A method and device for monitoring the state of a converter station based on a clustering neural network

Abstract

A method and apparatus for seismic status monitoring of converter stations based on clustering neural networks, relating to the field of seismic status monitoring. The method includes: responding to user operations for monitoring the operational status of different types of power equipment within a converter station under seismic conditions, acquiring parameter information for each power equipment in the pre-earthquake, mid-earthquake, and post-earthquake phases; performing time-series analysis of the parameter information using a multi-scale damage characterization system, and acquiring corresponding time-series data for the power equipment in the pre-earthquake, mid-earthquake, and post-earthquake phases; performing feature extraction on the time-series data according to a preset method, and acquiring parameter change trend information from the time-series data based on the feature extraction operation; acquiring status data for different types of power equipment under seismic conditions based on the parameter change trend information, the parameter information corresponding to the pre-earthquake, mid-earthquake, and post-earthquake phases, and using a group topology graph neural network; calculating the cluster centroid distance for each power equipment based on the input status data using a clustering statistical neural network; and calculating the comprehensive status assessment result for each power equipment within the converter station based on the cluster centroid distance through weighted calculation. The technical solution of this application solves the problems of insufficient decoupling of power equipment characteristics under complex seismic environments, lack of coupling effect, and inaccurate assessment of abnormal operating conditions that exist in traditional monitoring methods.

Description

Technical Field

[0001] This application relates to the field of seismic status monitoring, and in particular to a method and device for seismic status monitoring of converter stations based on clustering neural networks. Background Technology

[0002] Converter stations, as key facilities in the power system, are responsible for transmitting high-voltage direct current (HVDC) and ensuring the stable operation of the power grid. However, natural disasters such as earthquakes can damage or malfunction equipment within the converter station, affecting the security of power supply.

[0003] Equipment operation is typically monitored through manual inspections or local sensors, relying on the detection of single parameters such as vibration and temperature. However, due to the limitations of these monitoring methods, they cannot comprehensively capture multi-parameter information, and their processing methods are relatively traditional, resulting in slow response speeds and low accuracy. Consequently, when there are various types of power equipment in a converter station, not only is monitoring and analysis time-consuming, but it is also impossible to accurately determine the abnormal operating states of each type of power equipment. In other words, there is a problem of inaccurate assessment of abnormal operation of power equipment in complex environments.

[0004] Therefore, there is an urgent need for a method and device for monitoring the seismic status of converter stations based on clustering neural networks. Summary of the Invention

[0005] This application provides a method and device for monitoring the seismic status of converter stations based on clustering neural networks, which solves the problems of insufficient decoupling of power equipment characteristics under complex seismic environments, lack of coupling effect, and inaccurate assessment of abnormal operating status by traditional monitoring methods.

[0006] The first aspect of this application provides a method for monitoring the seismic status of a converter station based on a clustering neural network. The method includes: responding to a user's operation to monitor the operating status of different types of power equipment within the converter station under seismic conditions; acquiring parameter information for each power equipment in the pre-earthquake, mid-earthquake, and post-earthquake phases, including vibration, strain, acceleration, and temperature information; performing time-series analysis on the parameter information using a multi-scale damage characterization system, and acquiring corresponding time-series data for the power equipment in the pre-earthquake, mid-earthquake, and post-earthquake phases; performing feature extraction on the time-series data according to a preset method, and based on... Feature extraction is performed to obtain parameter variation trend information from time-series data, including vibration variation trend information, strain variation trend information, acceleration variation trend information, and temperature variation trend information. Based on the parameter variation trend information and the parameter information corresponding to the pre-earthquake, mid-earthquake, and post-earthquake stages, and using a group topology graph neural network, the state data corresponding to different types of power equipment under the earthquake environment are obtained. By inputting the state data, the cluster centroid distance corresponding to each power equipment is calculated using a clustering statistical neural network. Based on the cluster centroid distance, the comprehensive state assessment result corresponding to each power equipment in the converter station is calculated through weighted calculation.

[0007] Optionally, parameter information for each power device during the pre-earthquake, mid-earthquake, and post-earthquake phases can be obtained. Specifically, this includes: deploying a sensor array along the converter station using pre-set fiber optic sensing technology, including distributed fiber Bragg grating technology and Brillouin optical time-domain reflectometry; using pre-set fiber optic interferometry technology, including Michelson interferometry and Mach-Zehnder interferometry, to sense different types of power devices using sensitized optical fiber sensing arrays; and obtaining parameter information for each power device during the pre-earthquake, mid-earthquake, and post-earthquake phases based on the sensor array and the sensitized optical fiber sensing array.

[0008] Optionally, before performing time-series analysis of parameter information through a multi-scale damage characterization system and obtaining corresponding time-series data of power equipment in the pre-earthquake, mid-earthquake, and post-earthquake phases, it is necessary to construct a multi-scale damage characterization system. Specifically, this includes: establishing a cross-scale damage feature encoding model of vibration spectrum-strain gradient-temperature field, and realizing time-series evolution modeling of equipment damage state through gated cyclic units.

[0009] Optionally, a multi-scale damage characterization system is used to perform time-series analysis on the parameter information and obtain the corresponding time-series data of power equipment in the pre-earthquake, mid-earthquake, and post-earthquake phases. Specifically, this includes: performing time-series reconstruction on the parameter information, which is used to segment the parameter information according to different time windows; normalizing the parameter information after time-series reconstruction to reduce noise interference and enhance the comparability of the parameter information; and performing time-series analysis on the normalized parameter information based on a time step strategy to obtain the corresponding time-series data of power equipment in the pre-earthquake, mid-earthquake, and post-earthquake phases.

[0010] Optionally, state data corresponding to different types of power equipment under seismic conditions are obtained based on a group topology graph neural network. Specifically, this includes: obtaining the layout structure of different types of power equipment in the converter station, and encoding different types of power equipment as graph structure nodes in the group topology graph neural network based on the layout structure; modeling the seismic wave propagation path using a graph attention network based on the group topology graph neural network to capture the vibration coupling effect between different types of power equipment; and obtaining state data corresponding to different types of power equipment under seismic conditions based on the vibration coupling effect.

[0011] Optionally, by inputting state data, the cluster centroid distance corresponding to each power device is calculated according to a clustering statistical neural network. Specifically, this includes: dividing the state data of each power device into several clusters based on a clustering algorithm, with each cluster representing a type of abnormal state of the device; calculating the centroid value corresponding to each cluster, where the centroid value corresponds to the optimal center point of the state data in each cluster; and calculating the cluster centroid distance corresponding to each power device based on the centroid value.

