Cable channel pressure change monitoring system

Through variational autoencoder and graph convolution network technology, the problem of insufficient multi-dimensional and nonlinear data processing in the existing technology is solved, efficient analysis and data reconstruction of complex data environments are realized, and the system is significantly improved Accuracy and reliability.

CN118913378AInactive Publication Date: 2024-11-08STATE GRID SHANXI ELECTRIC POWER COMPANY TAIYUAN POWER SUPPLY COMPANY
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411412219.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-11
Publication Date
2024-11-08
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing pressure change monitoring technology has shortcomings in processing multi-dimensional and nonlinear data, especially in the precise analysis of complex data structures and dynamic changes, resulting in limited prediction and diagnostic accuracy, and lack of effective data recovery mechanisms, which increases system operation. risk.

Method used

Using variational autoencoder and graph convolution network technology, the processing and analysis of complex data is realized through potential spatial data conversion and dynamic graph structure update, and data reconstruction and repair are used to improve the accuracy of the system. Sex and reliability.

Benefits of technology

It significantly improves the processing capability of complex data environments, enhances the ability to capture and analyze dynamic changes in pressure distribution, improves the accuracy and reliability of the system, and performs outstandingly in preventing faults and identifying potential problems early.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118913378A_ABST
    Figure CN118913378A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of pressure change monitoring, in particular to a cable channel pressure change monitoring system which comprises a cable channel pressure sensing coding module, a cable channel pressure sensing coding module, a cable channel pressure sensing coding module, a cable channel pressure sensing coding module, a cable channel pressure sensing coding module, a cable channel pressure sensing coding module, a cable channel pressure sensing coding module, a cable channel pressure sensing coding module, a cable channel pressure sensing coding module, a cable channel pressure sensing coding module and a cable channel pressure sensing coding module, and obtaining a standardized potential vector. In the invention, through potential spatial data conversion and a graph convolutional network technology, the monitoring system realizes significant improvement on processing of a complex data environment, not only optimizes vectorization processing of data, but also improves capture and analysis capability on dynamic change of pressure distribution through dynamic graph structure updating, and therefore, the monitoring system can be used for monitoring the dynamic change of pressure distribution. The variational auto-encoder is used for reconstructing and repairing the data, so that the coping capacity of the system under the condition that the data is incomplete or wrong is further enhanced, the accuracy and reliability of the system are remarkably improved, and particularly, the system has outstanding performance in the aspects of preventing faults and recognizing potential problems in advance.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of pressure change monitoring, and in particular to a cable channel pressure change monitoring system. Background Art

[0002] The field of pressure change monitoring technology involves the development and application of methods and devices that can monitor and record pressure changes in a system, equipment or environment in real time. This technology is widely used in various fields, including industrial manufacturing, energy production, environmental monitoring, etc. By using pressure sensors or other sensor technologies, pressure changes can be monitored in real time and the data can be transmitted to the monitoring system for analysis and processing. The main purpose of pressure change monitoring technology is to provide timely and accurate pressure information in order to detect abnormal conditions in time, predict equipment failures, optimize system operation, and ensure the safe and stable operation of the production process.

[0003] Among them, the cable channel pressure change monitoring system is a monitoring system applied to cable channels or cable pipes, which is designed to monitor and record the pressure changes inside the channel in real time. Its main purpose is to ensure that the pressure inside the cable channel is within a safe range, detect pressure anomalies in time, and prevent equipment damage or accidents caused by pressure changes. This system usually uses technologies such as pressure sensors, which can continuously monitor pressure and provide real-time monitoring results and alarm information through data transmission and processing systems, thereby improving the safety and reliability of the cable channel.

