Rapid prediction method and device for oil and gas exploitation stratum, storage medium and equipment

Abstract

The invention discloses an oil and gas exploitation stratum rapid prediction method and device, a storage medium and equipment, and belongs to the technical field of oil and gas field development. Comprises: constructing a data set; the data set comprises logging parameters of multiple wells in the target work area; constructing a deep learning model; the deep learning model comprises an encoder and a decoder, the encoder comprises a set number of encoder layers, and each encoder layer comprises a multi-head self-attention sub-layer and a full-connection feedforward network sub-layer; the decoder comprises a set number of decoder layers, and each decoder layer comprises a mask multi-head self-attention sub-layer, a multi-head self-attention sub-layer and a full-connection feedforward network sub-layer; training the constructed deep learning model through the data set, and iteratively updating parameters to obtain optimal model parameters; and obtaining logging parameters of a to-be-logged well and inputting the logging parameters into the trained deep learning model to obtain a stratum prediction result. The method has the advantages of higher prediction precision, higher prediction speed and higher applicability.

Description

Technical Field

[0001] The present invention relates to the technical field of oil and gas field development, and in particular to a method, device, storage medium and equipment for rapid prediction of oil and gas production strata. Background Art

[0002] In the oil and gas well production process in the field of oil and gas drilling and production, reliable prediction of stratum structure is crucial to resource development and management. At present, the prediction of stratum structure is mainly based on geological principles and mathematical models, and the physical characteristics and geological structure of the stratum are modeled and simulated based on mathematical physics methods to predict the stratum structure. However, these mathematical and physical models have problems such as high complexity, slow calculation speed, low prediction accuracy, and lack of diversity.

[0003] In response to the above problems, researchers began to explore data-driven methods, trying to achieve stratigraphic prediction through machine learning and deep learning techniques. However, traditional data-driven methods face challenges when processing high-dimensional stratigraphic data. They are often only applicable to data in low-dimensional space and have difficulty capturing complex relationships in the data, resulting in unsatisfactory prediction results. Summary of the invention

[0004] In order to solve the defects of the prior art, the present invention provides a method, device, storage medium and equipment for rapid prediction of oil and gas production strata, which has higher prediction accuracy, faster prediction speed and stronger applicability.

[0005] The present invention provides the following technical solutions:

[0006] In a first aspect, the present invention provides a method for rapid prediction of oil and gas production strata, the method comprising:

[0007] Build a dataset;

[0008] Wherein, the data set includes logging parameters of multiple wells in the target work area;

[0009] Build deep learning models;

[0010] The deep learning model includes an encoder and a decoder arranged in sequence, the encoder includes a set number of encoder layers arranged in sequence, each encoder layer includes a multi-head self-attention sublayer and a fully connected feedforward network sublayer; the decoder includes a set number of decoder layers arranged in sequence, each decoder layer includes a masked multi-head self-attention sublayer, a multi-head self-attention sublayer and a fully connected feedforward network sublayer;

[0011] The constructed deep learning model is trained using the data set, and the parameters are iteratively updated to obtain the optimal model parameters;

[0012] The logging parameters of the well to be logged are obtained and input into the trained deep learning model to obtain the formation prediction results.

[0013] Furthermore, constructing the data set includes:

[0014] Acquire multi-source heterogeneous logging data of effective wells in the target work area;

[0015] Performing pre-processing including structural processing, filtering, and outlier detection on the multi-source heterogeneous logging data, and unifying the precision to obtain structured data with uniform precision;

[0016] Through geological layering data and structural surface trends, a geological model is established to achieve the correspondence between the geographical location coordinates in the structured data with uniform accuracy and the geological model coordinates of the geological model;

[0017] The geographical location coordinates and geological model coordinates, as well as drilling time data, single well GR data and cuttings logging data contained in the structured data of uniform accuracy are used as the logging parameters;

[0018] The logging parameters at each depth value of each well are combined into a one-dimensional vector, and the one-dimensional vectors at all depth values ​​of all wells are combined into a two-dimensional vector to construct a data set;

[0019] The dataset is divided into a training set and a test set in a ratio of 8:2.