[0012] Optionally, the cluster centroid distance for each power device is calculated based on the centroid value, specifically including: calculating the cluster centroid distance for each power device based on the centroid value using the following formula: ; in, For the first The power equipment and the first The centroid distance between clusters For the first The electrical equipment in the first State data across each feature dimension For the first The weight coefficients corresponding to each feature dimension Indicates the number of feature dimensions. For the first The cluster in the th order of ... Centroid values ​​on each feature dimension Indicates the first The standard deviation corresponding to each feature dimension The normalization coefficient is... This indicates the number of abnormal features in the parameter change trend information. Indicates the first Reference values ​​corresponding to each abnormal feature information These are the sensitivity adjustment parameters.

[0013] Optionally, based on the cluster centroid distance, a weighted comprehensive status assessment result is calculated for each power device in the converter station. Specifically, this includes: determining whether the cluster centroid distance is greater than a preset abnormal threshold; if the cluster centroid distance is greater than the preset abnormal threshold, then the comprehensive status assessment result for the power device is confirmed as an abnormal operating state; if the cluster centroid distance is less than or equal to the preset abnormal threshold, then the comprehensive status assessment result for the power device is confirmed as a normal operating state.

[0014] Optionally, after confirming that the comprehensive status assessment result of the power equipment is in an abnormal operating state if the cluster centroid distance is greater than a preset anomaly threshold, the method further includes: obtaining the anomaly type corresponding to the power equipment based on the cluster centroid distance; obtaining the equipment maintenance plan corresponding to the anomaly type from a preset anomaly handling plan database, wherein the preset anomaly handling plan database is used to store the correspondence between anomaly types and equipment maintenance plans; and sending the equipment maintenance plan to the user's terminal device.

[0015] A second aspect of this application provides a seismic status monitoring device for converter stations based on clustering neural networks. The device includes an acquisition module and a processing module, wherein... The acquisition module responds to user requests for monitoring the operational status of different types of power equipment within the converter station under seismic conditions. It acquires parameter information for each power equipment during the pre-earthquake, mid-earthquake, and post-earthquake phases, including vibration, strain, acceleration, and temperature information. It performs time-series analysis of the parameter information using a multi-scale damage characterization system, acquiring corresponding time-series data for each power equipment during these phases. It then performs feature extraction on the time-series data according to a preset method, obtaining parameter change trends, including vibration, strain, acceleration, and temperature trends. Based on these trends and the corresponding parameter information for each phase of the earthquake, it uses a group topology neural network to acquire the status data for different types of power equipment under seismic conditions.

[0016] The processing module is used to calculate the cluster centroid distance for each power device based on the input status data and a clustering statistical neural network; and to calculate the comprehensive status assessment result for each power device in the converter station based on the cluster centroid distance through weighted calculation.

[0017] A third aspect of this application provides an electronic device including a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described above.

[0018] A fourth aspect of this application provides a computer-readable storage medium storing a computer program, which is executed by a processor using the method described in any of the foregoing descriptions.

[0019] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. Obtain parameter information for each power device at each earthquake stage; obtain time-series data for each power device at each earthquake stage; extract parameter change trend information from the time-series data based on feature extraction; obtain state data for different types of power devices under earthquake conditions based on parameter change trend information and parameter information corresponding to each earthquake stage; calculate the cluster centroid distance for each power device based on clustering statistical neural network by inputting state data; calculate the comprehensive state assessment result for each power device in the converter station based on the cluster centroid distance by weighted calculation, thereby improving the accuracy of abnormal state identification of power devices in complex environments within the converter station by effectively integrating multi-parameter information, overcoming the limitations of traditional monitoring methods such as slow response speed, single monitoring means, and inaccurate anomaly assessment, and improving the intelligence and reliability of power device operation status monitoring under earthquake conditions.

[0020] 2. By pre-setting fiber optic sensing technology, a sensor array is deployed along the converter station. Furthermore, by pre-setting fiber optic interferometry technology, an enhanced fiber optic sensing array is installed around different types of power equipment. Based on the sensor array and the enhanced fiber optic sensing array, parameter information corresponding to each power equipment in the early, middle, and late stages of an earthquake can be obtained. This enables high-precision, real-time monitoring of different types of power equipment within the converter station, enhancing the ability to acquire power equipment status data under earthquake conditions.

[0021] 3. Based on clustering algorithms, the status data of each power device is divided into several clusters, each cluster representing a type of abnormal device status. The centroid value corresponding to each cluster is calculated, which is the value corresponding to the optimal center point of the status data in each cluster. Based on the centroid value, the cluster centroid distance corresponding to each power device is calculated, thereby realizing refined classification and anomaly identification of the operating status of power devices. This adapts to different equipment types and their complex operating characteristics, improves the reliability of equipment status assessment in complex seismic environments, and provides key support for subsequent comprehensive status assessment and maintenance decisions. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating a method for monitoring the seismic status of a converter station based on a clustering neural network, provided in an embodiment of this application. Figure 2 This is a schematic diagram of a converter station seismic status monitoring device based on a clustering neural network, provided in an embodiment of this application. Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0023] Explanation of reference numerals in the attached figures: 21. Acquisition module; 22. Processing module; 301. Processor; 302. Communication bus; 303. User interface; 304. Network interface; 305. Memory. Detailed Implementation

[0024] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0025] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to and includes any or all possible combinations of one or more of the listed items.

[0026] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0027] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0028] Please refer to Figure 1 The flowchart illustrates a method for monitoring the seismic status of a converter station based on a clustering neural network, provided in an embodiment of this application. The method is applied to a server, and the flowchart mainly includes the following steps: steps S101 to S105.

[0029] Step S101: In response to the user's operation to monitor the operating status of different types of power equipment in the converter station under earthquake conditions, obtain the parameter information of each power equipment in the early stage, middle stage and late stage of the earthquake.

