BiLSTM-based while-drilling electromagnetic wave logging data rapid inversion method and system

Abstract

The invention discloses a BiLSTM-based while-drilling electromagnetic wave logging data rapid inversion method and system, and the method achieves the efficient processing of logging data through constructing a deep network structure composed of four BiLSTM layers and a full-connection layer. According to the method, ideal magnetic dipole response is adopted as training data, and the problem that local minimum is likely to occur in traditional data interpretation is solved. And the recurrent dropout technology is utilized to reduce overfitting, so that the generalization ability of the model is improved. The technology not only can accurately classify the stratum model, but also can predict various stratum parameters, so that the accuracy and processing speed of data interpretation are remarkably improved, and powerful technical support is provided for geological exploration and stratum evaluation.

Description

Technical Field

[0001] The present invention relates to the technical fields of geophysical exploration and information technology, and particularly relates to a method and system for rapid inversion of electromagnetic logging while drilling data based on BiLSTM. Background Art

[0002] In modern logging technology, electromagnetic logging while drilling is an important geological exploration means. Electromagnetic logging while drilling data refers to the underground rock formation information and exploration data of formation boundaries obtained in real time by an electromagnetic wave instrument during the drilling process. These data are crucial for accurately evaluating formation properties, oil and gas reserves, and geological structures.

[0003] Its principle mainly utilizes the propagation and reflection characteristics of electromagnetic waves in the formation to detect formation information. The application scenarios mainly include the exploration and development processes of mineral resources such as oil and natural gas, providing important bases for geological steering and formation evaluation.

[0004] BiLSTM, that is, bidirectional long short-term memory unit, is a recurrent neural network structure in a deep learning model. It can handle long-term dependencies in sequence data, capture the context information in the sequence simultaneously through LSTM units in both forward and backward directions, and thus has been widely used in fields such as natural language processing, speech recognition, and time series analysis. In the inversion of logging data, BiLSTM can learn complex patterns in the logging data and achieve efficient interpretation of the data.

[0005] Currently, when interpreting data by an electromagnetic wave forward-looking long-range detection instrument while drilling, traditional deterministic methods or statistical methods are often used. Deterministic methods such as the Gauss-Newton method, etc., approximate the true formation model through iterative optimization, but are prone to falling into local minima, resulting in inaccurate interpretation results. Although statistical methods can avoid the local minimum problem to a certain extent, they have a large amount of calculation and are difficult to meet the requirements of real-time geological steering.

[0006] For example, in a complex geological environment, the local minimum problem is particularly prominent. For example, during the exploration process of an oilfield, when using a deterministic method for data interpretation, due to the influence of the multi-layer structure and anisotropy of the formation, the interpretation results repeatedly fell into local minima, resulting in incorrect judgment of the drilling direction and causing waste of resources and time.

[0007] In order to avoid the problem of falling into local minima and achieve timely guidance for geological steering, a strategy of combining multiple methods is often adopted in the industry. For example, introducing prior geological information to constrain the inversion process, or using more complex statistical models to improve the global search ability. However, these methods often increase the computational complexity and even require high-performance computers and long-time operations, which are impractical in on-site operations.

[0008] For example, in a certain geological exploration, in order to obtain a more accurate formation model, a complex statistical inversion method was adopted. Although a relatively satisfactory result was finally obtained, the inversion process took several days, far from meeting the requirements of real-time drilling guidance.

[0009] Therefore, providing a real-time electromagnetic logging while drilling data inversion method that can avoid falling into local minima and meet real-time requirements has become the technical problem to be solved by the present invention. Summary of the Invention

[0010] The technical problem solved by the present invention is to provide a fast inversion method and system for electromagnetic logging while drilling data based on BiLSTM to solve the problems of easy local minima and large computational amount in data interpretation and inability to meet real-time geological steering requirements in the above-mentioned background technology.