[0020] Furthermore, the number of encoder layers and decoder layers is 6 each, the multi-head self-attention sublayer and the fully connected feedforward network sublayer contained in each encoder layer are arranged in sequence, and the masked multi-head self-attention sublayer, the multi-head self-attention sublayer and the fully connected feedforward network sublayer contained in each decoder layer are arranged in sequence; each masked multi-head self-attention sublayer, multi-head self-attention sublayer and the fully connected feedforward network sublayer are respectively connected through the residual structure and then normalized and output.

[0021] Furthermore, the two-dimensional vector is used as an input sequence of the encoder, and the encoder processes the input sequence and converts it into an abstract representation; the input of the decoder is the output of the encoder, and the output of the decoder is the probability ranking of the predicted stratum categories;

[0022] The number of heads of the masked multi-head self-attention sublayer and the multi-head self-attention sublayer is 3. The multi-head attention mechanism of the masked multi-head self-attention sublayer and the multi-head self-attention sublayer is used to establish the interactive relationship between the oil and gas field well networks. The mask of the masked multi-head self-attention sublayer is used to mask future moments to ensure that the prediction of the current position will not depend on the future position.

[0023] Furthermore, the deep learning model constructed by the data set is trained and the parameters are iteratively updated to obtain the optimal model parameters, including:

[0024] Training the deep learning model using the training set, and testing the trained deep learning model using the test set;

[0025] Among them, the ADAM optimizer is used during training, and the cosine decay learning rate is adopted. The initial learning rate is 0.00001. The model is trained and the parameters are iteratively updated to obtain the optimal model parameters;

[0026] Save the trained deep learning model as the corresponding model file.

[0027] Furthermore, the step of obtaining the logging parameters of the well to be logged and inputting them into the trained deep learning model to obtain the formation prediction result includes:

[0028] Load the trained deep learning model by loading the saved model file;

[0029] Obtain the logging parameters of the well to be logged and input them into the loaded deep learning model, and output the probability ranking of the predicted formation categories;

[0030] Select the stratum category corresponding to the maximum probability to obtain the stratum prediction result.

[0031] In a second aspect, the present invention provides a rapid prediction device for oil and gas production strata, the device comprising:

[0032] Data preparation module, used to build data sets;

[0033] Wherein, the data set includes logging parameters of multiple wells in the target work area;

[0034] Model building module, used to build deep learning models;

[0035] The deep learning model includes an encoder and a decoder arranged in sequence, the encoder includes a set number of encoder layers arranged in sequence, each encoder layer includes a multi-head self-attention sublayer and a fully connected feedforward network sublayer; the decoder includes a set number of decoder layers arranged in sequence, each decoder layer includes a masked multi-head self-attention sublayer, a multi-head self-attention sublayer and a fully connected feedforward network sublayer;

[0036] A model training module is used to train the constructed deep learning model through the data set and iteratively update the parameters to obtain the optimal model parameters;

[0037] The prediction module is used to obtain the logging parameters of the well to be logged and input the trained deep learning model to obtain the formation prediction results.

[0038] Furthermore, the data preparation module includes:

[0039] A well logging data acquisition unit, used to acquire multi-source heterogeneous well logging data of effective wells in the target work area;

[0040] A data pre-processing unit, used to perform pre-processing including structural processing, filtering, and outlier detection on the multi-source heterogeneous logging data, and unify the precision to obtain structured data with uniform precision;

[0041] A geological model building unit is used to build a geological model through geological layering data and structural surface trends, so as to achieve the correspondence between the geographical location coordinates in the structured data with uniform accuracy and the geological model coordinates of the geological model;

[0042] A well logging selection unit is used to use the geographic location coordinates and geological model coordinates, as well as drilling time data, single well GR data and cuttings logging data contained in the structured data of uniform accuracy as the well logging parameters;

[0043] A data set construction unit is used to combine the logging parameters at each depth value of each well into a one-dimensional vector, and to combine the one-dimensional vectors at all depth values ​​of all wells into a two-dimensional vector, so as to construct a data set;

[0044] The data set division unit is used to divide the data set into a training set and a test set in a ratio of 8:2.

