, and the geological transformation probability feature map is used as the second channel.

[0056] S110, inputting the well logging input data into a pre-trained well logging curve prediction model to predict a well logging measurement curve.

[0057] It should be noted that the various layers of well logging data to be input mentioned in the embodiments of this specification may represent the overall well logging data to be input.

[0058] In an embodiment of the present specification, the loss function of the logging curve prediction model can be obtained in the following manner: first, the predicted logging measurement curve and the actual logging measurement curve are obtained; then, the Gaussian distribution parameters of the predicted logging measurement curve of each layer and the Gaussian distribution parameters of the actual logging measurement curve are determined respectively; then, the loss function of the logging curve prediction model is determined according to the similarity between the Gaussian distribution of the predicted logging measurement curve and the Gaussian distribution of the actual logging measurement curve; finally, the logging curve prediction model is obtained by training according to the loss function.

[0059] Furthermore, the above technical content is the application process after the well logging curve prediction model is trained. The well logging curve prediction model is further explained below:

[0060] 1. Read the original logging curve data and pre-process the data

[0061] The data to be processed includes 5 logging curves, namely average mechanical drilling speed, gamma, resistivity, rock skeleton density and porosity. We will use 4 of the curves as input data and the 5th curve as label data to train the network model so that it can learn the geological formation information contained in these 5 curves and then make predictions.

[0062] In the embodiments of this specification, average mechanical drilling speed, gamma data, resistivity data and porosity data can be used as examples to predict the rock skeleton density curve.

[0063] Taking the actual logging while drilling file of an oil well in an oil field as an example, after reading the logging data, the data of each logging curve is obtained. The logging data contains N depth sampling points, covering a depth of 900 meters.

[0064] 1.1 Perform median smoothing and normalization on the data.

[0065] For normalization processing, the embodiment of this specification proposes a layered data normalization method based on formation lithology, and the specific steps are as follows:

[0066] 1) Perform a preliminary stratum division based on the well logging curve, assuming it is divided into k layers.

[0067] 2) For the logging curve response of each layer, a normal distribution is approximated, and the expected μ and variance σ of each layer are calculated respectively.

[0068] 3) Perform the following normalization processing on the logging curve data of each layer:

[0069] (Nk is the number of points in the kth layer)

[0070] The traditional method is to normalize to the interval [0, 1] without considering the stratigraphic stratification characteristics and the Gaussian distribution characteristics of different lithologies to logging responses. This method normalizes the data through geological stratification and Gaussian model parameters, retaining the distribution characteristics of logging curve data for different lithologies. After comparing the results, this method can better predict lithology changes at depths where lithology changes are more drastic.

[0071] 1.2 Divide the dataset.

[0072] The embodiment of this specification divides the data set into a training set and a test set, wherein the training set is used to train the network model and the test set is used to test the network prediction effect. The training set data accounts for 70% of the total data, and the test set data accounts for 30% of the total data. The input data is the average mechanical drilling speed, gamma, resistivity and porosity data, and the label data and prediction data are rock skeleton density.

[0073] 1.3 Data segmentation.

[0074] Since the formation has the characteristics of slow change and uniform characteristics within a certain range of vertical depth, in order to better obtain the formation characteristics, we divide the data according to a certain depth. In this embodiment of the specification, ns data points are used as a group of data, and the entire data set is divided into samples. Among them:

[0075]

[0076] 2 Two-dimensional transformation processing of a single logging curve

[0077] 2.1 Multi-scale resampling

[0078] A single well logging curve contains complex geological responses, including both thin and thick geological responses, as well as homogeneous and inhomogeneous formation responses. In order to improve the ability of the network model to learn geological features, the well logging curve is first resampled at multiple scales to extract the formation responses of different layer thicknesses. The thick layer response corresponds to the low-frequency component in the curve, and the thin layer response corresponds to the high-frequency component in the curve. By designing multi-scale resampling parameters, the well logging curve is resampled, the logging curve responses at different scales are extracted, and combined into an image as one of the input data for the next network model. It is mainly divided into the following four steps:

[0079] 2.1.1 Designing resampling parameters

[0080] Through resampling, the logging response of different formation thicknesses can be extracted. The sampling interval is small (corresponding to high frequency) and the number of sampling points is large. For example, the total number of sampling points is 900, the sampling interval is set to 2, and the sampling point numbers are 2, 4, 6...900. The extracted information reflects the response of thick and thin layers coupled together. When the sampling interval is large (corresponding to low frequency), for example, the total number of sampling points is 900, the sampling interval is set to 10, and the sampling point numbers are 10, 20, 30...900, the sampling points of thin layers may be skipped during sampling, and the extracted information mainly reflects the information of thick layers. In order to fully explore the geological information under different sampling intervals, the resampling parameter sequence designed by this method is dx = [N-1, N-2,..., k], where N is the number of curve points and k is the number of geological layers in the target area, indicating that the minimum number of sampling points is greater than or equal to the number of layers.

