Electrical equipment operation state monitoring system and method

Abstract

The invention relates to the technical field of electrical equipment analysis, and particularly discloses an electrical equipment operation state monitoring system and method, and the method comprises the steps: the system is deployed at an electrical equipment site, integrates a low-power-consumption processor and an FPGA acceleration module, and is used for collecting multi-source data in real time, and executing lightweight model calculation and local early warning; and storing full data and operating a global analysis model, and communicating with the edge computing node to realize data collaborative analysis. According to the electrical equipment operation state monitoring system and method provided by the embodiment of the invention, through edge computing lightweight deployment, data preprocessing and local early warning are completed on the equipment site, and high efficiency of monitoring response and low-power-consumption hardware adaptation are realized; by means of an electromagnetic interference dynamic evaluation module, an acquisition strategy is intelligently adjusted, noise is suppressed, the reliability of multi-source data is guaranteed, the efficiency, precision and intelligent capability of electrical equipment state monitoring are remarkably improved, and technical support is provided for reliable operation of power system equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment analysis, and in particular to a system and method for monitoring the operating status of electrical equipment. Background Art

[0002] In power systems, electrical equipment such as transformers, circuit breakers, and switchgear operate under complex and variable conditions, often facing dynamic conditions such as load fluctuations, changes in ambient temperature and humidity, and electromagnetic interference. Existing condition monitoring technologies face the following bottlenecks: Traditional monitoring systems rely on cloud servers to process all data, resulting in high real-time response latency (over a few seconds), making it difficult to detect sudden faults such as partial discharge and overheating. Furthermore, when large-scale equipment clusters upload raw data simultaneously, network congestion can easily occur, resulting in poor system scalability. Summary of the Invention

[0003] The present invention aims to solve at least one of the technical problems in the related art to a certain extent. To this end, the present invention aims to provide an electrical equipment operating status monitoring system and method to improve the monitoring efficiency and accuracy of the electrical equipment.

[0004] To achieve the above-mentioned objectives, the first embodiment of the present invention proposes an electrical equipment operation status monitoring system, including: an edge computing node (EN): deployed at the electrical equipment site, integrating a low-power processor and an FPGA acceleration module, for real-time collection of multi-source data, execution of lightweight model calculations and local early warning; a cloud server (CS): stores full data and runs a global analysis model, communicates with the edge computing node to realize data collaborative analysis, the global analysis model constructs a device cluster association model based on a graph neural network to analyze the state propagation effect between devices, predict system-level failure risks, and use differential update technology to push model optimization parameters to the edge computing node; a multi-source sensor group: including voltage / current transformers, infrared sensors, ultrasonic probes and electromagnetic signal monitoring modules, for collecting operating parameters and environmental parameters of electrical equipment; a lightweight digital twin module: runs a simplified version of the device model on the edge computing node to map the status of key components of the equipment in real time.

[0005] In some embodiments of the present invention, the edge computing node has a built-in electromagnetic interference suppression module for collecting the real-time magnetic field strength of the operating environment of the electrical equipment. , electric field strength and electromagnetic wave frequency , through the formula:

[0006] Calculate the electromagnetic interference index ( ),in, 、 、 is the weight coefficient, represents the standard deviation operation, is the standard frequency; when When , the FPGA hardware filtering is triggered and / or the sensor sampling frequency is adjusted, where is the preset interference threshold.

[0007] In some embodiments of the present invention, the edge computing node has a built-in data pre-processing module for: performing time synchronization on the collected raw data with microsecond accuracy; Calculate the data collection compliance index:

[0008] in, For the The actual collection parameters of class data, To preset standard parameters, is the number of parameter categories; The data is resampled or marked as suspicious data, where The default compliance threshold.

[0009] In some embodiments of the present invention, the edge computing node has a built-in lightweight risk assessment module for: Extract real-time deviation array ,in, For the Group data The deviation value of the parameter, is the measured value, is the standard value; By formula Calculate real-time risk value ,in, is the parameter weight coefficient; when When the local sound and light alarm is triggered and an early warning log is generated, The preset risk threshold.

[0010] In some embodiments of the present invention, the cloud server has a built-in global analysis module for: Receive key feature values uploaded by edge computing nodes ,in, is the real-time risk value, is the rate of change of risk value, is the electromagnetic interference index, Collect compliance indices for data; A device cluster association model is constructed based on a graph neural network, where the node feature is the key feature value, and the edge weight is determined according to the electrical connection strength and space between devices. Analyze the state propagation effect between devices and predict system-level failure risks, including: For nodes Features, For the edge Attributes, For the message function: Index of the system-level failure risk ,in, Contribute risk values to each node; regularly push model optimization parameters to edge computing nodes, and use differential update technology to transmit parameter increments.

[0011] To achieve the above-mentioned object, a second embodiment of the present invention provides a method for monitoring the operating status of an electrical device, comprising the following steps: S1. Device initialization and edge computing node configuration: Build a lightweight digital twin model and load it to the edge computing node, configure sensor acquisition parameters and edge task priorities; S2. Multi-source data edge acquisition and preprocessing: Collect operating parameters and environmental parameters through the sensor group, and perform electromagnetic interference suppression and data compliance verification at the edge computing node; S3. Edge layer risk screening and local early warning: Extract real-time feature parameters, calculate real-time risk values and trigger local early warnings; S4. Cloud-based collaborative analysis and global early warning: Upload key feature values to the cloud, generate a comprehensive early warning report through the global model and send it to the edge computing node for execution.

[0012] In some embodiments of the present invention, the device initialization and edge computing node configuration steps include: loading a lightweight digital twin model, which includes the physical parameters and simplified mathematical models of key components of the device; configuring the edge computing task priority, and setting the real-time data processing ratio to ≥90%.

[0013] In some embodiments of the present invention, the multi-source data edge acquisition and preprocessing steps include: synchronously acquiring the voltage, current, temperature, vibration acceleration and electromagnetic signal of the electrical equipment to generate a raw data set with a microsecond time stamp; The data is filtered and the FPGA real-time filtering algorithm is enabled to suppress high-frequency noise.

