Lithology and lithofacies identification method and system

Abstract

The invention provides a lithology and lithofacies identification method and system, and solves the problem that the existing method cannot realize lithology and lithofacies intelligent identification. The lithology and lithofacies identification method comprises the following steps: obtaining lithology and lithofacies classification data, and obtaining first logging curve data based on the classification data; constructing a first layer sample set based on the classification data and the first logging curve data; performing normalization processing on the first logging curve data of each layer of samples in the first layer of sample set; carrying out one-hot coding on the classified data of each layer of samples after normalization processing; constructing a convolutional neural network model based on the classification data of each layer of samples after one-hot coding, and training the convolutional neural network model; and inputting the normalized second-layer sample set of the well to be identified into a trained convolutional neural network model, and outputting a lithology and lithofacies classification result corresponding to each layer of sample.

Description

Technical Field

[0001] The present invention relates to the technical field of petroleum geophysical exploration, and particularly relates to a lithology and lithofacies identification method and system. Background Art

[0002] With the great success of artificial intelligence in the fields of image recognition, natural language processing, etc., some scholars have gradually introduced artificial intelligence technology into the field of logging reservoir evaluation to carry out work such as lithology and lithofacies identification based on logging data. From the perspective of current research, most of the artificial intelligence lithology and lithofacies identification based on logging curves uses point samples, that is, one sample for each logging sampling depth point, the input attribute is the logging response value at this depth point or the derived parameter derived from the logging response value, and the sample label is the lithology and lithofacies result obtained from geological research at this depth point. However, from the perspective of logging theory, the logging response at each depth point is a comprehensive reflection of the physical properties of the formation within a certain depth range before and after this depth point, and the physical properties of the formation at each depth point will affect the logging response within a certain depth range before and after this depth point. Therefore, to identify the formation lithology and lithofacies at a certain depth point based on logging data, the logging data within a certain range before and after this depth point should be input. Currently, there is no intelligent identification of lithology and lithofacies using the above method. Summary of the Invention

[0003] In view of this, an embodiment of the present invention provides a lithology and lithofacies identification method and system, which solves the problem that the existing method cannot achieve intelligent identification of lithology and lithofacies.

[0004] In a first aspect, a lithology and lithofacies identification method provided by an embodiment of the present invention includes:

[0005] Obtain lithology and lithofacies classification data, and obtain first logging curve data based on the classification data;

[0006] Construct a first-layer sample set based on the classification data and the first logging curve data;

[0007] Perform normalization processing on the first logging curve data of each layer sample in the first-layer sample set;

[0008] Perform one-hot encoding on the classification data of each layer sample after normalization processing;

[0009] Construct a convolutional neural network model based on the classification data of each layer sample after one-hot encoding and train it;

[0010] Obtain the sampling data of the well to be identified, and obtain second logging curve data based on the sampling data;

[0011] Construct a second-layer sample set based on the sampling data and the second logging curve data;

[0012] Normalize the second logging curve data of each layer sample in the second layer sample set;

[0013] Input the normalized second layer sample set of the well to be identified into the trained convolutional neural network model, and output the lithology and lithofacies classification results corresponding to each layer sample.

[0014] In one embodiment, obtaining the first logging curve data based on the classification data includes: obtaining the data of a preset number of logging sampling points before and after the current depth point from the lithology and lithofacies classification data, and forming the first logging curve data.

[0015] In one embodiment, constructing the first layer sample set based on the classification data and the first logging curve data includes: associating each lithology and lithofacies classification data with the logging data of the corresponding logging sampling point as a sample label to form the first layer sample set; the first layer sample set is a single-point label layer sample set.

[0016] In one embodiment, performing one-hot encoding on the classification data of each layer sample after normalization processing includes: generating a vector for each layer sample label according to lithology and lithofacies classification, where the size of the vector is the same as the number of lithology and lithofacies classifications, setting the corresponding classification in the vector to 1, and setting other positions to 0.

[0017] In one embodiment, constructing and training a convolutional neural network model based on the classification data of each layer sample after one-hot encoding includes:

[0018] Set the input dimension of the convolutional neural network model; among them, the number of channels is the number of logging curves, the convolutional direction is the depth direction, and the output dimension is the number of lithology and lithofacies classifications;

[0019] For the layer sample sets obtained from multiple wells in the work area, set the first layer sample set in some wells as the training set, and set the first layer sample set in other wells as the test set;

[0020] Train the convolutional neural network model with the training set and evaluate the training effect with the test set in real time; among them, when the training cost gradually decreases and the test error is minimized, the optimal convolutional neural network model is obtained.

[0021] In one embodiment, normalizing the first logging curve data of each layer sample in the first layer sample set includes: for the first logging curve data using linear scale, perform normalization processing using the following formula,

[0022]

[0023] where, gvalue igvalue is the logging curve data corresponding to the i-th depth point after normalization, and LSCA and RSCA are the default left and right scales of this curve in the work area respectively.

[0024] In one implementation, the normalization process for the first logging curve data of each layer sample in the first layer sample set includes: for the first logging curve data using logarithmic scale, the following formula is used for normalization.

[0025]

[0026] where gvalue i is the logging curve data corresponding to the i-th depth point after normalization, and LSCA and RSCA are the default left and right scales of this curve in the work area respectively.

[0027] In a second aspect, a lithology and lithofacies identification system provided by an embodiment of the present invention includes:

[0028] An acquisition and analysis module, configured to acquire lithology and lithofacies classification data, and obtain first logging curve data based on the classification data; acquire sampling data of a well to be identified, and obtain second logging curve data based on the sampling data;

[0029] A sample set construction module, configured to construct a first layer sample set based on the classification data and the first logging curve data; construct a second layer sample set based on the sampling data and the second logging curve data;

[0030] A data processing module, configured to normalize the first logging curve data of each layer sample in the first layer sample set; normalize the second logging curve data of each layer sample in the second layer sample set;

[0031] An encoding module, configured to perform one-hot encoding on the classification data of each layer sample after normalization;

[0032] A model construction and training module, configured to construct a convolutional neural network model based on the classification data of each layer sample after one-hot encoding and train it;

[0033] A model analysis module, configured to input the second layer sample set after normalization of the well to be identified into the trained convolutional neural network model, and output the lithology and lithofacies classification results corresponding to each layer sample.