[0011] To solve the above technical problems, the technical solutions adopted by the present invention are as follows:

[0012] A fast inversion method for electromagnetic logging while drilling data based on BiLSTM includes the following steps:

[0013] Step 1: Obtain the electromagnetic wave signals received by the antenna and perform preprocessing to obtain logging data;

[0014] Step 2: Take the preprocessed logging data as input and input it into the BiLSTM layer and the neural network. The BiLSTM layer includes a rectified linear unit ReLU as an activation function, and the recurrent dropout technique is used in the BiLSTM layer to reduce overfitting;

[0015] Step 3: The neural network performs feature extraction, convolution, and fusion on the input logging data, and outputs a classification result and a prediction result. The classification result is any one of the preset models, and the prediction result is multiple formation parameters corresponding to the model category;

[0016] Step 4: Perform geological steering and formation evaluation according to the classification result and the prediction result;

[0017] As a further solution of the present invention, in step 1, before obtaining the electromagnetic wave signal received by the antenna, first, the propagation matrix method is used as the forward modeling algorithm to calculate the response of the magnetic dipole in the one-dimensional layered medium. During this process, the influence of the borehole environment is ignored, and the response of the ideal magnetic dipole in the layered medium is used to simulate the instrument response to generate the training dataset and the validation dataset. The training dataset will be used for the training of the subsequent neural network weights, while the validation dataset is used to verify the training results during the training process. After the preparation of the training dataset and the validation dataset, the electromagnetic wave signal received by the antenna is obtained and preprocessed to obtain the logging data for the input of the neural network.

[0018] As a further solution of the present invention, the neural network in step 3 is a recurrent neural network, and the input information of the neurons in its hidden layer not only comes from the output of the neurons in the input layer, but also from the output of the neurons in the hidden layer at the current moment; and each layer of this recurrent neural network shares the parameters U, V, and W to reduce the number of parameters and improve the network training speed.

[0019] As a further solution of the present invention, the output value st of the neurons in the hidden layer of the recurrent neural network at time t is calculated according to the input information It at the current moment and the output value st-1 of the neurons in the hidden layer at the previous moment through the activation function and the shared parameters U, V, and W; where the input information It includes the current logging data input and the possible feedback of the output layer neurons at the previous moment; the output ot of the output layer neurons at time t is calculated according to the output value st of the neurons in the hidden layer at time t through the connection weights and the activation function.

[0020] As a further solution of the present invention, the electromagnetic wave signal received by the antenna in step 1 specifically refers to the electromagnetic wave response signal from the formation received by the antenna on the logging tool during the logging-while-drilling process. These signals include but are not limited to formation resistivity, boundary position, anisotropy coefficient, and formation dip angle. These signals are preprocessed and converted into logging data that can be used for the input of the neural network.

[0021] As a further solution of the present invention, in step 2, the neural network uses the RMSprop optimizer to train the logging data whose loss function is defined as the mean square error.

[0022] As a further solution of the present invention, the output classification result in step 3 is any one of four preset formation models, and these four formation models respectively correspond to different formation characteristics; the prediction result is the parameters of various specific formation characteristics corresponding to the classified formation model. Among them, the parameters of the formation characteristics include, but are not limited to, formation thickness, resistivity, porosity, permeability, and water saturation. These parameters are obtained through the feature extraction, convolution, and fusion processes of the neural network to provide detailed data support for geosteering and formation evaluation.

[0023] An inversion system for electromagnetic wave forward-looking logging while drilling data based on a recurrent neural network, comprising:

[0024] A data preprocessing module, configured to obtain the electromagnetic wave signals received by three antennas and perform preprocessing to obtain logging data;

[0025] A neural network module, configured to receive the preprocessed logging data as input, perform feature extraction, convolution, and fusion through a neural network composed of four bidirectional long short-term memory (BiLSTM) layers and a fully connected layer, and output a classification result and a prediction result;

[0026] A result output module, configured to perform geosteering and formation evaluation according to the output of the neural network.

[0027] As a further solution of the present invention, the neural network module uses the RMSprop optimizer to train the logging data for which the loss function is defined as the mean square error.

