A method for predicting wellbore stability based on deep learning
Through the deep learning-based well wall stability prediction method, the Bi-LSTM model and SE attention mechanism are used to solve the problem of insufficient accuracy of well wall stability prediction in the existing technology, and more efficient well wall stability prediction and drilling cost control are achieved.
Patent Information
- Application Number
- CN202411798211.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-09
- Publication Date
- 2025-04-01
AI Technical Summary
The existing technology is difficult to effectively predict the stability of the well wall, resulting in frequent drilling fluid leakage and low sealing success rate, which increases drilling cost and construction time.
The well wall stability prediction method based on deep learning is adopted, and a large leak-proof and plugging database is established by obtaining drilling engineering data. The Bi-LSTM model is combined with the SE attention mechanism to train and optimize the model to improve prediction accuracy.
Through the training and optimization of deep learning models, the weight parameters and bias parameters in the model can be effectively optimized, and the accuracy of well wall stability prediction can be improved, and the problems of insufficient flexibility of traditional methods and difficult to capture complex relationships can be overcome.
Smart Images
Figure CN119740606A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of wellbore stability prediction, and in particular to a wellbore stability prediction method based on deep learning. Technical Background
[0002] In the digital information age, it is becoming more and more a trend to use computers and automation technology to deal with various problems encountered in oil and gas drilling. The continuous leakage of drilling fluid increases the loss of on-site working time. In the face of downhole leakage problems, improper treatment measures will lead to a low plugging success rate and even well abandonment. Frequent well leakage problems consume a lot of construction time. The plugging construction will increase the drilling cycle and greatly increase the drilling cost, and it cannot meet the strategic needs of low-cost development.
[0003] Data-driven machine learning, especially deep learning methods, seem to be a solution. Data-driven machine learning models rely on various parameters collected during oil field drilling, including drilling parameters, geological parameters, engineering parameters, drilling fluid parameters, etc. Through data processing, feature extraction, model training, model evaluation and other steps, a prediction model for lost formations is established. Machine learning models may become inflexible or insufficient to capture patterns in the data when faced with large-scale data. Deep learning models have advantages over machine learning models when dealing with large amounts of data because deep learning models can automatically learn high-level feature representations of data, which means they can extract useful features from raw data without manual feature engineering. Deep learning models are usually composed of multiple levels, which gradually extract abstract information from the data through multi-level representations and can process a large number of parameters, which enables them to adapt to large-scale data and learn complex relationships. Summary of the invention
[0004] In view of the above defects or deficiencies in the prior art, the present application aims to provide a method for predicting wellbore stability based on deep learning. The method comprises the following steps:
[0005] S1. Obtain drilling engineering data and establish a large database for leak prevention and plugging. Preprocess the data, screen the characteristic parameters based on the preprocessed data, and perform LOESS noise reduction on the characteristic parameters.
[0006] S2. Establish a deep learning model through Bi-LSTM;
[0007] S3, carrying the attention module through the SE attention mechanism;
[0008] S4, model training, the training steps include:
[0009] S41. Parameter initialization: Parameters such as weights and learning rates need to be set before they can be correctly applied to the model;
[0010] S42. Establish the loss function and set the stopping condition: The loss function is an important learning criterion, which can be used to measure the performance of a network and is often related to the optimization amount. The mean square error loss function is used to calculate the mean square error between them. The calculation formula is as follows:
[0011]
[0012] Where N is the length of the input sequence, is the predicted value of the model, y i is the true value;
[0013] S43. By applying the Adam optimizer, the weight parameters and bias parameters in the model can be effectively optimized;
[0014] S44. By carefully adjusting the parameters and functions, an effective network training process can be achieved, that is, the parameters of the model are continuously adjusted according to the pre-set number of training times until a certain threshold is reached. In this way, even if the expected number of training times is not reached, the model can stop running when the threshold is met. After 3 epochs, if the loss function has not been significantly improved, the deep learning model will be terminated.
[0015] Among them, the implementation process of the SE attention mechanism in S3 is as follows:
[0016] 1) Squeeze: It is a method used to compress two-dimensional features in order to obtain global features;
[0017] 2) Excitation: The process of generating different weight values aims to build correlation between channels, which is mainly achieved through the fully connected layer. The number of output weights in this process is equal to the number of input channels;
[0018] 3) Scale: Assign the previously calculated weights to each channel so that its features can be effectively evaluated.