[0045] Furthermore, the number of encoder layers and decoder layers is 6 each, the multi-head self-attention sublayer and the fully connected feedforward network sublayer contained in each encoder layer are arranged in sequence, and the masked multi-head self-attention sublayer, the multi-head self-attention sublayer and the fully connected feedforward network sublayer contained in each decoder layer are arranged in sequence; each masked multi-head self-attention sublayer, multi-head self-attention sublayer and the fully connected feedforward network sublayer are respectively connected through the residual structure and then normalized and output.

[0046] Furthermore, the two-dimensional vector is used as an input sequence of the encoder, and the encoder processes the input sequence and converts it into an abstract representation; the input of the decoder is the output of the encoder, and the output of the decoder is the probability ranking of the predicted stratum categories;

[0047] The number of heads of the masked multi-head self-attention sublayer and the multi-head self-attention sublayer is 3. The multi-head attention mechanism of the masked multi-head self-attention sublayer and the multi-head self-attention sublayer is used to establish the interactive relationship between the oil and gas field well networks. The mask of the masked multi-head self-attention sublayer is used to mask future moments to ensure that the prediction of the current position will not depend on the future position.

[0048] Furthermore, the model training module includes:

[0049] A model training unit, used to train the deep learning model using the training set, and to test the trained deep learning model using the test set;

[0050] Among them, the ADAM optimizer is used during training, and the cosine decay learning rate is adopted. The initial learning rate is 0.00001. The model is trained and the parameters are iteratively updated to obtain the optimal model parameters;

[0051] The model saving unit is used to save the trained deep learning model as a corresponding model file.

[0052] Furthermore, the prediction module includes:

[0053] Model loading unit, used to load the trained deep learning model by loading the saved model file;

[0054] A model prediction unit is used to obtain the logging parameters of the well to be logged and input the loaded deep learning model, and output the probability ranking of the predicted formation categories;

[0055] The prediction result acquisition unit is used to select the formation category corresponding to the maximum probability and obtain the formation prediction result.

[0056] In a third aspect, the present invention provides a computer-readable storage medium for rapid prediction of oil and gas production formations, comprising a memory for storing processor-executable instructions, which, when executed by the processor, implements the steps of the method for rapid prediction of oil and gas production formations described in the first aspect.

[0057] In a fourth aspect, the present invention provides a device for rapid prediction of oil and gas production strata, characterized in that it includes at least one processor and a memory storing computer-executable instructions, and when the processor executes the instructions, it implements the steps of the method for rapid prediction of oil and gas production strata described in the first aspect.

[0058] The present invention has the following beneficial effects:

[0059] The present invention constructs a deep learning model based on the encoder-decoder architecture, and introduces methods such as multi-head self-attention mechanism and masked multi-head attention. By using the self-attention mechanism and encoder-decoder architecture, the formation data is effectively encoded and decoded, thereby realizing the rapid and accurate prediction of formation properties in the actual oil and gas field production process under a large-scale well network, thereby improving the accuracy of the deep learning model. Compared with traditional physical-based models or data-driven models, the present invention has higher prediction accuracy, faster prediction speed and stronger applicability, and can provide more reliable and efficient formation prediction technology for the field of oil and gas production, and provides a new solution for formation prediction in the field of oil and gas production. In addition, the self-attention mechanism used in the model in this technical solution can be calculated in parallel, which improves the training efficiency of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 This is a flow chart of the method for rapid prediction of oil and gas production strata of the present invention;

[0061] Figure 2 Schematic diagram of the established geological model;

[0062] Figure 3 A schematic diagram of a deep learning model used in the present invention;

[0063] Figure 4 Schematic diagram of the attention mechanism;

[0064] Figure 5 It is a schematic diagram of the rapid prediction device for oil and gas production strata of the present invention. DETAILED DESCRIPTION

[0065] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0066] The embodiment of the present invention provides a method for rapid formation prediction in oil and gas production, which is used for rapid formation prediction in the production process of oil and gas fields under large-scale well networks in oil and gas production. Figure 1 As shown, the method includes:

[0067] S1: constructing a data set; wherein the data set includes logging parameters of multiple wells in the target work area;

[0068] In one example, the implementation of this step includes:

[0069] S11: Acquire multi-source heterogeneous logging data of effective wells in the target work area according to actual conditions.