[0081] 2.1.2 Resampling curve

[0082] According to the determined sampling point sequence, the logging curve is uniformly resampled to obtain Nk logging curves.

[0083] 2.1.3 Interpolation to construct equal-length sequences

[0084] In order to obtain logging curves of uniform length and facilitate graphic processing, it is necessary to standardize the lengths of the above Nk logging curves. The curves are uniformly interpolated into N points through the linear interpolation method. This step only increases the number of points in the curve sequence. Due to the use of the linear interpolation method, the formation characteristic response represented by it has not changed.

[0085] 2.1.4 Constructing multi-sampling rate curve graphs

[0086] The resampled logging sequences of different scales constructed in the above steps are merged to form a multi-sampling scale logging curve feature map corresponding to the logging curve, which is used as one of the input data of the subsequent network model.

[0087] 2.2 Geological transformation probability characteristic map

[0088] Investigate the transformation probability between each point of the logging curve. The change probability between each point is related to the properties of the underground medium. The main steps are as follows:

[0089] 2.2.1 Numerical Partitioning

[0090] According to the physical properties of the rock, the logging curve is divided into intervals. For example, different lithologies have different gamma values. Generally, mudstone has a higher gamma value, and sandstone has a lower gamma value. The specific value can be determined according to the geological statistics of the work area. Taking the gamma value as an example, the gamma value distribution interval corresponding to different lithologies is determined, assuming [e 1 , e 2 ,…,e m ] where e i is a gamma value, e m ~e m+1 is the gamma value distribution interval of a certain lithology. The basis of this numerical partition is rock physical statistical data, not logging curve value, and m represents the number of lithology types. For k strata, the lithology of each stratum is l[i], i = 1, 2, ... k.

[0091] 2.2.2 Numerical Statistics

[0092] Perform numerical statistics on the gamma logging curve and count the number of points q in each interval i , such as q 1 represents the total number of points on the curve falling between [0, e1], and at the same time determines the formation number corresponding to each point on the logging curve, denoted as {b i , i = 1, 2, ... N} where b i Represents the formation number corresponding to the i-th point, and N is the number of logging curve points.

[0093] 2.2.3 Calculating interval transformation probability

[0094] Among all the strata, the lithology combination of two adjacent layers is (m i , m j ) is c_ij. For k layers, the total number of combinations of two adjacent layers is k-1, so the lithology combination is defined as (m i , m j ) is:

[0095]

[0096] Among them, λ is the weight coefficient, which is used to weight different combinations. It can be set according to the geological conditions of the work area. When the geological characteristics are unclear, it can be set to 0.

[0097] On this basis, the transformation probability between any two strata is defined as:

[0098] f[i,j]=p[l[i],l[j]], i=1,2,…k; j=1,2,…k;

[0099] On this basis, the transformation probability between any two points on the curve is defined as:

[0100] M[i,j]=f[b[i],b[j]], i=1,2,…N; j=1,2,…N;

[0101] That is, first define the transformation probability of different lithologies, then calculate the transformation probability between different strata based on the lithology of each stratum, and then calculate the transformation probability between different data points in the logging curve based on the stratum to which each curve point belongs.

[0102] 2.2.5 Imaging

[0103] The transformation probability between different data points in a single logging curve constructed in the above steps is the geological transformation probability characteristic map, which is used as one of the input data of the subsequent model. The characteristic map shows the stratigraphic change information from a probability perspective.

[0104] 3. Multi-channel merging of data

[0105] Construct a 5-dimensional tensor, where the first dimension represents the number of samples, the second dimension represents the time step, the third dimension represents the image width, the fourth dimension represents the image height, and the fifth dimension represents the number of image channels. In this example, the number of samples is samples, the time step is 4, the image width and height are both preset values, and the number of image channels is 2. The multi-sampling rate transformation result is used as the first channel, and the geological probability transformation result is used as the second channel.