[0028] As a further solution of the present invention, the formation parameters include the horizontal resistivity of each layer, the distance to the boundary, the anisotropy coefficient of the formation where the instrument is located, and the formation dip angle.

[0029] Compared with the prior art, the beneficial effects of the present invention are as follows: By constructing a deep network structure composed of four BiLSTM layers and a fully connected layer, the efficient processing of logging data is realized. The present invention uses the ideal magnetic dipole response as the training data, solving the problem of being easily trapped in local minima in traditional data interpretation. The recurrent dropout technology is used to reduce overfitting and improve the generalization ability of the model. This technology can not only accurately classify the formation model but also predict various formation parameters, significantly improving the accuracy and processing speed of data interpretation and providing strong technical support for geological exploration and formation evaluation.

[0030] The additional aspects and advantages of the present invention will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present invention. Description of the Drawings

[0031] 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 for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0032] Figure 1 It is a schematic diagram of the principle of a recurrent neural network.

[0033] Figure 2 It is a schematic diagram of the principle of LSTM.

[0034] Figure 3 It is a neural network structure for inversion of azimuthal electromagnetic logging while drilling data.

[0035] Figure 4 It is a schematic diagram of instrument detection.

[0036] Figure 5 It is a formation model setting diagram.

[0037] Figure 6 It is a deep learning inversion flow chart. Detailed implementation manners

[0038] The following will clearly and completely describe the technical solutions 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 of them. 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.

[0039] Please refer to Figure 1 ——6. In the embodiment of the present invention, a fast inversion method for electromagnetic logging while drilling data based on BiLSTM is characterized by including the following steps:

[0040] Step 1, obtain electromagnetic wave signals received by, for example, three antennas, and perform preprocessing to obtain logging data;

[0041] Among them, in step 1, the propagation matrix method is also adopted as the forward modeling algorithm for calculating the response of a magnetic dipole in a one-dimensional layered medium. Ignoring the influence of the borehole environment, the response of an ideal magnetic dipole in the layered medium is used to simulate the tool response, thereby generating a training data set and a validation data set. The training data set is used to train the weights of the neural network, and the validation data set is used to verify the training results during the training process. Specifically, in step 1, before obtaining the electromagnetic wave signal received by the antenna, first, the propagation matrix method is adopted as the forward modeling algorithm to calculate the response of the magnetic dipole in the one-dimensional layered medium. To simplify the model and improve the calculation efficiency, the influence of the borehole environment is ignored in this process, and the response of the ideal magnetic dipole in the layered medium is used to simulate the tool response. In this way, a training data set and a validation data set are generated. The training data set will be used for the subsequent training of the neural network weights, while the validation data set is used to verify the training results during the training process. After completing the preparation of the data set, then obtain the electromagnetic wave signal received by the antenna and perform preprocessing to obtain the well logging data for the input of the neural network.

[0042] Step 2, use the preprocessed well logging data as the input and input it into a neural network composed of four bidirectional long short-term memory (BiLSTM) layers and a fully connected layer. Each BiLSTM layer contains a rectified linear unit (ReLU) as the activation function, and the recurrent dropout technique is used in the BiLSTM layer to reduce the overfitting phenomenon.

[0043] Step 3, the neural network performs feature extraction, convolution, and fusion on the input well logging data, and outputs a classification result and a prediction result. The classification result is any one of the four set models, and the prediction result is multiple formation parameters corresponding to the model category.

[0044] Step 4, perform geological steering and formation evaluation based on the classification result and the prediction result.

[0045] Among them, the neural network in step 3 adopts a recurrent neural network. The input information of the hidden layer neurons not only comes from the output of the input layer neurons but also from the output of the hidden layer neurons at the current moment. And each layer of this recurrent neural network shares the parameters U, V, and W to reduce the number of parameters and improve the network training speed.

[0046] Among them, "time point t" refers to a specific time point when the recurrent neural network processes sequential data. In the recurrent neural network, data is processed in time series, and each time point corresponds to an input data and a state update. Therefore, "time point t" is a specific time point in this sequence.