[0014] In some embodiments of the present invention, the edge layer risk initial screening and local early warning step includes: calculating the temperature growth rate within a 5-minute sliding time window:

[0015] in, 、 is the extreme temperature in the window, Minutes; perform fast Fourier transform on the vibration signal to extract the main frequency component , calculate the mean of the historical main frequency distribution Deviation rate , identify periodic shock characteristics; comprehensive real-time risk value , temperature growth rate and main frequency deviation rate Generate edge layer warning signals, where: and hour, The weight is increased by 20% and a Level II warning is triggered; when When the alarm is triggered, the local sound and light alarm will be triggered and an early warning log including temperature field cloud map and vibration spectrum waterfall map will be generated.

[0016] In some embodiments of the present invention, the cloud collaborative analysis and global warning step includes: when the edge computing node detects or When , upload the feature vector to the cloud ,in, It is the maximum allowable rate of change of risk value per unit time; the cloud simulates the status of equipment cluster through the global twin model and graph neural network to generate a comprehensive early warning report including fault level, hidden danger area and maintenance priority; after receiving the instruction, the edge computing node links the on-site equipment to adjust the operating parameters or trigger the maintenance process, and updates the local model parameters.

[0017] To achieve the above-mentioned purpose, the third aspect of the present invention proposes an electronic device, comprising a memory, a processor and a computer program stored in the memory. When the computer program is executed by the processor, the above-mentioned method for monitoring the operating status of electrical equipment is implemented.

[0018] The electrical equipment operating status monitoring system and method of the embodiments of the present invention complete data preprocessing and local early warning at the equipment site through lightweight deployment of edge computing, thereby achieving high efficiency of monitoring response and low-power hardware adaptation; with the help of the electromagnetic interference dynamic assessment module, it intelligently adjusts the collection strategy and suppresses noise to ensure the reliability of multi-source data; through the lightweight digital twin model and multi-dimensional feature fusion analysis, it maps the equipment status in real time and predicts risk trends, significantly improving the efficiency, accuracy and intelligence capabilities of electrical equipment status monitoring, and providing technical support for the reliable operation of power system equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The disclosure of the present invention is described with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. In the drawings, the same reference numerals are used to refer to the same components. Among them: Figure 1 Schematic diagram of the framework of an electrical equipment operating status monitoring system according to an embodiment of the present invention; Figure 2This is a time domain signal comparison diagram before and after the electromagnetic interference suppression module performs interference suppression in one embodiment of the present invention; Figure 3 1 is a comparison diagram (logarithmic coordinates) of the power spectrum density before and after the electromagnetic interference suppression module performs interference suppression in one embodiment of the present invention; Figure 4 This is a comparison diagram (linear coordinates) of the frequency domain suppression effect of the electromagnetic interference suppression module before and after interference suppression in one embodiment of the present invention; Figure 5 It is a structural diagram of a device cluster association model in one embodiment of the present invention; Figure 6 is a flow chart of a method for monitoring the operating status of electrical equipment according to another embodiment of the present invention; Figure 7 It is a structural diagram of an electronic device according to another embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0021] The following describes an electrical equipment operating status monitoring system and method, and an electronic device according to embodiments of the present invention with reference to the accompanying drawings.

[0022] In order to realize the automation of the whole process of "real-time monitoring - intelligent analysis - accurate early warning" of electrical equipment status and reduce the dependence on manual inspection, the present invention proposes an electrical equipment operation status monitoring system. Figure 1 This system framework diagram shows the electrical equipment operating status monitoring system. This system utilizes a distributed architecture based on edge computing and cloud collaboration. Through lightweight hardware deployment and layered data processing, it addresses core issues of traditional monitoring systems, such as poor real-time performance, low data reliability, and high hardware costs.

[0023] like Figure 1 As shown, an electrical equipment operation status monitoring system includes: edge computing nodes (EN): deployed at the electrical equipment site, such as near switch cabinets and transformers, so as to be close to the data source and shorten the data processing link.

[0024] Integrating low-power processors (such as ARM architectures) and FPGA acceleration modules within edge computing nodes balances computing efficiency and energy consumption, making them suitable for long-term operation in industrial environments. Furthermore, edge computing nodes only process key features of real-time data, such as risk values and interference indices, thus avoiding network pressure caused by full data uploads.

[0025] The ideal operating state of an edge computing node is: 90% of routine data is processed at the edge computing node, and only 10% of high-risk features are uploaded to the cloud, reducing network traffic by 90% and avoiding network congestion during large-scale device cluster monitoring.

[0026] Core functions of edge computing nodes include: multi-source data acquisition: synchronously acquiring operating parameters such as voltage, current, temperature, and vibration, as well as environmental parameters such as magnetic field strength, electric field strength, and electromagnetic wave frequency; local intelligent analysis: performing lightweight model calculations, such as risk assessment and electromagnetic interference suppression, to achieve millisecond-level local warnings without relying on the cloud. The low-power design allows edge computing nodes to operate without an external power source, such as outdoor equipment in distribution networks, broadening their application scenarios.

[0027] Cloud Server (CS): Stores all data and runs global analysis models, communicating with edge computing nodes to achieve collaborative data analysis.

[0028] The cloud server serves as the system's "brain," storing all historical data and running complex analytical models to provide global risk predictions. It encompasses two key functions: Graph Neural Network (GNN) cluster analysis: This builds a model of electrical connections and spatial associations between devices to analyze the cascading effects of a single device failure on the cluster, such as a busbar failure overloading surrounding switchgear. Differential update technology: This pushes incremental model parameters, such as optimized weight coefficients, only to edge computing nodes, reducing network load by less than 30 seconds.

[0029] The cloud server uses the feature data uploaded by edge computing nodes, combined with historical data to train a global model and reversely optimize the local algorithms of the edge computing nodes, such as adjusting the risk value weight coefficient, to form an "edge-cloud closed-loop optimization." Therefore, the cloud server's core tasks are: receiving key feature values uploaded by edge computing nodes, such as the risk value change rate and electromagnetic interference index, generating device health rankings and system-level risk reports; and regularly optimizing the local models of edge computing nodes to achieve "dynamic evolution of edge computing capabilities."

