Risk early warning method and device based on strain data of transmission tower

Abstract

The invention provides a risk early warning method and device based on transmission tower strain data, and the method comprises the steps: obtaining a plurality of groups of initial strain value time sequences corresponding to a plurality of weak part towers on a transmission tower, carrying out the data preprocessing, obtaining a plurality of groups of effective initial strain value time sequences, and building a risk prediction model, wherein the graph structure construction module is used for constructing a feature matrix for a plurality of groups of effective initial strain value time sequences, the multi-layer GCN module is used for extracting a time feature sequence of the feature matrix, and the Transform model module is used for processing the time feature sequence to output a predicted strain value time sequence. And formulating a risk threshold range corresponding to the plurality of risk levels and obtaining a risk level prediction time sequence corresponding to the prediction strain value time sequence. According to the method, the advantages of the multi-layer GCN module and the Transform model are utilized, the spatial features and the time features of the strain data are effectively extracted, and accurate risk prediction is achieved.

Description

Technical Field

[0001] The present invention relates to the field of power transmission tower system monitoring and safety, and in particular to a risk early warning method and device based on power transmission tower strain data. Background Art

[0002] As an important structure for power transmission, the safety of transmission towers is directly related to the reliability and stability of power supply. In recent years, with the increase of power grid load and the frequent occurrence of extreme weather, transmission towers are facing more and more potential safety hazards. Traditional transmission tower monitoring methods mainly rely on manual inspections and regular testing, which is not only time-consuming but also has limited ability to detect potential risks. The development of modern technology has provided new means for monitoring transmission towers. By installing strain sensors, the strain data of the towers can be obtained in real time to evaluate their health status. However, how to effectively analyze and utilize these complex time series data and quickly identify potential risk points is still a problem that needs to be solved. Summary of the invention

[0003] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a risk warning method and device based on transmission tower strain data, which utilizes the advantages of multi-layer graph convolutional neural network and Transformer model to effectively extract the spatial and temporal characteristics of strain data, thereby achieving accurate risk prediction.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0005] A risk warning method based on transmission tower strain data comprises the following steps:

[0006] Obtaining a number of groups of initial strain value time series corresponding to a number of weak position towers on the transmission tower, and performing data preprocessing on the number of groups of initial strain value time series to obtain a number of groups of valid initial strain value time series;

[0007] A risk prediction model is established to train several groups of valid initial strain value time series to obtain a trained risk prediction model, wherein the risk prediction model includes a graph structure construction module, a multi-layer GCN module, and a Transformer model module, a feature matrix is ​​constructed for several groups of valid initial strain value time series through the graph structure construction module, a time feature sequence matrix of the feature matrix is ​​extracted through the multi-layer GCN module, and the time feature sequence is processed through the Transformer model module to output a predicted strain value time series matrix;

[0008] A risk threshold range corresponding to multiple risk levels is formulated, and a risk level prediction time series matrix corresponding to several weak-position towers is obtained based on the predicted strain value time series matrix and the multiple risk threshold ranges.

[0009] Furthermore, the data preprocessing includes a missing data filling process and a data normalization process. The missing data filling process is as follows: continuity verification is performed on several groups of initial strain value time series collected, and the missing data is filled by a mean interpolation method.

[0010] Furthermore, the method for obtaining several groups of initial strain value time series corresponding to several weak position towers on the transmission tower is as follows:

[0011] A finite element model is established for the transmission tower to obtain several weak position towers corresponding to the weak positions where the strain of the transmission tower is concentrated. Corresponding fiber grating sensors are installed one by one on the several weak position towers, and several groups of initial strain value time series are collected through the fiber grating sensors.

[0012] Furthermore, the method in which the graph structure building module is used to build a feature matrix for several groups of valid initial strain value time series is:

[0013] The time when the initial strain value time series of the corresponding group of each weak part tower is generated is taken as a node, and the initial strain value time series of each weak part tower is taken as the feature vector corresponding to the node. The relationship between two nodes is represented by an edge, and a feature matrix is ​​constructed based on the nodes, the feature vectors corresponding to the nodes, and the relationship between the two nodes.

[0014] Furthermore, the method for extracting features of the feature matrix by the multi-layer GCN module is:

[0015] The multi-layer GCN module is set to an input layer, a hidden layer and an output layer, the hidden layer is set to the first GCN layer and the second GCN layer, the input layer is used to input a feature matrix, and the time feature sequence of the feature matrix is ​​obtained by training the first GCN layer and the second GCN layer, and the output layer is used to output the time feature sequence.