[0034] In a third aspect, an electronic device provided by an embodiment of the present invention includes a memory and a processor, where the memory is used to store one or more computer instructions, and when the one or more computer instructions are executed by the processor, the lithology and lithofacies identification method as described above is implemented.

[0035] Fourthly, a computer-readable storage medium provided by an embodiment of the present invention is characterized in that a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, it is used to implement the lithology and lithofacies identification method as described above.

[0036] A lithology and lithofacies identification method and system provided by an embodiment of the present invention, aiming at the problem of lithology and lithofacies identification in logging reservoir evaluation, proposes to establish a single-point label layer sample set that associates continuous logging curve data with lithology and lithofacies classification data, constructs an intelligent model and trains it, and then uses the intelligent model to predict the lithology and lithofacies of unknown wells. This method makes full use of the reflection ability of logging information on lithology and lithofacies, can meet the needs of lithology and lithofacies identification in logging reservoir evaluation, and has a simple and easy-to-operate invention process. Description of the Drawings

[0037] Figure 1 The figure shows a schematic flow chart of a lithology and lithofacies identification method provided by an embodiment of the present invention.

[0038] Figure 2 The figure shows a schematic structural diagram of a lithology and lithofacies identification system provided by an embodiment of the present invention. Detailed Embodiments

[0039] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0040] To address the above problems, the present invention generally adopts the following methods: 1. Multidimensional reconstruction of the original logging curves, using the original curve features, geological trend features, median filtering features, and clustering features as inputs, and constructing a sedimentary microfacies prediction model based on a bidirectional long short-term memory network; 2. Using data such as core observation and description, thin section analysis, etc. to divide the lithofacies of tight oil formations, taking seven logging parameters as inputs, and establishing a KNN machine learning model to predict lithofacies; 3. Constructing a deep convolutional autoencoder neural network lithology classification model and using a run-length smoothing algorithm to improve the effect of lithology classification; 4. Constructing a random forest lithofacies identification model on an imbalanced shale dataset and comparing the effects with other machine learning algorithms; 5. A lithology intelligent classification method based on the gradient boosting decision tree algorithm, constructing a lithology identification model based on the gradient boosting tree in the evaluation of tight sandstone reservoirs and comparing the application effects with algorithms such as Bayesian, random forest, and support vector machine.

[0041] Based on the current research, most of the artificial intelligence lithology and lithofacies recognition based on well logging curves adopt point samples, that is, one sample for each well logging sampling depth point. The input attribute is the well logging response value at this depth point or the derivative parameter derived from the well logging response value, and the sample label is the lithology and lithofacies result obtained from geological research at this depth point. However, from the perspective of well logging theory, the well logging response at each depth point is a comprehensive reflection of the physical properties of the formation within a certain depth range before and after this depth point. The physical properties of the formation at each depth point will affect the well logging response within a certain depth range before and after this depth point. Therefore, when identifying the lithology and lithofacies of the formation at a certain depth point based on well logging data, the well logging data within a certain range before and after this depth point should be input. Based on this understanding, the present invention proposes a new method for intelligent recognition of lithology and lithofacies based on well logging data. The specific implementation manner is as described in the following embodiments.

[0042] Example 1:

[0043] Lithology and lithofacies recognition is an important basic work in well logging reservoir evaluation. When applying artificial intelligence technology to well logging lithology and lithofacies recognition in the past, point samples were usually adopted, that is, one sample for each well logging sampling depth point. The input attribute is the well logging response value at this depth point or the derivative parameter derived from the well logging response value, and the sample label is the lithology and lithofacies result obtained from geological research at this depth point. However, from the perspective of well logging theory, the well logging response at each depth point is a comprehensive reflection of the physical properties of the formation within a certain depth range before and after this depth point. The physical properties of the formation at each depth point will affect the well logging response within a certain depth range before and after this depth point. Therefore, when identifying the lithology and lithofacies of the formation at a certain depth point based on well logging data, the well logging data within a certain range before and after this depth point should be input. Based on this understanding, this embodiment provides a lithology and lithofacies recognition method for intelligent recognition of lithology and lithofacies based on well logging data.

[0044] Figure 1 The following shows a schematic flow chart of a lithology and lithofacies recognition method provided by an embodiment of the present invention.

[0045] As Figure 1 shown, the lithology and lithofacies recognition method includes:

[0046] Step 01: Obtain lithology and lithofacies classification data, and obtain first well logging curve data based on the classification data.

[0047] Specifically, obtaining the first logging curve data based on the classification data includes: acquiring the data of a preset number of logging sampling points before and after the current depth point from the lithology and lithofacies classification data, and forming the first logging curve data. Sort out the lithology and lithofacies classification data and the logging curve data, and read the logging curve data of N logging sampling points before and after the depth of the lithology and lithofacies classification data, where N≥0; adding the current depth point, there are a total of 2*N + 1 logging sampling points, which are continuous in depth.

[0048] Step 02: Construct the first-layer sample set based on the classification data and the first logging curve data.

[0049] Specifically, constructing the first-layer sample set based on the classification data and the first logging curve data includes: using each lithology and lithofacies classification data as a sample label, associating it with the logging data of the corresponding logging sampling point, and forming the first-layer sample set; the first-layer sample set is a single-point label layer sample set. Using each lithology and lithofacies classification data as a sample label, associating it with the logging data of these (2*N + 1) logging sampling points, and forming a single-point label layer sample set.