[0030] Multi-source sensor group: including voltage / current transformers, infrared sensors, ultrasonic probes and electromagnetic signal monitoring modules, used to collect operating parameters and environmental parameters of electrical equipment, including: sensor configuration includes: Operating parameters: voltage / current transformer (accuracy level 0.2), infrared thermal imager (temperature resolution ±0.5°C), vibration accelerometer (frequency range 10-10kHz).

[0031] Environmental parameters: electromagnetic signal monitoring module (magnetic field strength resolution 0.1mT, electric field strength resolution 1V / m).

[0032] The multi-source sensor group uses synchronized acquisition technology to add microsecond-level timestamps to all sensor data, ensuring multi-parameter time alignment and supporting subsequent cross-analysis, such as the correlation between temperature anomalies and electromagnetic interference. The multi-source sensor group integrates multi-dimensional electrical, mechanical, and environmental data to avoid misjudgments of single parameters, such as missing insulation material degradation by monitoring only temperature. Electromagnetic signal data also provides a basis for subsequent interference suppression, improving data reliability.

[0033] Lightweight digital twin module: Runs a simplified device model on edge computing nodes to map the status of key device components in real time, such as the physical parameters (resistance, thermal capacity) of transformer windings and switchgear contacts with simplified mathematical models (such as heat conduction equations). The file size is less than 10MB, making it suitable for running on edge computing nodes.

[0034] The mapping direction is dynamic mapping, which updates sensor data to the twin model in real time and visualizes the device status, such as contact temperature field distribution, vibration mode, etc.

[0035] When an edge computing node triggers an early warning, the twin model highlights the abnormal area to assist operation and maintenance personnel in quickly locating it.

[0036] As an intermediate layer between physical devices and algorithms, the twin model converts sensor data into understandable physical states, such as temperature field distribution, which not only supports rapid decision-making of edge computing nodes but also provides standardized feature input for cloud servers.

[0037] This system combines FPGA acceleration with low-power processors for the first time, implementing complex algorithms at the edge layer, such as electromagnetic interference suppression and risk value calculation, breaking through the computing power limitations of traditional edge devices; at the same time, it integrates electrical parameters, mechanical vibration, and electromagnetic environment data to build a "device-environment" collaborative analysis model to fill the gap in single parameter monitoring; ultimately, through hierarchical processing and dynamic model optimization, it balances real-time and global analysis capabilities, making it suitable for the intelligent monitoring needs of large-scale equipment clusters, systematically solving the bottleneck of traditional monitoring technology, and possessing significant technological progress and engineering application value.

[0038] In some embodiments of the present invention, the electromagnetic interference suppression module is deployed in the edge computing node. By real-time monitoring of the electromagnetic signal characteristic parameters of the electrical equipment operating environment, the electromagnetic interference degree is quantified and the adaptive suppression strategy is triggered to ensure the reliability of the collected data. The module includes a signal acquisition unit, an interference index calculation unit and a dynamic adjustment unit to realize the "monitoring-assessment-response" closed loop, in which: first, the real-time magnetic field strength of the electrical equipment operating environment needs to be collected. , electric field strength and electromagnetic wave frequency , calculate the electromagnetic interference index by the following formula ( ):

[0039] This formula comprehensively quantifies the degree of interference of the electromagnetic environment on data acquisition. It has no units and the obtained electromagnetic interference index ( ) The larger the value, the stronger the interference.

[0040] in, 、 、 is the weight coefficient, which is set according to the characteristics of the equipment operating environment (such as in industrial scenarios > To highlight the impact of magnetic field interference), determined through historical data training or industry standards, + + ; represents the standard deviation operation, It represents the standard deviation of the magnetic field intensity within the detection period, reflecting the degree of magnetic field fluctuation. The calculation formula is:

[0041] In the above formula, The first The magnetic field strength sampling value, is the mean, is the number of sampling points; It represents the standard deviation of the electric field strength within the detection period, and the calculation method is the same as above; For standard frequency, such as 50Hz or 60Hz, as frequency deviation Baseline value; real-time magnetic field strength The unit is millitesla (mT), which is collected in real time by the Hall effect sensor; the electric field strength The unit is volt / meter (V / m), which is monitored in real time by an electric field probe; the frequency of electromagnetic waves The unit is Hertz (Hz), which is collected by a frequency meter and reflects the spectrum characteristics of electromagnetic interference.

[0042] In this module, you need to set the preset interference threshold , preset interference threshold Definition: The critical value for determining whether electromagnetic interference significantly affects data collection, determined by the equipment manufacturer or power operation and maintenance unit based on on-site environmental testing, such as taking historical normal operation data Mean + 3 times standard deviation.

[0043] when When the current electromagnetic environment is judged to have significant interference with data acquisition, the following suppression strategies are triggered: Suppression strategies Implementation Application Scenario Beneficial effects FPGA hardware filtering The high-frequency noise of the sensor's original signal is attenuated through the FPGA's built-in digital filter (such as an FIR filter), and the cutoff frequency is dynamically adjusted based on the real-time electromagnetic wave frequency minus the standard frequency. Scenarios with strong pulse interference (such as lightning and motor start and stop). The noise suppression rate is increased to over 80%, and the signal-to-noise ratio (SNR) is improved by 10-15dB. Sensor sampling frequency adjustment If the real-time electromagnetic wave frequency is greater than the standard frequency plus the frequency deviation (such as frequency deviation = 10kHz), reduce the sampling frequency to 1 / 2 of the original frequency to reduce high-frequency aliasing; if the real-time electromagnetic wave frequency is approximately equal to the standard frequency, increase the sampling frequency to 2 times to enhance the resolution of the power frequency signal. Narrowband interference (such as communication frequency band interference) scenarios. Sampling efficiency and anti-aliasing capability are balanced, and data efficiency is improved by 30%. Sensor working mode switching Switch to anti-interference mode: the magnetic field sensor enables the magnetic shield; the electric field sensor uses a differential sampling circuit. Scenarios with continuous strong interference (such as high-voltage areas in substations). The sensor's anti-interference capability is improved by 50%, extending the hardware life. Redundant data interpolation Abnormal data points are repaired by cubic spline interpolation using undisturbed sampling values at adjacent moments. Single-point data jump scenario caused by sudden interference. Data integrity is restored to 99%, reducing the probability of false alarms. The electromagnetic interference suppression module combines the three elements of magnetic field, electric field, and frequency. Compared with single parameter monitoring (such as measuring only the magnetic field), the interference identification accuracy is improved. At the same time, FPGA is used to implement microsecond-level filtering calculations, which is more efficient than pure software solutions and meets real-time requirements. The suppression measures are dynamically adjusted according to the interference type to avoid the waste of resources caused by "one-size-fits-all" solutions. For example, the sampling frequency is not blindly increased in the case of low-frequency interference, and the power consumption of edge computing nodes is reduced.