[0016] Furthermore, the method for the Transformer model module to process the feature output prediction strain value time series is:

[0017] The Transformer model consists of an encoder stack and a decoder stack. The encoder stack is composed of multiple identical encoders. Each layer of encoders consists of a multi-head attention module and a fully connected feedforward network. Residual connections and standardization are introduced to prevent model degradation and improve convergence speed.

[0018] The decoder stack is composed of a plurality of identical decoders stacked together. A multi-head attention module is inserted into each layer of the decoder to perform a self-attention operation on the output of the encoder stack. The self-attention operation obtains the attention weight of the time feature of each moment to the time features of other moments by dynamically evaluating the relationship between the time features of each moment in the initial strain value time series and the one-to-one corresponding time features of multiple moments before and after, and then multiplies the attention weight of each moment by the corresponding time feature to obtain the risk value, and collects the risk values ​​of all moments to obtain the predicted strain value time series.

[0019] Furthermore, the mean square error is used as the loss function, the Adam optimizer is selected to update the model parameters, and the early stopping method and cross-validation method are adopted to prevent overfitting.

[0020] A risk warning device based on transmission tower strain data, comprising:

[0021] A data acquisition module is used to acquire a plurality of groups of initial strain value time series corresponding to a plurality of weak position towers on a transmission tower, and perform data preprocessing on the plurality of groups of initial strain value time series to obtain a plurality of groups of effective initial strain value time series;

[0022] A risk prediction model building module is used to build a risk prediction model to train several groups of valid initial strain value time series to obtain a trained risk prediction model, wherein the risk prediction model includes a graph structure building module, a multi-layer GCN module, and a Transformer model module. The graph structure building module is used to build a feature matrix for several groups of valid initial strain value time series, and the multi-layer GCN module is used to extract the time feature sequence matrix of the feature matrix. The Transformer model module is used to process the time feature sequence to output a predicted strain value time series matrix;

[0023] The prediction module is used to formulate risk threshold ranges corresponding to multiple risk levels, and obtain the risk level prediction time series matrix corresponding to several weak-position towers according to the multiple risk threshold ranges based on the predicted strain value time series matrix.

[0024] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned risk warning method based on transmission tower strain data when executing the program.

[0025] A non-transitory computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned risk warning method based on transmission tower strain data.

[0026] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0027] The present invention provides a risk warning method and device based on transmission tower strain data. The advantages of graph convolutional neural network (GCN) and Transformer model are used to effectively extract the spatial and temporal features of strain data, thereby achieving accurate risk prediction. GCN can capture complex relationships between nodes, not limited to directly adjacent nodes, which is very useful for time series data with implicit dependencies. It can aggregate the information of neighboring nodes and improve the representation ability of node features, thereby improving the prediction effect. In addition, there is no problem of gradient disappearance and gradient explosion during GCN retraining, but GCN itself does not have the ability to directly process time information, and time series data often changes dynamically. Standard GCN usually assumes that the graph structure is static. Capturing dynamic changes requires combining other models to model time dependencies. The Transformer model can automatically select which time steps are important through the self-attention mechanism. This ability to dynamically select features enables the model to focus on the historical data most relevant to the current strain state when predicting, but the Transformer model has problems of low efficiency, poor feature extraction ability, gradient explosion and inability to process in parallel when processing time series data. Therefore, the present invention combines the advantages of graph convolutional neural network (GCN) and Transformer model. GCN can capture the spatial dependencies between nodes, while Transformer can model long-term dependencies in time series. This combination enables the model to process transmission tower strain data with complex spatiotemporal dependencies and achieve accurate prediction of the changing trend of strain data; in addition, in the present invention, there are many weak poles on the transmission tower, which can generate a large amount of data in a short time. The multi-layer GCN module is used to extract the time feature sequence of the feature matrix. The complexity of the model can be expanded by increasing the number of GCN layers to cope with more complex tasks. For complex graphs containing a large number of nodes and edges, the multi-layer GCN can still maintain good performance and effectively extract time features. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0029] Figure 1 A schematic diagram of a risk warning method based on transmission tower strain data of the present invention;