[0047] "The output value \(s_t\)" represents the output state of the neurons in the hidden layer of the recurrent neural network at time \(t\). This state is calculated based on the current input and the state at the previous time step, and it contains all the relevant information in the sequence up to the current time point.

[0048] "The output value \(s_{t - 1}\)" represents the output state of the neurons in the hidden layer of the recurrent neural network at time \(t - 1\), i.e., the previous time step. This state is passed to the next time step (\(t\)) as part of the calculation of the new state.

[0049] "U" is the weight matrix connecting the neurons in the input layer to the neurons in the hidden layer. In a recurrent neural network, this matrix is used to transform and transmit the data from the input layer to the hidden layer, and it is an important parameter for the neural network to learn the features of the input data.

[0050] "V" is the weight matrix between the neurons in the hidden layer and the output layer. It is responsible for transforming the state of the hidden layer into the output of the output layer. During the training process, this matrix is adjusted to minimize the prediction error of the network.

[0051] "W" is the weight matrix connecting the neurons in the hidden layer at the previous time step to the neurons in the hidden layer at the current time step. In a recurrent neural network, this matrix is used to store and transmit temporal information, enabling the network to remember the previous state and influence the current and future outputs. This is also the key to the recurrent neural network's ability to process sequential data.

[0052] The output value \(s_t\) of the neurons in the hidden layer of the recurrent neural network at time \(t\) is calculated based on the input information \(I_t\) at the current time and the output value \(s_{t - 1}\) of the neurons in the hidden layer at the previous time step, through an activation function and shared parameters U, V, and W; where the input information \(I_t\) includes the logging data input at the current time and the possible feedback from the neurons in the output layer at the previous time step; the output \(o_t\) of the neurons in the output layer at time \(t\) is calculated based on the output value \(s_t\) of the neurons in the hidden layer at time \(t\) through the weight matrix and the activation function.

[0053] The electromagnetic wave signals received by the antenna in step 1 specifically refer to the electromagnetic wave response signals received by the antenna on the logging tool from the formation during the logging - while - drilling process. These signals include, but are not limited to, formation resistivity, boundary position, anisotropy coefficient, and formation dip angle. These signals are pre - processed and transformed into logging data that can be used as input for the neural network.

[0054] In step 2, the neural network uses the RMSprop optimizer to train the logging data for which the loss function is defined as the mean squared error.

[0055] The output classification result in Step 3 is any one of four preset formation models, and these four formation models respectively correspond to different formation characteristics; the prediction result is the parameters of various specific formation characteristics corresponding to the classified formation model.

[0056] The parameters of the formation characteristics include but are not limited to formation thickness, resistivity, porosity, permeability, and water saturation. These parameters are obtained through the feature extraction, convolution, and fusion processes of the neural network to provide detailed data support for geosteering and formation evaluation.

[0057] An inversion system for electromagnetic wave forward-looking logging data while drilling based on a recurrent neural network, comprising:

[0058] A data preprocessing module, configured to obtain the electromagnetic wave signals received by three antennas and perform preprocessing to obtain logging data;

[0059] A neural network module, configured to receive the preprocessed logging data as input, perform feature extraction, convolution, and fusion through a neural network composed of four bidirectional long short-term memory (BiLSTM) layers and a fully connected layer, and output a classification result and a prediction result;

[0060] A result output module, configured to perform geosteering and formation evaluation according to the output of the neural network.

[0061] The neural network module uses the RMSprop optimizer to train the logging data with the loss function defined as the mean squared error. The formation parameters include the horizontal resistivity of each layer, the distance to the boundary, the anisotropy coefficient of the formation where the instrument is located, and the formation dip angle.

[0062] The present invention realizes the efficient processing of logging data by constructing a deep network structure composed of four BiLSTM layers and a fully connected layer. The present invention uses the ideal magnetic dipole response as training data, solving the problem of being prone to falling into local minima in traditional data interpretation. The recurrent dropout technology is used to reduce overfitting and improve the generalization ability of the model. This technology can not only accurately classify the formation model, but also predict various formation parameters, significantly improving the accuracy and processing speed of data interpretation, and providing strong technical support for geological exploration and formation evaluation.