[0070] For example, there are 138 wells in work area A, and valid data are obtained from 134 wells, with a total of 367,028 rows of logging data. The logging data are multi-source heterogeneous logging data.

[0071] S12: Pre-processing including structural processing, filtering, outlier detection, etc. is performed on the multi-source heterogeneous logging data, and the precision is unified (for example, the precision is unified to 0.2m) to obtain structured data with uniform precision.

[0072] S13: Establishing a geological model through geological stratification data and structural surface trends, and achieving correspondence between geographical location coordinates in structured data with uniform accuracy and geological model coordinates of the geological model.

[0073] Among them, the geological model established is as follows Figure 2 As shown, the geographic location coordinate is X d ,Y d , Z d , the geological model coordinate is X m ,Y m , Z m .

[0074] S14: Set the geographic location coordinate X d ,Y d , Z d， and the geological model coordinate X m , Y m , Z m , as well as the drilling time data, single well GR (natural gamma) data and cuttings logging data contained in the structured data with unified accuracy as logging parameters.

[0075] S15: Combining the logging parameters at each depth value of each well into a one-dimensional vector, and combining the one-dimensional vectors at all depth values ​​of all wells into a two-dimensional vector, so as to construct a data set.

[0076] First, each row of data, that is, the logging parameters at each depth of each well, is combined into a one-dimensional vector M i*1 , where i=9 is the logging related parameters input by a single well. The logging related parameters include: drilling time data, single well GR data, cuttings logging data, geographic location coordinates X d ,Y d , Zd , geological model coordinate X m , Y m , Z m Then all the one-dimensional vector data M under all depth values ​​are combined into a two-dimensional vector N i*j , where j is the number of different depth values ​​taken.

[0077] S16: Divide the dataset into training set and test set in a ratio of 8:2.

[0078] S2: Build a deep learning model.

[0079] Among them, Figure 3 As shown in the figure, the deep learning model includes an encoder and a decoder, which are set in sequence. The encoder includes a set number N (for example, N=6) of encoder layers, each of which includes two sublayers, namely, a multi-head self-attention sublayer Multi-Head Attention and a fully connected feedforward network sublayer Feed forward. Each sublayer is connected through a residual structure and then normalized (i.e., Add & Norm) is output. The dimension of each layer is d_model=512 to ensure that the input and output dimensions remain unchanged.

[0080] The decoder includes a set number N (for example, N=6) of decoder layers arranged in sequence, each decoder layer includes three sublayers, namely, the masked multi-head attention sublayer Masked Multi-head attention sublayer, the multi-head self-attention sublayer Multi-Head Attention sublayer and the fully connected feedforward network sublayer Feed forward, and each sublayer is connected through a residual structure and then normalized and output.

[0081] Two-dimensional vector N i*j As the input sequence of the encoder, the encoder processes the input sequence and converts it into an abstract representation. The decoder is used to generate the stratum category prediction, and its input is the output of the encoder, and the output is the probability size ranking of the predicted stratum category.

[0082] Optionally, the number of heads of the masked multi-head self-attention sublayer and the multi-head self-attention sublayer is 3. The multi-head attention mechanism of the masked multi-head self-attention sublayer and the multi-head self-attention sublayer is used to establish the interaction relationship between the well networks of the oil and gas fields. The masked multi-head self-attention sublayer is used to perform multi-head attention on the output of the encoder stack, and the mask is used for masking at future moments to ensure that the prediction of the current position does not depend on the future position.

[0083] The attention mechanism maps the query vector Q and a set of key-value pair vectors KV to the output, which is the weighted sum of the output values, and the weight is calculated by the similarity between Q and K.

[0084] The attention mechanism includes scaled dot-product attention mechanism Scaled Dot-Product Attention and multi-head attention mechanism Multi-Head Attention, such as Figure 4 As shown. The input of the scaled dot product attention mechanism consists of the vector query (dk), key (dk) and value (dv). The Q and K vectors are processed by dot product to calculate the similarity. The scaling factor sqrt (dk) (used to find the square root of dk) is used to avoid the situation where the variance of the inner product of Q and K is too large to make it difficult to learn. The softmax function is used to obtain the weight to obtain the weight of the value. Mask processing prevents the decoder from obtaining future information that affects prediction during training. The formula is as follows.