[0030] Figure 2 is a schematic diagram of the risk prediction model of the present invention;

[0031] Figure 3Schematic diagram of a multi-layer GCN module of the present invention;

[0032] Figure 4 Schematic diagram of the Transformer model module of the present invention;

[0033] Figure 5 Schematic diagram of the self-attention mechanism of the present invention;

[0034] Figure 6 is the actual strain curve in a specific embodiment of the present invention;

[0035] Figure 7 This is a comparison diagram of strain prediction value and true value in a specific embodiment of the present invention;

[0036] Figure 8 The diagram is a schematic diagram of a risk warning device based on transmission tower strain data according to the present invention. DETAILED DESCRIPTION

[0037] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0038] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "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 accompanying 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, the terms "first", "second", and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0039] Example 1

[0040] Embodiment 1 provides a risk warning method based on transmission tower strain data, such as Figure 1 As shown, the following steps are included:

[0041] Step S1: obtaining a plurality of groups of initial strain value time series corresponding to a plurality of weak position towers on the transmission tower, and performing data preprocessing on the plurality of groups of initial strain value time series to obtain a plurality of groups of effective initial strain value time series;

[0042] Step S2: Establish a risk prediction model to train several groups of valid initial strain value time series to obtain a trained risk prediction model, wherein the risk prediction model includes a graph structure construction module, a multi-layer GCN module, and a Transformer model module, construct a feature matrix for several groups of valid initial strain value time series through the graph structure construction module, extract a time feature sequence matrix of the feature matrix through the multi-layer GCN module, and process the time feature sequence through the Transformer model module to output a predicted strain value time series matrix;

[0043] Step S3: formulate a plurality of risk threshold ranges corresponding to the risk levels, and obtain a risk level prediction time series matrix corresponding to a plurality of weak-position towers according to the plurality of risk threshold ranges based on the predicted strain value time series matrix.

[0044] This embodiment provides a risk warning method based on transmission tower strain data, which uses the advantages of graph convolutional neural networks (GCN) and transformer models to effectively extract spatial and temporal features of strain data, thereby achieving accurate risk prediction.

[0045] GCN can capture complex relationships between nodes, not just directly adjacent nodes, which is very useful for time series data with implicit dependencies, and can aggregate information from neighboring nodes, improve the ability to represent node features, and thus improve prediction results. Moreover, there is no problem of gradient vanishing and gradient explosion during GCN retraining, but GCN itself does not have the ability to directly process time information, and time series data often changes dynamically, and standard GCN usually assumes that the graph structure is static. Capturing dynamic changes requires combining other models to model time dependencies. The Transformer model, through the self-attention mechanism, can automatically select which time step information is important. This ability to dynamically select features enables the model to focus on the historical data most relevant to the current strain state when predicting, but the Transformer model has problems of low efficiency, poor feature extraction ability, gradient explosion, and inability to process in parallel when processing time series data. Therefore, the present invention combines the advantages of graph convolutional neural network (GCN) and Transformer model. GCN can capture spatial dependencies between nodes, while Transformer can model long-term dependencies in time series. This combination enables the model to process transmission tower strain data with complex temporal and spatial dependencies and to achieve accurate prediction of the changing trend of strain data.

[0046] In addition, in this embodiment, there are many weak points on the transmission tower, which can generate a large amount of data in a short time. The multi-layer GCN module is used to extract the time feature sequence of the feature matrix. The complexity of the model can be expanded by increasing the number of GCN layers to cope with more complex tasks. For complex graphs containing a large number of nodes and edges, the multi-layer GCN can still maintain good performance and effectively extract time features.

[0047] In this embodiment, the method for obtaining a plurality of groups of initial strain value time series corresponding to a plurality of weak position towers on the transmission tower is:

[0048] A finite element model is established for the transmission tower to obtain several weak parts of the transmission tower corresponding to the weak parts where the strain is concentrated. Corresponding fiber grating sensors are installed on several weak parts of the tower, and several groups of initial strain value time series are collected through the fiber grating sensors. Since the strain of the weak parts of the tower changes quickly and the deformation is small, the fiber grating sensor can work stably under complex environmental conditions and collect strain data of the tower structure in real time. By analyzing the wavelength change of the light wave, the slight deformation of the tower can be reflected.

