Prediction method of resistivity curve of logging while drilling based on residual fully connected network

Abstract

The present invention discloses a method for predicting a while-drilling (LWD) resistivity curve based on a residual fully connected network, comprising: establishing a data set based on seismic background data, geological background data, and logging data of a study area to be drilled; wherein the data set includes a training data set and a test data set; inputting the training data set into a prediction model for the LWD resistivity curve ahead of the drill bit based on the residual fully connected network, outputting a predicted resistivity result, and substituting the predicted resistivity result and the actual resistivity value in the training data set into a loss function to calculate a loss value; calculating a gradient based on the loss value, and correcting the weight of the prediction model using the gradient and back propagation; inputting a test data set based on the corrected prediction model to obtain a prediction result for the LWD resistivity curve ahead of the drill bit. The method of the present invention predicts the LWD resistivity curve ahead of the drill bit, guides lithology identification work in the area to be drilled based on the prediction result, further reduces drilling risks, and improves the basis for while-drilling decision-making.

Description

Technical Field

[0001] The present invention belongs to the fields of marine oil and gas geological exploration and artificial intelligence and deep learning technology, and specifically relates to a method for predicting while-drilling resistivity curves based on a residual fully connected network. Background Art

[0002] With the advancement of information technology, artificial intelligence (AI) has been widely applied in various fields. AI uses machine learning and deep learning algorithms to mine historical big data. Deep learning (DL) is a new research direction in the field of machine learning (ML). It is an algorithmic model that learns the inherent laws of data and represents and learns data.

[0003] Fully Connected Residual Networks (FCRNs) introduce residual structures into fully connected networks to address issues such as vanishing or exploding gradients and network degradation during backpropagation in deep neural networks. While this network structure is rarely used in geophysics, it has widespread applications in military threat assessment, monitoring abnormal behavior in the power industry, and recognizing dynamic gestures in image processing.

[0004] In current projects, mud flushing fluids contaminate the formation during drilling, leading to errors in logging data. This has led to an increasing demand for logging while drilling (LWD), but the parameters measured by LWD are limited to resistivity and natural gamma ray. Furthermore, in the exploration of abnormally high-pressure reservoirs and those with complex structures, the accuracy of depth and pressure prediction ahead of the drill bit is low, and unclear structural profiling can easily lead to drilling disasters. Summary of the Invention

[0005] Purpose of the invention: In order to overcome the shortcomings of the existing technology, the present invention provides a method for predicting the resistivity curve of the logging while drilling based on a residual fully connected network. The method of the present invention predicts the resistivity curve of the logging while drilling ahead of the drill bit, and guides the lithology identification work of the area to be drilled based on the prediction results, further reducing the drilling risk and improving the basis for decision-making while drilling.

[0006] Technical solution: In the first aspect, the present invention provides a method for predicting a well logging resistivity curve based on a residual fully connected network, comprising:

[0007] Obtain seismic background data, geological background data and well logging data for the area to be drilled in the study area;

[0008] A data set is established based on the seismic background data, geological background data and well logging data of the area to be drilled in the study area; the data set includes a training data set and a test data set;

[0009] The training data set is input into the LWD resistivity curve prediction model based on the residual fully connected network ahead of the drill bit, and the predicted resistivity result is output;

[0010] Substitute the predicted resistivity result and the actual resistivity value in the training data set into the loss function to calculate the loss value;

[0011] Calculate the gradient according to the loss value, use the gradient to obtain the optimized loss value, and perform backpropagation based on the optimized loss value to correct the weight of the prediction model to obtain the corrected prediction model;

[0012] The test data set was input based on the revised prediction model to obtain the prediction results of the LWD resistivity curve ahead of the drill bit.

[0013] In a further embodiment, the seismic background data includes: seismic attributes, seismic tectonic interpretation model, and seismic inversion results; wherein the seismic attributes include interval velocity, instantaneous frequency, and seismic amplitude;

[0014] The geological background data includes a lithologic framework model;

[0015] The logging data includes time series and resistivity curves.