[0044] like Figure 2 、 Figure 3 、 Figure 4 As shown, the instant effect of interference suppression by the electromagnetic interference suppression module is intuitively demonstrated.

[0045] Figure 2 In the figure, the red line is the original signal containing electromagnetic interference, and the blue line is the clean signal after algorithm processing. Figure 2 Middle: Red curve (before suppression): The signal amplitude fluctuates violently, with peaks reaching ±4.2V, and contains obvious high-frequency oscillations and irregular pulses. Blue curve (after suppression): The signal becomes smooth and regular, with the amplitude controlled within the ±2.8V range, showing clear sine wave characteristics.

[0046] above Figure 2 In the figure, continuous time-domain processing is performed without delay, which meets the technical requirements of real-time monitoring and processing. The μs-level response speed of the edge computing node is shown in the figure as an instant purification effect of the signal, and ultimately a 33.3% amplitude reduction rate of 4.2V→2.8V is achieved.

[0047] Figure 3 In the figure, the power spectrum in the frequency range 0-1500Hz is analyzed: Before suppression: Obvious power peaks appeared at 120Hz, 300Hz, 800Hz, and 1200Hz, reaching , , , V² / Hz; after suppression: all interference peaks are reduced to V² / Hz or less.

[0048] Figure 4 Spectrum comparison in dB, with key data points: Data 1: At 120Hz, it drops from -10dB to -45dB (35dB suppression). Data 2: At 300Hz, it drops from -12dB to -48dB (36dB suppression). Data3: At 800Hz, it drops from -18dB to -50dB (32dB suppression). Data4: At 1200Hz, it drops from -20dB to -52dB (32dB suppression).

[0049] Figure 4 In the test, the suppression depth of 30-36dB at each frequency point quantitatively demonstrates the technical advantage and proves the suppression effect.

[0050] As an example, in a 10kV distribution room at a steel plant, when an arc furnace is started, the following is monitored: Fluctuation up to ±50mT (normal ≤5mT), (Deviation ), calculated ; At this time, the FPGA high-pass filter (cutoff frequency 50kHz) is triggered and the sensor sampling frequency is reduced to 5kHz, effectively suppressing the high-frequency noise generated by the arc. The temperature data fluctuation is reduced from ±8°C to ±1°C, avoiding "overheating" false alarms caused by interference.

[0051] In some embodiments of the present invention, a data preprocessing module is deployed in an edge computing node to ensure the temporal consistency and reliability of the data input into the analysis model by performing time synchronization, compliance assessment, and exception processing on the raw data collected by multi-source sensors.

[0052] This module includes a time synchronization unit, a compliance calculation unit, and an abnormal data processing unit, forming a data processing link with the electromagnetic interference suppression module and the lightweight risk assessment module. First, the collected raw data is time synchronized with microsecond accuracy.

[0053] As an example, GPS clock synchronization or IEEE 1588 precision clock protocol is used to uniformly calibrate the sensor sampling clock through the edge computing node FPGA hardware timer, with a synchronization accuracy of ≤1μs.

[0054] For cross-sensor data, linear interpolation is used to compensate for transmission delays. For example, automatic calibration is performed when the voltage and temperature data transmission delay difference is ≤5μs.

[0055] Then calculate the data collection compliance index:

[0056] This formula quantifies the degree of agreement between all acquisition parameters and standard values, with a value range of [0,1]. Indicates that all parameters are fully compliant.

[0057] Among them, the numerator Reflects the compliance of a single type of parameter. The smaller the difference, the higher the compliance. Denominator is the number of parameter categories, such as Corresponding to five types of parameters including voltage, current, temperature, vibration, and electromagnetic signal, the overall data quality is comprehensively evaluated through mean calculation; For the Real-time operating parameters or environmental parameters collected by sensors, such as voltage, temperature, and magnetic field strength. The unit depends on the parameter type, such as voltage in V and temperature in °C. It is the preset standard parameter, that is, the normal operating reference value of the i-th category parameter, determined by the equipment manufacturer or the power industry standard, such as the rated voltage, ambient temperature reference value, etc.

[0058] In this module, you need to set the preset compliance threshold , preset compliance thresholds Definition: The critical value for determining whether data intervention is required is set to 0.8 by default (which can be adjusted dynamically according to the importance of the equipment). It is determined by the operation and maintenance unit through historical data training (e.g., taking normal operation data). The mean of -0.2σ).

[0059] When the module is running, When the data is resampled, marked as suspicious data, or redundant data is fused, the specific compliance assessment and exception handling steps are as follows: Treatment strategy Implementation Related modules Technical Effects Resampling For data whose data collection compliance index is less than the preset compliance threshold, re-collect at twice the original sampling frequency (e.g., original 10kHz→20kHz); real-time resampling is achieved using FPGA parallel computing, with a delay of ≤10ms. Electromagnetic interference suppression module (filtered data) Data integrity has been improved to 99.5%, reducing missed data and misjudgments. Suspicious data marking For data with a compliance index less than the preset compliance threshold for three consecutive data collections, a timestamp is added and marked as "pending verification"; the marked data enters the edge computing node cache and waits for the cloud to reissue instructions for processing. Lightweight risk assessment module (skipping labeled data) Avoid abnormal data from participating in risk calculations and reduce the false alarm rate by 30%. Redundant data fusion For multi-sensor homologous data (such as dual-channel temperature sensors), Kalman filter fusion calculation is used to replace single sensor outliers with data with a confidence level ≥ 90%. Multi-source sensor group (redundant configuration) Data reliability is improved by 40%, suitable for critical equipment. It should be noted that the signal after FPGA filtering is first involved in DA calculation. , then the weight of the electromagnetic signal parameters in the DA calculation formula is automatically increased, that is, The coefficient is adjusted dynamically.