[0049] In this embodiment, the data preprocessing includes a missing data filling process and a data normalization process. The missing data filling process is as follows: the continuity of the collected initial strain value time series is verified, and the missing data is filled by the mean interpolation method. The missing data filling process and the data normalization process are performed to facilitate subsequent model training. According to a certain time window, the cleaned data is constructed into a time series to form a training model.

[0050] In this embodiment, the method in which the graph structure building module is used to build a feature matrix for a plurality of groups of valid initial strain value time series is:

[0051] The time when the initial strain value time series of the corresponding group of each weak part tower is generated is taken as a node, and the initial strain value time series of each weak part tower is taken as the feature vector corresponding to the node. The relationship between two nodes is represented by an edge, and a feature matrix is ​​constructed based on the nodes, the feature vectors corresponding to the nodes, and the relationship between the two nodes.

[0052] In this embodiment, Figure 2 and Figure 3 As shown, the method for extracting the features of the feature matrix by the multi-layer GCN module is:

[0053] The multi-layer GCN module is set to an input layer, a hidden layer and an output layer, the hidden layer is set to the first GCN layer and the second GCN layer, the input layer is used to input a feature matrix, and the time feature sequence of the feature matrix is ​​obtained by training the first GCN layer and the second GCN layer, and the output layer is used to output the time feature sequence.

[0054] In this embodiment, in GCN, data is organized in the form of a graph, where nodes represent individuals of strain data. i There is a eigenvector h i , which contains relevant information about the node, such as several sets of historical initial strain value time series. Edges represent the relationship between nodes. The core idea of ​​GCN is to aggregate node information through graph convolution operations. When processing time series data, the update of node features depends not only on its own features, but also on the features of its neighboring nodes.

[0055] Since the strain of the weak part of the tower changes quickly and the deformation is small, and a large amount of data can be generated in a short time, this embodiment only sets the first GCN layer and the second GCN layer, which can accurately extract the time feature sequence without increasing too much training time.

[0056] Specifically, a multi-layer GCN is designed to better capture features at different levels. A ReLU activation function is applied after each layer to increase the nonlinear expression ability of the model. The features extracted by GCN are processed using a Transformer model to capture the long-term dependencies of time series data. The self-attention mechanism is applied in the Transformer to allow the model to focus on the influence of different time steps when processing time series data. The input sequence is positionally encoded to maintain the order information of the time series.

[0057] In this embodiment, Figure 2 and Figure 4 As shown, the method for the Transformer model module to process the feature output prediction strain value time series is:

[0058] The Transformer model consists of an encoder stack and a decoder stack. The encoder stack is composed of multiple identical encoders. Each layer of encoders consists of a multi-head attention module and a fully connected feedforward network. Residual connections and standardization are introduced to prevent model degradation and improve convergence speed.

[0059] The decoder stack is composed of a plurality of identical decoders stacked together. A multi-head attention module is inserted into each layer of the decoder to perform a self-attention operation on the output of the encoder stack. The self-attention operation obtains the attention weight of the time feature of each moment to the time features of other moments by dynamically evaluating the relationship between the time features of each moment in the initial strain value time series and the one-to-one corresponding time features of multiple moments before and after, and then multiplies the attention weight of each moment by the corresponding time feature to obtain the risk value, and collects the risk values ​​of all moments to obtain the predicted strain value time series.

[0060] The application of the Transformer model can effectively capture complex time series features. Relying on a powerful self-attention mechanism, it solves the problems of inefficiency, gradient explosion, and parallelism that occur when traditional models process long sequences. Transformer relies on the self-attention mechanism to capture long-distance dependencies in sequences, has better time series modeling capabilities, and has great advantages in processing time series data. Figure 5 As shown in the figure, the core part of Transformer is the self-attention operation module. The self-attention mechanism enables the model to dynamically evaluate the relationship between each time step in the time series when processing strain data. In the strain data of the transmission tower, the strain value at a certain moment may be affected by multiple moments before and after. Through the self-attention mechanism, the model can learn these complex dependencies and assign different weights to each time step, thereby improving the accuracy of the prediction. The self-attention mechanism can be regarded as a fully connected layer that can dynamically update the weights. Due to the small number of parameters, multi-head attention is designed to learn richer information. The self-attention mechanism is a technology that can dynamically pay attention to information at different positions in the input sequence. The core idea is to determine the attention weight of each element to other elements by calculating the correlation between the elements in the input sequence. As shown in Figure 5 The following is a schematic diagram of the self-attention mechanism. First, the input time feature sequence is converted into a vector representation. This is usually achieved through an embedding layer or a linear transformation. For each element in the input sequence, three different vectors are generated: query, key, and value. The dot product between the query and the key is calculated to obtain the attention score. These scores represent the correlation between the elements in the sequence, and the calculation formula is shown below. In order to capture different features, a multi-head attention mechanism can be used to perform multiple independent attention calculations on the input sequence through multiple different projections, and then concatenate and project the results back to the original dimension.