[0016] In a further embodiment, a method for establishing a data set based on seismic background data, geological background data, and well logging data of a region to be drilled in a study area includes:

[0017] The seismic background data, geological background data and well logging data of the area to be drilled in the study area are subjected to outlier removal and standardization, and the preferred seismic attribute data are set according to the main controlling factors to obtain a data set with outliers removed and standardized to sort the seismic attribute data according to priority.

[0018] The data in the dataset is divided into training dataset and test dataset in a ratio of 70%:30%.

[0019] In a further embodiment, the seismic background data, geological background data, and well logging data of the area to be drilled in the study area are subjected to outlier removal and standardization, and preferred seismic attribute data are set according to the main controlling factors to obtain a data set with outlier removal and standardization sorted according to the priority of the seismic attribute data, including:

[0020] Delete the abnormal mutation points in the seismic background data, geological background data and logging data of the area to be drilled in the study area, or interpolate based on the upper and lower points of the abnormal points to obtain the data of the area to be drilled in the study area without the abnormal values;

[0021] The data of the area to be drilled in the study area were normalized after removing outliers to obtain a standardized data set of the area to be drilled in the study area;

[0022] The main controlling factors of the seismic attributes and resistivity curves in the standardized data set of the area to be drilled in the study area were analyzed to obtain the dependent and independent variables, among which the dependent variable is the resistivity curve value and the independent variables are the seismic attributes including interval velocity, instantaneous frequency and seismic amplitude.

[0023] Calculate the independent variables to obtain the grey correlation coefficient between each independent variable, i.e., different seismic attributes and resistivity curve values;

[0024] The grey correlation coefficients are sorted from large to small, and based on the sorting results, three categories of grey correlation coefficients with higher rankings than others are selected as the earthquake attribute data actually used in training.

[0025] In a further embodiment, the data of the area to be drilled in the study area is normalized after outliers are removed. The expression of the normalization is:

[0026]

[0027] In the formula Refers to a dataset; Refers to the mean of the data set; Refers to the standard deviation of the data set.

[0028] In a further embodiment, the network structure in the drill bit front-end logging while drilling resistivity curve prediction model based on the residual fully connected network is composed of multiple dual residual blocks, a single dual residual block contains two fully connected layers, and the fully connected layer includes an input layer, a hidden layer, a BN layer, and an output layer.

[0029] In a further embodiment, the predicted resistivity result and the actual resistivity value in the training data set are substituted into the loss function to calculate the loss value. The expression of the loss function is:

[0030] Loss y =ω η [J x -J(y)] 2

[0031] Where, is the input parameter vector; is the resistivity curve label; Represents the predicted resistivity value obtained by the residual fully connected network; represents the objective function; Represents the weight of the predicted resistivity value, that is, the weight of the layer before the output layer in the residual fully connected network.

[0032] In a further embodiment, a method for calculating a gradient based on a loss value and using the gradient to perform back propagation to correct the weights of the prediction model to obtain a corrected prediction model includes:

[0033] In a further embodiment, the input of the test data set based on the modified prediction model to obtain the prediction result of the LWD resistivity curve ahead of the drill bit further includes:

[0034] Compare the predicted resistivity curve of the logging while drilling ahead of the drill bit with the actual resistivity in the test data set, and evaluate the prediction model effect based on the comparison results;

[0035] The prediction model that has passed the effect evaluation will be used in actual production. In actual production, the resistivity curve of the study area will be obtained in real time to update the training data set for online learning and optimization of the prediction model.

[0036] In a second aspect, the present invention provides a processing device including a memory and a processor, wherein the memory stores a computer program, and when the program is executed by the processor, a method for predicting a well logging resistivity curve based on a residual fully connected network is implemented.

[0037] In a third aspect, the present invention provides a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0038] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0039] In this method, by introducing the residual fully connected neural network into the field of oil and gas exploration and drilling operations, using the seismic and geological background information of the study area, and continuously training the residual fully connected network, it can be applied to the prediction of the well drilling resistivity curve in front of the drill bit in the entire study area. Based on the prediction results, it assists in lithology identification, thereby improving the drilling plan and reducing drilling risks and costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is an overall flow chart of the method for predicting the resistivity curve of the logging while drilling ahead of the drill bit based on the residual fully connected network of the present invention;

[0041] Figure 2 It is a schematic diagram of the training process of the residual fully connected network and the structural diagram of the dual residual module in the method for predicting the resistivity curve of the logging while drilling ahead of the drill bit based on the residual fully connected network of the present invention;

[0042] Figure 3 It is a schematic diagram of earthquake amplitude in the earthquake attributes of the study area of this invention. DETAILED DESCRIPTION

[0043] In order to more fully understand the technical content of the present invention, the technical solution of the present invention is further introduced and illustrated in conjunction with specific embodiments below, but is not limited thereto.

