An intelligent sliding sleeve pressure wave signal identification method based on LSTM

Abstract

The application discloses an intelligent sliding sleeve pressure wave signal identification method based on LSTM, and relates to the field of communication.The method comprises the following steps: establishing an intelligent sliding sleeve device model; using a pressure pump to provide mud flow, obtaining pressure wave signals in a wellbore, and dividing the pressure wave signals into N pressure wave segments; inputting the first pressure wave segment into a trained edge classifier based on a long short-term memory network to obtain the label of the first pressure wave segment; for the nth pressure wave segment, inputting a preset label and the (n-1)th pressure wave segment into a trained waveform generator based on a long short-term memory network to obtain the label of the nth pressure wave segment; and identifying the pressure wave signals based on the label of the first pressure wave segment and the labels of the second pressure wave segment to the Nth pressure wave segment. The application accurately identifies and predicts the label of each pressure wave segment, and realizes efficient decoding based on the label.

Description

Technical Field

[0001] This application relates to the field of communications, and in particular to an LSTM-based method for identifying intelligent sliding sleeve pressure wave signals. Background Technology

[0002] With the deepening exploration and development of unconventional oil and gas resources, intelligent well completion technology has gradually become an important development direction for oil and gas field construction. Horizontal well drilling and completion technology can effectively develop unconventional oil and gas reservoirs such as tight oil and gas reservoirs. As one of the key tools of intelligent well completion systems, intelligent sliding sleeves can achieve remote layered control of multi-branch, multi-section wellbore, thereby effectively improving the level of intelligent production management. To optimize traditional fracturing technology, several methods exist for controlling the sliding sleeve from the surface, including RFID tag balls, hydraulic control, and electromagnetic wave communication technology. However, these control technologies all have varying degrees of limitations in deep wells, long horizontal sections, and high-pressure wellbores. For example, RFID requires pre-opening at the end of the sliding sleeve and has a high risk of runaway; electromagnetic communication suffers severe signal attenuation and is difficult to identify when formation resistivity is high or well depth is large. Therefore, there is an urgent need for a communication solution that can still transmit stably and accurately control the opening of the sliding sleeve under high pressure, high depth, and high noise environments.

[0003] In recent years, pressure wave communication technology based on wellbore fluid media has become a research hotspot in the field of downlink communication due to its advantages such as being cable-free, flexible in deployment, low in cost, and strong anti-interference capability. This technology achieves signal encoding, transmission, and decoding control by injecting pressure pulses with specific modulation methods into the wellbore. However, due to the complex downhole conditions, pressure wave signals are affected by various factors during transmission, such as wellbore structure, fluid column disturbance, and throttling noise, resulting in nonlinear distortion of the signal waveform, blurred edge features, and the presence of splicing segments further introduces significant sequence discontinuities.

[0004] Deep learning has demonstrated powerful expressive capabilities in the field of time series modeling. Traditional machine learning algorithms, such as Support Vector Machines (SVM), Decision Trees, and Random Forests, can learn and predict data. However, these models suffer from slow training and sensitivity to noise when dealing with time series data. In this context, Long Short-Term Memory Networks (LSTM), widely used in speech recognition and biosignal analysis due to their ability to capture long-term dependencies, have limited application in downhole pressure wave communication recognition and prediction. Considering the periodic edge structure and nonlinear perturbation characteristics in pressure wave sequences, LSTM possesses theoretical advantages, but its training and prediction strategies still need to be optimized to suit downhole operating conditions. Summary of the Invention

[0005] The purpose of this application is to provide an LSTM-based intelligent sliding sleeve pressure wave signal identification method, which can accurately identify pressure wave signals. In practical applications, this application can control the intelligent sliding sleeve device based on the identified pressure wave signal. In addition, when applied to equipment, this application can also improve the computing efficiency of edge computing devices and servers.

[0006] To achieve the above objectives, this application provides the following solution:

[0007] In a first aspect, this application provides an LSTM-based intelligent sliding sleeve pressure wave signal identification method, including:

[0008] A model of an intelligent sliding sleeve device is established; the intelligent sliding sleeve device model includes a pressure pump and a wellbore connected by pipelines.