[0061]

[0062] Among them, Q represents the query vector, which indicates the location information that the model is currently paying attention to, while K represents the key vector, which helps the model calculate the importance relative to the query vector, V represents the value vector, and d is the dimension of the key vector. The correlation between the query Q and the corresponding key K defines the self-attention matrix A. The output Z of the self-attention mechanism is the dot product of the value V and A. Self-attention is better at capturing the internal correlation of the features of the strain data.

[0063] In this embodiment, mean square error (MSE) is used as the loss function, and Adam optimizer is selected to update the model parameters. Early stopping and cross-validation are used to prevent overfitting and ensure the generalization ability of the model.

[0064] In this embodiment, several groups of initial strain value time series are divided into a training set and a test set. The training set is used to train the risk prediction model, and the performance of the risk prediction model is evaluated on the test set to calculate indicators such as the mean square error and the determination coefficient.

[0065] In this embodiment, multiple risk threshold ranges corresponding to risk levels are formulated, and the risks of the tower are divided into multiple levels (such as normal, warning, dangerous, etc.), which helps relevant managers to take necessary repair or reinforcement measures in time to ensure the long-term safe and stable operation of the tower.

[0066] Specifically, the mechanical simulation of the transmission tower is performed by finite element software to obtain the strain data of the transmission tower under various risks such as low risk, medium risk and high risk, and set it as the risk threshold of the transmission tower strain data.

[0067] In this embodiment, the real-time tower strain data obtained by the detection device is divided into a training set and a test set, and the strain value obtained by the prediction model proposed in the present invention is compared and analyzed with the true value. In order to further analyze the accuracy of the prediction results of the GCN+Transformer model, the prediction model is evaluated based on the following four indicators:

[0068] (1) Mean Squared Error (MSE): This indicator is the mean of the sum of squares of the errors between the predicted data and the true data. The smaller the value, the higher the accuracy.

[0069] (2) Root Mean Squared Error: RMSE (Root Mean Squared Error), which is the square root of MSE to eliminate the impact of dimension;

[0070] (3) Mean Absolute Error: MAE (Mean Absolute Error), which is used to evaluate the closeness between the predicted results and the real data set. The smaller the value, the better the effect.

[0071] (4) Coefficient of determination: R2 (R-Square), whose value range is [0,1]. The closer it is to 1, the stronger the ability of the predicted value to explain the true value, and the better the accuracy of the prediction model. The calculation formulas are as follows:

[0072]

[0073]

[0074]

[0075]

[0076] Where yi , are the measured value, simulated value and average value of the samples respectively; n is the number of samples.

[0077] Based on the strain data monitoring device of a transmission tower, the strain monitoring data with relatively complete data is selected as the experimental basic data. The time interval is 4 days from October 10, 2024 to October 14, 2024. After processing: the time interval is 5 seconds, 68217 data, and the effective initial strain value time series is obtained as follows Figure 6 As shown. The maximum strain at this point does not exceed 220με, and there is a certain regularity in the change of the effective initial strain value over time. 80% of the samples in the data set are used as training sets, and the remaining 20% ​​are used as test sets.

[0078] Figure 7 This is a visualization comparison of the predicted strain value and the true value after the GCN+Transformer model is trained. It can be observed from the chart that the change trend of the true value and the predicted strain value is roughly consistent, and the predicted strain value fluctuates between [172με, 176με], without a large deviation, and does not reach the risk threshold, indicating that the algorithm in this paper can effectively realize the strain data prediction of transmission towers, thereby realizing early warning of transmission tower risks.