[0106] 4. Build a network model

[0107] Input layer: The input layer is a constructed 5-dimensional tensor X_C[samples, 4, ns, ns, channels]

[0108] The first to fourth layers can be encapsulated using the TimeDistributed encapsulator.

[0109] The first layer: two-dimensional convolution layer Conv2D_L1, the number of convolution output filters is 256, the convolution kernel size is 2*2, and relu is used as the activation function.

[0110] The second layer: two-dimensional pooling layer Pool_L2, using the maximum pooling method, the pooling size is 2*2.

[0111] The third layer: two-dimensional convolution layer Conv2D_L3, the number of convolution output filters is 256, the convolution kernel size is 2*2, and Relu is used as the activation function.

[0112] The fourth layer: two-dimensional pooling layer Pool_L4, using the maximum pooling method, the pooling size is 2*2.

[0113] The fifth layer: Flatten_L5, which flattens the output results of the first four layers of the network.

[0114] The sixth layer: bidirectional LSTM_L6 layer, with 256 neurons.

[0115] The seventh layer: fully connected layer Dense_L7, the number of neurons is 256, and no activation function is used.

[0116] The eighth layer: fully connected layer Dense_L8, the number of neurons is ns, and no activation function is used.

[0117] 5. Training the network model

[0118] Define the optimizer: Use the Adam optimizer and set the learning rate to 0.01.

[0119] Custom loss function:

[0120] From the perspective of geological analysis, the accuracy of the prediction curve is reflected in whether it can reflect the lithology characteristics of the formation at the correct depth. The loss function can be defined using a geostatistical strategy. The rock physical properties of a certain lithology conform to the Gaussian distribution.

[0121] 5.1 Divide the depth into several layers.

[0122] 5.2 For each layer, the Gaussian distribution probability curve of the statistical prediction value is calculated, and its expectation is μ 1 , with standard deviation σ 1 , the Gaussian probability distribution of the statistical measured value, whose expectation is μ 2 , with standard deviation σ 2 .

[0123] 5.3 Calculate the similarity of Gaussian distribution between the predicted curve and the measured curve.

[0124]

[0125] 5.4 Calculate the similarity of all layers and take the average similarity as the error function.

[0126]

[0127] Run the model, perform forward calculations, and obtain predicted values;

[0128] The number of batch samples Batch_size is set to 20. According to the predicted value and label value, the loss function is calculated, back propagation is performed, and the network parameters are updated.

[0129] Set the number of Epochs for repeated training to 75, observe the convergence results of the loss value, and stop iteration when the loss value converges to the expected target. Name the model Forecast_Model and save the model.

[0130] 6 Model prediction

[0131] Take the average drilling speed, gamma, resistivity, and porosity logging curves from the test data set as input sample data, divide the input data into samples, and divide each sample into ns points, for a total of samples_test_data samples, where:

[0132]

[0133] The input data is organized into a three-dimensional tensor x_test with dimensions [samples_test_data, 4, ns].

[0134] Perform the above two transformations on x_test to obtain two types of image data, and then perform multi-channel merging.

[0135] Construct a 5-dimensional tensor, where the first dimension represents the number of samples, the second dimension represents the time step, the third dimension represents the image width, the fourth dimension represents the image height, and the fifth dimension represents the number of image channels. In the embodiment of this specification, the number of samples is samples_test_data, the time step is 4, the image width and height are both ns, and the number of image channels is 2. The resampling transformation result is used as the first channel, and the probability transformation result is used as the second channel.

[0136] Load the model Forecast_Model, take the above 5-dimensional tensor as input data, and output the predicted data Y after model calculation. The predicted data is the predicted lithology density value.

[0137] By comparing the predicted lithology density values ​​with the actual logging lithology density values ​​of the test set, it is found that the two curves have a good correlation and are basically consistent in the corresponding main layer characteristics. When the true value is missing, the predicted value can be used as a substitute to provide relevant formation physical property information and improve the success rate of geosteering and logging interpretation.

[0138] Further, the above technical features can be found in Figure 2 The data processing flow chart is shown.

[0139] Furthermore, Figure 3 A schematic diagram of a natural gamma logging curve provided in an embodiment of this specification.