[0030] Specifically, in response to user monitoring of the operational status of different types of power equipment within the converter station under seismic conditions, parameter information for each power device is acquired during the pre-earthquake, mid-earthquake, and post-earthquake phases. Specifically, a distributed fiber optic sensing system and an enhanced optical fiber sensing array are deployed at the converter station to monitor various parameters of the power equipment in real time, including vibration, strain, acceleration, and temperature. This parameter information helps to comprehensively assess the status changes of the equipment at different seismic stages, providing fundamental data for subsequent equipment health monitoring and anomaly detection. In the embodiments of this application, the power equipment includes, but is not limited to: converter transformers, valve cooling systems, DC field equipment, converter valves, and smoothing reactors.

[0031] In one possible implementation, step S101 further includes: deploying a sensor array along the converter station using preset fiber optic sensing technology, including distributed fiber Bragg grating technology and Brillouin optical time-domain reflectometry; using preset fiber optic interferometry technology, including Michelson interferometry and Mach-Zehnder interferometry; and acquiring parameter information for each power device in the pre-earthquake, mid-earthquake, and post-earthquake phases based on the sensor array and the enhanced fiber optic interferometry array.

[0032] Specifically, distributed fiber Bragg grating technology, by embedding Bragg gratings as sensing elements within optical fibers, can measure the strain and temperature changes of equipment. For example, through fiber optic sensing systems or sensitized optical cable arrays, it can specifically monitor coolant flow rate, valve temperature distribution, and thyristor stress in converter valves. This technology provides high-resolution spatial distribution monitoring, acquiring strain information along the entire length of the fiber, suitable for monitoring the overall stress and local deformation of equipment. Simultaneously, Brillouin optical time-domain reflectometry (BDR) analyzes the scattering characteristics of light waves in optical fibers to obtain high-precision strain and temperature information, enabling real-time monitoring over long distances, suitable for long-term monitoring of large-scale power equipment. By combining these technologies, fiber optic sensing arrays can perceive the multi-dimensional state of equipment in real time, providing reliable raw data for subsequent analysis.

[0033] Furthermore, fiber optic interferometry is pre-configured by deploying enhanced optical fiber sensing arrays around different types of power equipment to further improve the accuracy and sensitivity of monitoring. Fiber optic interferometry includes Michelson interferometry and Mach-Zehnder interferometry. Michelson interferometry compares the phase difference between two beams of light using the principle of interference, making it suitable for measuring minute displacement changes, such as vibration changes on the surface of equipment. Mach-Zehnder interferometry can effectively measure changes in temperature and strain, and can simultaneously acquire the responses of multiple signal sources, exhibiting high anti-interference capability and accuracy.

[0034] Step S102: Perform time-series analysis on parameter information through a multi-scale damage characterization system, and obtain time-series data of power equipment in the early, middle and late stages of an earthquake.

[0035] Specifically, time-series analysis is performed on the parameter information to obtain time-series data for each power device during the pre-earthquake, mid-earthquake, and post-earthquake phases. Specifically, physical parameters such as vibration, strain, acceleration, and temperature, as well as electrical parameters such as DC voltage fluctuations at the power station, AC bus voltage distortion rate, and current imbalance, are collected from the power devices at different earthquake stages (pre-earthquake, mid-earthquake, and post-earthquake). Time-series analysis methods (such as Fast Fourier Transform and filtering techniques) are then used to process this data. Through time-series analysis, the trends of parameter changes at different time points can be extracted, the response characteristics of the equipment during the earthquake can be identified, and time-series data for each stage can be obtained, providing a data foundation for subsequent condition assessment and anomaly detection.

[0036] In one possible implementation, step S102 further includes: establishing a cross-scale damage feature encoding model of vibration spectrum-strain gradient-temperature field, and realizing the temporal evolution modeling of equipment damage state through a gated loop unit.

[0037] Specifically, based on frequency domain feature extraction of vibration spectrum signals, spatiotemporal coupling analysis of strain gradient distribution, and modeling of temperature field evolution trends, a multidimensional damage characterization feature vector with cross-scale fusion is constructed. This feature vector is input into a gated recurrent unit (GRU) network, and dynamic modeling of the potential damage process of the equipment is achieved through a memory mechanism and a state update mechanism for damage feature sequences within historical time windows. During the modeling process, an attention mechanism is used to enhance the responsiveness to changes in key time segments, and a damage evolution representation with deep time dependence is obtained through multi-layer stacked GRU units. Finally, a high-dimensional temporal feature tensor representing the equipment state evolution trajectory throughout the earthquake is output, providing accurate and engineering-adjudicable basic data support for subsequent feature extraction operations.

[0038] In one possible implementation, step S102 further includes: performing a time series reconstruction operation on the parameter information, wherein the time series reconstruction operation is used to segment the parameter information according to different time windows; normalizing the parameter information after the time series reconstruction operation to reduce noise interference and enhance the comparability of the parameter information; and performing time series analysis on the normalized parameter information based on a time step strategy to obtain the corresponding time series data of the power equipment in the early, middle and late stages of the earthquake.

[0039] Specifically, a time-series reconstruction operation is performed on the parameter information to segment the time-series data according to different time windows, ensuring that the dynamic changes of the equipment during the earthquake can be effectively captured. In the time-series reconstruction operation, parameter information (such as vibration, strain, acceleration, temperature, etc.) in the pre-earthquake, mid-earthquake, and post-earthquake phases is first divided into multiple small time segments, typically according to preset time windows, including but not limited to every 10 seconds, 30 seconds, or 1 minute. This segmentation of the raw data allows for better processing and analysis of the detailed dynamic responses of the equipment at different earthquake stages.

[0040] Next, the parameter information after time series reconstruction is normalized. The purpose of normalization is to convert different types of parameter data (vibration, strain, acceleration, temperature, etc.) into the same standard range, typically normalized to the interval [0,1] or [-1,1]. The embodiments in this application do not limit the normalization interval and can be selected according to actual needs. This step helps reduce scale differences between different parameters, eliminates interference caused by different data units, enhances the comparability of various parameters in subsequent analyses, and ensures that the influence of each parameter on the final analysis results is balanced.

[0041] After normalization, time-series analysis is performed on the normalized parameter information based on a time-step strategy. The time-step strategy determines the data length at each time point and the data segmentation step size (e.g., an analysis window every 5 or 10 seconds), which helps capture subtle changes in equipment status at different time points. Time-series analysis methods can use common techniques such as sliding window method, Fourier transform, and time-domain analysis to analyze the variation patterns of various parameters over different time periods and identify the characteristic performance of equipment in the pre-, mid-, and post-earthquake phases.

[0042] Step S103: Perform feature extraction on the time series data according to a preset method, and obtain parameter change trend information in the time series data based on the feature extraction operation.