[0050] Step 03: Normalize the first logging curve data of each layer sample in the first-layer sample set.

[0051] Specifically, normalize the logging curve data of each layer sample. For curves such as natural gamma and three porosity curves that often use linear scales, perform linear normalization using Equation (1). For resistivity curves such as dual induction, dual laterolog, and microspherical focusing, logarithmic scales are often used, so perform logarithmic normalization using Equation (2).

[0052]

[0053]

[0054] Where gvalue i is the data of a certain logging curve corresponding to the i-th depth point after normalization processing, and LSCA and RSCA are the default left and right scales of this curve in the work area respectively.

[0055] Step 04: Perform one-hot encoding on the classification data of each layer sample after normalization processing.

[0056] Performing one-hot encoding on the classification data of each layer sample after normalization processing includes: generating a vector for each layer of sample labels according to the lithology and lithofacies classification, where the size of the vector is the same as the number of lithology and lithofacies classifications, assigning 1 to the corresponding classification in the vector, and assigning 0 to other positions.

[0057] Taking four types of lithology and lithofacies as an example, the encodings corresponding to each lithology and lithofacies classification are shown in Table 1.

[0058] Table 1 One-hot Encoding for Lithology and Lithofacies Classification (Four Categories)

[0059] Lithology and lithofacies classification Coding Class 1 [1 0 0 0] Class 2 [0 1 0 0] Class 3 [0 0 2 0] Class 4 [0 0 0 3]

[0060] Step 05: Construct and train a convolutional neural network model based on the classification data of each layer sample after one-hot encoding.

[0061] The construction and training of the convolutional neural network model based on the classification data of each layer sample after one-hot encoding includes:

[0062] Step 051: Set the input dimension of the convolutional neural network model; among them, the number of channels is the number of logging curves, the convolutional direction is the depth direction, and the output dimension is the number of lithology and lithofacies classifications. Optionally, set the input dimension of the network model to (2*N + 1, the number of logging curves).

[0063] Step 052: For the layer sample set obtained from multiple wells in the work area, set the first-layer sample set in some wells as the training set, and set the first-layer sample set in other wells as the test set.

[0064] Step 053: Train the convolutional neural network model with the training set and evaluate the training effect in real time with the test set; among them, when the training cost gradually decreases and the test error is minimized, the optimal convolutional neural network model is obtained. Among them, the training cost is the sum of the prediction errors of the training set layer samples; the test error is the sum of the prediction errors of the test set layer samples.

[0065] Step 06: Obtain the sampling data of the well to be identified, and obtain the second logging curve data based on the sampling data.

[0066] Step 07: Construct a second-layer sample set based on the sampling data and the second logging curve data.

[0067] Step 08: Normalize the second logging curve data of each layer sample in the second-layer sample set.

[0068] Step 09: Input the second-layer sample set of the well to be identified after normalization into the trained convolutional neural network model, and output the lithology and lithofacies classification results corresponding to each layer sample.

[0069] Specifically, in steps 06 to 09, a layer sample set is constructed for the unknown well, and a one-dimensional convolutional neural network model is used to predict the lithology and lithofacies of the unknown well. For wells without lithology and lithofacies data, according to the methods in steps 01 and 02, for the data of each logging sampling point, a layer sample is constructed by associating it with the subsequent 2*N logging sampling points. That is, assuming there are M logging sampling depths in the logging data file, then (M - 2*N) layer samples can be constructed; where N≥0 and M≥0. The logging curve data of the layer samples are normalized according to the method in step 03. The logging curves and parameters for normalization should correspond to those in step 03. The dimension of each layer sample is (2*N + 1, the number of logging curves), which is consistent with the input attribute dimension of the one-dimensional convolutional neural network model in step 05. The layer sample set of the unknown well is input into the trained intelligent model, and the lithology and lithofacies code corresponding to each layer sample can be predicted. The code is converted into the lithology and lithofacies classification result. The depth of the predicted lithology and lithofacies classification result is set as the depth of the (N + 1)-th logging sampling point of the layer sample (i.e., the central depth of the layer sample), and the predicted lithology and lithofacies classification result is obtained as (depi, resi), where resi is the lithology and lithofacies classification result predicted for the i-th layer sample of the unknown well, and depi is the depth corresponding to the predicted lithology and lithofacies classification result of the i-th layer sample of the unknown well.

[0070] In this embodiment, aiming at the problem of lithology and lithofacies identification in logging reservoir evaluation, a method is proposed to establish a single-point label layer sample set that associates continuous logging curve data with lithology and lithofacies classification data, construct an intelligent model and train it, and then use the intelligent model to predict the lithology and lithofacies of unknown wells. This fully utilizes the reflection ability of logging information on lithology and lithofacies and can meet the needs of lithology and lithofacies identification in logging reservoir evaluation. The invention process is simple and easy to operate.

[0071] Example 2:

[0072] This embodiment takes the identification of igneous rock lithology as an example.

[0073] First, a layer sample set is constructed for 8 wells with lithology and lithofacies data in the work area: organize the lithology classification data file, which has a total of 416 lithology classification data obtained from core observation and description, core thin section identification experiments, etc. Read the logging curve data of the corresponding wells according to the depth of the lithology classification data. Set N = 16, that is, read the logging curve data at the depth corresponding to the lithology classification data and 16 logging sampling points before and after it. Associate each lithology classification data with the logging curve data of the 33 logging sampling points read to obtain a single-point label layer sample set.

[0074] Normalize the well logging data of layer samples. There are 5 well logging curves in the layer sample set, namely natural gamma (GR), neutron porosity (CNL), density (DEN), acoustic travel time (AC), and deep lateral resistivity (RD). Usually, GR, CNL, DEN, and AC use linear scales, and RD uses a logarithmic scale. The default left and right scales in the work area are: GR: 0, 300; CNL: 45, -15; DEN: 1.85, 2.85; AC: 450, 150; ILD: 10, 10000. Perform well logging normalization on the well logging data of layer samples according to equations (1) and (2).