[0063] Among them, the ideal magnetic dipole model refers to, for example, an object or system with a magnetic moment. Under the action of an external magnetic field, this magnetic moment will produce a specific response. In this model, the specific shape and size of the magnetic dipole are usually ignored, and only the magnitude and direction of its magnetic moment are concerned. Doing so can simplify the problem and make the mathematical model easier to handle.

[0064] The generation of response data is achieved by calculating the responses of an ideal magnetic dipole under different external magnetic field conditions, which can generate a set of response data. These data describe the behavior of the magnetic dipole in different magnetic field environments. The response data may include physical quantities such as the magnetic field strength and direction generated by the magnetic dipole, which are closely related to the characteristics of the external magnetic field.

[0065] Moreover, the application of the ideal magnetic dipole model as training data, such as in electromagnetic logging while drilling or magnetic target detection and positioning, requires fast and accurate processing of data related to magnetic dipoles. To achieve this, machine learning-based methods, such as neural networks, can be used. When training these machine learning models, the "ideal magnetic dipole response" is used as training data. Through these data, the model can learn the response laws of the magnetic dipole under different magnetic field conditions. After training, the model can quickly and accurately predict the state or related parameters of the magnetic dipole based on actual measurement data, thus enabling the rapid detection and positioning of magnetic targets.

[0066] Example 1:

[0067] This example demonstrates the application of the fast inversion method for electromagnetic logging while drilling data based on BiLSTM in actual oil exploration.

[0068] During the exploration of an oilfield, due to the complex formation structure with multi-layer structures and anisotropic effects, traditional deterministic methods such as the Gauss-Newton method repeatedly fell into local minima during data interpretation, resulting in incorrect judgment of the drilling direction and wasting resources and time. To solve this problem, the fast inversion method for electromagnetic logging while drilling data based on BiLSTM proposed in the present invention was adopted.

[0069] First, electromagnetic wave response signals from the formation are received by three antennas on the logging while drilling instrument. These signals contain information such as formation resistivity, boundary position, anisotropy coefficient, and formation dip angle. After preprocessing, these signals are converted into logging data that can be used as input for the neural network.

[0070] Next, the preprocessed logging data is input into a neural network composed of four bidirectional long short-term memory (BiLSTM) layers and a fully connected layer. Each BiLSTM layer contains a rectified linear unit (ReLU) as the activation function, and the recurrent dropout technique is used in the BiLSTM layer to reduce overfitting. Through feature extraction, convolution, and fusion, the neural network outputs classification results and prediction results. The classification results are any one of four preset models, and these four formation models correspond to different formation characteristics respectively. The prediction results are parameters of various specific formation characteristics corresponding to the classified formation model, including formation thickness, resistivity, porosity, permeability, water saturation, etc.

[0071] The RMSprop optimizer is used to train the logging data with the mean squared error defined as the loss function to improve the prediction accuracy of the neural network. Through the efficient interpretation of the neural network, the problem of getting stuck in local minima is successfully avoided, and due to the fast processing ability of the neural network, this method meets the requirements of real-time geological steering.

[0072] During the exploration process, the real-time obtained geological steering and formation evaluation data are used to guide the selection of the drilling direction. Compared with the traditional deterministic method, the method proposed in the present invention significantly improves the success rate of drilling and reduces the waste of resources and time.

[0073] In summary, this embodiment demonstrates the successful application of the fast inversion method of electromagnetic logging while drilling data based on BiLSTM in oil exploration. This method can not only avoid getting stuck in local minima but also meet the real-time requirement, providing accurate and efficient data support for geological steering and formation evaluation.

[0074] Embodiment 2:

[0075] In this embodiment, we will elaborate in detail on how the fast inversion method of electromagnetic logging while drilling data based on BiLSTM combines with Figure 1 the shown recurrent neural network structure to achieve efficient and accurate geological steering and formation evaluation.