[0043] Specifically, in this step, representative and important features are first extracted from the time-series data obtained in step S102. These features can reflect the dynamic response of power equipment in the early, middle, and late stages of an earthquake. Feature extraction methods include, but are not limited to, the following: Time-domain feature extraction: By calculating the basic statistical characteristics of the time-series data, such as mean, standard deviation, maximum, minimum, skewness, and kurtosis, the basic distribution of the data is obtained; Frequency-domain feature extraction: Using methods such as Fourier transform (FFT), the time-series data is converted into the frequency domain to analyze the intensity and distribution of different frequency components in the data, and to identify information such as the vibration frequency and amplitude changes of the equipment; Time-frequency feature extraction: By using methods such as wavelet transform, the local time-frequency features of the time-series data are extracted. The above methods are very effective in capturing the instantaneous changes of signals such as vibration and acceleration, and are especially suitable for the analysis of non-stationary signals; Trend analysis features: By using trend detection methods (such as moving average, linear regression, etc.), the long-term trend of the equipment over different time periods is extracted. For example, monitoring the long-term trend of temperature or strain helps to identify whether the equipment is gradually becoming abnormal.

[0044] After obtaining the temporal characteristics of each parameter (such as vibration, strain, acceleration, temperature, etc.) through the above feature extraction operations, the changing trend information of each parameter is further analyzed. This trend information reflects the response of the equipment during the earthquake and can reveal whether the operating status of the equipment has undergone abnormal changes. The parameter changing trend information can be analyzed from the following perspectives: Vibration changing trend: By extracting and analyzing the features of vibration signals, the vibration amplitude and frequency changes of the equipment can be identified. During an earthquake, the vibration response of the equipment will change to a certain extent, possibly showing a short-term increase in amplitude or frequency change, thus reflecting the abnormal state of the equipment; Strain changing trend: The strain change of power equipment reflects the stress state of the equipment. By analyzing the trend of strain data, it is possible to identify whether the equipment has been subjected to excessive external force leading to deformation, and thus determine whether there is a potential risk of damage to the equipment; Acceleration changing trend: Acceleration signals are used to reflect the impact and vibration of the equipment during an earthquake. By analyzing the trend of acceleration changes, the dynamic response of the equipment can be understood, and it can be identified whether there is excessive vibration or irregular change pattern; Temperature changing trend: The temperature change of the equipment is an important indicator for judging its operating status. During an earthquake, equipment temperature changes may be caused by external vibrations, load changes, and other factors. Therefore, monitoring temperature change trends can help identify whether the equipment is affected or has overheated or other abnormal phenomena.

[0045] For different parameters, weighted integration can be performed based on their changing trends to obtain comprehensive trend information. For example, different trend information can be merged according to the relative importance or weight of each parameter to form a comprehensive equipment status change trend assessment model. This comprehensive assessment can not only reflect the changes of individual parameters, but also identify the overall changes in equipment status through multi-dimensional data fusion. Finally, the parameter changing trend information extracted from the time-series data is summarized to generate an equipment status change trend report. This report details the operating status of the equipment in the pre-earthquake, middle-earthquake, and post-earthquake phases, as well as the changing trend of each parameter. This provides important data support and reference for subsequent status assessment, anomaly detection, and fault diagnosis.

[0046] Step S104: Based on the parameter change trend information, the parameter information corresponding to the pre-earthquake, mid-earthquake, and post-earthquake stages, and using the group topology graph neural network, obtain the state data corresponding to different types of power equipment under the earthquake environment.

[0047] Specifically, based on the parameter change trend information and the parameter information corresponding to the early, middle and late stages of an earthquake, status data of different types of power equipment under earthquake conditions are obtained. The status data not only includes the change trend information of parameter information, but also includes snapshots of the equipment's operating status at specific moments. Status data generally includes: instantaneous parameter values ​​(such as vibration acceleration, temperature, current, etc. at the current moment), historical change trends (such as the root mean square value of vibration in the past 10 seconds, the temperature growth rate in the past hour), abnormal characteristics (such as the magnitude of strain data deviating from normal values), clustering statistics (such as the status clustering results of multiple sensors, whether it belongs to an abnormal cluster), etc.

[0048] In one possible implementation, step S104 further includes: obtaining the layout structure of different types of power equipment in the converter station, and encoding the different types of power equipment as graph structure nodes in the group topology graph neural network according to the layout structure; using the group topology graph neural network to model the seismic wave propagation path and capture the vibration coupling effect between different types of power equipment; and obtaining the state data corresponding to different types of power equipment under the seismic environment according to the vibration coupling effect.

[0049] Specifically, firstly, based on the converter station design drawings and operational deployment data, a spatial distribution map of the power equipment is constructed, and the physical adjacency relationships, functional correlations, and electrical connection paths between the equipment are extracted. According to the layout structure, each equipment node is mapped to a graph structure node in a graph neural network. Simultaneously, category attribute labels and initial feature vectors are assigned to different types of power equipment within the graph structure. Subsequently, based on the edge weights in the group topology graph neural network representing the vibration coupling strength between equipment, a graph attention mechanism is introduced to weighted model the dynamic response differences between adjacent equipment at different stages of seismic wave propagation, thereby identifying key coupling paths and typical vibration response patterns. To improve the model's ability to analyze complex vibration propagation paths, a multi-level graph attention module is introduced to capture cross-level vibration propagation effects, and a space-time joint feature enhancement mechanism is combined to dynamically update the node states. On this basis, combining the structural response sensitivity, coupling strength spectrum distribution, and historical vibration feedback data of each type of power equipment, a node-level state representation vector is calculated, forming state data reflecting the comprehensive response characteristics of various types of power equipment under the influence of the seismic environment, which serves as input to the subsequent clustering statistical neural network.

[0050] Step S105: By inputting the status data, the cluster centroid distance corresponding to each power device is calculated according to the clustering statistical neural network.

[0051] Specifically, the equipment status data generated in step S104 (including instantaneous parameter values, historical trends, abnormal features, clustering statistics, etc.) is input into the clustering statistical neural network. Based on the input data, the clustering statistical neural network groups the equipment status data into multiple clusters, each cluster representing a group of devices in similar states. The clustering algorithm calculates the distance between each power device and the centroid (i.e., the center point of the cluster) of its respective cluster, obtaining the cluster centroid distance. The cluster centroid distance reflects the degree of deviation between the equipment status and the cluster center; a larger value indicates a more abnormal equipment status, while a smaller value indicates a more normal equipment status. Finally, the cluster centroid distance calculated by the clustering statistical neural network will serve as one of the bases for equipment status assessment, providing data support for subsequent equipment health status judgment and fault diagnosis.

