A pipeline damage behavior prediction and protection method based on AI big model

Abstract

The present invention relates to the technical field of pipeline destruction behavior prediction, and discloses a pipeline destruction behavior prediction and protection method based on an AI large model, the method comprising: using sensors to collect power pipeline data in real time, splicing the extracted image feature vector and the pre-processed power pipeline data into a fused feature vector, using a pipeline destruction behavior early warning model to predict the pipeline state value of the fused feature vector, comparing the pipeline state value prediction result output by the model with a preset threshold, and triggering a real-time alarm system when the prediction result exceeds the preset threshold, taking safety protection measures to protect the pipeline. The present invention uses the linear feature representing the pipeline deformation as an indicator of the projection embedding amount, uses weighted processing to increase the feature representation of the pipeline deformation part in the image feature vector, effectively captures the deformation information, performs time series flow mapping processing on the fused feature vector, forms continuous flow information, and performs pipeline state prediction and protection.

Description

Technical Field

[0001] The present invention relates to the technical field of pipeline destructive behavior prediction, and in particular to a pipeline destructive behavior prediction and protection method based on an AI large model. Background Art

[0002] As an important part of the power system, power pipelines undertake key functions such as transmission, distribution and power supply. However, the damage of power pipelines may lead to power outages, safety accidents and even economic losses. In order to ensure the reliable operation of the power system, it is imperative to identify the potential destructive behavior of power pipelines in advance and take corresponding protective measures. To address this problem, the present invention proposes a pipeline destructive behavior prediction and protection method based on an AI large model, which achieves accurate prediction and effective protection of destructive behavior through deep learning and intelligent analysis of a large amount of data. Summary of the invention

[0003] In view of this, the present invention provides a pipeline destruction behavior prediction and protection method based on AI large model, the purpose is: 1) using sensors to collect a large amount of multi-modal power pipeline data, and adopting different preprocessing methods for data of different modes, using an improved visual converter model to perform linear projection and nonlinear projection on image data respectively, to obtain linear features and nonlinear features of image data, using linear features as indicators of projection embedding amount, calculating feature weights of nonlinear projection, realizing weighted calculation of nonlinear projection based on linear features, obtaining image feature vectors integrating linear features and nonlinear features, linear features corresponding to deformation in pipelines, using weighted processing to increase feature representation of pipeline deformation part in image feature vectors, effectively capturing pipeline deformation, and preventing pipeline damage. Pipe deformation information; 2) The extracted image feature vector and the current, voltage, temperature and power in the pre-processed power pipeline data are spliced ​​into a fused feature vector, and the pipeline damage behavior warning model is used to perform time series flow mapping on the fused feature vector. The more obvious the change of the image feature vector is, the higher the time series weight at the corresponding moment is, which helps to capture abnormal information of the pipeline image. In the process of time series flow mapping, the valid information of the previous moment is retained to form continuous flow information, and the pipeline state value prediction result output by the model is compared with the preset threshold. When the prediction result exceeds the preset threshold, the real-time alarm system is triggered, and safety protection measures are taken to protect the pipeline, so as to realize multimodal feature extraction and time series feature prediction of multimodal large data combined with AI large models.

[0004] To achieve the above purpose, the present invention provides a pipeline damage behavior prediction and protection method based on an AI large model, comprising the following steps:

[0005] S1: using sensors to collect power pipeline data in real time, wherein the power pipeline data includes current, voltage, power, temperature and image data, and preprocessing the collected power pipeline data to obtain preprocessed power pipeline data, wherein the preprocessing includes filtering and data normalization processing;

[0006] S2: extracting features from the image data in the preprocessed power pipeline data to obtain an image feature vector, wherein the improved visual converter model is the main implementation method of the feature extraction;

[0007] S3: splicing the extracted image feature vector and the current, voltage, temperature and power in the pre-processed power pipeline data into a fused feature vector, and constructing a pipeline damage behavior warning model, wherein the pipeline damage behavior warning model takes the fused feature vector as input and takes the pipeline state value prediction result as output;

[0008] S4: Use the pipeline damage behavior warning model to predict the pipeline state value of the fused feature vector, compare the pipeline state value prediction result output by the model with the preset threshold, and trigger the real-time alarm system when the prediction result exceeds the preset threshold, and take safety protection measures to protect the pipeline.