[0076] First, we clarify Figure 1 the meanings of the symbols in it: x represents the input of the hidden layer neurons, o represents the output of the network, s represents the output of the hidden layer neurons, and U, V, and W represent different connection weights respectively. These connection weights play a crucial role in the recurrent neural network. They are responsible for converting the input information into the state of the hidden layer and then converting the state of the hidden layer into the final output.

[0077] During the inversion process of electromagnetic logging-while-drilling data, we first preprocess the electromagnetic wave signals received by three antennas and convert them into an input format acceptable to the neural network. These preprocessed logging data are input into a neural network composed of four BiLSTM layers and a fully connected layer.

[0078] Each BiLSTM layer contains a ReLU activation function to increase the network's non-linear expression ability. At the same time, we use the recurrent dropout technique in the BiLSTM layer to reduce overfitting and improve the generalization ability of the model.

[0079] The hidden layer of the neural network adopts the structure of a recurrent neural network, whose characteristic is that the input information of the hidden layer neurons not only comes from the output of the input layer neurons but also from the output of the hidden layer neurons at the current moment. This structure enables the neural network to capture long-term dependencies in sequential data, thus better understanding the internal laws of logging data.

[0080] In addition, each layer of the recurrent neural network shares parameters U, V, and W, which greatly reduces the number of parameters and improves the network training speed. Let st be the output value of the hidden layer neurons at time t, which is calculated based on the current input information It and the output value st-1 of the hidden layer neurons at the previous moment. This information transmission method enables the neural network to remember the previous state and affect the current and future outputs, thus better processing sequential data.

[0081] In this embodiment, we perform feature extraction, convolution, and fusion on the input logging data through the neural network, and finally output classification results and prediction results. The classification result is any one of the four set models, and the prediction result is multiple formation parameters corresponding to the model category. These output results provide accurate and efficient data support for geosteering and formation evaluation.

[0082] Compared with traditional deterministic methods, the method proposed in the present invention combines the advantages of recurrent neural networks, can learn complex patterns in logging data, and realizes efficient interpretation of data. At the same time, due to the adoption of the BiLSTM structure, our method can not only avoid falling into local minima but also meet the real-time requirements. This has important practical significance in on-site operations, can significantly improve the success rate of drilling, and reduce waste of resources and time.

[0083] In summary, this embodiment demonstrates the successful application of the fast inversion method for electromagnetic logging-while-drilling data based on BiLSTM combined with the recurrent neural network structure in geological exploration. This method provides strong technical support for geosteering and formation evaluation through efficient data processing and accurate prediction results.

[0084] Embodiment 3:

[0085] When training a recurrent neural network, in view of the problem that the vanishing gradient and the exploding gradient limit the ability of the network to utilize long-term historical information, thus affecting the training effect, the long short-term memory unit (LSTM) is adopted.

[0086] LSTM is an improvement to the traditional recurrent neural network. It carries and crosses information over multiple time steps by adding an additional data stream (c_t), also known as the cell state. This design allows past information to flow continuously in the network and re-enter the calculation when needed, thus effectively solving the problem of vanishing gradient, as Figure 2 shown.

[0087] To further improve the performance of LSTM, we adopt the bidirectional LSTM (BiLSTM) structure. The bidirectional LSTM contains two standard LSTM networks, which process the input sequence in the forward and reverse orders respectively. By combining the representations in these two directions, the bidirectional LSTM can capture complex patterns and context information that a unidirectional LSTM might ignore.

[0088] In our implementation, the bidirectional LSTM is applied to process time series data, such as electromagnetic logging while drilling data. Through training, the network learns to extract meaningful features from the data and generates accurate prediction and classification results. These results have important guiding significance for geological exploration and formation evaluation.