[0060] As an example, in order to improve the adaptability of this module, the cloud server analyzes the DA distribution of cluster devices through the GNN model and regularly pushes the optimized , such as seasonal adjustment: summer temperature parameter Reduced by 0.05.

[0061] As an example, for 5 consecutive For a device, the cloud server triggers the full parameter resampling instruction of the device. After the edge computing node executes it, it uploads the original data to the cloud for in-depth verification, such as temperature-current correlation analysis.

[0062] The data preprocessing module combines GPS and FPGA hardware clock to solve the problem of asynchronous multi-sensor data and provide an accurate time reference for fault location. The DA formula is used to quantify data quality. Compared with traditional threshold judgment, it can identify "sub-healthy" data, such as parameter drift but not exceeding the threshold, and the warning accuracy is improved. In addition, the cloud server dynamically adjusts the DA threshold according to the device cluster data to adapt to the personalized needs of different devices, such as old devices. Reduce and minimize the contradictions caused by misjudgment or missed judgment.

[0063] As an example, assume that the main transformer monitoring system of a 220kV substation is: Input raw data acquisition: voltage , oil temperature , ; Calculate DA: ; because , triggering the resampling mechanism: voltage resampling value 219kV, oil temperature 61℃, recalculation , data compliance, thereby avoiding false "overvoltage" warnings caused by temporary offsets of the voltage sensor.

[0064] In some embodiments of the present invention, a lightweight risk assessment module is integrated into edge computing nodes. By extracting real-time deviations from multi-source parameters, calculating a comprehensive risk value, and comparing it with preset thresholds, this module enables rapid assessment of electrical equipment operational risks and provides local early warning. This module works closely with the data preprocessing module and the electromagnetic interference suppression module to ensure the reliability of input data and provide core decision support for equipment status monitoring.

[0065] When performing risk assessment, first extract the real-time deviation array , which is defined as: The real-time deviation array of the group data contains the The deviation value of the parameter.

[0066] in: For the In the group data The deviation value of the parameter, Here, it represents the index of the data group, which is used to distinguish the monitoring data groups at different times or different sampling periods. For example, a group of data is generated every 5 minutes ( );and Indicates the index of the parameter, corresponding to different monitoring parameters of electrical equipment, such as 1 is voltage, 2 is the current, 3 is temperature, etc.; For the In the group data The measured values of the parameters are derived from the sensor output after verification by the data preprocessing module and processing by the electromagnetic interference suppression module; For the The standard values of various parameters, such as the rated voltage of the equipment and the normal operating temperature range, are provided by the equipment manufacturer or determined according to industry standards; Then by the formula Calculate real-time risk value , which means the The real-time risk value of group data is calculated by weighted square root and square root to comprehensively reflect the degree of equipment operation risk; in, is the parameter weight coefficient, indicating the The weight coefficient of each parameter can be determined through the analytic hierarchy process (AHP) or machine learning training to reflect the degree of influence of different parameters on equipment risk. As an example, when high temperature has a great impact on transformer risk, The value is higher.

[0067] Set preset risk thresholds in this module The preset risk threshold is set by the power operation and maintenance unit based on the equipment's historical operating data and industry standards to determine whether to trigger an early warning. When the alarm is triggered, the local sound and light alarm will be triggered and an early warning log will be generated.

[0068] As an example, the preset risk threshold is Mean + 3 times standard deviation.

[0069] As an example, the lightweight risk assessment module only receives Calculate risk value based on data ,If the data is marked as suspicious, a 15% uncertainty coefficient will be automatically added when calculating the risk value, i.e. .

[0070] As an example, the cloud server regularly analyzes the historical risk data of the device cluster and optimizes it through machine learning algorithms. , and sent to the edge computing node. For example, when a certain type of transformer is running in summer, the temperature parameter weight Automatically increase by 10% to more accurately assess thermal overload risks.

[0071] It should be noted that the module input After time synchronization and compliance verification in the data preprocessing module and noise filtering in the electromagnetic interference suppression module, the deviation calculation is ensured to be based on reliable data. In addition to the local sound and light alarm, the generated early warning log contains detailed parameter deviation information (such as The data is then uploaded to the cloud server. The cloud server combines historical device data with cluster status to further analyze risk trends, such as determining whether there is a persistent overheating risk.

[0072] As an example, taking a transformer as an example, the monitoring parameters include voltage , current , oil temperature ,in: No. Group data: (Standard value ), (Standard value ), (Standard value ); Calculate the deviation: , , ; Assumed weight coefficient , , , calculate the risk value:

[0073] At this time, if the preset threshold ,but , triggering local sound and light alarms, and uploading logs to the cloud for further analysis to see if there is overload or heat dissipation anomaly.

[0074] In some embodiments of the present invention, the cloud server has a built-in global analysis module for: 1. Receive key feature values uploaded by edge computing nodes ,in, is the real-time risk value, is the electromagnetic interference index, To collect compliance index for data, is the risk value change rate, that is, the real-time risk value between two adjacent calculation cycles The rate of change is calculated as follows:

[0075] in, is the real-time risk value of the previous period, is the time interval between two adjacent calculation cycles (unit: seconds). For example, if , , , ,but .

[0076] 2. Build a device cluster association model based on graph neural network (GNN), analyze the state propagation effect between devices, and predict system-level failure risks.

[0077] In the process of building a graph neural network (GNN) model, we first need to define nodes and edges: Node: Each electrical device corresponds to a node in the graph, and the node feature vector is , comprehensively reflecting the current risks of the equipment, risk change trends, electromagnetic environment and data quality; Edges: Physical connections (such as power cable connections) or logical associations (such as load sharing relationships or geographic proximity) between devices form the edges of a graph. Edge weights can be assigned based on the strength of the connection or the degree of association (e.g., high-voltage transmission line connections have a higher weight than low-voltage lines).