[0009] As a further improvement method of the present invention:

[0010] Optionally, in step S1, using sensors to collect power pipeline data in real time includes:

[0011] Sensors are deployed on the surface of the power pipeline. In the embodiment of the present invention, the deployed sensors include an electric meter, a camera device, and a temperature sensor. The deployed sensors are used to collect power pipeline data in real time, wherein the power pipeline data includes current, voltage, power, temperature, and image data. The collected power pipeline data is expressed in the form of:

[0012] x=(x 1 ,x 2 ,x 3 ,x 4 ,x 5 )

[0013]

[0014] in:

[0015] x represents the collected power pipeline data, x i Indicates the data corresponding to the i-th power pipeline indicator, where the 1st to 5th power pipeline indicators are current, voltage, power, temperature and image indicators respectively;

[0016] x 1 ,x 2 ,x 3,x 4 ,x 5 The current, voltage, power, temperature and image data in the power pipeline data are represented in sequence;

[0017] It represents the time series representation of the data corresponding to the i-th power pipeline indicator, Indicates the index of the i-th power pipeline at t n The data corresponding to the moment, Indicates that the power pipeline is at t n The current difference between the two ends of the pipeline at the time, Indicates that the power pipeline is at t n The voltage difference between the two ends of the pipeline at the moment, Indicates that the power pipeline is at t n The power difference between the two ends of the pipeline at the moment, Indicates that the power pipeline is at t n The temperature of the moment, Indicates that the power pipeline is at t n The image of the moment, where is a grayscale image after grayscale processing; in the embodiment of the present invention, the current difference, voltage difference and power difference at both ends of the pipeline are the difference between the current inflow end and the current outflow end in the power pipeline;

[0018] The collected power pipeline data is preprocessed to obtain preprocessed power pipeline data.

[0019] Optionally, preprocessing the collected power pipeline data to obtain preprocessed power pipeline data includes:

[0020] The preprocessing process of power pipeline data is as follows:

[0021] S11: Normalize the current, voltage, power and temperature data in the power pipeline data, where the jth power pipeline index corresponds to data x j The normalization formula is:

[0022]

[0023]

[0024] in:

[0025] X j Indicates the data x corresponding to the j-th power pipeline index j The normalized result of , j∈[1,4];

[0026] Indicates the jth power pipeline index at t n Data corresponding to the time The normalized processing result of

[0027] represents the preset maximum value of the j-th power pipeline indicator, represents the preset minimum value of the j-th power pipeline indicator;

[0028] S12: Filter the image data in the power pipeline data, where the image The filtering formula is:

[0029]

[0030]

[0031]

[0032] in:

[0033] For images The filtered image of

[0034] W 1 is the filter matrix;

[0035] * is the convolution calculation;

[0036] Representing images The gray value of the pixel in the pth row and qth column in the image, P is The number of pixel rows of the image The number of pixel columns;

[0037] Display image A gray value matrix of 3×3 pixels centered on the gray value of the pixel in the pth row and qth column;

[0038] Represents the image after filtering The gray value of the pixel in the pth row and qth column;

[0039] S13: Constructing image data after filtering:

[0040] S14: Constructing pre-processed power pipeline data: X=(X 1 ,X 2 ,X 3 ,X 4 ,X 5 ).

[0041] Optionally, in step S2, feature extraction is performed on the image data in the preprocessed power pipeline data, including:

[0042] Feature extraction is performed on the image data in the preprocessed power pipeline data, wherein the improved visual converter model is the main implementation method of the feature extraction, wherein the image data X 5 Medium Image The feature extraction process is:

[0043] S21: Image The image blocks are divided into K equal-sized and non-overlapping blocks, where the set of image blocks obtained by division is:

[0044]

[0045] in:

[0046] represents the kth image block obtained by division;

[0047] S22: Perform linear projection and nonlinear projection on any image block obtained by division, wherein the image block The linear projection and nonlinear projection formulas of (k) are:

[0048]

[0049]

[0050] in:

[0051] Represents an image block Linear projection of

[0052] Represents an image block The flattened vector after flattening;

[0053] Represents an image block Nonlinear projection of ;

[0054] W 2 represents the linear projection basis, and T represents the transpose;

[0055] exp(·) represents an exponential function with a natural constant as the base;

[0056] S23: Perform projection embedding transformation on the linear projection, where the linear projection The projection embedding transformation formula is:

[0057]

[0058] in:

[0059] Represents a linear projection The projected embedding obtained by transformation;

[0060] dropout(·) represents the dropout layer, Represents a linear projection Any value in is discarded with a probability of 10%;

[0061] S24: Based on the projection embedding result, generate the feature weights corresponding to the nonlinear projection, where the nonlinear projection The feature weight calculation formula is:

[0062]

[0063] in:

[0064] Represents projected embedding The average pooling result is Represents projected embedding The maximum pooling result of ;

[0065] W 3 Represents multi-layer perception parameters;

[0066] Represents nonlinear projection The feature weights of

[0067] S25: weighting the nonlinear projection, wherein the nonlinear projection The weighted processing formula is:

[0068]

[0069] in:

[0070] Represents matrix dot multiplication calculation;

[0071] Represents nonlinear projection The weighted result of

[0072] S26: Composing the image The eigenvector of:

[0073]

[0074] in:

[0075] Representing images The eigenvector of

[0076] concat(·) represents a concatenation operation.