[0089] Example 4:

[0090] The present invention adopts a specific neural network structure to process azimuthal electromagnetic logging while drilling data and realizes fast inversion of the data. The following is a detailed description of the neural network structure used in the present invention and its working process, and a deep analysis of Figure 3 , Figure 4 , Figure 5 and Figure 6 contents.

[0091] First of all, the neural network structure used in the present invention is as Figure 3 shown. This is a deep network composed of four BiLSTM layers and a fully connected layer. Each BiLSTM layer is equipped with a rectified linear unit (ReLU) as the activation function to increase the nonlinearity of the network. The BiLSTM layer is characterized by being able to process sequence data in both the forward and reverse directions simultaneously, so as to capture information that a unidirectional LSTM might ignore. In addition, to reduce overfitting, the present invention implements the recurrent dropout technique in the BiLSTM layer. This technique can randomly discard some connections in the network, thereby improving the generalization ability of the model.

[0092] Figure 3 The structure of this neural network is shown in detail, with each layer clearly visible, including the input layer, four BiLSTM layers, a fully connected layer, and an output layer. Data enters the network from the input layer, is processed by the BiLSTM layers, integrated in the fully connected layer, and finally obtains the result from the output layer.

[0093] In terms of data preparation, the present invention uses the response of an ideal magnetic dipole in a layered medium as the instrument response and uses the propagation matrix method to generate training datasets and validation datasets. These datasets will be used for the training and validation of the neural network.

[0094] Figure 4 The schematic diagram of the instrument used in the present invention is shown. The instrument has a detection depth of up to 30 m. Therefore, in setting the model in the present invention, the situation of detecting at most one interface above or below the instrument is considered.

[0095] Based on this consideration, the present invention sets four different models, as Figure 5 shown. These models represent different formation structures and detection situations. The inversion targets of formation parameters include the horizontal resistivity of each layer, the distance to the boundary, the anisotropy coefficient of the formation where the instrument is located, and the formation dip angle, etc. These parameters are of great significance for geological exploration and formation evaluation.

[0096] During the training process, the present invention uses the RMSprop optimizer and defines the loss function as the mean squared error. This setting helps the network learn better and optimize the model parameters. At the same time, in order to prevent overfitting and improve the training efficiency, the present invention also adopts the EarlyStopping callback method and sets the corresponding tolerance.

[0097] After the network training is completed, it can be used to invert the actual logging data. The inversion process is as Figure 6 shown. First, the preprocessed logging data is input into the trained neural network; then, the network will perform operations such as feature extraction, convolution, and fusion on these data; finally, the network will output the classification result and the prediction result. The classification result indicates which of the four set models the data belongs to, and the prediction result includes various formation parameters corresponding to the model category.

[0098] Generally speaking, the present invention realizes the rapid inversion of azimuthal electromagnetic logging while drilling data by adopting a specific neural network structure and training strategy. This method not only improves the accuracy of data interpretation but also greatly improves the processing speed, providing strong technical support for geological exploration and formation evaluation. At the same time, the present invention also has a certain degree of flexibility and scalability and can be adjusted and optimized according to actual needs.

[0099] In the present invention, unless otherwise clearly specified or limited, the terms "installed", "set", "connected", "fixed", "swivel-connected", etc. shall be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or integrated; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the internal connection of two components or the interaction relationship between two components. Unless otherwise clearly defined, for those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0100] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention.

Claims

1. A fast inversion method for electromagnetic logging while drilling data based on BiLSTM, characterized in that, It includes the following steps: Step 1: Obtain the electromagnetic wave signals received by the antenna and perform preprocessing to obtain logging data; Step 2: Use the preprocessed logging data as input and input it into the BiLSTM layer and the neural network. The BiLSTM layer contains a rectified linear unit ReLU as the activation function, and the recurrent dropout technique is used in the BiLSTM layer to reduce the overfitting phenomenon; Step 3: The neural network performs feature extraction, convolution, and fusion on the input logging data, and outputs classification results and prediction results. The classification results are any one of the preset models, and the prediction results are multiple formation parameters corresponding to the model category; Step 4: Perform geological steering and formation evaluation based on the classification results and prediction results.