[0009] The pressure pump provides mud flow, pressure wave signals are acquired in the wellbore, and the pressure wave signals are divided into N pressure bands; the N pressure bands include the first pressure band to the Nth pressure band; 1 <N;

[0010] The first pressure band is input into a trained edge classifier based on a long short-term memory network to obtain the label of the first pressure band; the label is "1" or "0"; where "1" represents a falling edge and "0" represents a rising edge;

[0011] For the nth pressure band, the preset label and the (n-1)th pressure band are input into a trained waveform generator based on a long short-term memory network to obtain the label for the nth pressure band; 1 <n≤N;

[0012] The pressure wave signal is identified based on the tags of the first pressure band and the tags of the second to Nth pressure bands.

[0013] According to the specific embodiments provided in this application, this application has the following technical effects:

[0014] This application provides an LSTM-based intelligent sliding sleeve pressure wave signal recognition method. First, the pressure wave signal is divided into multiple pressure bands, and each band is processed separately, avoiding noise interference and information attenuation problems in long-sequence signals, thus improving recognition accuracy. Furthermore, a pre-trained edge classifier based on a Long Short-Term Memory (LSTM) network and a pre-trained waveform generator based on an LSTM network are used to accurately identify and predict the labels of the pressure bands. This not only achieves efficient label-based decoding but also avoids the high computational overhead of end-to-end decoding of the entire pressure wave signal. This application achieves high-precision recognition and efficient decoding of complex pressure wave signals, with significant advantages such as high recognition accuracy, strong robustness, and excellent decoding efficiency. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating an LSTM-based intelligent sliding sleeve pressure wave signal identification method in one embodiment of this application.

[0017] Figure 2 A schematic diagram of an intelligent sliding sleeve device model in an LSTM-based intelligent sliding sleeve pressure wave signal identification method provided in an embodiment of this application;

[0018] Figure 3 for Figure 2 A schematic diagram of the tee in the model of the intelligent sliding sleeve device;

[0019] Figure 4 for Figure 2 A schematic diagram of the throttle valve in the model of the intelligent sliding sleeve device;

[0020] Figure 5 A schematic diagram of the process for obtaining pressure wave signal tags in an LSTM-based intelligent sliding sleeve pressure wave signal identification method provided in another embodiment of this application;

[0021] Figure 6 A training diagram of an edge classifier based on a long short-term memory network and a waveform generator based on a long short-term memory network in an LSTM-based intelligent sliding sleeve pressure wave signal recognition method provided in another embodiment of this application;

[0022] Figure 7A schematic diagram illustrating the propagation of pressure wave signals using an intelligent sliding sleeve device model in another embodiment of this application, which is a method for identifying intelligent sliding sleeve pressure wave signals based on LSTM.

[0023] Figure 8 In another embodiment of this application, a method for identifying intelligent sliding sleeve pressure wave signals based on LSTM is provided, showing the throttle valve opening change diagram and the wellbore pressure wave signal change with the throttle valve opening when the control command is "01111000".

[0024] Figure 9 A schematic diagram illustrating the insertion of blank bands between adjacent evacuation pressure wave signals in an LSTM-based intelligent sliding sleeve pressure wave signal identification method provided in another embodiment of this application;

[0025] Figure 10 A schematic diagram of the training results of an edge classifier based on a long short-term memory network in an LSTM-based intelligent sliding sleeve pressure wave signal recognition method provided in another embodiment of this application;

[0026] Figure 11 A schematic diagram of the training results of a waveform generator based on a long short-term memory network in an LSTM-based intelligent sliding sleeve pressure wave signal recognition method provided in another embodiment of this application;

[0027] Figure 12 This is a schematic diagram of the functional modules of an LSTM-based intelligent sliding sleeve pressure wave signal recognition system in one embodiment of this application;

[0028] Figure 13 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application.