[0077] Optionally, in step S3, the extracted image feature vector and the current, voltage, temperature and power in the pre-processed power pipeline data are spliced ​​into a fused feature vector, including:

[0078] The extracted image feature vector and the current, voltage, temperature and power in the pre-processed power pipeline data are spliced ​​into a fused feature vector, and a pipeline damage behavior warning model is constructed. The splicing process of the fused feature vector is as follows:

[0079] Obtain the current, voltage, temperature and power data (X) of the pre-processed power pipeline data X 1 ,X 2 ,X 3 ,X 4 );

[0080] Combine the image feature vectors for splicing to obtain a fused feature vector, where the fused feature vector is represented as:

[0081]

[0082]

[0083] in:

[0084] y represents the fused feature vector, Indicates t n The fusion characteristics of the moment;

[0085] Construct an early warning model for pipeline damage behavior.

[0086] Optionally, the constructing of a pipeline damage behavior early warning model includes:

[0087] A pipeline destruction behavior warning model is constructed, which takes the fused feature vector as input and the pipeline state value prediction result as output, wherein the pipeline destruction behavior warning model includes an input layer, a feature mapping layer and a prediction layer, wherein the input layer is used to receive the fused feature vector, the feature mapping layer is used to perform time series flow mapping processing on the fused feature vector, and the prediction layer performs prediction according to the time series flow mapping processing result to obtain the pipeline state value prediction result.

[0088] Optionally, in step S4, using the pipeline damage behavior warning model to predict the pipeline state value of the fused feature vector includes:

[0089] The pipeline damage behavior warning model is used to predict the pipeline state value of the fused feature vector. The pipeline state value prediction result output by the model is compared with the preset threshold. When the prediction result exceeds the preset threshold, the real-time alarm system is triggered and safety protection measures are taken to protect the pipeline. The process of the pipeline damage behavior warning model predicting the pipeline state value of the fused feature vector is as follows:

[0090] S41: the input layer receives the fused feature vector y;

[0091] S42: The feature mapping layer performs time series flow mapping processing on the fused feature vector y, wherein the formula for the time series flow mapping processing is:

[0092]

[0093]

[0094]

[0095]

[0096] in:

[0097] y represents the time series flow mapping processing result of the fused feature vector y;

[0098] Represents the fused features in the fused feature vector y The timing weight of

[0099] Represents fusion features The timing flow information, w j represents the information weight of the j-th power pipeline indicator; in the embodiment of the present invention,

[0100] S43: The prediction layer performs prediction based on the time series flow mapping processing result y to obtain the pipeline state value prediction result y * :

[0101] y * =W * *y

[0102] in:

[0103] * indicates convolution processing, W * Represents the connection matrix of the prediction layer.

[0104] In order to solve the above problem, the present invention provides an electronic device, the electronic device comprising:

[0105] A memory storing at least one instruction;

[0106] Communication interface, enabling electronic equipment to communicate; and

[0107] The processor executes the instructions stored in the memory to implement the pipeline damage behavior prediction and protection method based on the AI ​​big model mentioned above.

[0108] In order to solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned pipeline destructive behavior prediction and protection method based on the AI ​​big model.

[0109] Compared with the prior art, the present invention proposes a pipeline damage behavior prediction and protection method based on an AI large model, which has the following advantages:

[0110] First, this scheme proposes an image feature extraction method. The image blocks are divided into K equal-sized and non-overlapping blocks, where the set of image blocks obtained by division is:

[0111]

[0112] in: represents the kth image block obtained by division; linear projection and nonlinear projection are performed on any image block obtained by division, where the image block The linear projection and nonlinear projection formulas are:

[0113]

[0114]

[0115] in: Represents an image block Linear projection of Represents an image block The flattened vector after flattening; Represents an image block Nonlinear projection of W 2 represents the linear projection basis, T represents the transpose; exp(·) represents the exponential function with the natural constant as the base; the linear projection is projected and embedded, where the linear projection The projection embedding transformation formula is:

[0116]

[0117] in: Represents a linear projection The projected embedding obtained by transformation; dropout(·) represents the dropout layer, Represents a linear projection Any value in is discarded with a discard probability of 10%; based on the projection embedding result, the feature weights corresponding to the nonlinear projection are generated, where the nonlinear projection The feature weight calculation formula is:

[0118]

