Prediction method, system and electronic device for lithology category and rock layer thickness of drilling

Abstract

The present invention discloses a method, system and electronic device for predicting the lithology category and rock layer thickness of a borehole, relating to the technical field of virtual borehole simulation. The method includes: obtaining a borehole coordinate sequence of a virtual borehole; the borehole coordinate sequence includes a plurality of borehole coordinates; the virtual borehole is an unprobed borehole; inputting the borehole coordinate sequence of the virtual borehole into a lithology category prediction model to obtain a lithology category sequence corresponding to the virtual borehole; the lithology category prediction model is obtained based on a bidirectional gated recurrent unit network, and the lithology category sequence includes the lithology category corresponding to each borehole coordinate; inputting the borehole coordinate sequence of the virtual borehole and the corresponding lithology category sequence into a rock layer thickness prediction model to obtain a rock layer thickness sequence corresponding to the virtual borehole; the rock layer thickness prediction model is obtained based on a gated recurrent unit network, and the rock layer thickness sequence includes the rock layer thickness corresponding to each borehole coordinate. The present invention realizes the simulation of the stratigraphic structure in the borehole blank area.

Description

Technical Field

[0001] The present invention relates to the technical field of virtual drilling simulation, and particularly to a method, a system and an electronic device for predicting the lithology type and the formation thickness of a borehole. Background Art

[0002] The stratum structure is the result of long-term geological processes and has statistical regularity macroscopically. Understanding the stratum structure and its laws is an important topic in 3D geological modeling. Among them, borehole data is an important data source for studying the stratum structure. In terms of acquisition, the accuracy of borehole data is high, but the acquisition cost is large and the time-consuming for collecting borehole data is long. In terms of spatial distribution, boreholes are discretely distributed in the study area, and due to the high cost of boreholes, the number of boreholes is relatively small. Therefore, how to extract the information in borehole data to obtain the geological information in the borehole blank area is an urgent problem to be solved in mineral exploration.

[0003] Currently, the main method for simulating the stratum situation in a specific area is to select different interpolation methods to connect borehole data and draw a 2D geological profile or establish a 3D geological model. Among them, there are many interpolation methods to choose from, such as inverse distance interpolation, Kriging interpolation, etc. These interpolation methods have their own advantages and disadvantages, and there are certain differences in the final effects. Inevitably, subjective factors are involved in the process of selecting interpolation methods, lacking scientificity. Summary of the Invention

[0004] The purpose of the present invention is to provide a method, a system and an electronic device for predicting the lithology type and the formation thickness of a borehole, which realizes the simulation of the stratum structure in the borehole blank area.

[0005] To achieve the above purpose, the present invention provides the following solutions:

[0006] A method for predicting the lithology type and the formation thickness of a borehole, comprising:

[0007] Obtaining a borehole coordinate sequence of a virtual borehole; the borehole coordinate sequence includes a plurality of borehole coordinates; the virtual borehole is an unprobed borehole;

[0008] Inputting the borehole coordinate sequence of the virtual borehole into a lithology type prediction model to obtain a lithology type sequence corresponding to the virtual borehole; the lithology type prediction model is obtained based on a bidirectional gated recurrent unit network, and the lithology type sequence includes the lithology type corresponding to each borehole coordinate;

[0009] Inputting the borehole coordinate sequence of the virtual borehole and the corresponding lithology type sequence into a formation thickness prediction model to obtain a formation thickness sequence corresponding to the virtual borehole; the formation thickness prediction model is obtained based on a gated recurrent unit network, and the formation thickness sequence includes the formation thickness corresponding to each borehole coordinate.

[0010] Optionally, the training process of the lithology type prediction model and the formation thickness prediction model includes:

[0011] Obtain multiple sets of historical borehole data; the historical borehole data includes the coordinate sequence of the actual borehole, the corresponding lithology type sequence, and the corresponding formation thickness sequence;

[0012] Based on the bidirectional gated recurrent unit network, construct an initial lithology type prediction network;

[0013] Based on the gated recurrent unit network, construct an initial formation thickness prediction network;

[0014] Use the coordinate sequence of each actual borehole and the corresponding lithology type sequence to train the initial lithology type prediction network to obtain the lithology type prediction model;

[0015] Use the coordinate sequence of each actual borehole, the corresponding lithology type sequence, and the corresponding formation thickness sequence to train the initial formation thickness prediction network to obtain the formation thickness prediction model.