[0019] The present invention has the following beneficial effects:
[0020] This method can effectively optimize the weight parameters and bias parameters in the model by applying the Adam optimizer, thereby achieving the purpose of minimizing the loss. At the same time, by training the model, an effective model can be obtained, and these models can be used to evaluate their capabilities for comparison with other models. This method can overcome the defects of theoretical models that adopt multiple assumptions, have poor universality, and are difficult to calibrate advanced model parameters. By processing data information with a bidirectional structure, it can process forward information at the same time, greatly improving model performance. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] In order to more clearly illustrate the scheme in the embodiment of the present invention, the following is a brief introduction to the drawings required for describing the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without paying creative work.
[0022] Figure 1 It is a block diagram of the LSTM unit structure of the present invention;
[0023] Figure 2 This is a structural block diagram of the SE module of the present invention.
[0024] Figure 3 This is a structural diagram of the bidirectional LSTM model of the present invention. DETAILED DESCRIPTION
[0025] The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the invention are shown in the accompanying drawings.
[0026] The accompanying drawings in the embodiments of the present invention: different types of section lines in the drawings are not marked according to national standards, nor do they impose requirements on the materials of the components. Instead, they are used to distinguish the cross-sectional views of the components in the drawings.
[0027] The wellbore stability prediction method based on Bi-LSTM fusion SE module includes the following steps:
[0028] S1. Obtain drilling engineering data and establish a large database for leak prevention and plugging. Preprocess the data, screen the characteristic parameters based on the preprocessed data, and perform LOESS noise reduction on the characteristic parameters.
[0029] S2. Establish a deep learning model through Bi-LSTM;
[0030] Figure 2As shown in the figure, the first layer is the input layer, which inputs data; then there are two one-dimensional convolutional layers, each consisting of 16 convolutional kernels of size 4, with a step size of 1 and a linear activation function; after the two convolutional layers extract features from the data, it passes through two bidirectional LSTM layers, which are also the core of the Bi-LSTM model. Finally, there are two fully connected layers. The first fully connected layer combines all the features of the upper layer, and finally outputs the decomposition results through the fully connected layer of N=1.
[0031] Figure 3 As shown, preferably, the Bi-LSTM in S2 uses RNN and its variant LSTM to process data information, which has a bidirectional structure. The basic architecture of the LSTM consists of five parts: input x t 、output h t , Update gate i t 、Forget gate t , output gate o t ,in:
[0032] The expression of the forget gate ft is as follows:
[0033] f t =σ(W f *[h t-1 , x t ]+b f )
[0034] Update Gate t The expression of is as follows:
[0035] i t =σ(W i *[h t-1 , x t ]+b i )
[0036] c′ t =tanh(W c *[h t-1 , x t ]+b c )
[0037] c t =f t *c t-1 +i t *c′ t
[0038] Output gate o t The expression of is as follows:
[0039] o t =σ(W o *[h t-1 , xt ]+b o )
[0040] h t =o t *tanh(c t )
[0041] The above content describes the basic structure of a unidirectional LSTM unit, which can only process information before the current moment. In contrast, RNN and its variants (such as bidirectional LSTM) have a bidirectional structure and can process forward information and backward information at the same time, providing a more comprehensive data understanding and analysis, so that each moment can determine the final output based on the previous information and the subsequent information. This bidirectional structure greatly improves the model performance because they can process information at two moments at the same time. The borehole stability prediction method used in the present invention is based on Bi-LSTM (Bidirectional LSTM), which can capture the information of the previous and next moments, overcoming the defects of the traditional neural network method.
[0042] S3, carrying the attention module through the SE attention mechanism;
[0043] S31, Squeeze: is a method used to compress two-dimensional features in order to obtain global features;
[0044] S32, Excitation: The process of generating different weight values aims to build correlation between channels, which is mainly achieved through the fully connected layer. The number of output weights in this process is equal to the number of input channels;
[0045] S33, Scale: Assign the previously calculated weights to each channel so that its features can be effectively evaluated.