[0119] in: Represents projected embedding The average pooling result is Represents projected embedding The maximum pooling result of 3 Represents multi-layer perception parameters; Represents nonlinear projection feature weights; weighted processing of nonlinear projection, where nonlinear projection The weighted processing formula is:

[0120]

[0121] in: Represents matrix dot multiplication calculation; Represents nonlinear projection The weighted result of The eigenvector of:

[0122]

[0123] in: Representing images The feature vector of the image data is obtained by using sensors to collect a large amount of multi-modal power pipeline data, and different pre-processing methods are adopted for data of different modes. The improved visual converter model is used to perform linear projection and nonlinear projection on the image data respectively, and the linear features and nonlinear features of the image data are obtained. The linear features are used as indicators of the projection embedding amount, and the feature weights of the nonlinear projection are calculated. The weighted calculation of the nonlinear projection based on the linear features is realized, and the image feature vector integrating the linear features and the nonlinear features is obtained. The linear features correspond to the deformation in the pipeline. The weighted processing is used to increase the feature representation of the pipeline deformation part in the image feature vector, and the pipeline deformation information is effectively captured.

[0124] At the same time, this scheme proposes a pipeline state prediction method, which uses the pipeline damage behavior warning model to predict the pipeline state value of the fused feature vector, compares the pipeline state value prediction result output by the model with the preset threshold, and triggers the real-time alarm system when the prediction result exceeds the preset threshold, and takes safety protection measures to protect the pipeline. The process of the pipeline damage behavior warning model predicting the pipeline state value of the fused feature vector is as follows: the input layer receives the fused feature vector y; the feature mapping layer performs time series flow mapping processing on the fused feature vector y, and the formula for time series flow mapping processing is:

[0125]

[0126]

[0127]

[0128]

[0129] Where: y represents the time series flow mapping processing result of the fused feature vector y; Represents the fused features in the fused feature vector y The timing weight of Represents fusion features The timing flow information, w j represents the information weight of the jth power pipeline indicator; the prediction layer predicts according to the time series flow mapping processing result y, and obtains the pipeline state value prediction result y * :

[0130] y * =W * *y

[0131] in:

[0132] * indicates convolution processing, W * Represents the connection matrix of the prediction layer. This solution concatenates the extracted image feature vector and the current, voltage, temperature and power in the pre-processed power pipeline data into a fused feature vector, and uses the pipeline damage behavior warning model to perform time series flow mapping on the fused feature vector. The more obvious the change of the image feature vector, the higher the time series weight at the corresponding moment, which helps to capture abnormal information of the pipeline image. In the process of time series flow mapping, the valid information of the previous moment is retained to form continuous flow information, and the pipeline state value prediction result output by the model is compared with the preset threshold. When the prediction result exceeds the preset threshold, the real-time alarm system is triggered, and safety protection measures are taken to protect the pipeline, realizing multimodal feature extraction and time series feature prediction of multimodal large data combined with AI large models. BRIEF DESCRIPTION OF THE DRAWINGS

[0133] Figure 1 A flowchart of a pipeline damage behavior prediction and protection method based on an AI large model provided in one embodiment of the present invention;

[0134] Figure 2 A schematic diagram of the structure of an electronic device for implementing a pipeline damage behavior prediction and protection method based on an AI large model provided in one embodiment of the present invention.

[0135] In the figure: 1 electronic device, 10 processor, 11 memory, 12 program, 13 communication interface.

[0136] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0137] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0138] The embodiment of the present application provides a pipeline damage behavior prediction and protection method based on an AI big model. The execution subject of the pipeline damage behavior prediction and protection method based on the AI ​​big model includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided in the embodiment of the present application. In other words, the pipeline damage behavior prediction and protection method based on the AI ​​big model can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0139] Embodiment 1:

[0140] S1: Use sensors to collect power pipeline data in real time, where the power pipeline data includes current, voltage, power, temperature and image data, and preprocess the collected power pipeline data to obtain preprocessed power pipeline data, wherein the preprocessing includes filtering and data normalization processing.

[0141] In step S1, the sensors are used to collect power pipeline data in real time, including:

[0142] Sensors are deployed on the surface of the power pipeline. In the embodiment of the present invention, the deployed sensors include an electric meter, a camera device, and a temperature sensor. The deployed sensors are used to collect power pipeline data in real time, wherein the power pipeline data includes current, voltage, power, temperature, and image data. The collected power pipeline data is expressed in the form of:

[0143] x=(x 1 ,x 2 ,x 3 ,x 4 ,x5 )

[0144]

[0145] in:

[0146] x represents the collected power pipeline data, x i Indicates the data corresponding to the i-th power pipeline indicator, where the 1st to 5th power pipeline indicators are current, voltage, power, temperature and image indicators respectively;