[0085]

[0086] The multi-head attention mechanism linearly projects Q, K, and V into the scaled dot product attention mechanism, performs h attention function operations after projection h times, generates dv-dimensional output values ​​through parallel calculation, and concatenates each output value and performs another projection to obtain the final value. Through the multi-head attention mechanism, different subspace information from different positions can be jointly paid attention to.

[0087] The deep learning model structure of the present invention has a self-attention mechanism and an encoder-decoder architecture, which can effectively process high-dimensional data and automatically learn complex nonlinear relationships between data during training. The attention mechanism abandons the traditional cyclic or convolutional neural network structure, solving the problem that the traditional neural network structure is difficult to handle long-term dependencies and has low computational efficiency. Using the self-attention mechanism and encoder-decoder architecture, high-dimensional stratigraphic data can be effectively processed, thereby improving prediction accuracy and diversity.

[0088] S3: Train the constructed deep learning model through the data set and iteratively update the parameters to obtain the optimal model parameters.

[0089] The specific process of training includes:

[0090] S31: Train the deep learning model through the training set, and test the trained deep learning model through the test set.

[0091] Among them, the ADAM optimizer is used during training, and the cosine decay learning rate is adopted. The initial learning rate is 0.00001. The model is trained and the parameters are iteratively updated to obtain the optimal model parameters.

[0092] S32: Save the trained deep learning model as a corresponding model file.

[0093] S4: Obtain the logging parameters of the well to be logged and input them into the trained deep learning model to obtain the formation prediction results.

[0094] As a specific implementation method, this step includes:

[0095] S41: Load the trained deep learning model by loading the saved model file.

[0096] S42: Obtain the logging parameters of the well to be logged, normalize them, and input them into the loaded deep learning model, and output the probability ranking of the predicted formation categories.

[0097] S43: Select the formation category corresponding to the maximum probability to obtain the formation prediction result, and finally obtain the formation prediction related information through inverse normalization.

[0098] The present invention constructs a deep learning model based on the encoder-decoder architecture, and introduces methods such as multi-head self-attention mechanism and masked multi-head attention. By using the self-attention mechanism and encoder-decoder architecture, the formation data is effectively encoded and decoded, thereby realizing the rapid and accurate prediction of formation properties in the actual oil and gas field production process under a large-scale well network, thereby improving the accuracy of the deep learning model. Compared with traditional physical-based models or data-driven models, the present invention has higher prediction accuracy, faster prediction speed and stronger applicability, and can provide more reliable and efficient formation prediction technology for the field of oil and gas production, and provides a new solution for formation prediction in the field of oil and gas production. In addition, the self-attention mechanism used in the model in this technical solution can be calculated in parallel, which improves the training efficiency of the model.

[0099] The following table is based on the field logging data results, and the accuracy verification results of the model of the present invention for blind well testing.

[0100] Table 1 Blind well test results

[0101]

[0102] Based on the results of field logging data, the model of the present invention, the traditional physical model prediction, and the traditional Subspace k-NN machine learning method were compared. By comparison, it was found that the average relative error of the model proposed in the present invention for formation prediction was 7.67% (see the table above), while the average relative error of the traditional physical model was 16.1%, and the traditional Subspacek-NN error was 15.8%. In terms of speed, the model proposed in the present invention takes 0.2s to complete a prediction, the physical model takes 5.2s, and the LSTM model takes 0.3s. Therefore, taking all factors into consideration, the present technical solution has greater advantages in speed and accuracy. The formation prediction method proposed in the present invention is in a leading position in terms of theoretical methods, solving key technical problems and main technical indicators.

[0103] The embodiment of the present invention also provides a rapid prediction device for oil and gas production strata, such as Figure 5 As shown, the device comprises:

[0104] Data preparation module 1 is used to construct the dataset.

[0105] The data set includes logging parameters of multiple wells in the target work area.