[0004] Although existing pressure change monitoring technologies have real-time monitoring capabilities, their capabilities are limited when processing multi-dimensional and nonlinear data, especially in the accurate analysis of complex data structures and dynamic changes. These technologies are often unable to effectively process and analyze data from multiple sources, resulting in limitations in prediction and diagnostic accuracy. This limitation is more obvious when data is lost or damaged. The lack of an effective data recovery mechanism may lead to inaccurate monitoring results and increase system operation risks. For example, in environmental monitoring or energy production, insufficient data analysis may miss key warning information, affecting the stable operation and safety of the entire system. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose a cable channel pressure change monitoring system.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: A cable channel pressure change monitoring system comprises: The cable channel pressure sensing encoding module collects the pressure readings and temperature data of the pressure cover plate and the pressure-resistant pipe fittings, calls the encoder part of the variational autoencoder, converts the data into a vector representation of the latent space, obtains a standardized latent vector, and uses the standardized latent vector to encode the pressure state of the seal and the bracket to obtain the node state encoding result; The cable channel spatial feature analysis module constructs a graph structure based on the node state encoding result, connects the pressure transmission path between compensators through edges, executes the graph convolution network to update the node state, obtains the updated node feature, analyzes the spatial distribution of pressure on the fixing parts, and obtains the spatial correlation analysis result; The cable channel anomaly identification module calculates the deviation between the pressure reading of each bracket node in the cable channel and the expected model based on the spatial correlation analysis result, compares the deviation with the threshold, identifies the compensator node with the deviation exceeding the threshold, analyzes the spatial diffusion trend of the node, and obtains the anomaly detection result; The cable channel data integrity reconstruction module determines the data reconstruction area of ​​the pressure-resistant pipe based on the abnormal detection results, uses the decoder part of the variational autoencoder to generate estimated pressure data from the latent space, fills in the seal data gaps or repairs errors, and obtains the reconstructed pressure data.

[0007] As a further solution of the present invention, the step of obtaining the standardized latent vector is specifically as follows: Collect the pressure readings and temperature data of the pressure cover and pressure-resistant pipe fittings, using the formula: ; ; The pressure and temperature are range normalized respectively to obtain standardized data sets of pressure and temperature; in, represents the raw pressure reading, represents the original temperature data, , Standardized data representing pressure and temperature; The normalized data set of pressure and temperature is input into the encoder part of the variational autoencoder using the formula: ; Generate latent vector; in, represent and A collection of represents the weight matrix of the encoder, represents the encoder bias vector, represents the generated latent vector; Using the latent vector, the formula is adopted: ; Normalize to generate a standardized latent vector; in, represents the modulus of the latent vector, Represents the generated normalized latent vector.

[0008] As a further solution of the present invention, the step of obtaining the node status encoding result is specifically: Using the normalized latent vector, the pressure state of the seal and the bracket is encoded using the formula: ; Use the square root modulus length method to improve data stability and obtain preliminary state encoding; in, represents the normalized latent vector, represents the weight matrix of the encoding network, represents the bias vector of the encoding network, Represents the initial state encoding generated; The preliminary state encoding is optimized using the formula: ; Adjust the size and stability of the encoding vector to obtain an optimized state encoding; in, represents the modulus length of the state encoding, , is a parameter used to adjust the encoding scale and stability. Represents the optimized state code; Apply non-linear activation function The optimized state code is processed using the formula: ; Get the node status encoding result; in, represents the activation function, is the nonlinear adjustment coefficient, Represents the node status encoding result.

[0009] As a further solution of the present invention, the step of obtaining the spatial association analysis result is specifically: Starting from the node state encoding result, a graph structure is constructed by connecting the pressure transmission paths between compensators, using the formula: ; Generate a pressure transmission path diagram model; in, Representation Node and The pressure transmission efficiency between , Adjust the decay rate and distance sensitivity of pressure transmission, Represents the physical distance between nodes, , Indicates the state code of the node; The graph convolutional network is used to update the node status on the pressure transfer graph model, using the formula: ; Get updated node features; in, represents the updated node features, represents the original node feature of the kth neighbor, , are graph convolutional network parameters, is the activation function; Based on the updated node characteristics, the spatial distribution of pressure on the fixture is analyzed using the formula: ; Calculate the weighted sum of node features in the entire graph structure to obtain the spatial correlation analysis results; in, For location The weight of Represents the results of spatial association analysis.

[0010] As a further solution of the present invention, the step of obtaining the abnormality detection result is specifically: Starting from the results of the spatial correlation analysis, the deviation of the pressure reading of each bracket node in the cable channel from the expected model is calculated using the formula: ; Get the pressure deviation of each node; in, represents the pressure deviation at node i, is the actual pressure reading, is the pressure value of the expected model, Small positive numbers to avoid division by zero; Compare the pressure deviation of each node with the threshold, using the formula: ; Obtain node identifications where deviation exceeds threshold; in, is the deviation threshold, represents the deviation sensitivity, Indicates the probability of whether node i is abnormal; To analyze the spatial diffusion trend in node identifications with deviations exceeding the threshold, the formula is used: ; Generate anomaly detection results; in, represents the weight of the node, Representation Node Location, is the smoothing parameter, Trend results indicating anomalous diffusion.