[0052] In one possible implementation, step S105 further includes: dividing the status data of each power device into several clusters based on a clustering algorithm, with each cluster representing a device abnormal state; calculating the centroid value corresponding to each cluster, where the centroid value corresponds to the optimal center point of the status data in each cluster; and calculating the cluster centroid distance corresponding to each power device based on the centroid value.

[0053] Specifically, firstly, the state data of each power device is clustered according to a pre-defined clustering algorithm (such as K-means, DBSCAN, spectral clustering, etc.). The purpose of clustering is to group device data with similar states into the same cluster, while device data with different states are assigned to different clusters. Each cluster represents an abnormal state of a device, including but not limited to: failure of the flow valve cooling system, breakage of the reactor winding, deformation of the converter transformer tank, etc. The states of devices within a cluster are relatively similar. After clustering, the next step is to calculate the centroid value of each cluster. The centroid value is the optimal center point of all device state data within a cluster, representing the "average" or "center" value of the device state in that cluster. The centroid is usually calculated as the mean or weighted average of all device state data within the cluster. For example, for multidimensional state data (such as vibration, temperature, acceleration, etc.), the centroid value is the mean of these parameters in the device state data within that cluster. For each power device, the distance between its state data and the centroid value of its cluster is calculated; this distance is the cluster centroid distance. The cluster centroid distance can be calculated using Euclidean distance. The cluster centroid distance for each power device is calculated based on its centroid value using the following formula: ; in, For the first The power equipment and the first The centroid distance between clusters For the first The electrical equipment in the first State data across each feature dimension For the first The weight coefficients corresponding to each feature dimension Indicates the number of feature dimensions. For the first The cluster in the th order of ... Centroid values ​​on each feature dimension Indicates the first The standard deviation corresponding to each feature dimension The normalization coefficient is... This indicates the number of abnormal features in the parameter change trend information. Indicates the first Reference values ​​corresponding to each abnormal feature information This is a sensitivity adjustment parameter. The calculated cluster centroid distance reflects the difference between the equipment status data and the normal operating state (or other abnormal state). If the cluster centroid distance is small, it indicates that the equipment status is close to the cluster center, indicating that the equipment is operating normally; if the cluster centroid distance is large, it indicates that the equipment status deviates from the normal range, and there may be an abnormality or malfunction. These cluster centroid distances can be used to assess the equipment status and determine whether it is in normal operating condition or abnormal. Based on the calculated cluster centroid distance, further equipment status assessments can be performed. When the cluster centroid distance of the equipment exceeds a certain threshold, it indicates that the equipment may be in an abnormal state, requiring further fault diagnosis or maintenance measures. Different equipment may have different abnormality thresholds; cluster analysis helps to dynamically adjust these thresholds to adapt to the different operating conditions and characteristics of different equipment.

[0054] Step S106: Based on the cluster centroid distance, calculate the comprehensive status assessment results corresponding to each power equipment in the converter station through weighted calculation.

[0055] Specifically, the cluster centroid distance of each power device represents the distance between the device's state and its normal or abnormal state. The system determines whether the cluster centroid distance exceeds a preset abnormality threshold. If the cluster centroid distance exceeds the preset abnormality threshold, the overall state assessment result for the power device is confirmed as an abnormal operating state; if the cluster centroid distance is less than or equal to the preset abnormality threshold, the overall state assessment result for the power device is confirmed as a normal operating state. Typically, the preset abnormality threshold can be set in, but is not limited to, the following ways: Based on statistical methods: for example, a cluster centroid distance exceeding a certain standard deviation of the mean cluster centroid distance can be defined as abnormal; Based on historical data: by analyzing the distribution of cluster centroid distances during normal operation, a normal range can be set, and anything exceeding this range is considered abnormal; Based on machine learning models: using supervised or unsupervised learning methods to train a model to automatically identify abnormalities. The overall state assessment result is obtained by weighting the cluster centroid distances, resulting in a comprehensive health score for each power device, expressed by the formula: ; in, A comprehensive health score for all electrical equipment. This represents the total number of different types of electrical equipment. For the first The weights corresponding to different types of power equipment are used to evaluate the overall operating status of the equipment. Higher values ​​indicate that the equipment is in an abnormal state, while lower values ​​indicate that the equipment is in a good state.

[0056] In one possible implementation, step S106 further includes: obtaining the anomaly type corresponding to the power equipment based on the cluster centroid distance; obtaining the equipment maintenance plan corresponding to the anomaly type from a preset anomaly handling plan database, wherein the preset anomaly handling plan database is used to store the correspondence between anomaly types and equipment maintenance plans; and sending the equipment maintenance plan to the user's terminal device.