[0079] From the evaluation indicators in Table 1, it can be seen that the mean square error MSE, root mean square error RMSE, and mean absolute error MAE of the algorithm proposed in the present invention are all at a relatively low level, indicating that the algorithm of the present invention has a high prediction accuracy. The determination coefficient R2 value is 0.99. The closer it is to 1, the stronger the ability of the predicted value to explain the true value, and the better the accuracy of the prediction model, indicating that the predicted value of the present invention has a strong fitting ability with the true value. From the above analysis, it can be seen that the transmission tower strain data prediction method based on the GCN+Transformer model can effectively realize the data prediction of the transmission tower in the future time period, and realize risk warning by comparing with the risk threshold.

[0080] Table 1 Model evaluation indicators

[0081] Evaluation indicators Risk early warning model Mean Squared Error (MSE) 7.17 RMSE (Root Mean Squared Error) 2.68 Mean Absolute Error (MAE) 1.72 Coefficient of determination R2(R-Square) 0.99

[0082] Example 2

[0083] Embodiment 2 provides a risk warning device based on transmission tower strain data, including:

[0084] A data acquisition module is used to acquire a plurality of groups of initial strain value time series corresponding to a plurality of weak position towers on a transmission tower, and perform data preprocessing on the plurality of groups of initial strain value time series to obtain a plurality of groups of effective initial strain value time series;

[0085] A risk prediction model building module is used to build a risk prediction model to train several groups of valid initial strain value time series to obtain a trained risk prediction model, wherein the risk prediction model includes a graph structure building module, a multi-layer GCN module, and a Transformer model module. The graph structure building module is used to build a feature matrix for several groups of valid initial strain value time series, and the multi-layer GCN module is used to extract the time feature sequence matrix of the feature matrix. The Transformer model module is used to process the time feature sequence to output a predicted strain value time series matrix;

[0086] The prediction module is used to formulate risk threshold ranges corresponding to multiple risk levels, and obtain the risk level prediction time series matrix corresponding to several weak-position towers according to the multiple risk threshold ranges based on the predicted strain value time series matrix.

[0087] Example 3

[0088] Embodiment 3 provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned risk warning method based on transmission tower strain data when executing the program.

[0089] Example 4

[0090] Embodiment 4 provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the risk warning method based on transmission tower strain data is implemented.

[0091] The memory in the embodiment of the present invention is used to store various types of data to support the operation of the electronic device. Examples of such data include: any computer program used to operate on the electronic device.

[0092] The risk warning method based on the strain data of the transmission tower disclosed in the embodiment of the present invention can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capability. In the implementation process, each step of the risk warning method based on the strain data of the transmission tower can be completed by the hardware integrated logic circuit or software instructions in the processor. The above-mentioned processor may be a general-purpose processor, a digital signal processor (DSP, Digital Signal Processor), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor can implement or execute the methods, steps and logic block diagrams disclosed in the embodiment of the present invention. The general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the method disclosed in the embodiment of the present invention can be directly embodied as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software module may be located in a storage medium, which is located in a memory. The processor reads the information in the memory and completes the steps of the risk warning method based on the strain data of the transmission tower provided in the embodiment of the present invention in combination with its hardware.

[0093] In an exemplary embodiment, the electronic device may be implemented by one or more application specific integrated circuits (ASIC), DSP, programmable logic device (PLD), complex programmable logic device (CPLD), FPGA, general purpose processor, controller, microcontroller (MCU), microprocessor, or other electronic components to execute the aforementioned method.

[0094] It can be understood that the memory can be a volatile memory or a non-volatile memory, and can also include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a flash memory, a magnetic surface memory, an optical disk, or a compact disc read-only memory (CD-ROM); the magnetic surface memory can be a disk memory or a tape memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory (SRAM), synchronous static random access memory (SSRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), direct memory bus random access memory (DRRAM). The memory described in the embodiments of the present invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0095] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A risk warning method based on transmission tower strain data, characterized in that: The following steps are involved: Obtaining a number of groups of initial strain value time series corresponding to a number of weak position towers on the transmission tower, and performing data preprocessing on the number of groups of initial strain value time series to obtain a number of groups of valid initial strain value time series; A risk prediction model is established to train several groups of valid initial strain value time series to obtain a trained risk prediction model, wherein the risk prediction model includes a graph structure construction module, a multi-layer GCN module, and a Transformer model module, a feature matrix is ​​constructed for several groups of valid initial strain value time series through the graph structure construction module, a time feature sequence matrix of the feature matrix is ​​extracted through the multi-layer GCN module, and the time feature sequence is processed through the Transformer model module to output a predicted strain value time series matrix; A risk threshold range corresponding to multiple risk levels is formulated, and a risk level prediction time series matrix corresponding to several weak-position towers is obtained based on the predicted strain value time series matrix and the multiple risk threshold ranges.