[0011] As a further solution of the present invention, the step of obtaining the reconstructed pressure data is specifically as follows: Based on the abnormal detection results, the data reconstruction area of ​​the pressure-resistant pipe fittings is determined using the formula: ; Generate data reconstruction area; in, Represents the data reconstruction area of ​​the node, Represents the set of neighbor nodes of a node, is the influence coefficient, represents the interaction strength between nodes, is the nonlinear adjustment parameter; Based on the data reconstruction area, the decoder part of the variational autoencoder is used, using the formula: ; generating estimated pressure data; in, Represents the estimated pressure data of the node, represents the representation of the latent space, is the decoder parameter set, is the noise factor; The output of the estimated pressure data is combined with the actual pressure data to reconstruct the data, fill in the gaps in the seal data or repair the errors, using the formula: ; Obtain reconstructed pressure data; in, represents the reconstructed pressure data, is the actual pressure data, is the balance coefficient, is the error correction term.

[0012] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, through latent space data conversion and graph convolutional network technology, the monitoring system has achieved significant improvement in processing complex data environments. Not only has the vectorization of data been optimized, but also the ability to capture and analyze dynamic changes in pressure distribution has been improved through dynamic graph structure updates. The use of variational autoencoders for data reconstruction and repair further enhances the system's ability to cope with incomplete or erroneous data, thereby significantly improving the accuracy and reliability of the system, especially in preventing faults and identifying potential problems early. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is a system flow chart of the present invention; Figure 2 A flowchart of the steps for obtaining a standardized latent vector according to the present invention; Figure 3 A flowchart of the steps for obtaining the node status encoding result of the present invention; Figure 4 A flowchart of the steps for obtaining the spatial correlation analysis results of the present invention; Figure 5 A flowchart of the steps for obtaining abnormality detection results of the present invention; Figure 6 This is a flow chart of the steps for obtaining the reconstructed pressure data of the present invention. DETAILED DESCRIPTION

[0014] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0015] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0016] Example 1: Please refer to Figure 1 , a cable channel pressure change monitoring system comprises: The cable channel pressure sensing encoding module collects the pressure readings and temperature data of the pressure cover and the pressure-resistant pipe, calls the encoder part of the variational autoencoder, converts the data into a vector representation of the latent space, obtains a standardized latent vector, and uses the standardized latent vector to encode the pressure state of the seal and the bracket to obtain the node state encoding result; The cable channel spatial feature analysis module builds a graph structure based on the node state encoding results, connects the pressure transfer paths between compensators through edges, executes the graph convolution network to update the node state, obtains the updated node features, analyzes the spatial distribution of pressure on the fixtures, and obtains the spatial correlation analysis results; The cable channel anomaly identification module calculates the deviation between the pressure reading of each bracket node in the cable channel and the expected model based on the spatial correlation analysis results, compares the deviation with the threshold, identifies the compensator nodes with deviations exceeding the threshold, analyzes the spatial diffusion trend of the nodes, and obtains the anomaly detection results; The cable channel data integrity reconstruction module determines the data reconstruction area of ​​the pressure-resistant pipe based on the anomaly detection results, uses the decoder part of the variational autoencoder to generate estimated pressure data from the latent space, fills in the seal data gaps or repairs errors, and obtains the reconstructed pressure data.

[0017] The node state encoding results are specifically standardized vectors and pressure state encodings. The spatial correlation analysis results include the update characteristics of the nodes and the spatial distribution of pressure. The anomaly detection results are specifically deviation quantification, abnormal node identification, and diffusion trend analysis. The reconstructed pressure data includes estimated pressure values, data filling, and error repair.

[0018] See also Figure 2 , the specific steps for obtaining the standardized potential vector are: Collect the pressure readings and temperature data of the pressure cover and pressure-resistant pipe fittings, using the formula: ; ; The pressure and temperature are range normalized respectively to obtain standardized data sets of pressure and temperature; in, represents the raw pressure reading, represents the original temperature data, , Standardized data representing pressure and temperature; The normalized data set of pressure and temperature is input into the encoder part of the variational autoencoder using the formula: ; Generate latent vector; in, represent and A collection of represents the weight matrix of the encoder, represents the encoder bias vector, represents the generated latent vector; Using the latent vector, the formula is: ; Normalize to generate a standardized latent vector; in, represents the modulus of the latent vector, Represents the generated normalized latent vector.