[0029] Figure label:

[0030] Pressure pump-1, throttle valve-2, wellbore-3, bottom hole-4. Detailed Implementation

[0031] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] To make the objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0033] The Long Short-Term Memory Network (LSTM) is a special type of Recurrent Neural Network (RNN) structure specifically designed to handle and predict long-term dependencies in time series data. Traditional RNNs tend to suffer from the problem of weakened input influence when dealing with long sequences, while LSTM effectively retains important historical information and suppresses interference from irrelevant information by introducing gating mechanisms (including input gates, forget gates, and output gates). This structure enables LSTM to perform excellently in tasks such as speech recognition, natural language processing, and industrial time series signal modeling.

[0034] In a non-cabled measurement-while-drilling system, downhole information is often transmitted and encoded in the form of pressure waves. Identifying key edge features (such as rising edges and falling edges) in mud pressure wave signals is of great significance for achieving high-precision signal decoding. However, in actual working conditions, pressure wave signals have obvious non-linear, non-stationary, and strong noise background characteristics, making it difficult for traditional feature engineering or classification methods based on static statistics to accurately capture their dynamic change trends.

[0035] Therefore, this application introduces the Long Short-Term Memory (LSTM) network to model and identify pressure wave signals. Compared with traditional methods, LSTM can make full use of historical pressure wave information and automatically learn potential pattern differences during the signal change process, thereby achieving accurate identification of rising edges and falling edges.

[0036] In an exemplary embodiment, as Figure 1 shown, a method for identifying intelligent sliding sleeve pressure wave signals based on LSTM is provided. This method is executed by a computer device, which can be specifically executed by a computer device such as a terminal or a server alone, or jointly executed by a terminal and a server, and includes the following steps 101 to step 105. Among them:

[0037] Step 101, establish an intelligent sliding sleeve device model; the intelligent sliding sleeve device model includes a pressure pump 1 and a wellbore 3 connected by a pipeline.

[0038] Step 102, use the pressure pump 1 to provide mud flow, obtain the pressure wave signal in the wellbore 3, and divide the pressure wave signal into N pressure wave bands; the N pressure wave bands include the first pressure wave band to the Nth pressure wave band; 1 < N. In this application, a sliding window is used to divide the pressure wave signal into N pressure wave bands.

[0039] Step 103, input the first pressure wave band into a trained edge classifier based on the long short-term memory network to obtain the label of the first pressure wave band; the label is "1" or "0"; where "1" represents a falling edge and "0" represents a rising edge.

[0040] Step 104: For the nth pressure band, input the preset label and the (n-1)th pressure band into the trained waveform generator based on a long short-term memory network to obtain the label for the nth pressure band; <n≤N。

[0041] Step 105: Identify the pressure wave signal based on the tags of the first pressure band and the tags of the second to Nth pressure bands.

[0042] By implementing steps 101 to 105 above, tag-based efficient decoding is achieved through tag identification and prediction for each band, avoiding the high computational overhead of end-to-end decoding of the entire signal. Furthermore, the decoded pressure wave signal can be used to control the intelligent sliding sleeve device.

[0043] In step 101, the intelligent sliding sleeve device model further includes a tee and a throttle valve 2; the tee is connected to the pressure pump 1, the throttle valve 2, and the wellbore 3 respectively via pipes, such as... Figure 2 As shown, the tee is connected to the pressure pump 1, the choke valve 2, and the bottom hole 4 of the wellbore 3, respectively. The pressure pump 1 provides high-pressure mud flow, which serves as a carrier of pressure wave signals; the tee divides the high-pressure flow into a main channel and a signal modulation channel; the choke valve 2 controls the flow rate to generate continuous pressure wave signals (opening / closing creates pressure fluctuations); the wellbore 3 serves as a channel for the pressure wave signals to be transmitted to the bottom hole 4 via the drill string or to return from the bottom hole 4; the bottom hole 4 is the installation location for signal modulation devices or measuring instruments, determines the final endpoint of the pressure wave signal propagation, and in this application serves as the location for pressure wave signal sampling and decoding.

[0044] Regarding the model of the intelligent sliding sleeve device, this application obtains the initial values ​​of the boundary conditions for pressure pump 1, three-way valve, and throttle valve 2 through steady-state equations. Ignoring the influence of the piston stroke of pressure pump 1, the displacement of pressure pump 1 can be described as follows: .in, This refers to the displacement of pressure pump 1. This is the baseline displacement.