[0106] Model building module 2 is used to build a deep learning model.

[0107] Among them, the deep learning model includes an encoder and a decoder arranged in sequence, the encoder includes a set number of encoder layers arranged in sequence, each encoder layer includes a multi-head self-attention sublayer and a fully connected feedforward network sublayer; the decoder includes a set number of decoder layers arranged in sequence, each decoder layer includes a masked multi-head self-attention sublayer, a multi-head self-attention sublayer and a fully connected feedforward network sublayer.

[0108] The model training module 3 is used to train the constructed deep learning model through the data set and iteratively update the parameters to obtain the optimal model parameters.

[0109] The prediction module 4 is used to obtain the logging parameters of the well to be logged and input the trained deep learning model to obtain the formation prediction results.

[0110] As an example, the aforementioned data preparation module includes:

[0111] The logging data acquisition unit is used to acquire multi-source heterogeneous logging data of effective wells in the target work area.

[0112] The data pre-processing unit is used to perform pre-processing including structural processing, filtering, and outlier detection on multi-source heterogeneous logging data, and unify the accuracy to obtain structured data with uniform accuracy.

[0113] The geological model building unit is used to build a geological model through geological stratification data and structural surface trends, so as to achieve the correspondence between the geographical location coordinates in the structured data with uniform accuracy and the geological model coordinates of the geological model.

[0114] The well logging selection unit is used to use the geographic location coordinates and geological model coordinates, as well as the drilling time data, single well GR data and cuttings logging data contained in the structured data of uniform accuracy as well logging parameters.

[0115] The data set construction unit is used to combine the logging parameters at each depth value of each well into a one-dimensional vector, and to combine the one-dimensional vectors at all depth values ​​of all wells into a two-dimensional vector to construct a data set.

[0116] The data set partitioning unit is used to partition the data set into a training set and a test set in a ratio of 8:2.

[0117] In the deep learning model of the present invention, the number of encoder layers and decoder layers is 6, the multi-head self-attention sublayer and the fully connected feedforward network sublayer contained in each encoder layer are arranged in sequence, and the masked multi-head self-attention sublayer, the multi-head self-attention sublayer and the fully connected feedforward network sublayer contained in each decoder layer are arranged in sequence; each masked multi-head self-attention sublayer, the multi-head self-attention sublayer and the fully connected feedforward network sublayer are respectively connected through a residual structure and then normalized and output.

[0118] The aforementioned two-dimensional vector is used as the input sequence of the encoder, and the encoder processes the input sequence and converts it into an abstract representation; the input of the decoder is the output of the encoder, and the output of the decoder is the probability ranking of the predicted stratum categories.

[0119] The number of heads of the masked multi-head self-attention sublayer and the multi-head self-attention sublayer is 3. The multi-head attention mechanism of the masked multi-head self-attention sublayer and the multi-head self-attention sublayer is used to establish the interactive relationship between the well networks of oil and gas fields. The mask of the masked multi-head self-attention sublayer is used to mask the future moments to ensure that the prediction of the current position does not depend on the future position.

[0120] The model training module of the present invention specifically includes:

[0121] The model training unit is used to train the deep learning model through the training set and test the trained deep learning model through the test set.

[0122] Among them, the ADAM optimizer is used during training, and the cosine decay learning rate is adopted. The initial learning rate is 0.00001. The model is trained and the parameters are iteratively updated to obtain the optimal model parameters.

[0123] The model saving unit is used to save the trained deep learning model as a corresponding model file.

[0124] As a specific implementation, the aforementioned prediction module includes:

[0125] The model loading unit is used to load the trained deep learning model by loading the saved model file.

[0126] The model prediction unit is used to obtain the logging parameters of the well to be logged and input the loaded deep learning model, and output the probability size ranking of the predicted formation category.

[0127] The prediction result acquisition unit is used to select the formation category corresponding to the maximum probability and obtain the formation prediction result.