[0075] Perform one-hot encoding on the lithology classification data in the layer samples. The lithology classification in this area includes basalt, trachyte, andesite, and rhyolite. Therefore, the encoding is: basalt [1 0 0 0], trachyte [0 1 0 0], andesite [0 0 1 0], rhyolite [0 0 0 1].

[0076] Construct and train a one-dimensional convolutional neural network model: Since each layer sample well logging data has 33 well logging sampling points, and there are 5 well logging curves at each well logging sampling point, the input attribute dimension of the designed network model is (33, 5), that is, the number of input channels is 5, and the convolution direction is the depth direction. Design the first hidden layer of the network model as a one-dimensional convolutional layer: The parameters of this layer are: 16 one-dimensional convolutional kernels, the size of the convolutional kernel is 8, the stride is 1, and the activation function is the RELU function; Design the second hidden layer of the network model as a pooling layer: The parameters of this layer are: the pooling method is max pooling, the filter size is 2, and the stride is 2; Design the third hidden layer of the network model as a one-dimensional convolutional layer: The parameters of this layer are the same as those of the first hidden layer. Design the fourth hidden layer of the network model as a pooling layer: The parameters of this layer are the same as those of the second hidden layer; Design the fifth hidden layer of the network model as a flattening layer: Flatten the output of the fourth hidden layer to output a one-dimensional vector; Design the output layer of the network model as a fully connected layer, with the number of neurons being 4, and the activation function being the SoftMax function, and output the lithology classification encoding predicted by the network model.

[0077] Among the obtained layer samples, 267 layer samples from 5 wells were used as the training set, and 149 layer samples from another 3 wells were used as the test set. Using the Keras tool of the TensorFlow machine learning platform, the aforementioned one-dimensional convolutional neural network model was constructed. The training set was input for iterative training. Each iteration updated the weights of the network model once. The network model with updated weights was used to predict the training set, compared with the training set labels (lithology classification data encoding), and the training cost was calculated and output. At the same time, the network model with updated weights was used to predict the test set, compared with the test set labels, and the error was calculated and the model with the smallest error was saved. After 10,000 times of training, it was found that the training cost continued to decrease during the training process, while the trend of the prediction error for the test set was to decrease first and then increase. The error was the smallest at the 3136th time of training. The network model at this time can be considered the optimal model.

[0078] The optimal network model was applied to unknown wells: The lithology of the Yingcheng Formation of Well XX6 and Well XX11 in this area was identified. The logging data of the Yingcheng Formation of Well XX6 and Well XX11 were processed at each depth point. The logging curve data of each depth point and the 32 (2*N = 32) depth points behind it formed a layer sample. The logging curve data of each layer sample was of dimension (33,5). The logging curve data of the layer sample was normalized and input into the network model to obtain the predicted lithology classification code. Its depth was set as the center depth of the layer sample, and the code was restored to the lithology classification result. Well XX6 and Well XX11 did not participate in the training of the intelligent model and were blind wells. The identification results of the intelligent model were compared with the lithology classification results obtained by methods such as core experiments. Among the 123 lithology classification result data, 101 lithology identifications by the intelligent model were correct. The lithology identification accuracy of the blind well test was 89.4%, which confirmed the effectiveness of this method. The comparison results are shown in Table 2.

[0079] Table 2 Comparison of the predicted igneous rock lithology and the actual lithology by the intelligent model:

[0080]

[0081] This embodiment proposes a method for establishing a single-point label layer sample set that associates continuous logging curve data with lithology and lithofacies classification data, constructing an intelligent model and training it, and then using the intelligent model to predict the lithology and lithofacies of unknown wells. It makes full use of the reflection ability of logging information on lithology and lithofacies and can meet the needs of lithology and lithofacies identification in logging reservoir evaluation. The invention process is simple and easy to operate.

[0082] Example 3:

[0083] This embodiment provides a lithology and lithofacies identification system 100. Figure 2 The following shows the structural schematic diagram of a lithology and lithofacies identification system provided by an embodiment of the present invention; refer to Figure 2As shown in the figure, the lithology and lithofacies identification system includes an acquisition and analysis module 10, a sample set construction module 20, a data processing module 30, an encoding module 40, a model construction and training module 50, and a model analysis module 60. Among them,

[0084] The acquisition and analysis module 10 is used to acquire lithology and lithofacies classification data, and obtain the first logging curve data based on the classification data; acquire the sampling data of the well to be identified, and obtain the second logging curve data based on the sampling data.

[0085] The sample set construction module 20 is used to construct the first-layer sample set based on the classification data and the first logging curve data; construct the second-layer sample set based on the sampling data and the second logging curve data.

[0086] The data processing module 30 is used to perform normalization processing on the first logging curve data of each layer sample in the first-layer sample set; perform normalization processing on the second logging curve data of each layer sample in the second-layer sample set.

[0087] The encoding module 40 is used to perform one-hot encoding on the classification data of each layer sample after normalization processing.

[0088] The model construction and training module 50 is used to construct a convolutional neural network model based on the classification data of each layer sample after one-hot encoding and train it.

[0089] The model analysis module 60 is used to input the second-layer sample set after normalization processing of the well to be identified into the trained convolutional neural network model, and output the lithology and lithofacies classification results corresponding to each layer sample.

[0090] Furthermore, the acquisition and analysis module 10 is also used to acquire the data of a preset number of logging sampling points before and after the current depth point from the lithology and lithofacies classification data, and form the first logging curve data. Organize the lithology and lithofacies classification data and the logging curve data, read the logging curve data of N logging sampling points before and after the depth of the lithology and lithofacies classification data, where N≥0; add the current depth point, a total of 2*N + 1 logging sampling points, and the depth is continuous.