[0045] The tee is installed between the pressure pump 1, the throttle valve 2, and the bottom 4 of the wellbore 3, and is connected by a pipe. As the name suggests, the tee has three flow paths, as follows: Figure 3 As shown: Under ideal conditions, the consistency equation for a tee is: ; .in, P1 represents the flow rate in the pipeline where pressure pump 1 is located, and P1 represents the pressure in the pipeline where pressure pump 1 is located. P2 represents the flow rate in the pipe where throttle valve 2 is located, and P2 represents the pressure in the pipe where throttle valve 2 is located. P3 represents the flow rate at wellbore 3, and P3 represents the pressure in the pipeline where throttle valve 2 is located.

[0046] When throttle valve 2 is applied, the link boundary is as follows: Figure 4 As shown, the rule governing the opening change of throttle valve 2 can be described as follows: ;in, It is the opening ratio of throttle valve 2. t It is time. t c This is the time required for throttle valve 2 to close completely. m It is the opening index of throttle valve 2.

[0047] The pressure loss at throttle valve 2 is related to the degree of opening and closing of throttle valve 2, and can be expressed as follows:

[0048] .

[0049] in, Input flow rate to throttle valve 2. The input and output flow rates of throttle valve 2 For the flow rate through throttle valve 2, The initial flow rate of throttle valve 2, The initial pressure of throttle valve 2, The inlet pressure of throttle valve 2 This is the outlet pressure of throttle valve 2.

[0050] In another exemplary embodiment of this application, after dividing the pressure wave signal into N pressure bands, the method further includes: normalizing any one of the pressure bands using the following formula:

[0051] .

[0052] in, For any pressure band, This represents the average value of the pressure band. The standard deviation of the pressure band. It is a tiny constant. This is the normalized pressure band.

[0053] In another exemplary embodiment of this application, to decode the pressure wave signal in the intelligent sliding sleeve device model, an edge classifier based on a long short-term memory network is constructed. This edge classifier takes the first pressure band as input and determines the edge type (rising edge, falling edge, or blank band) corresponding to the first pressure band based on its features, thereby achieving accurate identification of the pressure wave signal. Compared to traditional feature engineering methods, the edge classifier based on a long short-term memory network has significant advantages in modeling nonlinear dynamic behavior and capturing long-term dependencies.

[0054] Meanwhile, to simulate the actual pressure response in the intelligent sliding sleeve device, a waveform generator based on a long short-term memory network was further constructed. This waveform generator uses historical pressure bands and preset labels as dual inputs to learn the evolution patterns of different types of pressure bands in the wellbore 3, and outputs the pressure band corresponding to the preset label. Step 104 specifically includes:

[0055] For the nth pressure band, the preset label and the (n-1)th pressure band are input into the trained waveform generator based on the long short-term memory network to obtain the rising edge pressure band and the falling edge pressure band; the preset label includes "1" and "0".

[0056] The residual values ​​of the rising pressure band and the falling pressure band are calculated with the nth pressure band to obtain the residual values ​​of the rising pressure band and the falling pressure band.

[0057] Determine the minimum value between the residual value of the rising edge pressure band and the residual value of the falling edge pressure band, and use the label corresponding to the minimum value as the label of the nth pressure band.

[0058] A complete pressure wave signal contains eight pressure bands, corresponding to eight labels. When predicting the labels corresponding to the pressure bands using the waveform generator, the first pressure band cannot be obtained from historical information and can only be classified using the edge classifier to obtain its label; the labels of the remaining bands are predicted by the waveform generator. The specific process is as follows: Figure 5 As shown, this method can be used to obtain the tag value corresponding to the complete pressure wave signal, thus achieving the purpose of decoding the time sequence signal.

[0059] In another exemplary embodiment of this application, before acquiring the pressure wave signal in the wellbore 3, the method further includes: generating multiple control commands using a random function.