[0044] Combine Figure 1 The method for predicting the well logging resistivity curve based on the residual fully connected network in this embodiment is further described, including:

[0045] Step 1: Obtain seismic background data, geological background data and well logging data for the area to be drilled in the study area respectively; in this embodiment, the seismic background data includes: seismic attributes, seismic tectonic interpretation model, and seismic inversion results; wherein, seismic attributes include interval velocity, instantaneous frequency and seismic amplitude, such as Figure 3 The earthquake amplitude attribute shown is represented;

[0046] The geological background data includes a lithologic framework model;

[0047] The logging data includes time series and resistivity curves.

[0048] Step 2: Create a dataset based on the seismic background data, geological background data, and well logging data of the area to be drilled in the study area; the dataset includes a training dataset and a test dataset;

[0049] Step 2.1: Remove outliers from the seismic background data, geological background data, and well logging data, perform standardization, and set preferred seismic attribute data based on the main controlling factors to obtain a data set that has been de-outliersed, standardized, and prioritized according to seismic attribute data;

[0050] Step 2.2: Divide the data in the dataset into training dataset and test dataset in a ratio of 70%:30%.

[0051] Step 2.1 includes:

[0052] Step 2.1.1: Delete the abnormal mutation points in the seismic background data, geological background data and well logging data of the area to be drilled in the study area or interpolate based on the upper and lower points of the abnormal points to obtain data without abnormal values;

[0053] Step 2.1.2: Standardize the data after removing outliers to obtain a unified standardized data set; preferably, in the standardization of the data after removing outliers, the expression of the standardization is:

[0054]

[0055] In the formula Refers to a dataset; Refers to the mean of the data set; Refers to the standard deviation of the data set.

[0056] Step 2.1.3: Perform a master factor analysis on the seismic attributes and resistivity curves in the standardized dataset of the area to be drilled in the study area to obtain dependent and independent variables. The dependent variable is the resistivity curve value, and the independent variables are seismic attributes including interval velocity, instantaneous frequency, and seismic amplitude.

[0057] Step 2.1.4: Calculate the independent variables to obtain the grey correlation coefficient between each independent variable, i.e., different seismic attributes and resistivity curve values;

[0058] Step 2.1.5: Sort the grey correlation coefficients from large to small, and based on the sorting results, select three categories of grey correlation coefficients that rank higher than the others as the earthquake attribute data actually used in training.

[0059] Step 3: Input the training data set into the LWD resistivity curve prediction model based on the residual fully connected network ahead of the drill bit, and output the predicted resistivity result;

[0060] like Figure 2 The network structure in the drill bit front logging while drilling resistivity curve prediction model based on the residual fully connected network in this embodiment is composed of multiple dual residual blocks. A single dual residual block contains two fully connected layers, and the fully connected layer includes an input layer, a hidden layer, a BN layer, and an output layer.

[0061] Step 4: Substitute the predicted resistivity result and the actual resistivity value in the training data set into the loss function to calculate the loss value; the loss function expression is:

[0062] Loss y =ω η [J x -J(y)] 2

[0063] Where, is the input parameter vector; is the resistivity curve label; Represents the predicted resistivity value obtained by the residual fully connected network; represents the objective function; Represents the weight of the predicted resistivity value, that is, the weight of the layer before the output layer in the residual fully connected network.

[0064] Step 5: Calculate the gradient according to the loss value, use the gradient to obtain the optimized loss value, and perform back propagation based on the optimized loss value to correct the weight of the prediction model to obtain the corrected prediction model; specifically, calculate the gradient according to the loss value, and perform back propagation in the above prediction model network structure to correct the weight of the model, repeat steps 3 to 5 to cyclically optimize the loss value until the loss value is small and stable, and output the optimized loss value; use back propagation to adjust the parameters of the loss value to obtain a stable and corrected prediction model.