[0147] x 1 ,x 2 ,x 3 ,x 4 ,x 5 The current, voltage, power, temperature and image data in the power pipeline data are represented in sequence;

[0148] It represents the time series representation of the data corresponding to the i-th power pipeline indicator, Indicates the index of the i-th power pipeline at t n The data corresponding to the moment, Indicates that the power pipeline is at t n The current difference between the two ends of the pipeline at the time, Indicates that the power pipeline is at t n The voltage difference between the two ends of the pipeline at the moment, Indicates that the power pipeline is at t n The power difference between the two ends of the pipeline at the moment, Indicates that the power pipeline is at t n The temperature of the moment, Indicates that the power pipeline is at t n The image of the moment, where is a grayscale image after grayscale processing; in the embodiment of the present invention, the current difference, voltage difference and power difference at both ends of the pipeline are the difference between the current inflow end and the current outflow end in the power pipeline;

[0149] The collected power pipeline data is preprocessed to obtain preprocessed power pipeline data.

[0150] The preprocessing of the collected power pipeline data to obtain the preprocessed power pipeline data includes:

[0151] The preprocessing process of power pipeline data is as follows:

[0152] S11: Normalize the current, voltage, power and temperature data in the power pipeline data, where the jth power pipeline index corresponds to data x j The normalization formula is:

[0153]

[0154]

[0155] in:

[0156] X j Indicates the data x corresponding to the j-th power pipeline index j The normalized result of , j∈[1,4];

[0157] Indicates the jth power pipeline index at t n Data corresponding to the time The normalized processing result of

[0158] represents the preset maximum value of the j-th power pipeline indicator, represents the preset minimum value of the j-th power pipeline indicator;

[0159] S12: Filter the image data in the power pipeline data, where the image The filtering formula is:

[0160]

[0161]

[0162]

[0163] in:

[0164] For images The filtered image of

[0165] W 1 is the filter matrix;

[0166] * is the convolution calculation;

[0167] Representing images The gray value of the pixel in the pth row and qth column in the image, P is The number of pixel rows of the image The number of pixel columns;

[0168] Display image A gray value matrix of 3×3 pixels centered on the gray value of the pixel in the pth row and qth column;

[0169] Represents the image after filtering The gray value of the pixel in the pth row and qth column;

[0170] S13: Constructing image data after filtering:

[0171] S14: Constructing pre-processed power pipeline data: X=(X 1 ,X 2 ,X 3 ,X 4 ,X 5 ).

[0172] S2: Extract features from the image data in the preprocessed power pipeline data to obtain an image feature vector.

[0173] In the step S2, feature extraction is performed on the image data in the pre-processed power pipeline data, including:

[0174] Feature extraction is performed on the image data in the preprocessed power pipeline data, wherein the improved visual converter model is the main implementation method of the feature extraction, wherein the image data X 5 Medium Image The feature extraction process is:

[0175] S21: Image The image blocks are divided into K equal-sized and non-overlapping blocks, where the set of image blocks obtained by division is:

[0176]

[0177] in:

[0178] represents the kth image block obtained by division;

[0179] S22: Perform linear projection and nonlinear projection on any image block obtained by division, wherein the image block The linear projection and nonlinear projection formulas of (k) are:

[0180]

[0181]

[0182] in:

[0183] Represents an image block Linear projection of

[0184] Represents an image block The flattened vector after flattening;

[0185] Represents an image block Nonlinear projection of ;

[0186] W 2 represents the linear projection basis, and T represents the transpose;

[0187] exp(·) represents an exponential function with a natural constant as the base;

[0188] S23: Perform projection embedding transformation on the linear projection, where the linear projection The projection embedding transformation formula is:

[0189]

[0190] in:

[0191] Represents a linear projection The projected embedding obtained by transformation;

[0192] dropout(·) represents the dropout layer, Represents a linear projection Any value in is discarded with a probability of 10%;

[0193] S24: Based on the projection embedding result, generate the feature weights corresponding to the nonlinear projection, where the nonlinear projection The feature weight calculation formula is:

[0194]

[0195] in:

[0196] Represents projected embedding The average pooling result is Represents projected embedding The maximum pooling result of ;

[0197] W 3 Represents multi-layer perception parameters;

[0198] Represents nonlinear projection The feature weights of

[0199] S25: weighting the nonlinear projection, wherein the nonlinear projection The weighted processing formula is:

[0200]

[0201] in:

[0202] Represents matrix dot multiplication calculation;

[0203] Represents nonlinear projection The weighted result of

[0204] S26: Composing the image The eigenvector of:

[0205]

[0206] in:

[0207] Representing images The eigenvector of

[0208] concat(·) represents a concatenation operation.