[0060] The throttle valve 2 is switched on and off sequentially according to multiple control commands to obtain multiple pressure wave signals within the wellbore 3.

[0061] Multiple pressure wave signals within the wellbore 3 are used as a sample dataset. This sample dataset is then used to train the edge classifier and waveform generator based on a long short-term memory (LSTM) network, resulting in the trained edge classifier and waveform generator. Figure 6 As shown.

[0062] The control commands employ zero-one control commands. To ensure good timing synchronization during transmission, signal transmission is based on Return-to-Zero (RZ) encoding. RZ encoding changes the signal level during each bit transmission, avoiding the possibility of prolonged continuous identical signal levels seen in traditional encoding methods, thus solving the clock synchronization problem. This encoding method exhibits strong anti-interference capabilities in noisy environments, ensuring that control commands are accurately and reliably transmitted to the receiving end.

[0063] like Figure 7 As shown, a smart sliding sleeve device model is used to simulate the propagation process of pressure wave signals. First, a control command is randomly generated using a random function, consisting of 8 bits of 0s and 1s. Each "0" or "1" represents a change in the opening of throttle valve 2: "0" corresponds to the closed state of throttle valve 2, while "1" corresponds to the open state. The default state of throttle valve 2 is half-open. By controlling the change in the opening of throttle valve 2 through the control command, the pressure inside wellbore 3 can be adjusted. Increasing the opening of throttle valve 2 will cause the amplitude of the pressure wave signal inside wellbore 3 to decrease, appearing as a falling edge within a unit symbol length time; conversely, decreasing the opening of throttle valve 2 will cause the amplitude of the pressure wave signal inside wellbore 3 to increase, appearing as a rising edge within a unit symbol length time. For example, when the control command is set to "1", throttle valve 2 is in its maximum open state, with an opening degree of 0.7; when the control command is "0", throttle valve 2 is in its minimum open state, with an opening degree of 0.3; and when there is no signal input, it is in a half-open state, with an opening degree of 0.5. Figure 8 As shown, when the control command is "01111000", the original code length is 30s, and the noise intensity is -20dB, the diagram shows the change in the opening of the throttle valve 2 and the change in the pressure wave signal in the wellbore 3 with the opening of the throttle valve 2. It can be clearly seen that the opening of the throttle valve 2 changes accurately with the input of the command signal.

[0064] The edge classifier and waveform generator based on the Long Short-Term Memory (LSTM) network are trained using the sample dataset to obtain the trained edge classifier and waveform generator based on the LSM network, specifically including:

[0065] Each pressure wave signal in the sample dataset is thinned to obtain multiple thinned pressure wave signals.

[0066] A blank band is inserted between adjacent dilution pressure wave signals, as follows: Figure 9 As shown, a sample dataset including the dilution pressure wave signal and the blank band was obtained.

[0067] Using a sample dataset including slack pressure wave signals and blank bands, the edge classifier and waveform generator based on the long short-term memory network are trained to obtain the trained edge classifier and waveform generator based on the long short-term memory network.

[0068] Because a single pressure wave signal contains a large amount of pressure wave data, it undergoes thinning. Simultaneously, by inserting blank bands between the thinned pressure wave signals, each complete thinned pressure wave signal is connected together using blank bands. This simulates the continuous transmission of commands in real-world scenarios and facilitates data extraction using a sliding window for training the edge classifier and waveform generator. When extracting data using the sliding window, blank bands are automatically skipped upon encountering them, and are not used for training.

[0069] In another exemplary embodiment of this application, in order to evaluate the accuracy and effectiveness of the trained edge classifier based on a long short-term memory network and the trained waveform generator based on a long short-term memory network, squared error (MSE) and coefficient of determination (R²) are used. 2 The performance of the edge classifier based on the long short-term memory network and the trained waveform generator based on the long short-term memory network are evaluated using four metrics: root mean square error (RMSE), mean absolute error (MAE), etc.

[0070] .

[0071] .

[0072] .

[0073] .