[0065] Step 6: Input the test data set based on the modified prediction model to obtain the prediction result of the LWD resistivity curve ahead of the drill bit. This embodiment also includes:

[0066] Compare the predicted resistivity curve of the logging while drilling ahead of the drill bit with the actual resistivity in the test data set, and evaluate the prediction model effect based on the comparison results;

[0067] The prediction model that has passed the effectiveness evaluation is applied to actual production. In actual production, the resistivity curve of the study area is obtained in real time to update the training data set and maintain the prediction model's online learning optimization state. It should be noted that the resistivity curve of the study area is obtained in real time during production. This data can be used as new training data for the model to perform online learning, achieving the goal of real-time optimization.

[0068] Example 2:

[0069] This embodiment provides a processing device including a memory and a processor. The memory stores a computer program. When the program is executed by the processor, the following method for predicting a LWD resistivity curve based on a residual fully connected network is implemented.

[0070] Obtain seismic background data, geological background data and well logging data for the area to be drilled in the study area;

[0071] A data set is established based on the seismic background data, geological background data and well logging data of the area to be drilled in the study area; the data set includes a training data set and a test data set;

[0072] The training data set is input into the LWD resistivity curve prediction model based on the residual fully connected network ahead of the drill bit, and the predicted resistivity result is output;

[0073] Substitute the predicted resistivity result and the actual resistivity value in the training data set into the loss function to calculate the loss value;

[0074] Calculate the gradient according to the loss value, and use the gradient and back propagation to correct the weight of the prediction model to obtain the corrected prediction model;

[0075] The test data set was input based on the revised prediction model to obtain the prediction results of the LWD resistivity curve ahead of the drill bit.

[0076] Example 3:

[0077] This embodiment describes a readable storage medium having a computer program stored thereon. When the program is executed by a processor, the steps of the method described in the claims are implemented.

[0078] Obtain seismic background data, geological background data and well logging data for the area to be drilled in the study area;

[0079] A data set is established based on the seismic background data, geological background data and well logging data of the area to be drilled in the study area; the data set includes a training data set and a test data set;

[0080] The training data set is input into the LWD resistivity curve prediction model based on the residual fully connected network ahead of the drill bit, and the predicted resistivity result is output;

[0081] Substitute the predicted resistivity result and the actual resistivity value in the training data set into the loss function to calculate the loss value;

[0082] Calculate the gradient according to the loss value, and use the gradient and back propagation to correct the weight of the prediction model to obtain the corrected prediction model;

[0083] The test data set was input based on the revised prediction model to obtain the prediction results of the LWD resistivity curve ahead of the drill bit.

[0084] In summary, this method introduces a residual fully connected neural network into the field of oil and gas exploration and drilling operations. By leveraging the seismic and geological background information of the study area and continuously training the residual fully connected network, it can be applied to the prediction of LWD resistivity curves ahead of the drill bit throughout the entire study area. This assists in lithology identification, thereby improving drilling plans and reducing drilling risks and costs. The generalization performance of this method depends on the structural complexity of the study area and the effectiveness of the online learning and training of the residual fully connected network.

[0085] The embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0086] The embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0087] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0088] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0089] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0090] The above is only a preferred embodiment of the present invention. Without departing from the technical principles of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be considered as the scope of protection of the present invention.

Claims

1. A method for predicting resistivity curves of logging while drilling based on a residual fully connected network, characterized in that: include: Obtain seismic background data, geological background data and well logging data for the area to be drilled in the study area; Establish a data set based on seismic background data, geological background data and well logging data of the area to be drilled in the study area; The data set includes training data set and test data set; The training data set is input into the LWD resistivity curve prediction model based on the residual fully connected network ahead of the drill bit, and the predicted resistivity result is output; Substitute the predicted resistivity result and the actual resistivity value in the training data set into the loss function to calculate the loss value; Calculate the gradient according to the loss value, use the gradient to obtain the optimized loss value, and perform backpropagation based on the optimized loss value to correct the weight of the prediction model to obtain the corrected prediction model; The modified prediction model is used as the input for the test data set to obtain the predicted results of the LWD resistivity curve ahead of the drill bit. The network structure of the drill bit front-end logging while drilling resistivity curve prediction model based on residual fully connected network consists of multiple dual residual blocks. A single dual residual block contains two fully connected layers, and the fully connected layer includes an input layer, a hidden layer, a BN layer, and an output layer.