2. The risk warning method based on transmission tower strain data according to claim 1 is characterized in that: The data preprocessing includes a gap data filling process and a data normalization process. The gap data filling process method is: verifying the continuity of several groups of initial strain value time series collected, and filling the gap data by a mean interpolation method.

3. The risk warning method based on transmission tower strain data according to claim 1 is characterized in that: The method for obtaining several groups of initial strain value time series corresponding to several weak position towers on the transmission tower is: A finite element model is established for the transmission tower to obtain several weak position towers corresponding to the weak positions where the strain of the transmission tower is concentrated. Corresponding fiber grating sensors are installed one by one on the several weak position towers, and several groups of initial strain value time series are collected through the fiber grating sensors.

4. The risk warning method based on transmission tower strain data according to claim 1 is characterized in that: The method for the graph structure building module to build a feature matrix for several groups of valid initial strain value time series is: The time when the initial strain value time series of the corresponding group of each weak part tower is generated is taken as a node, and the initial strain value time series of each weak part tower is taken as the feature vector corresponding to the node. The relationship between two nodes is represented by an edge, and a feature matrix is ​​constructed based on the nodes, the feature vectors corresponding to the nodes, and the relationship between the two nodes.

5. The risk warning method based on transmission tower strain data according to claim 4 is characterized in that: The method used by the multi-layer GCN module to extract the features of the feature matrix is: The multi-layer GCN module is set to an input layer, a hidden layer and an output layer, the hidden layer is set to the first GCN layer and the second GCN layer, the input layer is used to input a feature matrix, and the time feature sequence of the feature matrix is ​​obtained by training the first GCN layer and the second GCN layer, and the output layer is used to output the time feature sequence.

6. The risk warning method based on transmission tower strain data according to claim 5 is characterized in that: The method used by the Transformer model module to process the feature output prediction strain value time series is: The Transformer model consists of an encoder stack and a decoder stack. The encoder stack is composed of multiple identical encoders. Each layer of encoders consists of a multi-head attention module and a fully connected feedforward network. Residual connections and standardization are introduced to prevent model degradation and improve convergence speed. The decoder stack is composed of a plurality of identical decoders stacked together. A multi-head attention module is inserted into each layer of the decoder to perform a self-attention operation on the output of the encoder stack. The self-attention operation obtains the attention weight of the time feature of each moment to the time features of other moments by dynamically evaluating the relationship between the time features of each moment in the initial strain value time series and the one-to-one corresponding time features of multiple moments before and after, and then multiplies the attention weight of each moment by the corresponding time feature to obtain the risk value, and collects the risk values ​​of all moments to obtain the predicted strain value time series.

7. The risk warning method based on transmission tower strain data according to claim 1 is characterized in that: The mean square error is used as the loss function, the Adam optimizer is selected to update the model parameters, and the early stopping method and cross-validation method are used to prevent overfitting.

8. A risk warning device based on transmission tower strain data, characterized in that: include: A data acquisition module is used to acquire a plurality of groups of initial strain value time series corresponding to a plurality of weak position towers on a transmission tower, and perform data preprocessing on the plurality of groups of initial strain value time series to obtain a plurality of groups of effective initial strain value time series; A risk prediction model building module is used to build a risk prediction model to train several groups of valid initial strain value time series to obtain a trained risk prediction model, wherein the risk prediction model includes a graph structure building module, a multi-layer GCN module, and a Transformer model module. The graph structure building module is used to build a feature matrix for several groups of valid initial strain value time series, and the multi-layer GCN module is used to extract the time feature sequence matrix of the feature matrix. The Transformer model module is used to process the time feature sequence to output a predicted strain value time series matrix; The prediction module is used to formulate risk threshold ranges corresponding to multiple risk levels, and obtain the risk level prediction time series matrix corresponding to several weak-position towers according to the multiple risk threshold ranges based on the predicted strain value time series matrix.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the risk warning method based on transmission tower strain data as described in any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the risk warning method based on transmission tower strain data as described in any one of claims 1 to 7 is implemented.

Images