[0057] Specifically, anomaly types are determined through different feature values, using a feature library (multi-scale damage characterization system). In the embodiments of this application, the methods for confirming anomaly types include, but are not limited to, the following: Cluster-based labeling: If the centroid distance of a device's cluster exceeds a threshold, the characteristics of its respective cluster can be viewed, and the anomaly type determined based on these characteristics. For example, some clusters may represent temperature anomalies, while others may represent vibration anomalies; Distance-based analysis: By calculating the distance between the device's state and the centroids of multiple clusters, it can be determined which type of anomaly the device is closer to. For example, the device's state may be closer to the "overload," "overheat," or "vibration anomaly" clusters; Multi-dimensional anomaly analysis: The anomaly type is determined based on the changing trends of multiple parameters (such as temperature, vibration, acceleration, etc.). For example, an increase in temperature and acceleration may indicate device overload or component wear. A preset anomaly handling scheme database is a database that stores the mapping relationship between anomaly types and device maintenance schemes. Its purpose is to automatically obtain the most suitable maintenance scheme based on the detected anomaly type. The construction of this database involves the following aspects: First, the anomaly handling solution database needs to incorporate maintenance rules specific to the converter station and optimize the rule base based on expert experience. Therefore, it is necessary to define all possible anomaly types, such as vibration anomalies (including converter valve vibration, converter transformer vibration, smoothing reactor vibration, etc.); strain anomalies (including converter valve cooling system pipe strain, converter transformer casing strain, GIS (gas-insulated switchgear) strain, etc.); and temperature anomalies (including converter valve temperature anomalies, converter transformer oil temperature anomalies, busbar temperature anomalies, etc.). Each type may also be divided into different severity levels (e.g., minor, moderate, severe). For each anomaly type and its severity, corresponding maintenance plans are developed. For example, for minor vibration anomalies, it may be recommended to check the connections and perform preventative maintenance; while for severe temperature anomalies, it may be necessary to immediately shut down the system for inspection and replacement of faulty components.Establish a mapping relationship between anomaly types and equipment maintenance plans. For example: Vibration anomaly -> Check vibration sensors, clean equipment, check motors: Inspect the converter valve support structure, confirm the bolt tightness, and reinforce or adjust the support structure if necessary; Monitor the vibration mode of the converter transformer through fiber optic sensing systems or vibration sensors. If the vibration exceeds the standard, check the winding support structure to confirm whether reinforcement or replacement of buffer materials is needed; Inspect the fixing bracket of the smoothing reactor, adjust the fixing bolts and add damping measures to reduce the impact of vibration, etc.; Temperature anomaly -> Check the cooling system, check temperature control equipment, clean the radiator: Monitor the operating status of the converter valve cooling system, including coolant flow rate, cooling fan speed, etc. If anomalies are found, check the coolant replenishment, whether there is any blockage in the cooling system pipes, and replace the coolant or clean the cooling channels if necessary; Inspect the converter transformer... Analyze abnormal oil temperature in converters, check the oil circulation system for normal operation, and perform oil sample testing if necessary to confirm the presence of aging or impurity deposits. Use thermal imaging technology to monitor the busbar temperature distribution; if localized overheating is detected, check for loose joints and retighten or replace contacts. For abnormal strain, check structural components and repair or replace damaged parts: use fiber optic grating sensors to monitor the strain distribution in the converter valve cooling pipes; if localized strain exceeds the standard, check for loose pipe connections and replace damaged pipes or reinforce the support structure if necessary. Perform external stress testing on the converter transformer tank; if significant deformation is found, consider adjusting the external support structure and check the integrity of tank welds and sealing strips. For abnormal strain in GIS equipment, check the integrity of sealing parts and perform SF6 gas leak testing; if necessary, perform localized welding repair or replace seals. In the pre-set abnormality handling solution database, abnormality types and equipment maintenance solutions are established through pre-defined rules. For example, if the cluster centroid distance is greater than the preset abnormality threshold and the abnormality type is "vibration abnormality," the corresponding equipment maintenance plan might be "check the vibration sensor and equipment connection parts." If the cluster centroid distance is greater than the preset abnormality threshold and the abnormality type is "temperature abnormality," the corresponding equipment maintenance plan might be "check the cooling system and temperature control device." Once the abnormality type of the equipment is determined and the corresponding maintenance plan is obtained, the system will send the maintenance plan to the user's terminal device (such as the operator's mobile phone, monitoring system, or maintenance personnel's work terminal) through an appropriate communication channel. In this way, the user can obtain the equipment fault information and the maintenance measures to be taken in a timely manner, thereby responding quickly and taking action to ensure the continuous and stable operation of the equipment.

[0058] This application employs the aforementioned method to obtain parameter information corresponding to each power device at each earthquake stage; obtain time-series data corresponding to each power device at each earthquake stage; obtain parameter change trend information from the time-series data through feature extraction; obtain state data corresponding to different types of power devices under earthquake conditions based on the parameter change trend information and the parameter information corresponding to each earthquake stage; calculate the cluster centroid distance corresponding to each power device based on the input state data and a clustering statistical neural network; and calculate the comprehensive state assessment result corresponding to each power device in the converter station based on the cluster centroid distance through weighted calculation. This effectively integrates multi-parameter information, improves the accuracy of abnormal state identification of power devices in complex environments within the converter station, overcomes the limitations of traditional monitoring methods such as slow response speed, single monitoring means, and inaccurate abnormal assessment, and enhances the intelligence and reliability of power device operation status monitoring under earthquake conditions.

[0059] Please refer to Figure 2 The diagram illustrates a module schematic of a converter station seismic status monitoring device based on a clustering neural network, according to an embodiment of this application. The device is a server, which includes an acquisition module 21 and a processing module 22. The acquisition module 21 is used to respond to user operations for monitoring the operating status of different types of power equipment in the converter station under earthquake conditions. It acquires parameter information for each power equipment in the pre-earthquake, mid-earthquake, and post-earthquake phases, including vibration, strain, acceleration, and temperature information. It performs time-series analysis of the parameter information using a multi-scale damage characterization system and acquires corresponding time-series data for the power equipment in the pre-earthquake, mid-earthquake, and post-earthquake phases. It performs feature extraction on the time-series data according to a preset method and acquires parameter change trend information, including vibration, strain, acceleration, and temperature change trends. Based on the parameter change trend information and the corresponding parameter information in the pre-earthquake, mid-earthquake, and post-earthquake phases, it acquires the state data corresponding to different types of power equipment under earthquake conditions using a group topology graph neural network.

[0060] Processing module 22 is used to calculate the cluster centroid distance corresponding to each of the power devices by inputting the status data and using a clustering statistical neural network; and to calculate the comprehensive status assessment result corresponding to each of the power devices in the converter station by weighting the cluster centroid distance.

[0061] In one possible implementation, the acquisition module 21 is used to acquire parameter information corresponding to each power device in the early, middle, and late stages of an earthquake. Specifically, this includes: deploying a sensor array along the converter station using preset fiber optic sensing technology, including distributed fiber Bragg grating technology and Brillouin optical time-domain reflectometry; using preset fiber optic interferometry technology, including Michelson interferometry and Mach-Zehnder interferometry; and acquiring parameter information corresponding to each power device in the early, middle, and late stages of an earthquake based on the sensor array and the enhanced fiber optic interferometry array.

[0062] In one possible implementation, the acquisition module 21 is used to construct a multi-scale damage characterization system before performing time-series analysis of parameter information through the multi-scale damage characterization system and acquiring the corresponding time-series data of power equipment in the early, middle and late stages of an earthquake. Specifically, this includes: establishing a cross-scale damage feature encoding model of vibration spectrum-strain gradient-temperature field, and realizing time-series evolution modeling of equipment damage state through a gated loop unit.