2. The method for predicting resistivity curve of logging while drilling based on residual fully connected network according to claim 1, characterized in that: The seismic background data includes: seismic attributes, seismic tectonic interpretation model, and seismic inversion results; wherein the seismic attributes include layer velocity, instantaneous frequency, and seismic amplitude; The geological background data includes a lithologic framework model; The logging data includes time series and resistivity curves.

3. The method for predicting resistivity curve of logging while drilling based on residual fully connected network according to claim 1, characterized in that: The method of establishing a data set based on seismic background data, geological background data and well logging data of the area to be drilled in the study area includes: The seismic background data, geological background data and well logging data of the area to be drilled in the study area are subjected to outlier removal and standardization, and seismic attribute data are set according to the main controlling factors to obtain a data set with outliers removed, standardized and sorted according to the priority of seismic attribute data; The data in the dataset is divided into training dataset and test dataset in a ratio of 70%:30%.

4. The method for predicting resistivity curves of logging while drilling based on residual fully connected networks according to claim 3, characterized in that: The method of removing outliers and performing standardization processing on seismic background data, geological background data, and well logging data, and setting seismic attribute data according to main controlling factors, to obtain a data set with outliers removed and standardized so as to be sorted according to seismic attribute data priority includes: Delete the abnormal mutation points in the seismic background data, geological background data and logging data of the area to be drilled in the study area, or interpolate based on the upper and lower points of the abnormal points to obtain the data of the area to be drilled in the study area without the abnormal values; The data of the area to be drilled in the study area are standardized after removing outliers to obtain the standardized data set of the area to be drilled in the study area; The main controlling factors of the seismic attributes and resistivity curves in the standardized data set of the area to be drilled in the study area were analyzed to obtain the dependent and independent variables, among which the dependent variable is the resistivity curve value and the independent variables are the seismic attributes including interval velocity, instantaneous frequency and seismic amplitude. Calculate the independent variables to obtain the grey correlation coefficient between each independent variable, i.e., different seismic attributes and resistivity curve values; The grey correlation coefficients are sorted from large to small, and based on the sorting results, three categories of grey correlation coefficients with higher rankings than others are selected as the earthquake attribute data actually used in training.

5. The method for predicting resistivity curve of logging while drilling based on residual fully connected network according to claim 4, characterized in that: The data of the area to be drilled in the study area is standardized after outliers are removed. The expression of the standardized processing is: ； In the formula Refers to a dataset; Refers to the mean of the data set; Refers to the standard deviation of the data set.

6. The method for predicting resistivity curve of logging while drilling based on residual fully connected network according to claim 1, characterized in that: Substitute the predicted resistivity result and the actual resistivity value in the training data set into the loss function to calculate the loss value. The expression of the loss function is: ； Where, is the input parameter vector; is the resistivity curve label; Represents the predicted resistivity value obtained by the residual fully connected network; represents the objective function; Represents the weight of the predicted resistivity value.

7. The method for predicting resistivity curves of logging while drilling based on residual fully connected networks according to claim 1, characterized in that: Based on the modified prediction model and inputting the test data set, the prediction results of the LWD resistivity curve ahead of the drill bit are obtained, which also include: Compare the predicted resistivity curve of the LWD ahead of the drill bit with the actual resistivity in the test data set, and evaluate the prediction model effect based on the comparison results; The prediction model that has passed the effect evaluation will be used in actual production. In actual production, the resistivity curve of the study area will be obtained in real time to update the training data set for online learning and optimization of the prediction model.

8. A processing device comprising a memory and a processor, characterized in that: The memory stores a computer program, which, when executed by the processor, implements the method for predicting a while-drilling resistivity curve based on a residual fully connected network according to any one of claims 1 to 7.

9. A readable storage medium, characterized in that: A computer program is stored thereon, and when the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images