[0074] in, This represents the total number of pressure bands. This refers to the pressure waveband number. This is the predicted value for the i-th pressure band. This represents the true value of the i-th pressure band. The mean value of the pressure band is the actual value. MSE measures the "energy" of the prediction error; RMSE is the square root of the MSE result, more intuitively reflecting the typical magnitude of the "mean prediction deviation"; the coefficient of determination R... 2 R represents the proportion of the variance in the real data explained by the model. 2 When the value is 1, the prediction is completely correct; MAE calculates the absolute average of the errors at all points, reflecting the magnitude of the "average deviation".

[0075] In another exemplary embodiment of this application, the training parameters of the edge classifier based on the Long Short-Term Memory network are set as follows: the unit symbol length corresponding to a pressure wave signal after thinning is 20s, the number of pressure wave data points corresponding to the unit symbol is 125, the sliding window is selected as the length of a complete symbol segment, and the sliding step size is 125. The trained edge classifier is used for label judgment, and the result is as follows. Figure 10 As shown, the edge classifier determines the label of rising or falling edge based on the change of pressure wave. It can clearly distinguish the rising or falling of the waveform and obtain the predicted label. By comparing the predicted label with the real label, it can be clearly seen that the classifier's judgment is very accurate and accurately skips the number of splicing segments.

[0076] The training parameters for the waveform generator based on a Long Short-Term Memory (LSTM) network are set as follows: the unit symbol length corresponding to a pressure wave signal after thinning is 20 seconds, the number of pressure wave data points corresponding to the unit symbol is 125, the sliding window is selected as the length of a complete symbol segment, and the sliding step size is 125. The results are as follows. Figure 11 As shown in the first subplot, the predicted curve accurately tracks the overall upward and downward trend of the true curve, with only a slight lag at minor high-frequency fluctuations. The second subplot is the residual time series plot, where the prediction error is mainly distributed within ±0.5 and does not exhibit systematic drift over time. The third subplot is the probability distribution histogram, which is approximately symmetrical with peaks concentrated in the ±0.2 range, indicating that the error is random and without significant bias. In terms of quantitative metrics, the waveform generator based on the Long Short-Term Memory network achieved MSE=0.1111, MAE=0.2701, RMSE=0.3333, and R0.05 in the test segment. 2 =0.888, indicating that the waveform generator based on the long short-term memory network can explain about 89% of the fluctuation variance, further verifying the efficient fitting capability of the pressure wave signal dynamics.

[0077] Based on the same inventive concept, such as Figure 12 As shown in the figure, this application embodiment also provides an intelligent sliding sleeve pressure wave signal identification system based on LSTM, the system comprising:

[0078] The intelligent sliding sleeve device model building module 120 is used to build an intelligent sliding sleeve device model; the intelligent sliding sleeve device model includes a pressure pump 1 and a well barrel 3 connected by a pipeline.

[0079] Pressure wave acquisition module 121 is used to acquire pressure wave signals in the wellbore 3 by using the mud flow provided by the pressure pump 1, and to divide the pressure wave signals into N pressure wave bands; the N pressure wave bands include the first pressure wave band to the Nth pressure wave band; 1 <N。

[0080] The first pressure band label acquisition module 122 is used to input the first pressure band into a trained edge classifier based on a long short-term memory network to obtain the label of the first pressure band; the label is "1" or "0"; "1" represents a falling edge; and "0" represents a rising edge.

[0081] The nth pressure band label acquisition module 123 is used to input a preset label and the (n-1)th pressure band into a trained waveform generator based on a long short-term memory network to obtain the label of the nth pressure band; <n≤N。

[0082] The pressure wave signal identification module 124 is used to identify the pressure wave signal based on the label of the first pressure wave band and the labels of the second to Nth pressure wave bands.

[0083] This application addresses the problem of pressure wave signal identification and prediction in intelligent sliding sleeve communication, constructing an edge classifier and a waveform generator based on a Long Short-Term Memory (LSTM) network. First, the propagation process of continuous pressure wave signals caused by changes in the opening of throttle valve 2 is simulated, generating a high-quality sample dataset suitable for deep learning. For pressure wave signal identification, a sliding window extraction strategy and a Z-score normalization scheme are designed, and classification learning is performed based on the rising and falling edges of the LSM network edge classifier. The classifier achieves 100% accuracy on the validation set. For pressure wave signal generation and prediction, a waveform generator based on an LSM network is constructed, achieving recursive identification of subsequent labels through a minimum residual decision method.