[0063] In one possible implementation, the acquisition module 21 is used to perform time-series analysis on parameter information through a multi-scale damage characterization system, and to acquire time-series data of power equipment in the pre-earthquake, mid-earthquake, and post-earthquake phases. Specifically, this includes: performing time-series reconstruction on the parameter information, which is used to segment the parameter information according to different time windows; normalizing the parameter information after time-series reconstruction to reduce noise interference and enhance the comparability of parameter information; and performing time-series analysis on the normalized parameter information based on a time step strategy, and acquiring time-series data of power equipment in the pre-earthquake, mid-earthquake, and post-earthquake phases.

[0064] In one possible implementation, the processing module 22 is used to obtain state data corresponding to different types of power equipment under seismic conditions based on a group topology graph neural network. Specifically, this includes: obtaining the layout structure of different types of power equipment in the converter station, and encoding different types of power equipment as graph structure nodes in the group topology graph neural network based on the layout structure; modeling the seismic wave propagation path using a graph attention network based on the group topology graph neural network to capture the vibration coupling effect between different types of power equipment; and obtaining state data corresponding to different types of power equipment under seismic conditions based on the vibration coupling effect.

[0065] In one possible implementation, the processing module 22 is used to calculate the centroid distance of each power device based on the input state data and a clustering statistical neural network. Specifically, this includes: dividing the state data of each power device into several clusters based on a clustering algorithm, with each cluster representing a device abnormal state; calculating the centroid value corresponding to each cluster, where the centroid value is the value corresponding to the optimal center point of the state data in each cluster; and calculating the centroid distance of each power device based on the centroid value.

[0066] In one possible implementation, the processing module 22 is used to calculate the cluster centroid distance corresponding to each power device based on the centroid value, specifically including: calculating the cluster centroid distance corresponding to each power device based on the centroid value using the following formula: ; in, For the first The power equipment and the first The centroid distance between clusters For the first The electrical equipment in the first State data across each feature dimension For the first The weight coefficients corresponding to each feature dimension Indicates the number of feature dimensions. For the first The cluster in the th order of ... Centroid values ​​on each feature dimension Indicates the first The standard deviation corresponding to each feature dimension The normalization coefficient is... This indicates the number of abnormal features in the parameter change trend information. Indicates the first Reference values ​​corresponding to each abnormal feature information These are the sensitivity adjustment parameters.

[0067] In one possible implementation, the processing module 22 is used to calculate the comprehensive status assessment result corresponding to each power device in the converter station based on the cluster centroid distance through weighted calculation. Specifically, it includes: determining whether the cluster centroid distance is greater than a preset abnormal threshold; if the cluster centroid distance is greater than the preset abnormal threshold, then confirming that the comprehensive status assessment result corresponding to the power device is an abnormal operating state; if the cluster centroid distance is less than or equal to the preset abnormal threshold, then confirming that the comprehensive status assessment result corresponding to the power device is a normal operating state.

[0068] In one possible implementation, after the processing module 22 confirms that the comprehensive status assessment result corresponding to the power equipment is an abnormal operating state if the cluster centroid distance is greater than a preset abnormal threshold, the method further includes: obtaining the abnormal type corresponding to the power equipment based on the cluster centroid distance; obtaining the equipment maintenance plan corresponding to the abnormal type from a preset abnormal handling plan database, wherein the preset abnormal handling plan database is used to store the correspondence between abnormal types and equipment maintenance plans; and sending the equipment maintenance plan to the user's terminal device.

[0069] It should be noted that the above embodiments of the apparatus are only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0070] This application also provides an electronic device. (See reference...) Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: at least one processor 301, at least one communication bus 302, a user interface 303, at least one network interface 304, and a memory 305.

[0071] The communication bus 302 is used to enable communication between these components.

[0072] The user interface 303 may include a display screen and a camera. Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0073] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0074] The processor 301 may include one or more processing cores. The processor 301 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 305, and by calling data stored in memory 305. Optionally, the processor 301 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 301 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 301 and may be implemented as a separate chip.

[0075] The memory 305 may include random access memory (RAM) or read-only memory. Optionally, the memory 305 may include a non-transitory computer-readable storage medium. The memory 305 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 305 may also be at least one storage device located remotely from the aforementioned processor 301. (Refer to...) Figure 3 The memory 305, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application for monitoring the seismic status of converter stations based on a clustering neural network.

[0076] exist Figure 3In the illustrated electronic device, the user interface 303 is primarily used to provide an input interface for the user and acquire user input data; while the processor 301 can be used to call the application program for converter station seismic status monitoring based on clustering neural networks stored in the memory 305. When executed by one or more processors 301, the electronic device performs one or more of the methods described in the above embodiments. It should be noted that, for the foregoing method embodiments, for the sake of simplicity, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0077] This application also provides a computer-readable storage medium storing instructions. When executed by one or more processors, these instructions cause an electronic device to perform one or more of the methods described in the above embodiments.

[0078] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0079] In the various embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between apparatuses or units may be electrical or other forms.

[0080] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0081] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0082] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0083] The above description is merely an exemplary embodiment disclosed in this application and should not be construed as limiting the scope of this application. Any equivalent changes and modifications made in accordance with the teachings of this application shall still fall within the scope of this application. Those skilled in the art, upon considering the disclosure of the specification and practical truths, will readily conceive of other embodiments disclosed in this application.

[0084] This application is intended to cover any variations, uses, or adaptations disclosed herein that follow the general principles disclosed herein and include common knowledge or customary technical means in the art that are not described in this application.

Claims

1. A method for monitoring the seismic status of a converter station based on a clustering neural network, characterized in that, The method includes: In response to the user's operation to monitor the operating status of different types of power equipment in the converter station under earthquake conditions, the system acquires the parameter information of each power equipment in the early, middle and late stages of the earthquake, including vibration information, strain information, acceleration information and temperature information. The parameter information was analyzed over time using a multi-scale damage characterization system, and the corresponding time-series data of the power equipment in the early, middle and late stages of the earthquake were obtained. The time series data is subjected to feature extraction operation according to a preset method, and the parameter change trend information in the time series data is obtained according to the feature extraction operation. The parameter change trend information includes vibration change trend information, strain change trend information, acceleration change trend information and temperature change trend information. Based on the parameter change trend information, the parameter information corresponding to the pre-earthquake, mid-earthquake, and post-earthquake stages, and using a group topology graph neural network, the state data corresponding to different types of power equipment under earthquake conditions are obtained. By inputting the state data, the cluster centroid distance corresponding to each of the power devices is calculated according to the clustering statistical neural network; Based on the cluster centroid distance, the comprehensive status assessment results corresponding to each power equipment in the converter station are calculated by weighted calculation.