[0019] Step 1: Data normalization; Collecting pressure readings and temperature data Assume that the following sample data is obtained from the pressure cover and pressure-resistant pipe fittings: Pressure readings : The unit is ; Temperature readings : The unit is Kelvin; formula: ; Parameter calculation: : Normalized pressure; : Normalized temperature; , are the minimum and maximum values ​​of the pressure data respectively.

[0020] , are the minimum and maximum values ​​of the temperature data respectively.

[0021] Calculation example: For pressure data: ; ; for Standardization: ; For temperature data: ; ; for Standardization: ; Step 2: Latent vector generation; formula: ; Parameter calculation: : Encoder weight matrix, assumed to be ; : Encoder bias, assumed to be ; : Input standardized data set ; Calculation example: ; Normalized latent vector Calculation: (unit vector); Step 3: Normalize the latent vector; formula: ; Parameter calculation: : The latent vector obtained in step 2; Calculation example: because is already a unit vector, and its modulus is 1, so Unchanged, still

[0022] See also Figure 3 , the specific steps for obtaining the node status encoding result are: The normalized latent vector is used to encode the pressure state of the seal and the bracket using the formula: ; Use the square root modulus length method to improve data stability and obtain preliminary state encoding; in, represents the normalized latent vector, represents the weight matrix of the encoding network, represents the bias vector of the encoding network, Represents the initial state encoding generated; The preliminary state encoding is optimized using the formula: ; Adjust the size and stability of the encoding vector to obtain an optimized state encoding; in, represents the modulus length of the state encoding, , is a parameter used to adjust the encoding scale and stability. Represents the optimized state code; Apply non-linear activation function The optimized state encoding is processed using the formula: ; Get the node status encoding result; in, represents the activation function, is the nonlinear adjustment coefficient, Represents the node status encoding result.

[0023] Detailed explanation and derivation of the formula in step 1; formula: ; Parameter explanation: : The weight matrix of the encoding network.

[0024] : Bias vector for the encoding network.

[0025] : The normalized latent vector of the input.

[0026] c: Generated preliminary state encoding.

[0027] Derivation process: Assumptions: is a Matrix, assuming its elements are .

[0028] is a vector, assuming its elements are .

[0029] is a vector, assuming its elements are .

[0030] calculate : ; calculate : ; Calculate the modulus length: ; final : ; What the results mean: It is a coded representation of the pressure state of the seal and the bracket, where the numerical value indicates the relative strength and direction of the code.

[0031] Detailed explanation and derivation of the formula in step 2; formula: ; Parameter explanation: : The preliminary state code calculated in step 1.

[0032] , : Tuning parameter used to adjust the size and stability of the encoding vector.

[0033] : Optimized status encoding.

[0034] Assume setup: and ,These parameters are adjusted experimentally to find the best performance.

[0035] Calculation process: Known From step 1: Calculate the module length : ; calculate : ; What the results mean: Represents an optimized state encoding. A larger value reflects the amplification of the original encoding, making the features more prominent and easier to process later.

[0036] Detailed explanation and derivation of the formula in step 3; formula: ; Parameter explanation: : The optimized state code calculated in step 2.

[0037] : Nonlinear adjustment coefficient, selected to enhance the nonlinear characteristics of the encoding.

[0038] : The final node state encoding result.

[0039] : Activation function, the commonly used one is sigmoid function.

[0040] Assume setup: , for the sake of example, a common coefficient is chosen.

[0041] Calculation process: Known : Apply the sigmoid function: ; calculate : ; What the results mean: Represents the encoding result after processing by the activation function. Its value is between 0 and 1, reflecting the final form of the encoding vector after nonlinear transformation, which is suitable for downstream classification or other machine learning tasks.

[0042] See also Figure 4 , the specific steps for obtaining the results of spatial association analysis are: Starting from the node state encoding results, a graph structure is constructed by connecting the pressure transmission paths between compensators, using the formula: ; Generate a pressure transmission path diagram model; in, Representation Node and The pressure transmission efficiency between , Adjust the decay rate and distance sensitivity of pressure transmission, Represents the physical distance between nodes, , Indicates the state code of the node; The graph convolutional network is used to update the node status on the pressure transfer graph model using the formula: ; Get updated node features; in, represents the updated node features, represents the original node feature of the kth neighbor, , are graph convolutional network parameters, is the activation function; Based on the updated node characteristics, the spatial distribution of pressure on the fixture is analyzed using the formula: ; Calculate the weighted sum of node features in the entire graph structure to obtain the spatial correlation analysis results; in, For location The weight of Represents the results of spatial association analysis.