[0016] Optionally, after obtaining multiple sets of historical borehole data, it further includes:

[0017] Preprocess each set of historical borehole data; the preprocessing includes: feature selection, feature scaling, formation merging, borehole screening, sequence filling, and category encoding.

[0018] Optionally, the initial lithology type prediction network includes: a bidirectional gated recurrent unit layer, a non-linear activation function, and a Dropout layer.

[0019] Optionally, the initial formation thickness prediction network includes: a Masking layer, three gated recurrent unit modules, and a fully connected layer.

[0020] A prediction system for the lithology type and formation thickness of a borehole includes:

[0021] A coordinate sequence acquisition module for acquiring the borehole coordinate sequence of a virtual borehole; the borehole coordinate sequence includes multiple borehole coordinates; the virtual borehole is an unprobed borehole;

[0022] A first prediction module for inputting the borehole coordinate sequence of the virtual borehole into the lithology type prediction model to obtain the corresponding lithology type sequence of the virtual borehole; the lithology type prediction model is obtained based on the bidirectional gated recurrent unit network, and the lithology type sequence includes the lithology type corresponding to each borehole coordinate;

[0023] A second prediction module, configured to input the drilling coordinate sequence and the corresponding lithology category sequence of the virtual borehole into a formation thickness prediction model to obtain the formation thickness sequence corresponding to the virtual borehole; the formation thickness prediction model is obtained based on a gated recurrent unit network, and the formation thickness sequence includes the formation thickness corresponding to each drilling coordinate.

[0024] An electronic device includes a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the method for predicting the lithology category and formation thickness of the borehole described above.

[0025] Optionally, the memory is a readable storage medium.

[0026] According to the specific embodiments provided by the present invention, the following technical effects are disclosed by the present invention:

[0027] The present invention discloses a method, a system and an electronic device for predicting the lithology category and formation thickness of a borehole. First, a drilling coordinate sequence of a virtual borehole is obtained; the drilling coordinate sequence includes a plurality of drilling coordinates; the virtual borehole is an unprobed borehole; then, the drilling coordinate sequence of the virtual borehole is input into a lithology category prediction model to obtain the lithology category sequence corresponding to the virtual borehole; the lithology category prediction model is obtained based on a bidirectional gated recurrent unit network, and the lithology category sequence includes the lithology category corresponding to each drilling coordinate; finally, the drilling coordinate sequence and the corresponding lithology category sequence of the virtual borehole are input into a formation thickness prediction model to obtain the formation thickness sequence corresponding to the virtual borehole; the formation thickness prediction model is obtained based on a gated recurrent unit network, and the formation thickness sequence includes the formation thickness corresponding to each drilling coordinate, thereby realizing the simulation of the formation structure in the borehole blank area. Description of the Drawings

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0029] Figure 1 It is a schematic flowchart of the method for predicting the lithology category and formation thickness of a borehole provided in Embodiment 1 of the present invention;

[0030] Figure 2 It is a schematic flowchart of the virtual borehole simulation method based on a gated recurrent unit. Detailed Embodiments

[0031] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to 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. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0032] The object of the present invention is to provide a method, system and electronic device for predicting the lithology type and formation thickness of boreholes, aiming to realize the simulation of the formation structure in the borehole blank area.

[0033] To make the above objects, features and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0034] Embodiment 1

[0035] Figure 1 It is a schematic flow chart of the method for predicting the lithology type and formation thickness of boreholes provided in Embodiment 1 of the present invention. As Figure 1 shown, the method for predicting the lithology type and formation thickness of boreholes in this embodiment includes:

[0036] Step 101: Obtain the borehole coordinate sequence of the virtual borehole; the borehole coordinate sequence includes multiple borehole coordinates; the virtual borehole is an unprobed borehole.

[0037] Step 102: Input the borehole coordinate sequence of the virtual borehole into the lithology type prediction model to obtain the lithology type sequence corresponding to the virtual borehole; the lithology type prediction model is obtained based on the bidirectional gated recurrent unit network, and the lithology type sequence includes the lithology types corresponding to each borehole coordinate.