2. A fast inversion method for electromagnetic logging while drilling data based on BiLSTM according to claim 1, characterized in that In Step 1, before obtaining the electromagnetic wave signals received by the antenna, first use the propagation matrix method as the forward algorithm to calculate the response of the magnetic dipole in the one-dimensional layered medium. During this process, the influence of the borehole environment is ignored, and the response of the ideal magnetic dipole in the layered medium is used to simulate the instrument response to generate the training dataset and the validation dataset. The training dataset will be used for the training of the subsequent neural network weights, and the validation dataset is used to verify the training results during the training process. After preparing the training dataset and the validation dataset, then obtain the electromagnetic wave signals received by the antenna and perform preprocessing to obtain the logging data for neural network input.

3. A fast inversion method for electromagnetic logging while drilling data based on BiLSTM according to claim 1, characterized in that The neural network in Step 3 uses a recurrent neural network. The input information of the hidden layer neurons not only comes from the output of the input layer neurons but also from the output of the hidden layer neurons at the current moment; and each layer of this recurrent neural network shares the parameters U, V, and W to reduce the number of parameters and improve the network training speed.

4. A fast inversion method for electromagnetic logging while drilling data based on BiLSTM according to claim 2, characterized in that, The output value st of the hidden layer neurons of the recurrent neural network at time t is calculated based on the input information It at the current moment and the output value st-1 of the hidden layer neurons at the previous moment through the activation function and the shared parameters U, V, and W; among them, the input information It includes the logging data input at the current moment and the possible feedback of the output layer neurons at the previous moment. The output ot of the output layer neurons at time t is calculated based on the output value st of the hidden layer neurons at time t through the connection weights and the activation function.

5. A fast inversion method for electromagnetic logging while drilling data based on BiLSTM according to claim 1, characterized in that, The electromagnetic wave signals received by the antenna in Step 1 specifically refer to the electromagnetic wave response signals from the formation received by the antenna on the logging tool during the logging-while-drilling process. These signals include, but are not limited to, formation resistivity, boundary position, anisotropy coefficient, and formation dip angle. These signals are preprocessed and converted into logging data that can be used as input for the neural network.

6. A fast inversion method for electromagnetic logging while drilling data based on BiLSTM according to claim 1, characterized in that In Step 2, the neural network uses the RMSprop optimizer to train the logging data with the loss function defined as the mean squared error.

7. A fast inversion method for electromagnetic logging while drilling data based on BiLSTM according to claim 1, characterized in that, The output classification result described in step 3 is any one of four preset formation models, and these four formation models respectively correspond to different formation characteristics; the prediction result is the parameters of various specific formation characteristics corresponding to the classified formation model. Among them, the parameters of formation characteristics include, but are not limited to, formation thickness, resistivity, porosity, permeability, and water saturation. These parameters are obtained through the feature extraction, convolution, and fusion processes of the neural network to provide detailed data support for geosteering and formation evaluation.

8. An electromagnetic wave forward-looking logging while drilling data inversion system based on a recurrent neural network, characterized in that, It includes: A data preprocessing module, which is used to obtain the electromagnetic wave signals received by three antennas and perform preprocessing to obtain logging data; A neural network module, which is used to receive the preprocessed logging data as input, perform feature extraction, convolution, and fusion through a neural network composed of four bidirectional long short-term memory (BiLSTM) layers and a fully connected layer, and output a classification result and a prediction result; A result output module, which is used to perform geosteering and formation evaluation according to the output of the neural network.

9. The fast inversion system for electromagnetic logging while drilling data based on BiLSTM according to claim 8, characterized in that, The neural network module uses the RMSprop optimizer to train the logging data with the mean square error defined as the loss function.

10. A rapid inversion system for electromagnetic logging while drilling data based on BiLSTM according to claim 8, characterized in that, The formation parameters include the horizontal resistivity of each layer, the distance to the boundary, the anisotropy coefficient of the formation where the instrument is located, and the formation dip angle.

Images