[0209] S3: The extracted image feature vector and the current, voltage, temperature and power in the pre-processed power pipeline data are spliced ​​into a fused feature vector to construct a pipeline damage behavior warning model, which takes the fused feature vector as input and the pipeline state value prediction result as output.

[0210] In the step S3, the extracted image feature vector and the current, voltage, temperature and power in the pre-processed power pipeline data are spliced ​​into a fused feature vector, including:

[0211] The extracted image feature vector and the current, voltage, temperature and power in the pre-processed power pipeline data are spliced ​​into a fused feature vector, and a pipeline damage behavior warning model is constructed. The splicing process of the fused feature vector is as follows:

[0212] Obtain the current, voltage, temperature and power data (X) of the pre-processed power pipeline data X 1 ,X 2 ,X 3 ,X 4 );

[0213] Combine the image feature vectors for splicing to obtain a fused feature vector, where the fused feature vector is represented as:

[0214]

[0215]

[0216] in:

[0217] y represents the fused feature vector, Indicates t n The fusion characteristics of the moment;

[0218] Construct an early warning model for pipeline damage behavior.

[0219] The pipeline destruction behavior early warning model is constructed, including:

[0220] A pipeline destruction behavior warning model is constructed, which takes the fused feature vector as input and the pipeline state value prediction result as output, wherein the pipeline destruction behavior warning model includes an input layer, a feature mapping layer and a prediction layer, wherein the input layer is used to receive the fused feature vector, the feature mapping layer is used to perform time series flow mapping processing on the fused feature vector, and the prediction layer performs prediction according to the time series flow mapping processing result to obtain the pipeline state value prediction result.

[0221] S4: Use the pipeline damage behavior warning model to predict the pipeline state value of the fused feature vector, compare the pipeline state value prediction result output by the model with the preset threshold, and trigger the real-time alarm system when the prediction result exceeds the preset threshold, and take safety protection measures to protect the pipeline.

[0222] In the step S4, the pipeline damage behavior warning model is used to predict the pipeline state value of the fused feature vector, including:

[0223] The pipeline damage behavior warning model is used to predict the pipeline state value of the fused feature vector. The pipeline state value prediction result output by the model is compared with the preset threshold. When the prediction result exceeds the preset threshold, the real-time alarm system is triggered and safety protection measures are taken to protect the pipeline. The process of the pipeline damage behavior warning model predicting the pipeline state value of the fused feature vector is as follows:

[0224] S41: the input layer receives the fused feature vector y;

[0225] S42: The feature mapping layer performs time series flow mapping processing on the fused feature vector y, wherein the formula for the time series flow mapping processing is:

[0226]

[0227]

[0228]

[0229]

[0230] in:

[0231] y represents the time series flow mapping processing result of the fused feature vector y;

[0232] Represents the fused features in the fused feature vector y The timing weight of

[0233] Represents fusion features The timing flow information, w j represents the information weight of the j-th power pipeline indicator; in the embodiment of the present invention,

[0234] S43: The prediction layer performs prediction based on the time series flow mapping processing result y to obtain the pipeline state value prediction result y * :

[0235] y * =W * *y

[0236] in:

[0237] * indicates convolution processing, W * Represents the connection matrix of the prediction layer.

[0238] Embodiment 2:

[0239] like Figure 2 , is a schematic diagram of the structure of an electronic device for implementing a pipeline damage behavior prediction and protection method based on an AI big model provided by an embodiment of the present invention.

[0240] The electronic device 1 may include a processor 10 , a memory 11 , a communication interface 13 and a bus, and may also include a computer program stored in the memory 11 and executable on the processor 10 , such as a program 12 .

[0241] The memory 11 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a magnetic memory, a disk, an optical disk, etc. The memory 11 may be an internal storage unit of the electronic device 1 in some embodiments, such as a mobile hard disk of the electronic device 1. The memory 11 may also be an external storage device of the electronic device 1 in other embodiments, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. Further, the memory 11 may also include both an internal storage unit of the electronic device 1 and an external storage device. The memory 11 may be used not only to store application software and various types of data installed in the electronic device 1, such as the code of the program 12, etc., but also to temporarily store data that has been output or is to be output.

[0242] The processor 10 may be composed of an integrated circuit in some embodiments, for example, a single packaged integrated circuit, or a plurality of integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips. The processor 10 is the control core (Control Unit) of the electronic device, and uses various interfaces and lines to connect the various components of the entire electronic device, and executes or executes the programs or modules stored in the memory 11 (such as the program 12 for realizing pipeline damage behavior prediction and protection based on the AI ​​large model), and calls the data stored in the memory 11 to execute various functions of the electronic device 1 and process data.