[0038] Step 103: Input the borehole coordinate sequence of the virtual borehole and the corresponding lithology type sequence into the formation thickness prediction model to obtain the formation thickness sequence corresponding to the virtual borehole; the formation thickness prediction model is obtained based on the gated recurrent unit network, and the formation thickness sequence includes the formation thicknesses corresponding to each borehole coordinate.

[0039] As an optional implementation manner, the training process of the lithology type prediction model and the formation thickness prediction model includes:

[0040] (1) Obtain multiple groups of historical borehole data; the historical borehole data includes the coordinate sequence of the actual borehole, the corresponding lithology type sequence and the corresponding formation thickness sequence.

[0041] As an optional implementation manner, after obtaining multiple groups of historical borehole data, it further includes:

[0042] (2) Preprocess the historical drilling data for each group; the preprocessing includes: feature selection, feature scaling, formation merging, borehole screening, sequence filling, and category encoding.

[0043] (3) Based on the bidirectional gated recurrent unit network, construct an initial lithology category prediction network.

[0044] As an optional implementation, the initial lithology category prediction network includes: a bidirectional gated recurrent unit layer, a non-linear activation function, and a Dropout layer.

[0045] (4) Based on the gated recurrent unit network, construct an initial formation thickness prediction network.

[0046] As an optional implementation, the initial formation thickness prediction network includes: a Masking layer, three gated recurrent unit modules, and a fully connected layer.

[0047] (5) Use the coordinate sequence and the corresponding lithology category sequence of each actual borehole to train the initial lithology category prediction network to obtain a lithology category prediction model.

[0048] Specifically, use the coordinate sequence of each actual borehole as the input and the corresponding lithology category sequence as the output to train the initial lithology category prediction network to obtain a lithology category prediction model.

[0049] (6) Use the coordinate sequence, the corresponding lithology category sequence, and the corresponding formation thickness sequence of each actual borehole to train the initial formation thickness prediction network to obtain a formation thickness prediction model.

[0050] Specifically, use the coordinate sequence and the corresponding lithology category sequence of each actual borehole as the input and the corresponding formation thickness sequence as the output to train the initial formation thickness prediction network to obtain a formation thickness prediction model.

[0051] To implement the prediction method in Embodiment 1, a virtual borehole simulation method based on gated recurrent units as shown in Figure 2 is also provided, including the following steps:

[0052] S1. Obtain the original drilling data and preprocess the original drilling data, reconstruct the original drilling data into a coordinate sequence, a lithology category sequence, and a formation thickness sequence, where the coordinate sequence includes the x coordinate and the y coordinate, and divide the processed data into a training set and a test set, and the division ratio is 8:2.

[0053] The preprocessing includes six steps: feature selection, feature scaling, formation merging, borehole screening, sequence filling, and category encoding. Specifically, as follows:

[0054] S1.1: Select the borehole coordinates, lithology categories, and formation thicknesses of the boreholes in the original borehole data to form a data set.

[0055] S1.2: Perform feature scaling on the borehole coordinates in the data set. Use min-max normalization to process the borehole coordinate data so that the borehole coordinate data is on the same order of magnitude as other features.

[0056] S1.3: In the original borehole data, the same borehole usually contains repeated formations. These formations have the same coordinates and lithology. Merge adjacent formations with the same lithology and sum the formation thicknesses simultaneously.

[0057] S1.4: Screen the boreholes so that the number of layers of the boreholes in the data set is approximately the same to reduce the data imbalance problem caused by data filling.

[0058] S1.5: Fill 0 values into all coordinate sequences, lithology category sequences, and formation thickness sequences until the maximum sequence length Max_sec, and Max_sec = Mc + 1, where Mc is the maximum number of formations in the retained boreholes.

[0059] S1.6: The data set contains several different lithologies. It is necessary to convert the format of the lithology categories from strings to numerical types for easy input into the neural network model for training. Considering that there is no size relationship between lithology categories, finally convert the borehole data into one-hot encoding.

[0060] S2. Construct a virtual borehole simulation model. The model consists of a lithology category prediction model and a formation thickness prediction model. Among them, the lithology category prediction model is established based on the Bidirectional Gate Recurrent Unit (BiGRU) algorithm and includes a bidirectional GRU layer, a non-linear activation function, and a Dropout layer. The formation thickness prediction model is established based on the Gate Recurrent Unit (GRU) algorithm and includes a Masking layer, three GRU modules, and a Dense layer. Each GRU module consists of a GRU layer and a dropout layer.