[0078] like Figure 5 As shown, a total of 5 nodes are set, each node contains the feature vector ; The edge defines the connection relationship and weight between devices, and the weight indicates the strength of the association (for example, high-voltage lines have a higher weight).

[0079] In addition, the state propagation effect between devices can also be analyzed: Using the message passing mechanism of GNN, nodes Receive neighboring nodes Information , the formula is ,in, For nodes Features, For the edge Attributes, is a message function (such as a multilayer perceptron).

[0080] For example, a transformer node of Increased and , through message passing, the switch cabinet nodes connected to it It receives this information and updates its own risk prediction to analyze whether the abnormal load on the switchgear is caused by transformer overload, thereby predicting the risk of system-level failure, such as local power outage.

[0081] The system-level failure risk prediction process is as follows: First, receive the data uploaded by the edge computing node , stored in the cloud database; Then normalize the data: , etc., unified dimensions; Then input the GNN model, after L layers of message passing (e.g. L=3), output the system risk contribution value of each node ; Finally, all the , and obtain the system-level failure risk index ( Here is the total number of devices). (preset system risk threshold), an early warning report is generated to prompt operation and maintenance personnel to perform system-level maintenance.

[0082] 3. Regularly push model optimization parameters to edge computing nodes and support differential update technology.

[0083] The principle of differential update technology is to calculate the global model parameters on the cloud The model parameters last used by the edge computing node The difference , only sent to edge computing nodes .

[0084] The advantage is that it can significantly reduce the amount of data transmission. For example, the edge computing node model has parameters, only 10% of the parameters are updated, and the differential update only needs to transmit The change value of the parameter, not all Parameters, update efficiency is greatly improved.

[0085] The specific implementation process includes: The cloud server regularly trains the global model to obtain new parameters ; calculate ,right Perform compression coding (such as Huffman coding); Finally, it is sent to the edge computing node, and the node updates the local model after decoding. .

[0086] As a specific application example, this electrical equipment operating status monitoring system was used for a 110kV substation deployed with 10 high-voltage switchgear. While traditional monitoring solutions rely on centralized cloud-based analysis, this system employs an "edge computing + cloud collaboration" architecture. The monitoring period was one month, focusing on comparing overheating fault warning response time, data transmission volume, and fault prediction accuracy.

[0087] During monitoring, one day, sensors in a traditional monitoring solution detected a cabinet temperature rise of 85°C (threshold 80°C). Data upload to the cloud took 0.8 seconds, and analysis took 0.5 seconds, for a total delay of 1.3 seconds. However, because the electromagnetic environment was not monitored, the concurrent high-frequency electromagnetic interference was not detected, leading to a misjudgment of the temperature rise as normal and a potential risk being missed.

[0088] This system calculates the temperature growth rate in real time through edge computing nodes (Threshold 3°C / min), and electromagnetic interference index was detected at the same time , triggering local early warning; Then upload the characteristic value When the information is uploaded to the cloud, the cloud server discovered through GNN analysis that the temperature correlation between the cabinet and adjacent equipment had increased (correlation coefficient +0.6), determining it to be a potential risk of poor contact and issuing a three-day advance warning, thereby avoiding power outages and reducing losses.

[0089] like Figure 6 As shown, based on the above electrical equipment operating status monitoring system, a method for monitoring the operating status of electrical equipment is proposed here. This method solves the three core problems of traditional monitoring systems through an "edge-cloud" collaborative architecture: Insufficient real-time performance: Traditional solutions rely on the cloud to process all data, resulting in response delays of seconds and an inability to detect sudden failures. Poor data reliability: Electromagnetic interference and sensor errors lead to data distortion, and single parameter monitoring is prone to misjudgment; Lack of global analysis: Lack of correlation analysis of equipment clusters makes it impossible to predict systemic risks.

[0090] The method specifically comprises the following steps: S1. Device initialization and edge computing node configuration: Build a lightweight digital twin model and load it to the edge computing node, configure sensor acquisition parameters and edge task priorities; Lightweight digital twin model loading includes: The model includes key components of the equipment, such as the physical parameters (resistance and thermal capacity) of transformer windings and switchgear contacts; simplified mathematical models, such as the first-order differential equation for heat conduction, to achieve a file size of less than 10MB and an edge computing node loading time of less than 200ms.

[0091] As an example, the transformer model only retains the winding temperature-load mapping relationship and ignores non-critical structures such as casing heat sink details to ensure real-time calculation efficiency of edge computing nodes.

[0092] Edge task priority configuration: Set the real-time data processing ratio to ≥ 90%. Through task scheduling algorithms, such as giving priority to the earliest EDF deadline, it can ensure that real-time tasks such as electromagnetic interference suppression and risk calculation are executed first, and non-real-time tasks such as historical data archiving account for ≤ 10%.

[0093] S2. Multi-source data edge acquisition and preprocessing: Due to electromagnetic interference in industrial sites, such as inverters and arcs, which cause sensor data distortion, traditional filtering algorithms have high latency. In addition, the lack of synchronization of multi-sensor data will lead to cross-analysis failures, such as the inability to correlate voltage drops with mechanical vibration anomalies. Therefore, it is necessary to collect operating parameters and environmental parameters through a sensor group, and perform electromagnetic interference suppression and data compliance verification at the edge computing node. Specifically, this includes: Synchronous acquisition and timestamp synchronization: Using GPS clock synchronization (accuracy ±1μs) + FPGA hardware timer, it synchronously samples voltage (accuracy ±0.2%), current (±0.5A), temperature (±0.5℃), vibration (acceleration resolution 0.1m / s²), and electromagnetic signals (magnetic field strength 0.1mT), with a timestamp accuracy of 1μs.

[0094] Anti-interference measures: When the electromagnetic interference index When the high-frequency noise is detected, the FPGA real-time filtering algorithm is triggered to suppress the high-frequency noise, and the cut-off frequency is dynamically adjusted. For example, when the high-frequency interference is detected, it is set to 50kHz, and the noise suppression rate is ≥80%.

[0095] Data compliance verification: Calculating the Data Collection Compliance Index ,right Resample or mark the data to ensure that the proportion of valid data is ≥ 95%.