[0043] Step 1: Construction of pressure transmission path diagram model; formula: ; Parameter interpretation and derivation: : Pressure transmission efficiency between nodes i and j.

[0044] : Attenuation parameter, assumed to be 1.5, which is a coefficient that adjusts the overall pressure transmission intensity.

[0045] : Distance sensitivity parameter, assumed to be 0.1, which indicates the attenuation rate of pressure transmission efficiency with increasing distance.

[0046] : The physical distance between nodes i and j, assumed to be 5 units.

[0047] , : The state encoding value of nodes i and j, assuming and .

[0048] Calculation process and examples: Calculate the attenuation of the effect of distance between nodes on pressure: .

[0049] Compute the synthetic effect of node state encoding: .

[0050] The pressure transmission efficiency is obtained by combining the above calculations: .

[0051] Result description: represents node i and The pressure transfer efficiency between , which represents the efficiency of pressure transfer from node to node under a given state encoding and distance.

[0052] Step 2: The graph convolutional network updates the node status; formula: ; Parameter interpretation and derivation: , : They are the node features before and after the update respectively.

[0053] , : Weights and biases of the graph convolutional network, assuming the weight is 0.8 and the bias is 0.1.

[0054] : Activation function, the sigmoid function is used here.

[0055] Calculation process and examples: Assume there are 3 neighbor nodes, and the characteristic value of each node is .

[0056] Compute the sum of the products of the weights and the neighbor features: .

[0057] Add the bias and pass it through the activation function: .

[0058] Using the sigmoid activation function: .

[0059] Result description: It represents the feature value of the updated node, which is close to 1, showing a high activation level of the node state.

[0060] Step 3: Spatial correlation analysis; formula: ; Parameter interpretation and derivation: : Results of spatial association analysis.

[0061] : The updated features of the node at position x.

[0062] The weight of position x, assumed to be 1.1.

[0063] Calculation process and examples: Assume there are 3 nodes in the graph and the updated feature of each node is 0.999.

[0064] Compute the weighted sum of features: .

[0065] Result description: Represents the results of spatial correlation analysis of pressure distribution in the entire graph. This value reflects the overall activity of the overall node state. A higher value indicates a strong pressure correlation between nodes.

[0066] See also Figure 5 , the specific steps for obtaining anomaly detection results are: Starting from the results of the spatial correlation analysis, the deviation of the pressure readings of each bracket node in the cable channel from the expected model is calculated using the formula: ; Get the pressure deviation of each node; in, represents the pressure deviation at node i, is the actual pressure reading, is the pressure value of the expected model, Small positive numbers to avoid division by zero; Compare the pressure deviation of each node with the threshold, using the formula: ; Obtain node identifications where deviation exceeds threshold; in, is the deviation threshold, represents the deviation sensitivity, Indicates the probability of whether node i is abnormal; To analyze the spatial diffusion trend in node identifications with deviations exceeding the threshold, the formula is used: ; Generate anomaly detection results; in, represents the weight of the node, represents the location of the node, is the smoothing parameter, Trend results indicating anomalous diffusion.

[0067] Step 1: Calculate the pressure deviation ; formula: ; Parameter explanation and derivation process: : The actual pressure reading at the node.

[0068] : The pressure reading calculated by the node according to the expected model.

[0069] : A small positive number that avoids division by zero errors can take very small values, such as .

[0070] Assume the following example: Assumptions Unit pressure.

[0071] Assumptions Unit pressure.

[0072] set up .

[0073] Calculation process: Calculate the pressure difference: .

[0074] Calculate the denominator: .

[0075] calculate : .

[0076] Result interpretation: Pressure deviation The value of is 0.02398, which represents the standardized difference between the actual pressure and the expected pressure. This value is small, indicating that the pressure deviation is within the acceptable range.

[0077] Step 2: Compare the pressure deviation with the threshold and determine the abnormal node; formula: ; Parameter explanation and derivation process: : Pressure deviation obtained from step 1.

[0078] : Set the pressure deviation threshold, assumed to be 0.02.

[0079] : Pressure deviation sensitivity, assumed to be 0.005.

[0080] Calculation process: Calculate the exponential part: .