[0140] Furthermore, Figure 4The multi-sampling scale logging curve characteristic diagram (left side) and the geological transformation probability characteristic diagram (right side) provided in the embodiments of this specification.

[0141] Furthermore, Figure 5 This is a schematic diagram comparing the predicted results and actual measurement results provided in the embodiments of this specification.

[0142] Figure 6 A schematic diagram of a well logging curve prediction device provided for one or more embodiments of this specification includes:

[0143] at least one processor; and,

[0144] a memory communicatively connected to the at least one processor; wherein,

[0145] The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0146] Acquire well logging data to be input from the current well logging data, and perform geological stratification on the well logging data to be input according to geological stratification characteristics, so as to obtain well logging data to be input for each layer that retains the distribution characteristics of the well logging data for each layer;

[0147] Performing multi-scale resampling on the logging data to be input at each layer to obtain a multi-sampling scale logging characteristic map;

[0148] Determine the geological transformation probability characteristic map between the logging data to be input in each layer according to the geological attribute data acquired in advance;

[0149] The multi-sampling scale well logging characteristic map is combined with the geological transformation probability characteristic map to obtain well logging input data;

[0150] The logging input data is input into a pre-trained logging curve prediction model to predict a logging measurement curve.

[0151] One or more embodiments of this specification provide a non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions are configured to:

[0152] Acquire well logging data to be input from the current well logging data, and perform geological stratification on the well logging data to be input according to geological stratification characteristics, so as to obtain well logging data to be input for each layer that retains the distribution characteristics of the well logging data for each layer;

[0153] Performing multi-scale resampling on the logging data to be input at each layer to obtain a multi-sampling scale logging characteristic map;

[0154] Determine the geological transformation probability characteristic map between the logging data to be input in each layer according to the geological attribute data acquired in advance;

[0155] The multi-sampling scale well logging characteristic map is combined with the geological transformation probability characteristic map to obtain well logging input data;

[0156] The logging input data is input into a pre-trained logging curve prediction model to predict a logging measurement curve.

[0157] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment, and non-volatile computer storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0158] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0159] The above description is only one or more embodiments of this specification and is not intended to limit this specification. For those skilled in the art, one or more embodiments of this specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of one or more embodiments of this specification shall be included in the scope of the claims of this specification.

Claims

1. A method for predicting a well logging curve, characterized in that: The method comprises: Acquire well logging data to be input from the current well logging data, and perform geological stratification on the well logging data to be input according to geological stratification characteristics, so as to obtain well logging data to be input for each layer that retains the distribution characteristics of the well logging data for each layer; Performing multi-scale resampling on the logging data to be input at each layer to obtain a multi-sampling scale logging characteristic map; Determine the geological transformation probability characteristic map between the logging data to be input in each layer according to the geological attribute data acquired in advance; The multi-sampling scale well logging characteristic map is combined with the geological transformation probability characteristic map to obtain well logging input data; Inputting the well logging input data into a pre-trained well logging curve prediction model to predict a well logging measurement curve; The multi-scale resampling of the input well logging data of each layer to obtain a multi-sampling scale well logging characteristic map specifically includes: The well logging data to be inputted in each layer are sampled respectively according to a plurality of sampling intervals set in advance, so as to obtain sampling data corresponding to each sampling interval; The sampling data corresponding to each sampling interval are merged to obtain a multi-sampling scale logging characteristic map.

2. The method according to claim 1, characterized in that The step of geologically stratifying the well logging data to be input according to the geological stratification characteristics to obtain the well logging data to be input for each layer that retains the distribution characteristics of the well logging data for each layer specifically includes: geologically stratify the well logging data to be input according to geological stratification characteristics; Determine the Gaussian distribution parameters corresponding to the distribution of well logging data in each layer respectively; The well logging data to be input is normalized according to the Gaussian distribution parameters to obtain the well logging data to be input for each layer that retains the distribution characteristics of the well logging data for each layer.

3. The method according to claim 1, characterized in that The merging of the sampling data corresponding to each sampling interval to obtain a multi-sampling scale logging characteristic map specifically includes: The sampling data corresponding to each sampling interval are processed respectively by a linear interpolation method to unify the sampling data corresponding to each sampling interval into the same number of sampling points, and determine the sequence of the sampling data corresponding to each sampling interval; The sequences of sampling data corresponding to the sampling intervals are combined to obtain a multi-sampling scale logging characteristic diagram.