[0096] S3. Initial risk screening at the edge layer and local warnings: Traditional cloud-based analysis delays lead to delayed fault warnings, which can easily miss the optimal time for action. Single-parameter warnings (e.g., temperature exceeding a threshold) are prone to false alarms, such as false overheating caused by elevated ambient temperatures. Therefore, it is necessary to extract real-time feature parameters, calculate real-time risk values, and trigger local warnings.

[0097] S4. Cloud-based collaborative analysis and global early warning: Because single-device early warnings cannot predict system-level risks, and model updates rely on manual deployment, they cannot adapt to dynamic scenarios such as device aging and environmental changes. Therefore, key feature values must be uploaded to the cloud. A comprehensive early warning report is generated using the global model and sent to edge computing nodes for execution. After receiving cloud-based instructions, the edge computing nodes activate on-site equipment (such as adjusting transformer taps) or trigger maintenance processes, and simultaneously update local twin model parameters, forming a closed "monitoring-analysis-execution" loop.

[0098] This method systematically solves the real-time, reliability and global analysis problems of traditional monitoring through layered design and collaborative mechanism, from device initialization, data collection, local analysis to global optimization. Steps S1 and S2 provide data foundation, step S3 achieves rapid response, and step S4 provides global decision-making, forming a complete technical chain from "device-level precise monitoring" to "system-level risk prevention and control."

[0099] In some embodiments of the present invention, the edge layer risk initial screening and local early warning steps include: 1. Calculate the temperature growth rate within a 5-minute sliding window. This value represents the dynamic temperature change trend of key equipment components and is used to identify early signs of overheating failures:

[0100] in, 、 are the maximum and minimum temperatures in the window, The window duration is in minutes, fixed at 5 minutes, and synchronized with the edge computing node data collection cycle.

[0101] This mechanism can capture abnormal temperature rise of equipment in real time. For example, when a sudden increase in transformer load causes the oil temperature to rise from 60°C to 75°C within 5 minutes, , reaching the set temperature growth rate threshold At this time, the system automatically increases the weight of the temperature parameter in the risk assessment model (for example, from the default 0.4 to 0.5) to strengthen the response to overheating risks.

[0102] The sliding window mechanism is explained here, using a 5-minute overlapping window (overlap rate 50%), that is, the window data is updated every 2.5 minutes to ensure that sudden temperature changes can be captured in real time.

[0103] As an example, when the transformer load suddenly increases, It can rise from 1℃ / min to 5℃ / min, triggering a level one warning.

[0104] 2. Extract the main frequency component of the vibration signal (Collected through vibration acceleration sensors), it reflects the mechanical status of the equipment, such as bearing wear, loose components, etc., and then compares it with the historical main frequency distribution to identify periodic impact characteristics.

[0105] Main frequency calculation method: Perform 1024-point fast Fourier transform (FFT) on the vibration signal and take the frequency corresponding to the peak value of the power spectrum density as , the frequency resolution is ; Periodic shock feature recognition: Calculate the current and historical main frequency average Deviation rate: ; when When the power spectrum density amplitude exceeds 3 times the standard deviation of the historical mean, it is determined to be a periodic shock. For example, when the contacts of a switch cabinet are loose, the main vibration frequency shifts from the normal 120Hz to 180Hz. , the system immediately marks it as a potential mechanical failure.

[0106] 3. Comprehensive real-time risk value , temperature growth rate and main frequency deviation rate Generate edge layer warning signals.

[0107] Dynamic weight adjustment: When and Real-time risk value The weight coefficients of the medium temperature and vibration parameters are automatically increased by 20%, triggering a Level II yellow warning, indicating that "overheating with mechanical abnormalities may occur."

[0108] Linkage alarm mechanism: When When an overheating accident occurs, the edge computing node immediately activates the local sound and light alarm and generates an early warning log containing a temperature field cloud map (such as the overheated contact area captured by the infrared thermal imager) and a vibration spectrum waterfall map (showing the trend of the main frequency over time).

[0109] As an example, a transformer When the fault occurs, the system will simultaneously output a composite fault report of "winding overheating (78°C) + abnormal core vibration (main frequency deviation 35%)" to guide the operation and maintenance personnel to give priority to repair.

[0110] As an example, based on the Level II and Level III warnings, Level I and Level IV warnings are added to form a four-level response mechanism. The triggering conditions and handling strategies for each level are as follows: Level Ⅰ (red); trigger conditions: Corresponding measures: The edge node immediately cuts off non-critical equipment loads (e.g., cooling systems are prioritized for power supply); and pushes an emergency maintenance work order (priority factor × 2) containing location information to the operation and maintenance terminal. Level II (yellow); trigger conditions: Corresponding measures: Start redundant data collection with backup sensors (such as cross-validation of dual-channel temperature sensors); adjust the sampling frequency to 20kHz (vibration) + 5Hz (temperature); Level III (orange); trigger conditions: Corresponding measures: Edge nodes store enhanced logs (including raw waveform data) locally; upload feature vectors to the cloud (data volume is compressed to 1 / 10 of the original); Level IV (blue); trigger conditions: ; Corresponding measures: Trigger the sensor self-calibration process (such as infrared sensor blackbody calibration); start the device health trend analysis in the cloud.

[0111] In the initial screening and local early warning steps of edge layer risks, this solution is different from traditional single temperature or vibration monitoring. and The combined analysis greatly improves the detection rate of compound faults; and through historical fault data training, dynamic adjustment and For example, during high temperatures in summer, the temperature weight is automatically increased by 15% to avoid false alarms caused by ambient temperature fluctuations.

[0112] As an example, in a 110kV substation application, after a sudden load increase on the main transformer, the edge computing node monitors: °C, °C, calculated ; The main frequency of vibration shifts from the normal 100Hz to 130Hz, ; At this point, the system immediately triggered a Level II warning, and the infrared thermal imager located that the high-voltage side bushing was overheating (the temperature field cloud map showed a hotspot temperature of 85°C). At the same time, the vibration spectrum waterfall diagram showed a 130Hz main frequency with an increase in harmonics. It was ultimately verified that the oxidation of the bushing contacts caused the contact resistance to increase, thus avoiding the equipment burnout accident in advance.