[0081] calculate : .

[0082] Result interpretation: The value is 0.689, indicating that the node is approximately This is a probability value indicating that the node has a certain risk of abnormality.

[0083] Step 3: Analyze the spatial diffusion trend of abnormal nodes; formula: ; Parameter explanation and derivation process: : The weight of the node, assuming that the weight of all nodes is 1.

[0084] : The position coordinates of the node, assuming the node is located at 10 units.

[0085] : Smoothing parameter, assumed to be 1.

[0086] Calculation process: For an abnormal node, calculate: .

[0087] calculate : .

[0088] Result interpretation: The value of is 10.05, which indicates the spatial diffusion trend of abnormal nodes. A larger value may indicate that the anomaly has a wider impact range.

[0089] See also Figure 6 ,The specific steps for obtaining the reconstructed pressure data are: Based on the anomaly detection results, the data reconstruction area of ​​the pressure-resistant pipe fittings is determined using the formula: ; Generate data reconstruction area; in, Represents the data reconstruction area of ​​the node, Represents the set of neighbor nodes of a node, is the influence coefficient, represents the interaction strength between nodes, is the nonlinear adjustment parameter; Based on the data reconstruction area, the decoder part of the variational autoencoder is used, using the formula: ; generating estimated pressure data; in, Represents the estimated pressure data of the node, represents the representation of the latent space, is the decoder parameter set, is the noise factor; Using the output of the estimated pressure data, combined with the actual pressure data, the data is reconstructed to fill in the gaps in the seal data or repair errors, using the formula: ; Obtain reconstructed pressure data; in, represents the reconstructed pressure data, is the actual pressure data, is the balance coefficient, is the error correction term.

[0090] Step 1: Determination of data reconstruction area; formula: ; Parameter explanation: : The data reconstruction area of ​​the node; : The set of neighbor nodes of a node; :node The influence coefficient of :node and The interaction strength; : nonlinear adjustment parameter; Calculation process: Assume that the node has two neighbors j and k, where: ; ; ; calculate Specific values: ; here Represents the comprehensive score of the reconstruction area of ​​the node. A higher score means that the node is more important in data reconstruction.

[0091] Step 2: Generation of estimated pressure data; formula: ; Parameter explanation: : Estimated pressure data of the node; : The representation of node i in the latent space; : parameter set of the decoder; : Noise coefficient; Calculation process: Assume the decoder output is , noise figure , random noise : ; ; ; here For Node Estimated pressure value after accounting for random noise.

[0092] Step 3: Final construction of pressure data; formula: ; Parameter explanation: : Final pressure data of the node; : Actual observed pressure data; : Estimated pressure data; : balance coefficient; : Error correction term; Calculation process: Assumptions: ; (from step 2); ; ; ; here It represents the final pressure data of the adjusted node, which is corrected after considering the weighted average of the actual observed data and the estimated data.

[0093] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A cable channel pressure change monitoring system, characterized in that: The system comprises: The cable channel pressure sensing encoding module collects the pressure readings and temperature data of the pressure cover plate and the pressure-resistant pipe fittings, calls the encoder part of the variational autoencoder, converts the data into a vector representation of the latent space, obtains a standardized latent vector, and uses the standardized latent vector to encode the pressure state of the seal and the bracket to obtain the node state encoding result; The cable channel spatial feature analysis module constructs a graph structure based on the node state encoding result, connects the pressure transmission path between compensators through edges, executes the graph convolution network to update the node state, obtains the updated node feature, analyzes the spatial distribution of pressure on the fixing parts, and obtains the spatial correlation analysis result; The cable channel anomaly identification module calculates the deviation between the pressure reading of each bracket node in the cable channel and the expected model based on the spatial correlation analysis result, compares the deviation with the threshold, identifies the compensator node with the deviation exceeding the threshold, analyzes the spatial diffusion trend of the node, and obtains the anomaly detection result; The cable channel data integrity reconstruction module determines the data reconstruction area of ​​the pressure-resistant pipe based on the abnormal detection results, uses the decoder part of the variational autoencoder to generate estimated pressure data from the latent space, fills in the seal data gaps or repairs errors, and obtains the reconstructed pressure data.