[0091] Furthermore, the sample set construction module 20 is also used to use each lithology and lithofacies classification data as a sample label, associate it with the logging data of the corresponding logging sampling point, and form the first-layer sample set; the first-layer sample set is a single-point label layer sample set. Use each lithology and lithofacies classification data as a sample label, associate it with the logging data of these (2*N + 1) logging sampling points, and form a single-point label layer sample set.

[0092] Further, the encoding module 40 is further configured to generate a vector for each layer of sample labels classified by lithology and lithofacies, where the size of the vector is the same as the number of lithology and lithofacies classifications, assign 1 to the corresponding classification in the vector, and assign 0 to other positions.

[0093] Further, the model construction and training module 50 is further configured to set the input dimension of the convolutional neural network model; wherein, the number of channels is the number of logging curves, the convolutional direction is the depth direction, and the output dimension is the number of lithology and lithofacies classifications; for the layer sample sets obtained from multiple wells in the work area, set the first layer sample set in a part of the wells as the training set, and set the first layer sample set in another part of the wells as the test set; train the convolutional neural network model with the training set and evaluate the training effect in real time with the test set; wherein, when the training cost gradually decreases and the test error is minimized, the optimal convolutional neural network model is obtained. Wherein, the training cost is the sum of the prediction errors of the layer samples in the training set; the test error is the sum of the prediction errors of the layer samples in the test set.

[0094] Further, the data processing module 30 is further configured to:

[0095] For the first logging curve data with linear scale, the following formula (1) is used for normalization processing,

[0096]

[0097] For the first logging curve data with logarithmic scale, the following formula (2) is used for normalization processing,

[0098]

[0099] wherein, gvalue i is the logging curve data corresponding to the i-th depth point after normalization processing; LSCA and RSCA are the default left and right scales of the curve in the work area, respectively.

[0100] Further, the model analysis module 60 is further configured to construct a layer sample set for an unknown well and use a one-dimensional convolutional neural network model to predict the lithology and lithofacies of the unknown well. For a well without lithology and lithofacies data, according to the methods in step 01 and step 02, for the data of each logging sampling point, construct 1 layer sample with the data of the subsequent 2*N logging sampling points, that is, assuming there are M logging sampling depths in the logging data file, then (M - 2*N) layer samples can be constructed; where N≥0, M≥0. Normalize the logging curves of the layer samples according to the method in step 03. The logging curves and parameters for normalization need to correspond to those in step 03. The dimension of each layer sample is (2*N + 1, number of logging curves), which is consistent with the input attribute dimension of the one-dimensional convolutional neural network model in step 05. Input the unknown well layer sample set into the trained intelligent model, and the lithology and lithofacies code corresponding to each layer sample can be predicted. Convert the code into a lithology and lithofacies classification result, and set the depth of the predicted lithology and lithofacies classification result to the depth of the (N + 1)th logging sampling point of the layer sample (i.e., the central depth of the layer sample), and obtain the predicted lithology and lithofacies classification result as (depi, resi), where resi is the lithology and lithofacies classification result predicted for the i-th layer sample of the unknown well, and depi is the depth corresponding to the predicted lithology and lithofacies classification result of the i-th layer sample of the unknown well.

[0101] A lithology and lithofacies identification method and system provided by an embodiment of the present invention aims at the problem of lithology and lithofacies identification in logging reservoir evaluation, proposes to establish a single-point label layer sample set that associates continuous logging curve data with lithology and lithofacies classification data, constructs an intelligent model and trains it, and then uses the intelligent model to predict the lithology and lithofacies of unknown wells. It makes full use of the reflection ability of logging information on lithology and lithofacies, can meet the needs of lithology and lithofacies identification in logging reservoir evaluation, and the invention process is simple and easy to operate.

[0102] Example 4:

[0103] This embodiment provides an electronic device, which can be a mobile phone, a computer, a tablet computer, etc., including a memory and a processor. A computer program is stored on the memory, and when the computer program is executed by the processor, it implements the lithology and lithofacies identification method as described in Embodiment 1. It can be understood that the electronic device may further include an input / output (I / O) interface and a communication component.

[0104] Among them, the processor is used to execute all or part of the steps in the lithology and lithofacies identification method in Embodiment 1. The memory is used to store various types of data, which may include, for example, instructions of any application program or method in the electronic device, as well as data related to the application program.

[0105] The processor may be implemented by an Application Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Programmable Logic Device (PLD), a Field Programmable Gate Array (FPGA), a controller, a microcontroller, a microprocessor, or other electronic components, and is used to execute the lithology and lithofacies identification method in the first embodiment above.

[0106] Figure 1 The figure shows a schematic flowchart of a lithology and lithofacies identification method provided by an embodiment of the present invention.

[0107] As Figure 1 shown, specifically, the lithology and rock identification method includes:

[0108] Step 01: Obtain lithology and lithofacies classification data, and obtain first logging curve data based on the classification data.

[0109] Specifically, the obtaining of the first logging curve data based on the classification data includes: obtaining the data of a preset number of logging sampling points before and after the current depth point from the lithology and lithofacies classification data, and forming the first logging curve data. Sort out the lithology and lithofacies classification data and the logging curve data, and read the logging curve data of N logging sampling points before and after the depth of the lithology and lithofacies classification data, where N≥0; adding the current depth point, there are a total of 2*N + 1 logging sampling points, which are continuous in depth.

[0110] Step 02: Construct a first-layer sample set based on the classification data and the first logging curve data.

[0111] Specifically, the constructing of the first-layer sample set based on the classification data and the first logging curve data includes: using each lithology and lithofacies classification data as a sample label, associating it with the logging data of the corresponding logging sampling point, and forming the first-layer sample set; the first-layer sample set is a single-point label layer sample set. Use each lithology and lithofacies classification data as a sample label, associate it with the logging data of these (2*N + 1) logging sampling points, and form a single-point label layer sample set.