[0084] In summary, this application demonstrates outstanding performance in both accuracy and robustness, possesses significant engineering application value, and can provide reliable data support and algorithmic foundation for non-cable drilling measurement and remote sliding sleeve control.

[0085] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 13As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores pressure wave signals. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements an LSTM-based intelligent sliding sleeve pressure wave signal recognition method.

[0086] Those skilled in the art will understand that Figure 13 The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0087] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0088] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0089] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for identifying intelligent sliding sleeve pressure wave signals based on LSTM, characterized in that, The method includes: A model of an intelligent sliding sleeve device is established; the intelligent sliding sleeve device model includes a pressure pump and a wellbore connected by pipelines. The pressure pump provides mud flow, pressure wave signals are acquired in the wellbore, and the pressure wave signals are divided into N pressure bands; the N pressure bands include the first pressure band to the Nth pressure band; 1 <N; The first pressure band is input into a trained edge classifier based on a long short-term memory network to obtain a label for the first pressure band; the label is "1" or "0"; where "1" represents a falling edge and "0" represents a rising edge. For the nth pressure band, the preset label and the (n-1)th pressure band are input into the trained waveform generator based on a long short-term memory network to obtain the label for the nth pressure band; 1 <n≤N; The pressure wave signal is identified based on the tags of the first pressure band and the tags of the second to Nth pressure bands; For the nth pressure band, the preset label and the (n-1)th pressure band are input into the trained waveform generator based on a long short-term memory network to obtain the label for the nth pressure band, specifically including: For the nth pressure band, the preset label and the (n-1)th pressure band are input into the trained waveform generator based on the long short-term memory network to obtain the rising edge pressure band and the falling edge pressure band; the preset label includes "1" and "0"; The residual values ​​of the rising pressure band and the falling pressure band are calculated with the nth pressure band to obtain the residual values ​​of the rising pressure band and the falling pressure band. Determine the minimum value between the residual value of the rising edge pressure band and the residual value of the falling edge pressure band, and use the label corresponding to the minimum value as the label of the nth pressure band.

2. The LSTM-based intelligent sliding sleeve pressure wave signal identification method according to claim 1, characterized in that, After dividing the pressure wave signal into N pressure wave bands, the method further includes: The following formula is used to normalize each pressure band: ； in, For pressure band, This represents the average value of the pressure band. The standard deviation of the pressure band. It is a tiny constant. This is the normalized pressure band.

3. The LSTM-based intelligent sliding sleeve pressure wave signal identification method according to claim 1, characterized in that, The intelligent sliding sleeve device model also includes a tee and a throttle valve; the tee is connected to the pressure pump, the throttle valve, and the wellbore via pipes; before acquiring the pressure wave signal in the wellbore, it also includes: Multiple control commands are generated using a random function; The throttle valve is controlled sequentially according to multiple control commands to obtain multiple pressure wave signals within the wellbore; Multiple pressure wave signals within the wellbore are used as a sample dataset. The edge classifier and waveform generator based on the long short-term memory network are trained using the sample dataset to obtain the trained edge classifier and waveform generator based on the long short-term memory network.

4. The LSTM-based intelligent sliding sleeve pressure wave signal identification method according to claim 3, characterized in that, The edge classifier and waveform generator based on the Long Short-Term Memory (LSTM) network are trained using the sample dataset to obtain the trained edge classifier and waveform generator based on the LSM network, specifically including: Each pressure wave signal in the sample dataset is thinned to obtain multiple thinned pressure wave signals. By inserting blank bands between adjacent dilution pressure wave signals, a sample dataset including dilution pressure wave signals and blank bands is obtained. Using a sample dataset including slack pressure wave signals and blank bands, the edge classifier and waveform generator based on the long short-term memory network are trained to obtain the trained edge classifier and waveform generator based on the long short-term memory network.

Images