[0243] The communication interface 13 may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices, and to achieve connection and communication between internal components of the electronic device.

[0244] The bus may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize connection and communication between the memory 11 and at least one processor 10, etc.

[0245] Figure 2 Only an electronic device with components is shown, and those skilled in the art will understand that Figure 2 The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0246] For example, although not shown, the electronic device 1 may also include a power source (such as a battery) for supplying power to each component. Preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so that the power management device can realize functions such as charging management, discharging management, and power consumption management. The power source may also include any components such as one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, etc. The electronic device 1 may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0247] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), or a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode) touch device. The display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device 1 and to display a visual user interface.

[0248] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0249] The program 12 stored in the memory 11 in the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can achieve:

[0250] Using sensors to collect power pipeline data in real time, where the power pipeline data includes current, voltage, power, temperature and image data, and preprocessing the collected power pipeline data to obtain preprocessed power pipeline data;

[0251] Perform feature extraction on the image data in the preprocessed power pipeline data to obtain an image feature vector;

[0252] The extracted image feature vector and the current, voltage, temperature and power in the pre-processed power pipeline data are spliced ​​into a fused feature vector to build a pipeline damage behavior warning model;

[0253] The pipeline damage behavior warning model is used to predict the pipeline state value of the fused feature vector. The pipeline state value prediction result output by the model is compared with the preset threshold. When the prediction result exceeds the preset threshold, the real-time alarm system is triggered and safety protection measures are taken to protect the pipeline.

[0254] Specifically, the specific implementation method of the processor 10 for the above instructions can refer to Figure 1 to Figure 2 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0255] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the terms "including", "comprising" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0256] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0257] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A pipeline damage behavior prediction and protection method based on AI large model, characterized in that: The method comprises: S1: using sensors to collect power pipeline data in real time, wherein the power pipeline data includes current, voltage, power, temperature and image data, and preprocessing the collected power pipeline data to obtain preprocessed power pipeline data, wherein the preprocessing includes filtering and data normalization processing; S2: extracting features from the image data in the preprocessed power pipeline data to obtain an image feature vector; Feature extraction is performed on the image data in the preprocessed power pipeline data, where the image data X5 The feature extraction process is: S21: Image The image blocks are divided into K equal-sized and non-overlapping blocks, where the set of image blocks obtained by division is: in: represents the kth image block obtained by division; S22: Perform linear projection and nonlinear projection on any image block obtained by division, wherein the image block The linear projection and nonlinear projection formulas are: in: Represents an image block Linear projection of Represents an image block The flattened vector after flattening; Represents an image block Nonlinear projection of ; W2 represents the linear projection basis, and T represents the transpose; exp(·) represents an exponential function with a natural constant as the base; S23: Perform projection embedding transformation on the linear projection, where the linear projection The projection embedding transformation formula is: in: Represents a linear projection The projected embedding obtained by transformation; dropout(·) represents the dropout layer, Represents a linear projection Any value in is discarded with a probability of 10%; S24: Based on the projection embedding result, generate the feature weights corresponding to the nonlinear projection, where the nonlinear projection The feature weight calculation formula is: in: Represents projected embedding The average pooling result is Represents projected embedding The maximum pooling result of ; W3 represents multi-layer perception parameters; Represents nonlinear projection The feature weights of S25: weighting the nonlinear projection, wherein the nonlinear projection The weighted processing formula is: in: Represents matrix dot multiplication calculation; Represents nonlinear projection The weighted result of S26: Composing the image The eigenvector of: in: Representing images The eigenvector of concat(·) indicates concatenation operation; S3: splicing the extracted image feature vector and the current, voltage, temperature and power in the pre-processed power pipeline data into a fused feature vector, and constructing a pipeline damage behavior warning model, wherein the pipeline damage behavior warning model takes the fused feature vector as input and takes the pipeline state value prediction result as output; S4: Use the pipeline damage behavior warning model to predict the pipeline state value of the fused feature vector, compare the pipeline state value prediction result output by the model with the preset threshold, and trigger the real-time alarm system when the prediction result exceeds the preset threshold, and take safety protection measures to protect the pipeline.