[0061] In the lithology category prediction model, the number of hidden neurons in the bidirectional GRU layer is 42, the non-linear activation function is the ReLU layer, and the parameter of the dropout layer is 0.5. In the formation thickness prediction model, each GRU module contains a GRU layer with 64 hidden neurons and a dropout layer with a parameter of 0.5, and the activation function of the model is tanh.

[0062] S3. Feed the training set data into the virtual drilling simulation model for training to solve the model parameters; use the trained virtual drilling simulation model to predict the test set data and test the performance of the model.

[0063] Train the lithology classification model using the efficient ADAM optimization algorithm and the cross-entropy loss function. The number of samples grabbed each time for training is 128, and after 1000 rounds of training, confirm the optimal parameters of the model; train the rock formation thickness prediction model using the efficient ADAM optimization algorithm and the mean squared error loss function. The number of samples grabbed each time for training is 128, and after 500 rounds of training, confirm the optimal parameters of the model.

[0064] S4. Generate a virtual drill hole, input the virtual drill hole coordinate sequence into the virtual drilling simulation model, and the model predicts the lithology classification sequence and the rock formation thickness sequence corresponding to the virtual coordinates. The combination of the two sequences can obtain the specific structure of the virtual drill hole.

[0065] The virtual drill hole coordinates refer to the coordinates of the undetected points. To meet the requirements of the model, the virtual drill hole coordinates need to be filled to the required length; after inputting the virtual drill hole coordinates into the lithology classification model, the corresponding lithology sequence is obtained. After merging the virtual drill hole coordinate sequence and the lithology classification sequence and inputting them into the rock formation thickness prediction model, the corresponding rock formation thickness can be obtained.

[0066] The specific embodiment uses the drill hole data from the barite ore district in Tianzhu Dahebian area for virtual drill hole simulation, including the following steps:

[0067] Collect the original drill hole data in the test area and perform data preprocessing on the obtained data to generate the data set required for the experiment. The original drill hole data includes features such as drill hole number, x coordinate, y coordinate, drill hole depth, lithology classification, and formation classification. The specific steps of data preprocessing are as follows:

[0068] Select the drill hole x coordinate, y coordinate, lithology classification, and rock formation thickness in the original drill hole data to form the data set.

[0069] Perform feature scaling on the drill hole coordinates in the data set. Use min-max normalization to process the drill hole coordinate data so that the drill hole coordinate data is at the same order of magnitude as other features.

[0070] In the original data, the same drill hole usually contains repeated formations, and these formations have the same coordinates and lithology, as shown in Table 1. Merge the adjacent formations with the same lithology, and sum the rock formation thickness at the same time. The processed results are shown in Table 2.

[0071] Table 1 Original data table

[0072] stratum x coordinate y coordinate lithology thickness ZK03-1 0.306114 0.629323 shale 30.1 ZK03-1 0.306114 0.629323 conglomerate 0.25 ZK03-1 0.306114 0.629323 shale 9.29 ZK03-1 0.306114 0.629323 shale 1.44 ZK03-1 0.306114 0.629323 shale 0.17

[0073] Table 2 Processed Result Table

[0074]

[0075]

[0076] In the dataset used in this specific embodiment, the borehole with the most layers has 41 layers, and the borehole with the fewest layers has only 1 layer. To reduce the impact of data imbalance, it is necessary to screen the boreholes. After multiple experimental comparisons, the borehole data with the number of rock layers between 10 and 30 is finally selected for retention.

[0077] Data filling: Since the GRU algorithm is used in this invention to build the model, to meet the model requirements, the coordinate data, lithology category sequence, and rock layer thickness sequence of all samples need to be filled with 0 values to the maximum sequence length of 31, so as to facilitate data input into the model for training.

[0078] The dataset contains 10 different lithologies. It is necessary to convert the format of the lithology category from string to numerical type. Considering that there is no size relationship between lithology categories, the borehole data is finally converted into one-hot encoding.

[0079] The dataset is divided into a training set and a test set, and the division ratio is 8:2. Note that due to the particularity of the borehole data, the entire borehole data needs to be uniformly divided into the training set or the test set here.