[0128] The present invention constructs a deep learning model based on the encoder-decoder architecture, and introduces methods such as multi-head self-attention mechanism and masked multi-head attention. By using the self-attention mechanism and encoder-decoder architecture, the formation data is effectively encoded and decoded, thereby realizing the rapid and accurate prediction of formation properties in the actual oil and gas field production process under a large-scale well network, thereby improving the accuracy of the deep learning model. Compared with traditional physical-based models or data-driven models, the present invention has higher prediction accuracy, faster prediction speed and stronger applicability, and can provide more reliable and efficient formation prediction technology for the field of oil and gas production, and provides a new solution for formation prediction in the field of oil and gas production. In addition, the self-attention mechanism used in the model in this technical solution can be calculated in parallel, which improves the training efficiency of the model.

[0129] The device provided in the above embodiment, its implementation principle and the technical effect produced correspond to the embodiment of the above method one by one. For the sake of brief description, the part not mentioned in the embodiment of the device can refer to the corresponding content in the embodiment of the above method. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the modules and units described in the device can refer to the corresponding process in the embodiment of the above method, and will not be repeated here.

[0130] The method for rapid prediction of oil and gas production strata described in the above embodiment provided by the present invention can implement business logic through a computer program and record it on a storage medium, and the storage medium can be read and executed by a computer to achieve the effect of the scheme described in the method embodiment of this specification. Therefore, the embodiment of the present invention also provides a computer-readable storage medium for rapid prediction of oil and gas production strata, including a memory for storing processor-executable instructions, and when the instructions are executed by the processor, the steps of the method for rapid prediction of oil and gas production strata including the above embodiment are implemented.

[0131] The storage medium may include a physical device for storing information, which is usually a medium that digitizes the information and then stores it in an electrical, magnetic or optical manner. The storage medium may include: a device that stores information in an electrical energy manner, such as various memories, such as RAM, ROM, etc.; a device that stores information in a magnetic energy manner, such as a hard disk, a floppy disk, a magnetic tape, a magnetic core memory, a magnetic bubble memory, a USB flash drive; a device that stores information in an optical manner, such as a CD or a DVD. Of course, there are other readable storage media, such as quantum memory, graphene memory, etc.

[0132] The storage medium described above may also include other implementation methods according to the description of the method embodiment. The implementation principle and technical effects produced by this embodiment are the same as those of the aforementioned method embodiment. For details, please refer to the description of the relevant method embodiment, and no further description will be given here.

[0133] The embodiment of the present invention also provides a device for rapid prediction of oil and gas production strata, which may be a separate computer, or may include an actual operating device using one or more of the methods or one or more of the embodiments of this specification. The device for rapid prediction of oil and gas production strata may include at least one processor and a memory storing computer executable instructions, and when the processor executes the instructions, any one or more steps of the method for rapid prediction of oil and gas production strata are implemented.

[0134] The device described above may also include other implementation methods according to the description of the method embodiment. The implementation principle and technical effects produced by this embodiment are the same as those of the aforementioned method embodiment. For details, please refer to the description of the relevant method embodiment, and no further description will be given here.

[0135] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. They should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for rapid prediction of oil and gas production strata, characterized in that: The method comprises: Build a dataset; Wherein, the data set includes logging parameters of multiple wells in the target work area; Build deep learning models; The deep learning model includes an encoder and a decoder arranged in sequence, the encoder includes a set number of encoder layers arranged in sequence, each encoder layer includes a multi-head self-attention sublayer and a fully connected feedforward network sublayer; the decoder includes a set number of decoder layers arranged in sequence, each decoder layer includes a masked multi-head self-attention sublayer, a multi-head self-attention sublayer and a fully connected feedforward network sublayer; The constructed deep learning model is trained using the data set, and the parameters are iteratively updated to obtain the optimal model parameters; The logging parameters of the well to be logged are obtained and input into the trained deep learning model to obtain the formation prediction results.