2. A pipeline damage behavior prediction and protection method based on AI large model as claimed in claim 1, characterized in that: In step S1, the sensors are used to collect power pipeline data in real time, including: Sensors are deployed on the surface of the power pipeline, and the deployed sensors are used to collect power pipeline data in real time, where the power pipeline data includes current, voltage, power, temperature and image data. The collected power pipeline data is expressed in the form of: x=(x1,x2,x3,x4,x5) in: x represents the collected power pipeline data, x i Indicates the data corresponding to the i-th power pipeline indicator, where the 1st to 5th power pipeline indicators are current, voltage, power, temperature and image indicators respectively; x1, x2, x3, x4, and x5 represent the current, voltage, power, temperature, and image data in the power pipeline data respectively; It represents the time series representation of the data corresponding to the i-th power pipeline indicator, Indicates the index of the i-th power pipeline at t n The data corresponding to the moment, Indicates that the power pipeline is at t n The current difference between the two ends of the pipeline at the time, Indicates that the power pipeline is at t n The voltage difference between the two ends of the pipeline at the moment, Indicates that the power pipeline is at t n The power difference between the two ends of the pipeline at the moment, Indicates that the power pipeline is at t n The temperature of the moment, Indicates that the power pipeline is at t n The image of the moment, where It is the grayscale image after grayscale processing; The collected power pipeline data is preprocessed to obtain preprocessed power pipeline data.

3. A pipeline damage behavior prediction and protection method based on AI large model as claimed in claim 2, characterized in that: The preprocessing of the collected power pipeline data to obtain the preprocessed power pipeline data includes: The preprocessing process of power pipeline data is as follows: S11: Normalize the current, voltage, power and temperature data in the power pipeline data, where the jth power pipeline index corresponds to data x j The normalization formula is: in: X j Indicates the data x corresponding to the j-th power pipeline index j The normalized result of , j∈[1,4]; Indicates the jth power pipeline index at t n Data corresponding to the time The normalized processing result of represents the preset maximum value of the jth power pipeline indicator, represents the preset minimum value of the j-th power pipeline indicator; S12: Filter the image data in the power pipeline data, where the image The filtering formula is: in: For images The filtered image of W1 is the filter matrix; * is the convolution calculation; Representing images The gray value of the pixel in the pth row and qth column in the image, P is The number of pixel rows of the image The number of pixel columns; Display image A gray value matrix of 3×3 pixels centered on the gray value of the pixel in the pth row and qth column; Represents the image after filtering The gray value of the pixel in the pth row and qth column; S13: Constructing image data after filtering: S14: Construct pre-processed power pipeline data: X=(X1, X2, X3, X4, X5).

4. A pipeline damage behavior prediction and protection method based on AI large model as claimed in claim 1, characterized in that: In the step S3, the extracted image feature vector and the current, voltage, temperature and power in the pre-processed power pipeline data are spliced ​​into a fused feature vector, including: The extracted image feature vector and the current, voltage, temperature and power in the pre-processed power pipeline data are spliced ​​into a fused feature vector, and a pipeline damage behavior warning model is constructed. The splicing process of the fused feature vector is as follows: Obtain the current, voltage, temperature and power data (X1, X2, X3, X4) in the pre-processed power pipeline data X; Combine the image feature vectors for splicing to obtain a fused feature vector, where the fused feature vector is represented as: in: y represents the fused feature vector, Indicates t n The fusion characteristics of the moment; Construct an early warning model for pipeline damage behavior.

5. A pipeline damage behavior prediction and protection method based on AI large model as claimed in claim 4, characterized in that: The pipeline destruction behavior early warning model is constructed, including: A pipeline destruction behavior warning model is constructed, which takes the fused feature vector as input and the pipeline state value prediction result as output, wherein the pipeline destruction behavior warning model includes an input layer, a feature mapping layer and a prediction layer, wherein the input layer is used to receive the fused feature vector, the feature mapping layer is used to perform time series flow mapping processing on the fused feature vector, and the prediction layer performs prediction according to the time series flow mapping processing result to obtain the pipeline state value prediction result.

6. A pipeline damage behavior prediction and protection method based on AI large model as claimed in claim 1, characterized in that: In the step S4, the pipeline damage behavior warning model is used to predict the pipeline state value of the fused feature vector, including: The pipeline damage behavior warning model is used to predict the pipeline state value of the fused feature vector. The pipeline state value prediction result output by the model is compared with the preset threshold. When the prediction result exceeds the preset threshold, the real-time alarm system is triggered and safety protection measures are taken to protect the pipeline. The process of the pipeline damage behavior warning model predicting the pipeline state value of the fused feature vector is as follows: S41: the input layer receives the fused feature vector y; S42: The feature mapping layer performs time series flow mapping processing on the fused feature vector y, wherein the formula for the time series flow mapping processing is: in: y represents the time series flow mapping processing result of the fused feature vector y; Represents the fused features in the fused feature vector y The timing weight of Represents fusion features The timing flow information, w j represents the information weight of the j-th power pipeline indicator; S43: The prediction layer performs prediction based on the time series flow mapping processing result y to obtain the pipeline state value prediction result y * : y * =W * *y in: * indicates convolution processing, W * Represents the connection matrix of the prediction layer.

Images