[0046] S4, model training, the training steps include:
[0047] S41. Parameter initialization: Parameters such as weights and learning rates need to be set before they can be correctly applied to the model;
[0048] S42. Establish the loss function and set the stopping condition: The loss function is an important learning criterion, which can be used to measure the performance of a network and is often related to the optimization amount. The mean square error loss function is used to calculate the mean square error between them. The calculation formula is as follows:
[0049]
[0050] Where N is the length of the input sequence, is the predicted value of the model, y i is the true value;
[0051] S43. By applying the Adam optimizer, the weight parameters and bias parameters in the model can be effectively optimized;
[0052] S44. By carefully adjusting the parameters and functions, an effective network training process can be achieved, that is, the parameters of the model are continuously adjusted according to the pre-set number of training times until a certain threshold is reached. In this way, even if the expected number of training times is not reached, the model can stop running when the threshold is met. After 3 epochs, if the loss function has not been significantly improved, the deep learning model will be terminated.
[0053] The above contents are merely examples and explanations of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the invention or exceed the scope defined by the claims, they shall all fall within the protection scope of the present invention.
Claims
1. A method for predicting wellbore stability based on deep learning, characterized in that: The following steps are involved: S1. Obtain drilling engineering data, establish a large database for leak prevention and plugging, pre-process the data, screen the characteristic parameters based on the pre-processed data, and perform LOESS noise reduction on the characteristic parameters; S2. Establish a deep learning model through Bi-LSTM. First, there is the input layer, which inputs data. Then there are two one-dimensional convolutional layers, each consisting of 16 convolution kernels of size 4, with a step size of 1 and a linear activation function. After the two convolutional layers extract the features of the data, it passes through two bidirectional LSTM layers, which are also the core of the Bi-LSTM model. Finally, there are two fully connected layers. The first fully connected layer integrates all the features of the upper layer, and finally outputs the decomposition results through the fully connected layer of N=1. S3, carrying the attention module through the SE attention mechanism; S4, model training, the training steps include: S41. Parameter initialization: Parameters such as weights and learning rates need to be set before they can be correctly applied to the model; S42. Establish the loss function and set the stopping condition: The loss function is an important learning criterion, which can be used to measure the performance of a network and is often related to the optimization amount. The mean square error loss function is used to calculate the mean square error between them. The calculation formula is as follows: Where N is the length of the input sequence, is the predicted value of the model, y i is the true value; S43. By applying the Adam optimizer, the weight parameters and bias parameters in the model can be effectively optimized; S44. By carefully adjusting the parameters and functions, an effective network training process can be achieved, that is, the parameters of the model are continuously adjusted according to the pre-set number of training times until a certain threshold is reached. Even if the expected number of training times is not reached, the model can stop running when the threshold is met. After 3 epochs, if the loss function has not been significantly improved, the deep learning model will be terminated.
2. The method for identifying and dividing shale sedimentary cycles of different scales according to claim 1, characterized in that: In step S2, Bi-LSTM uses RNN and its variant LSTM to process data information, which has a bidirectional structure. The basic architecture of the LSTM consists of five parts: input x t 、output h t , Update gate i t 、Forget gate t , output gate O t ,in: Forget Gate t The expression of is as follows: f t =σ(W f *[h t-1 ,x t ]+b f ) Update Gate t The expression of is as follows: i t =σ(W i *[h t-1 ,x t ]+b i ) c' t =tanh(W c *[h t-1 ,x t ]+b c ) c t =f t *c t-1 +i t *c′ t Output Gate O t The expression of is as follows: the t =σ(W o *[h t-1 ,x t ]+b o ) h t =o t *tanh(c t ) Finally, the output value of the unit is determined, and the output gate calculates h t The sigmoid layer determines which part of the cell state is output, and the tanh layer processes it to obtain a value between -1 and 1, and finally determines the output.
3. The method for identifying and dividing shale sedimentary cycles of different scales according to claim 1, characterized in that: In step S3, the implementation process of the SE attention mechanism is as follows: S31, Squeeze compression: is a method used to compress two-dimensional features in order to obtain global features; S32, Excitatio n Motivation: The process of generating different weight values aims to build correlation between channels, which is mainly achieved through the fully connected layer. The number of output weights in this process is equal to the number of input channels; S33, Scal e : Assign the previously calculated weights to each channel so that its features can be effectively evaluated.