2. The method for rapid prediction of oil and gas production strata according to claim 1, characterized in that: The constructing of the data set comprises: Acquire multi-source heterogeneous logging data of effective wells in the target work area; Performing pre-processing including structural processing, filtering, and outlier detection on the multi-source heterogeneous logging data, and unifying the precision to obtain structured data with uniform precision; Through geological layering data and structural surface trends, a geological model is established to achieve the correspondence between the geographical location coordinates in the structured data with uniform accuracy and the geological model coordinates of the geological model; The geographical location coordinates and geological model coordinates, as well as drilling time data, single well GR data and cuttings logging data contained in the structured data of uniform accuracy are used as the logging parameters; The logging parameters at each depth value of each well are combined into a one-dimensional vector, and the one-dimensional vectors at all depth values ​​of all wells are combined into a two-dimensional vector to construct a data set; The dataset is divided into a training set and a test set in a ratio of 8:

2.

3. The rapid prediction method for oil and gas production strata according to claim 2 is characterized in that: The number of encoder layers and decoder layers is 6 each. The multi-head self-attention sublayer and the fully connected feedforward network sublayer contained in each encoder layer are arranged in sequence, and the masked multi-head self-attention sublayer, the multi-head self-attention sublayer and the fully connected feedforward network sublayer contained in each decoder layer are arranged in sequence; each masked multi-head self-attention sublayer, the multi-head self-attention sublayer and the fully connected feedforward network sublayer are respectively connected through a residual structure and then normalized and output.

4. The method for rapid prediction of oil and gas production strata according to claim 3, characterized in that: The two-dimensional vector is used as the input sequence of the encoder, and the encoder processes the input sequence and converts it into an abstract representation; the input of the decoder is the output of the encoder, and the output of the decoder is the probability ranking of the predicted stratum categories; The number of heads of the masked multi-head self-attention sublayer and the multi-head self-attention sublayer is 3. The multi-head attention mechanism of the masked multi-head self-attention sublayer and the multi-head self-attention sublayer is used to establish the interactive relationship between the oil and gas field well networks. The mask of the masked multi-head self-attention sublayer is used to mask future moments to ensure that the prediction of the current position will not depend on the future position.

5. The method for rapid prediction of oil and gas production strata according to claim 4, characterized in that: The deep learning model constructed by the data set is trained and the parameters are iteratively updated to obtain the optimal model parameters, including: Training the deep learning model using the training set, and testing the trained deep learning model using the test set; Among them, the ADAM optimizer is used during training, and the cosine decay learning rate is adopted. The initial learning rate is 0.00001. The model is trained and the parameters are iteratively updated to obtain the optimal model parameters; Save the trained deep learning model as the corresponding model file.

6. The method for rapid prediction of oil and gas production strata according to claim 5, characterized in that: The step of obtaining the logging parameters of the well to be logged and inputting the trained deep learning model to obtain the formation prediction result includes: Load the trained deep learning model by loading the saved model file; Obtain the logging parameters of the well to be logged and input them into the loaded deep learning model, and output the probability ranking of the predicted formation categories; Select the stratum category corresponding to the maximum probability to obtain the stratum prediction result.

7. A rapid prediction device for oil and gas production strata, characterized in that: The device comprises: Data preparation module, used to build data sets; Wherein, the data set includes logging parameters of multiple wells in the target work area; Model building module, used to build deep learning models; The deep learning model includes an encoder and a decoder arranged in sequence, the encoder includes a set number of encoder layers arranged in sequence, each encoder layer includes a multi-head self-attention sublayer and a fully connected feedforward network sublayer; the decoder includes a set number of decoder layers arranged in sequence, each decoder layer includes a masked multi-head self-attention sublayer, a multi-head self-attention sublayer and a fully connected feedforward network sublayer; A model training module is used to train the constructed deep learning model through the data set and iteratively update the parameters to obtain the optimal model parameters; The prediction module is used to obtain the logging parameters of the well to be logged and input the trained deep learning model to obtain the formation prediction results.

8. A computer-readable storage medium for rapid prediction of oil and gas production strata, characterized in that: It comprises a memory for storing processor executable instructions, and when the instructions are executed by the processor, the steps of the method for rapid prediction of oil and gas production formations according to any one of claims 1-6 are implemented.

9. A device for rapid prediction of oil and gas production strata, characterized in that: The method comprises at least one processor and a memory storing computer executable instructions, wherein when the processor executes the instructions, the steps of the method for rapid prediction of oil and gas production formations as claimed in any one of claims 1 to 6 are implemented.

Images