2. The method according to claim 1, characterized in that, The acquisition of parameter information for each of the aforementioned power devices during the pre-earthquake, mid-earthquake, and post-earthquake phases specifically includes: A sensor array is deployed along the converter station using a preset fiber optic sensing technology, which includes distributed fiber Bragg grating technology and Brillouin optical time domain reflectometry. Sensitized optical fiber sensing arrays are used around different types of power equipment using preset fiber optic interferometry techniques, including Michelson interferometry and Mach-Zehnder interferometry. Based on the sensor array and the enhanced optical fiber sensing array, the parameter information corresponding to each of the power devices in the early stage, middle stage and late stage of the earthquake is obtained.

3. The method according to claim 1, characterized in that, Before performing time-series analysis on the parameter information using a multi-scale damage characterization system and obtaining the corresponding time-series data of the power equipment in the pre-earthquake, mid-earthquake, and post-earthquake phases, the multi-scale damage characterization system needs to be constructed, specifically including: A cross-scale damage feature encoding model based on vibration spectrum, strain gradient, and temperature field is established, and the temporal evolution model of equipment damage state is realized through gated cyclic units.

4. The method according to claim 1, characterized in that, The step of performing time-series analysis on the parameter information using a multi-scale damage characterization system and obtaining the corresponding time-series data of the power equipment in the pre-earthquake, mid-earthquake, and post-earthquake phases specifically includes: A time series reconstruction operation is performed on the parameter information, wherein the time series reconstruction operation is used to segment the parameter information according to different time windows; The parameter information after the time series reconstruction operation is normalized to reduce noise interference and enhance the comparability of the parameter information; Based on the time step strategy, time series analysis is performed on the parameter information after the normalization process, and the time series data of the power equipment corresponding to the pre-earthquake, mid-earthquake, and post-earthquake events are obtained.

5. The method according to claim 1, characterized in that, The process of obtaining state data for different types of power equipment under seismic conditions based on a group topology graph neural network specifically includes: The layout structure of different types of power equipment in the converter station is obtained, and the different types of power equipment are encoded as graph structure nodes in the group topology graph neural network according to the layout structure. Based on the group topology graph neural network, a graph attention network is used to model the seismic wave propagation path and capture the vibration coupling effect between different types of power equipment. Based on the vibration coupling effect, the state data corresponding to different types of power equipment under seismic conditions are obtained.

6. The method according to claim 1, characterized in that, The step of calculating the cluster centroid distance for each power device by inputting the state data and using a clustering statistical neural network specifically includes: Based on a clustering algorithm, the status data of each power device is divided into several clusters, and each cluster represents an abnormal state of the device. Calculate the centroid value corresponding to each cluster, where the centroid value is the value corresponding to the optimal center point of the state data in each cluster; The cluster centroid distance for each of the power devices is calculated based on the centroid value.

7. The method according to claim 6, characterized in that, The calculation of the cluster centroid distance for each of the power devices based on the centroid value specifically includes: The cluster centroid distance for each of the power devices is calculated based on the centroid value using the following formula: ； in, For the first The power equipment mentioned above and the first The centroid distance between the clusters is given. For the first The aforementioned power equipment in the first The state data in each feature dimension For the first The weight coefficients corresponding to each of the aforementioned feature dimensions. This indicates the number of the feature dimensions. For the first The clusters mentioned above are in the first... The centroid values ​​on each feature dimension Indicates the first The standard deviations corresponding to each of the aforementioned feature dimensions The normalization coefficient is... This indicates the number of abnormal feature information in the parameter change trend information. Indicates the first Reference values ​​corresponding to the aforementioned abnormal feature information These are the sensitivity adjustment parameters.

8. The method according to claim 1, characterized in that, The step of calculating the comprehensive condition assessment result for each power device in the converter station based on the cluster centroid distance through weighted calculation specifically includes: Determine whether the cluster centroid distance is greater than a preset abnormal threshold; If the cluster centroid distance is greater than the preset abnormal threshold, then the comprehensive status assessment result corresponding to the power equipment is confirmed to be an abnormal operating state. If the cluster centroid distance is less than or equal to the preset abnormal threshold, then the comprehensive status assessment result corresponding to the power equipment is confirmed to be in normal operating condition.

9. The method according to claim 8, characterized in that, After confirming that the comprehensive status assessment result corresponding to the power equipment is in an abnormal operating state if the cluster centroid distance is greater than the preset abnormal threshold, the method further includes: Based on the cluster centroid distance, the anomaly type corresponding to the power equipment is obtained; Obtain the equipment maintenance plan corresponding to the anomaly type from the preset anomaly handling plan database, wherein the preset anomaly handling plan database is used to store the correspondence between the anomaly type and the equipment maintenance plan; The device maintenance plan is sent to the user's terminal device.

10. A converter station seismic status monitoring device based on clustering neural networks, characterized in that, The device includes an acquisition module and a processing module, wherein, The acquisition module is used to respond to user operations for monitoring the operating status of different types of power equipment in a converter station under earthquake conditions. It acquires parameter information for each power equipment in the pre-earthquake, mid-earthquake, and post-earthquake phases, including vibration, strain, acceleration, and temperature information. It performs time-series analysis on the parameter information using a multi-scale damage characterization system and acquires time-series data for each power equipment in the pre-earthquake, mid-earthquake, and post-earthquake phases. It performs feature extraction on the time-series data according to a preset method and acquires parameter change trend information in the time-series data based on the feature extraction. This parameter change trend information includes vibration change trend information, strain change trend information, acceleration change trend information, and temperature change trend information. Based on the parameter change trend information, the parameter information corresponding to the pre-earthquake, mid-earthquake, and post-earthquake phases, and using a group topology graph neural network, it acquires state data corresponding to different types of power equipment under earthquake conditions. The processing module is used to calculate the cluster centroid distance corresponding to each of the power devices by inputting the status data and using a clustering statistical neural network; and to calculate the comprehensive status assessment result corresponding to each of the power devices in the converter station by weighting the cluster centroid distance.

Images