4. The method according to claim 1, characterized in that The step of determining the geological transformation probability characteristic map between the well logging data to be inputted at each layer according to the geological attribute data acquired in advance specifically includes: Determine the attribute value distribution interval corresponding to each type of rock according to the geological attribute data; Determine the rock type of each layer of well logging data to be input according to the distribution range of attribute values ​​corresponding to each type of rock; Determine the transformation probability of any two types of rocks according to the rock types of the well logging data to be input in each layer; Determine the transformation probability between any two well logging data to be input according to the transformation probability of the any two types of rocks and the geological strata in which each well logging data to be input is located in each layer of well logging data to be input; According to the transformation probability between any two well logging data to be input, a geological transformation probability characteristic map between the well logging data to be input in each layer is determined.

5. The method according to claim 4, characterized in that Before determining the geological transformation probability characteristic map between the well logging data to be input at each layer, the method further includes: Determine the transformation probability between any two layers of well logging data to be input according to the transformation probability of the any two types of rocks and the rock types of the well logging data to be input at each layer; According to the transformation probability between any two layers of well logging data to be input, a geological transformation probability characteristic map between the layers of well logging data to be input is determined.

6. The method according to claim 1, characterized in that Before inputting the logging input data into the pre-trained logging curve prediction model, the method further includes: Obtaining predicted well logging measurement curves and actual well logging measurement curves; Determine the Gaussian distribution parameters of the predicted well logging measurement curve of each layer and the Gaussian distribution parameters of the actual well logging measurement curve; Determining a loss function of the well logging curve prediction model according to a similarity between a Gaussian distribution of the predicted well logging measurement curve and a Gaussian distribution of the actual well logging measurement curve; The logging curve prediction model is obtained by training according to the loss function.

7. The method according to claim 1, characterized in that The logging data to be input is one or more of an average drilling speed drilling curve, a gamma logging curve, a resistivity logging curve and a porosity logging curve, and the logging measurement curve is a lithology density logging curve.

8. A prediction device for well logging curves, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: Acquire well logging data to be input from the current well logging data, and perform geological stratification on the well logging data to be input according to geological stratification characteristics, so as to obtain well logging data to be input for each layer that retains the distribution characteristics of the well logging data for each layer; Performing multi-scale resampling on the logging data to be input at each layer to obtain a multi-sampling scale logging characteristic map; Determine the geological transformation probability characteristic map between the logging data to be input in each layer according to the geological attribute data acquired in advance; The multi-sampling scale well logging characteristic map is combined with the geological transformation probability characteristic map to obtain well logging input data; Inputting the well logging input data into a pre-trained well logging curve prediction model to predict a well logging measurement curve; The multi-scale resampling of the input well logging data of each layer to obtain a multi-sampling scale well logging characteristic map specifically includes: The well logging data to be inputted in each layer are sampled respectively according to a plurality of sampling intervals set in advance, so as to obtain sampling data corresponding to each sampling interval; The sampling data corresponding to each sampling interval are merged to obtain a multi-sampling scale logging characteristic map.

9. A non-volatile computer storage medium, characterized in that: A computer executable instruction is stored, wherein the computer executable instruction is configured to: Acquire well logging data to be input from the current well logging data, and perform geological stratification on the well logging data to be input according to geological stratification characteristics, so as to obtain well logging data to be input for each layer that retains the distribution characteristics of the well logging data for each layer; Performing multi-scale resampling on the logging data to be input at each layer to obtain a multi-sampling scale logging characteristic map; Determine the geological transformation probability characteristic map between the logging data to be input in each layer according to the geological attribute data acquired in advance; The multi-sampling scale well logging characteristic map is combined with the geological transformation probability characteristic map to obtain well logging input data; Inputting the well logging input data into a pre-trained well logging curve prediction model to predict a well logging measurement curve; The multi-scale resampling of the input well logging data of each layer to obtain a multi-sampling scale well logging characteristic map specifically includes: The well logging data to be inputted in each layer are sampled respectively according to a plurality of sampling intervals set in advance, so as to obtain sampling data corresponding to each sampling interval; The sampling data corresponding to each sampling interval are merged to obtain a multi-sampling scale logging characteristic map.

Images