[0112] Step 03: Perform normalization processing on the first logging curve data of each layer sample in the first-layer sample set.

[0113] Specifically, normalize the logging curve data of each layer sample. For curves such as natural gamma and triple porosity curves that often use linear scale, perform linear normalization using Equation (1). For resistivity curves such as dual induction, dual laterolog, and microspherical focusing, logarithmic scale is often used, so perform logarithmic normalization using Equation (2).

[0114]

[0115]

[0116] where gvalue i is the logging curve data corresponding to the i-th depth point after normalization, and LSCA and RSCA are the default left and right scales of this curve in the work area respectively.

[0117] Step 04: Perform one-hot encoding on the classification data of each layer sample after normalization.

[0118] The one-hot encoding of the classification data of each layer sample after normalization includes: generating a vector for each layer sample label classified by lithology and lithofacies, where the size of the vector is the same as the number of lithology and lithofacies classifications, setting the corresponding classification in the vector to 1, and setting other positions to 0.

[0119] Taking four types of lithology and lithofacies as an example, the encodings corresponding to each lithology and lithofacies classification are shown in Table 1.

[0120] Table 1 One-hot Encoding of Lithology and Lithofacies Classification (Four-classification)

[0121] Lithology and lithofacies classification Coding Class 1 [1 0 0 0] Class 2 [0 1 0 0] Class 3 [0 0 2 0] Class 4 [0 0 0 3]

[0122] Step 05: Build a convolutional neural network model based on the classification data of each layer sample after one-hot encoding and train it.

[0123] The building and training of the convolutional neural network model based on the classification data of each layer sample after one-hot encoding includes:[[]]

[0124] Step 051: Set the input dimension of the convolutional neural network model; among them, the number of channels is the number of logging curves, the convolutional direction is the depth direction, and the output dimension is the number of lithology and lithofacies classifications. Optionally, set the input dimension of the network model to (2*N + 1, the number of logging curves).

[0125] Step 052: For the layer sample set obtained from multiple wells in the work area, set the first layer sample set in a part of the wells as the training set, and set the first layer sample set in another part of the wells as the test set.

[0126] Step 053: Train the convolutional neural network model with the training set and evaluate the training effect in real time with the test set; among them, when the training cost gradually decreases and the test error is minimized, the optimal convolutional neural network model is obtained. Among them, the training cost is the sum of the prediction errors of the training set layer samples; the test error is the sum of the prediction errors of the test set layer samples.

[0127] Step 06: Obtain the sampling data of the well to be identified, and obtain the second logging curve data based on the sampling data.

[0128] Step 07: Construct a second-layer sample set based on the sampling data and the second logging curve data.

[0129] Step 08: Normalize the second logging curve data of each layer sample in the second-layer sample set.

[0130] Step 09: Input the normalized second-layer sample set of the well to be identified into the trained convolutional neural network model, and output the lithology and lithofacies classification results corresponding to each layer sample.

[0131] Specifically, for steps 06 to 09, construct a layer sample set for the unknown well and use the one-dimensional convolutional neural network model to predict the lithology and lithofacies of the unknown well. For wells without lithology and lithofacies data, according to the methods in steps 01 and 02, for the data of each logging sampling point, construct 1 layer sample with the data of the next 2*N logging sampling points. That is, assuming there are M logging sampling depths in the logging data file, then (M - 2*N) layer samples can be constructed; where N≥0, M≥0. Normalize the logging curve data of the layer samples according to the method in step 03. The logging curves and parameters for normalization need to correspond to those in step 03. The dimension of each layer sample is (2*N + 1, number of logging curves), which is consistent with the input attribute dimension of the one-dimensional convolutional neural network model in step 05. Input the unknown well layer sample set into the trained intelligent model, and the lithology and lithofacies coding corresponding to each layer sample can be predicted. Convert the coding into the lithology and lithofacies classification result, and set the depth of the predicted lithology and lithofacies classification result to the depth of the (N + 1)th logging sampling point of the layer sample (i.e., the central depth of the layer sample). The predicted lithology and lithofacies classification result is (depi, resi), where resi is the lithology and lithofacies classification result predicted for the i-th layer sample of the unknown well, and depi is the depth corresponding to the predicted lithology and lithofacies classification result of the i-th layer sample of the unknown well.

[0132] This embodiment aims at the problem of lithology and lithofacies identification in logging reservoir evaluation, and proposes a method of establishing a single-point tag layer sample set that correlates continuous logging curve data with lithology and lithofacies classification data, constructing an intelligent model and training it, and then using the intelligent model to predict the lithology and lithofacies of unknown wells. This method makes full use of the reflection ability of logging information on lithology and lithofacies, and can meet the needs of lithology and lithofacies identification in logging reservoir evaluation. The invention process is simple and easy to operate.

[0133] The memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM for short), electrically erasable programmable read-only memory (EEPROM for short), erasable programmable read-only memory (EPROM for short), programmable read-only memory (PROM for short), read-only memory (ROM for short), magnetic memory, flash memory, magnetic disk or optical disk.

[0134] Example 5:

[0135] This embodiment also provides a computer-readable storage medium. In each embodiment of the present invention, each functional unit can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.

[0136] Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention.

[0137] The aforementioned storage medium includes: flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, server, APP application mall, and various other media that can store program verification codes. A computer program is stored thereon, and when the computer program is executed by a processor, the following method steps can be implemented:

[0138] Step 01: Obtain lithology and lithofacies classification data, and obtain first logging curve data based on the classification data.