2. The cable channel pressure change monitoring system according to claim 1, characterized in that: The steps for obtaining the standardized potential vector are specifically as follows: Collect the pressure readings and temperature data of the pressure cover and pressure-resistant pipe fittings, using the formula: ; ; The pressure and temperature are range normalized respectively to obtain standardized data sets of pressure and temperature; in, represents the raw pressure reading, represents the original temperature data, , Standardized data representing pressure and temperature; The normalized data set of pressure and temperature is input into the encoder part of the variational autoencoder using the formula: ; Generate latent vector; in, represent and A collection of represents the weight matrix of the encoder, represents the encoder bias vector, represents the generated latent vector; Using the latent vector, the formula is adopted: ; Normalize to generate a standardized latent vector; in, represents the modulus of the latent vector, Represents the generated normalized latent vector.

3. The cable channel pressure change monitoring system according to claim 2, characterized in that: The steps for obtaining the node status encoding result are specifically as follows: Using the normalized latent vector, the pressure state of the seal and the bracket is encoded using the formula: ; Use the square root modulus length method to improve data stability and obtain preliminary state encoding; in, represents the normalized latent vector, represents the weight matrix of the encoding network, represents the bias vector of the encoding network, Represents the initial state encoding generated; The preliminary state encoding is optimized using the formula: ; Adjust the size and stability of the encoding vector to obtain an optimized state encoding; in, represents the modulus length of the state encoding, , is a parameter used to adjust the encoding scale and stability. Represents the optimized state code; Apply non-linear activation function The optimized state code is processed using the formula: ; Get the node status encoding result; in, represents the activation function, is the nonlinear adjustment coefficient, Represents the node status encoding result.

4. The cable channel pressure change monitoring system according to claim 3, characterized in that: The steps for obtaining the spatial association analysis results are specifically as follows: Starting from the node state encoding result, a graph structure is constructed by connecting the pressure transmission paths between compensators, using the formula: ; Generate a pressure transmission path diagram model; in, Representation Node and The pressure transmission efficiency between , Adjust the decay rate and distance sensitivity of pressure transmission, Represents the physical distance between nodes, , Indicates the state code of the node; The graph convolutional network is used to update the node status on the pressure transfer graph model, using the formula: ; Get updated node features; in, represents the updated node features, represents the original node feature of the kth neighbor, , are graph convolutional network parameters, is the activation function; Based on the updated node characteristics, the spatial distribution of pressure on the fixture is analyzed using the formula: ; Calculate the weighted sum of node features in the entire graph structure to obtain the spatial correlation analysis results; in, For location The weight of Represents the results of spatial association analysis.

5. The cable channel pressure change monitoring system according to claim 4, characterized in that: The steps for obtaining the abnormal detection result are specifically as follows: Starting from the results of the spatial correlation analysis, the deviation of the pressure reading of each bracket node in the cable channel from the expected model is calculated using the formula: ; Get the pressure deviation of each node; in, represents the pressure deviation at node i, is the actual pressure reading, is the pressure value of the expected model, Small positive numbers to avoid division by zero; Compare the pressure deviation of each node with the threshold, using the formula: ; Obtain node identifications where deviation exceeds threshold; in, is the deviation threshold, represents the deviation sensitivity, Indicates the probability of whether node i is abnormal; To analyze the spatial diffusion trend in node identifications with deviations exceeding the threshold, the formula is used: ; Generate anomaly detection results; in, represents the weight of the node, represents the location of the node, is the smoothing parameter, Trend results indicating anomalous diffusion.

6. The cable channel pressure change monitoring system according to claim 5, characterized in that: The steps for obtaining the reconstructed pressure data are specifically as follows: Based on the abnormal detection results, the data reconstruction area of ​​the pressure-resistant pipe fittings is determined using the formula: ; Generate data reconstruction area; in, Represents the data reconstruction area of ​​the node, Represents the set of neighbor nodes of a node, is the influence coefficient, represents the interaction strength between nodes, is the nonlinear adjustment parameter; Based on the data reconstruction area, the decoder part of the variational autoencoder is used, using the formula: ; generating estimated pressure data; in, Represents the estimated pressure data of the node, represents the representation of the latent space, is the decoder parameter set, is the noise factor; The output of the estimated pressure data is combined with the actual pressure data to reconstruct the data, fill in the gaps in the seal data or repair the errors, using the formula: ; Obtain reconstructed pressure data; in, represents the reconstructed pressure data, is the actual pressure data, is the balance coefficient, is the error correction term.

Citation Information

Cited By

  • Cable channel pressure change monitoring system based on sensing network

    CN119689176A