[0080] Use the bidirectional GRU algorithm to build a lithology category prediction model. The model includes a bidirectional GRU layer, an activation layer, and a Dense layer. Among them, the number of hidden neurons in the bidirectional GRU layer is 42, the non-linear activation function is the ReLU layer, and the parameter of the dropout layer is 0.5.

[0081] Use the GRU algorithm to build a rock layer thickness prediction model. The model contains a Masking layer, three GRU modules, and a Dense layer. Among them, each GRU module contains a GRU layer with 42 hidden neurons and a dropout layer with a parameter of 0.5, and the activation function of the model is tanh.

[0082] Model training process: When training the lithology category prediction model, the loss function uses the cross-entropy loss function, and the training optimizer uses the Adam optimization algorithm. The Adam optimization algorithm is an adaptive learning rate optimization algorithm. The number of samples grabbed each time for training is 128, and after 1000 rounds of training, the optimal parameters of the model are confirmed; use the efficient Adam optimization algorithm and the mean squared error loss function to train the rock layer thickness model. The number of samples grabbed each time for training is 128, and after 500 rounds of training, the optimal parameters of the model are confirmed.

[0083] In the model prediction process, the input data during the training of the rock formation thickness model is the real lithology category sequence. However, there is no lithology category sequence during the borehole simulation in the real scenario. Therefore, when making predictions, the results output by the lithology category model are used as the input data for the rock formation thickness model.

[0084] Embodiment 2

[0085] The lithology category and rock formation thickness prediction system for boreholes in this embodiment includes:

[0086] A coordinate sequence acquisition module, configured to acquire the borehole coordinate sequence of a virtual borehole; the borehole coordinate sequence includes a plurality of borehole coordinates; the virtual borehole is an unprobed borehole.

[0087] A first prediction module, configured to input the borehole coordinate sequence of the virtual borehole into the lithology category prediction model to obtain the lithology category sequence corresponding to the virtual borehole; the lithology category prediction model is obtained based on a bidirectional gated recurrent unit network, and the lithology category sequence includes the lithology category corresponding to each borehole coordinate.

[0088] A second prediction module, configured to input the borehole coordinate sequence of the virtual borehole and the corresponding lithology category sequence into the rock formation thickness prediction model to obtain the rock formation thickness sequence corresponding to the virtual borehole; the rock formation thickness prediction model is obtained based on a gated recurrent unit network, and the rock formation thickness sequence includes the rock formation thickness corresponding to each borehole coordinate.

[0089] Embodiment 3

[0090] An electronic device includes a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the method for predicting the lithology category and rock formation thickness of a borehole in Embodiment 1.

[0091] As an optional implementation manner, the memory is a readable storage medium.

[0092] In this specification, each embodiment is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0093] Specific examples are used in this article to elaborate on the principles and implementation manners of the present invention. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A method for predicting the lithology type and rock layer thickness of a borehole, characterized in that: The method comprises: Acquire a drilling coordinate sequence of a virtual drilling hole; the drilling coordinate sequence includes a plurality of drilling coordinates; the virtual drilling hole is an undetected drilling hole; Inputting the borehole coordinate sequence of the virtual borehole into the lithology category prediction model to obtain the lithology category sequence corresponding to the virtual borehole; the lithology category prediction model is obtained based on a bidirectional gated cyclic unit network, and the lithology category sequence includes the lithology category corresponding to each borehole coordinate; Inputting the drilling coordinate sequence and the corresponding lithology category sequence of the virtual drilling hole into the rock formation thickness prediction model to obtain the rock formation thickness sequence corresponding to the virtual drilling hole; the rock formation thickness prediction model is obtained based on the gated cyclic unit network, and the rock formation thickness sequence includes the rock formation thickness corresponding to each drilling coordinate; The training process of the lithology category prediction model and the rock layer thickness prediction model includes: Acquire multiple sets of historical drilling data; the historical drilling data includes a coordinate sequence of the actual drilling, a corresponding lithology category sequence, and a corresponding rock layer thickness sequence; Based on the bidirectional gated cyclic unit network, an initial network for lithology category prediction is constructed; Based on the gated recurrent unit network, an initial network for predicting rock layer thickness is constructed; The initial network for lithology category prediction is trained using the coordinate sequence of each actual borehole and the corresponding lithology category sequence to obtain a lithology category prediction model; The initial network for predicting rock thickness is trained using the coordinate sequence of each actual borehole, the corresponding lithology category sequence and the corresponding rock thickness sequence to obtain a rock thickness prediction model; After obtaining multiple sets of historical drilling data, it also includes: Preprocess each group of historical drilling data; preprocessing includes: feature scaling, stratigraphic merging, sequence filling and category coding; specifically includes: The drilling coordinates, lithology categories and rock layer thickness of the actual drilling holes in the historical drilling data are selected to form a data set; Use min-max normalization to process the drilling coordinates and perform feature scaling on the drilling coordinates in the dataset; The adjacent strata with the same lithology corresponding to the same actual borehole containing repeated strata in the historical drilling data are merged, and the thickness of the strata is summed; Fill all coordinate sequences, lithology sequence, and rock layer thickness sequence with 0 values ​​to the maximum sequence length Max_sec, and Max_sec=Mc+1, where Mc is the maximum number of strata in the borehole; Convert the lithology category from string to numeric format.