[0139] Specifically, obtaining the first logging curve data based on the classification data includes: obtaining data of a preset number of logging sampling points before and after the current depth point from the lithology and lithofacies classification data, and forming first logging curve data. Organize the lithology and lithofacies classification data and the logging curve data, and read the logging curve data of N logging sampling points before and after the depth of the lithology and lithofacies classification data, where N≥0; adding the current depth point, there are a total of 2*N + 1 logging sampling points, which are continuous in depth.

[0140] Step 02: Construct a first-layer sample set based on the classification data and the first logging curve data.

[0141] Specifically, constructing the first-layer sample set based on the classification data and the first logging curve data includes: using each lithology and lithofacies classification data as a sample label, associating it with the logging data of the corresponding logging sampling point, and forming a first-layer sample set; the first-layer sample set is a single-point label layer sample set. Use each lithology and lithofacies classification data as a sample label, associate it with the logging data of these (2*N + 1) logging sampling points, and form a single-point label layer sample set.

[0142] Step 03: Perform normalization processing on the first logging curve data of each layer sample in the first-layer sample set.

[0143] Specifically, perform normalization processing on the logging curve data of each layer sample. For curves such as natural gamma and three porosity curves that often use linear scales, perform linear normalization processing using Equation (1). For resistivity curves such as dual induction, dual lateral, and microspherical focusing, logarithmic scales are often used, so perform logarithmic normalization processing using Equation (2).

[0144]

[0145]

[0146] where gvalue i$x_{i}$ is the data of a certain logging curve corresponding to the $i$-th depth point after normalization, and LSCA and RSCA are the default left and right scales of this curve in the work area, respectively.

[0147] Step 04: Perform one-hot encoding on the classification data of each layer sample after normalization.

[0148] The one-hot encoding of the classification data of each layer sample after normalization includes: generating a vector for each layer sample label according to lithology and lithofacies classification, where the size of the vector is the same as the number of lithology and lithofacies classifications, setting the corresponding classification in the vector to 1, and setting other positions to 0.

[0149] Taking four types of lithology and lithofacies as an example, the encodings corresponding to each lithology and lithofacies classification are shown in Table 1.

[0150] Table 1 One-hot Encoding of Lithology and Lithofacies Classification (Four-Class Classification)

[0151]

[0152]

[0153] Step 05: Construct a convolutional neural network model based on the classification data of each layer sample after one-hot encoding and train it.

[0154] The construction and training of the convolutional neural network model based on the classification data of each layer sample after one-hot encoding include:

[0155] Step 051: Set the input dimension of the convolutional neural network model; where the number of channels is the number of logging curves, the convolutional direction is the depth direction, and the output dimension is the number of lithology and lithofacies classifications. Optionally, set the input dimension of the network model to (2*N + 1, the number of logging curves).

[0156] Step 052: For the layer sample set obtained from multiple wells in the work area, set the first layer sample set in some wells as the training set, and set the first layer sample set in other wells as the test set.

[0157] Step 053: Train the convolutional neural network model with the training set and evaluate the training effect with the test set in real time; when the training cost gradually decreases and the test error is minimized, the optimal convolutional neural network model is obtained. Among them, the training cost is the sum of the prediction errors of the layer samples in the training set; the test error is the sum of the prediction errors of the layer samples in the test set.

[0158] Step 06: Obtain the sampling data of the well to be identified and obtain the second logging curve data based on the sampling data.

[0159] Step 07: Construct a second layer sample set based on the sampling data and the second logging curve data.

[0160] Step 08: Normalize the second logging curve data of each layer sample in the second-layer sample set.

[0161] Step 09: Input the normalized second-layer sample set of the well to be identified into the trained convolutional neural network model, and output the lithology and lithofacies classification results corresponding to each layer sample.

[0162] Specifically, in Steps 06 to 09, a layer sample set is constructed for the unknown well and the lithology and lithofacies of the unknown well are predicted using a one-dimensional convolutional neural network model. For a well without lithology and lithofacies data, according to the methods in Steps 01 and 02, for the data of each logging sampling point, one layer sample is constructed by combining it with the subsequent 2*N logging sampling points. That is, assuming there are M logging sampling depths in the logging data file, then (M - 2*N) layer samples can be constructed; where N≥0, M≥0. The logging curve data of the layer samples is log-normalized according to the method in Step 03, and the logging curves and parameters for normalization need to correspond to those in Step 03. The dimension of each layer sample is (2*N + 1, number of logging curves), which is consistent with the input attribute dimension of the one-dimensional convolutional neural network model in Step 05. Input the unknown well layer sample set into the trained intelligent model, and the lithology and lithofacies coding corresponding to each layer sample can be predicted. Convert the coding into the lithology and lithofacies classification result, and set the depth of the predicted lithology and lithofacies classification result to the depth of the (N + 1)-th logging sampling point of the layer sample (i.e., the central depth of the layer sample). The predicted lithology and lithofacies classification result is (depi, resi), where resi is the lithology and lithofacies classification result predicted for the i-th layer sample of the unknown well, and depi is the depth corresponding to the predicted lithology and lithofacies classification result of the i-th layer sample of the unknown well.

[0163] In this embodiment, aiming at the problem of lithology and lithofacies identification in logging reservoir evaluation, a method is proposed to establish a single-point label layer sample set that associates continuous logging curve data with lithology and lithofacies classification data, construct an intelligent model and train it, and then use the intelligent model to predict the lithology and lithofacies of unknown wells. This method makes full use of the reflection ability of logging information on lithology and lithofacies and can meet the needs of lithology and lithofacies identification in logging reservoir evaluation. The invention process is simple and easy to operate.

[0164] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention. Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0165] In several embodiments provided in the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the couplings, direct couplings, or communication connections shown or discussed with each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0166] The basic principles of the present application have been described above in combination with specific embodiments. However, it should be noted that the advantages, benefits, effects, etc. mentioned in the present application are only examples and not limitations. It cannot be considered that these advantages, benefits, effects, etc. are essential for each embodiment of the present application. In addition, the specific details disclosed above are only for the purposes of illustration and easy understanding, rather than limitations. These details do not limit the present application to necessarily adopt the above specific details for implementation.