2. The method for predicting the lithology type and rock layer thickness of a borehole according to claim 1, characterized in that: The initial network for lithology category prediction includes: a bidirectional gated recurrent unit layer, a nonlinear activation function and a Dropout layer.

3. The method for predicting the lithology type and rock layer thickness of a borehole according to claim 1, characterized in that: The initial network for predicting rock thickness includes: a Masking layer, three gated recurrent unit modules and a fully connected layer.

4. A system for predicting the lithology type and rock thickness of a borehole, characterized in that: The system comprises: A coordinate sequence acquisition module, used to acquire a drilling coordinate sequence of a virtual drilling hole; the drilling coordinate sequence includes a plurality of drilling coordinates; the virtual drilling hole is an undetected drilling hole; A first prediction module is used to input the borehole coordinate sequence of the virtual borehole into a lithology category prediction model to obtain a lithology category sequence corresponding to the virtual borehole; the lithology category prediction model is obtained based on a bidirectional gated cyclic unit network, and the lithology category sequence includes the lithology category corresponding to each borehole coordinate; The second prediction module is used to input the drilling coordinate sequence and the corresponding lithology category sequence of the virtual drilling hole into the rock formation thickness prediction model to obtain the rock formation thickness sequence corresponding to the virtual drilling hole; the rock formation thickness prediction model is obtained based on the gated cyclic unit network, and the rock formation thickness sequence includes the rock formation thickness corresponding to each drilling coordinate; The training process of the lithology category prediction model and the rock layer thickness prediction model includes: Acquire multiple sets of historical drilling data; the historical drilling data includes a coordinate sequence of the actual drilling, a corresponding lithology category sequence, and a corresponding rock layer thickness sequence; Based on the bidirectional gated cyclic unit network, an initial network for lithology category prediction is constructed; Based on the gated recurrent unit network, an initial network for predicting rock layer thickness is constructed; The initial network for lithology category prediction is trained using the coordinate sequence of each actual borehole and the corresponding lithology category sequence to obtain a lithology category prediction model; The initial network for predicting rock thickness is trained using the coordinate sequence of each actual borehole, the corresponding lithology category sequence and the corresponding rock thickness sequence to obtain a rock thickness prediction model; After obtaining multiple sets of historical drilling data, it also includes: Preprocess each group of historical drilling data; preprocessing includes: feature scaling, stratigraphic merging, sequence filling and category coding; specifically includes: The drilling coordinates, lithology categories and rock layer thickness of the actual drilling holes in the historical drilling data are selected to form a data set; Use min-max normalization to process the drilling coordinates and perform feature scaling on the drilling coordinates in the dataset; The adjacent strata with the same lithology corresponding to the same actual borehole containing repeated strata in the historical drilling data are merged, and the thickness of the strata is summed; Fill all coordinate sequences, lithology sequence, and rock layer thickness sequence with 0 values ​​to the maximum sequence length Max_sec, and Max_sec=Mc+1, where Mc is the maximum number of strata in the retained borehole; Convert the lithology category from string to numeric format.

5. An electronic device, characterized in that: It comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the method for predicting the lithology category and rock layer thickness of a borehole as described in any one of claims 1 to 3.

6. An electronic device according to claim 5, characterized in that: The memory is a readable storage medium.

Images