[0167] The block diagrams of the devices, apparatuses, equipment, and systems involved in the present application are only illustrative examples and do not intend to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner.

[0168] It should also be noted that in the devices, equipment, and methods of the present application, each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent solutions of the present application.

[0169] The foregoing description of the disclosed aspects enables any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of this application. Thus, this application is not intended to be limited to the aspects shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0170] In the description of this application, the meaning of "a plurality of" is at least two, such as two, three, etc., unless otherwise specifically defined. All directional indications (such as up, down, left, right, front, back, top, bottom...) in the embodiments of this application are only used to explain the relative positional relationship and movement conditions between components in a specific posture (as shown in the drawings). If the specific posture changes, the directional indications will also change accordingly. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally further include steps or units not listed, or may optionally further include other steps or units inherent to these processes, methods, products or devices.

[0171] In addition, the mention of "embodiments" in this context means that the specific features, structures, or characteristics described in connection with the embodiments may be included in at least one embodiment of this application. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein may be combined with other embodiments.

[0172] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, and all of them should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims. The above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for lithology and lithofacies identification, characterized in that, Including: Obtain lithology and lithofacies classification data, and obtain first logging curve data based on the classification data; Construct a first-layer sample set based on the classification data and the first logging curve data; Normalize the first logging curve data of each layer sample in the first-layer sample set; Perform one-hot encoding on the classification data of each layer sample after normalization; Construct a convolutional neural network model based on the classification data of each layer sample after one-hot encoding and train it; Obtain sampling data of the well to be identified, and obtain second logging curve data based on the sampling data; Construct a second-layer sample set based on the sampling data and the second logging curve data; Normalize the second logging curve data of each layer sample in the second-layer sample set; Input the second-layer sample set of the well to be identified after normalization into the trained convolutional neural network model, and output the lithology and lithofacies classification results corresponding to each layer sample.

2. The lithology and lithofacies identification method according to claim 1, wherein The obtaining the first logging curve data based on the classification data includes: obtaining data of a preset number of logging sampling points before and after the current depth point from the lithology and lithofacies classification data, and forming the first logging curve data.

3. The lithology and lithofacies identification method according to claim 2, characterized in that The constructing the first-layer sample set based on the classification data and the first logging curve data includes: using each lithology and lithofacies classification data as a sample label, associating it with the logging data of the corresponding logging sampling point, and forming the first-layer sample set; the first-layer sample set is a single-point label layer sample set.

4. The lithology and lithofacies identification method according to claim 3, characterized in that The performing one-hot encoding on the classification data of each layer sample after normalization includes: generating a vector for each layer of sample labels according to lithology and lithofacies classification, where the size of the vector is the same as the number of lithology and lithofacies classifications, setting the corresponding classification in the vector to 1, and setting other positions to 0.

5. The lithology and lithofacies identification method according to claim 1, characterized in that The constructing a convolutional neural network model based on the classification data of each layer sample after one-hot encoding and training it includes: Setting the input dimension of the convolutional neural network model; among them, the number of channels is the number of logging curves, the convolutional direction is the depth direction, and the output dimension is the number of lithology and lithofacies classifications; For the layer sample sets obtained from multiple wells in the work area, set the first-layer sample sets in some wells as the training set, and set the first-layer sample sets in other wells as the test set; Train the convolutional neural network model with the training set and evaluate the training effect with the test set in real time; among them, when the training cost gradually decreases and the test error is minimized, the optimal convolutional neural network model is obtained.

6. The lithology and lithofacies identification method according to claim 1, characterized in that The normalizing the first logging curve data of each layer sample in the first-layer sample set includes: for the first logging curve data using linear scale, perform normalization processing using the following formula, Among them, gvalue i is the logging curve data corresponding to the i-th depth point after normalization; LSCA and RSCA are the default left and right scales of this curve in the work area, respectively.

7. The lithology and lithofacies identification method according to claim 1, characterized in that The normalizing the first logging curve data of each layer sample in the first-layer sample set includes: for the first logging curve data using logarithmic scale, perform normalization processing using the following formula, Among them, gvalue i is the data of a certain logging curve corresponding to the i-th depth point after normalization; LSCA and RSCA are the default left and right scales of this curve in the work area, respectively.

8. A lithology and lithofacies identification system, characterized in that, Including: An acquisition analysis module for obtaining lithology and lithofacies classification data and obtaining first logging curve data based on the classification data; Obtain sampling data of the well to be identified, and obtain second logging curve data based on the sampling data; A sample set construction module, configured to construct a first-layer sample set based on the classification data and the first logging curve data; and construct a second-layer sample set based on the sampling data and the second logging curve data; A data processing module, configured to perform normalization processing on the first logging curve data of each layer sample in the first-layer sample set; Perform normalization processing on the second logging curve data of each layer sample in the second-layer sample set; An encoding module, configured to perform one-hot encoding on the classification data of each layer sample after normalization processing; A model construction and training module, configured to construct and train a convolutional neural network model based on the classification data of each layer sample after one-hot encoding; A model analysis module, configured to input the second-layer sample set after normalization processing of the well to be identified into the trained convolutional neural network model, and output the lithology and lithofacies classification results corresponding to each layer sample.

9. An electronic device, characterized in that, It includes a memory and a processor, where the memory is used to store one or more computer instructions, and when the one or more computer instructions are executed by the processor, the lithology and lithofacies identification method according to any one of the above claims 1-7 is implemented.

10. A computer-readable storage medium, characterized in that, A computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, it is used to implement the lithology and lithofacies identification method according to any one of the above claims 1-7.

Images