[0113] In some embodiments of the present invention, the cloud-based collaborative analysis and global warning steps specifically include: 1. When the edge computing node detects or When , upload the feature vector to the cloud .

[0114] in, is the maximum allowable rate of change of risk value per unit time, defined as ( During normal operation standard deviation of ); It is the absolute value of the risk value change within a unit of time (such as 1 minute) and is used to capture abnormal fluctuations at the moment of equipment failure.

[0115] The triggering conditions for the data upload mechanism are: Emergency upload: When the edge computing node immediately uploads the feature vector ; Mutation upload: When When, even if ,Data is still uploaded and used in the cloud to analyze the propagation path of risk mutations, such as whether a sudden change in the load of a certain device will lead to a chain reaction.

[0116] Upload data format: Uses standardized JSON format, including timestamp, device ID, characteristic value and unit, to ensure cross-platform compatibility.

[0117] 2. The cloud simulates the status of the equipment cluster through a global twin model and graph neural network to generate a comprehensive early warning report that includes fault level, hidden danger area, and maintenance priority.

[0118] Among them, the global digital twin model includes building a digital twin of the equipment cluster, integrating real-time data of the edge layer (temperature, vibration, load) and physical models (heat conduction equation, vibration modal equation), and simulating the evolution of the cluster state; then outputting the fault diffusion path (device-level association diagram) and the real-time status of the affected equipment, such as temperature field distribution, vibration energy distribution, etc.

[0119] As an example, thermal coupling analysis is used to predict the fault chain of "transformer A overheating → adjacent switchgear B overload → abnormal vibration".

[0120] The graph neural network uses semi-supervised learning, inputs historical fault data (labeled with fault level and impact range), and trains GNN to predict the probability of fault propagation.

[0121] 3. Based on the generated comprehensive early warning report, the edge computing node links the on-site equipment to adjust the operating parameters or trigger the maintenance process, and updates the local model parameters.

[0122] Adjusting operating parameters includes: receiving cloud instructions, linking field equipment through edge control modules (such as PLC interface, Modbus protocol), and optimizing operating status in real time. As an example, reducing the transformer load from 120% to 90% reduces the temperature increase rate. .

[0123] The maintenance process is triggered by automatically generating a maintenance work order when the maintenance priority exceeds the preset threshold. This work order includes recommendations for detection tools, such as "using an infrared thermal imager to detect the equipment's D winding and a vibration analyzer to collect the 10-10kHz spectrum," and is pushed to the operation and maintenance terminal, such as a mobile app or work order system.

[0124] Updating local model parameters includes: the cloud encrypts and sends the optimized threshold and local correlation weight of GNN, and the edge computing node updates the local early warning model to achieve adaptability.

[0125] Corresponding to the above embodiment, the present invention further provides an electronic device.

[0126] like Figure 7FIG2 is a schematic diagram of the structure of an electronic device according to the present invention. The electronic device 200 includes a processor 201 and a memory 203. The processor 201 and the memory 203 are connected, for example, via a bus 202. Optionally, the electronic device 200 may further include a transceiver 204. It should be noted that in actual applications, the number of transceivers 204 is not limited to one, and the structure of the electronic device 200 does not constitute a limitation on the embodiments of the present invention.

[0127] Processor 201 may be a CPU (central processing unit), a general-purpose processor, a DSP (digital signal processor), an ASIC (application-specific integrated circuit), an FPGA (field programmable gate array), or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the present disclosure. Processor 201 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, or a combination of a DSP and a microprocessor.

[0128] The bus 202 may include a path for transmitting information between the above components. The bus 202 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industrial Standard Architecture) bus. The bus 202 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0129] The memory 203 is used to store a computer program corresponding to the method for monitoring the operating status of electrical equipment according to the above embodiment of the present invention, and the computer program is controlled and executed by the processor 201. The processor 201 is used to execute the computer program stored in the memory 203 to implement the contents of the above method embodiment.

[0130] The electronic device 200 includes, but is not limited to, mobile terminals such as laptop computers and PADs (tablet computers), and fixed terminals such as desktop computers. Figure 7 The electronic device 200 shown is merely an example and should not limit the functions and scope of use of the embodiments of the present invention.

[0131] The electronic device 200 of the embodiment of the present invention is lightweightly deployed through edge computing, completes data preprocessing and local early warning at the equipment site, and achieves high efficiency of monitoring response and low-power hardware adaptation; with the help of the electromagnetic interference dynamic assessment module, it intelligently adjusts the collection strategy and suppresses noise to ensure the reliability of multi-source data; through the lightweight digital twin model and multi-dimensional feature fusion analysis, it maps the equipment status in real time and predicts risk trends, significantly improving the efficiency, accuracy and intelligence capabilities of electrical equipment status monitoring, and providing technical support for the reliable operation of power system equipment.

[0132] It should be noted that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, analyze, propagate, or transmit a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic device), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or otherwise processing it in a suitable manner if necessary, and then stored in a computer memory.

[0133] It should be understood that various components of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof may be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0134] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0135] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0136] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. An electrical equipment operating status monitoring system, characterized in that: include: Edge computing nodes: Deployed at electrical equipment sites, they integrate low-power processors and FPGA acceleration modules to collect multi-source data in real time, perform lightweight model calculations, and provide local early warnings. Cloud servers: Store all data and run global analysis models, communicating with edge computing nodes for collaborative data analysis. The global analysis model builds a device cluster association model based on a graph neural network to analyze the state propagation effect between devices and predict system-level failure risks. It also uses differential update technology to push model optimization parameters to edge computing nodes. Multi-source sensor group: including voltage / current transformers, infrared sensors, ultrasonic probes and electromagnetic signal monitoring modules, used to collect operating parameters and environmental parameters of electrical equipment; Lightweight digital twin module: Runs a simplified device model on the edge computing node to map the device status in real time.

2. The system according to claim 1, wherein: The edge computing node has a built-in electromagnetic interference suppression module for: Collect the real-time magnetic field strength of the operating environment of electrical equipment , electric field strength and electromagnetic wave frequency , through the formula: Calculate the electromagnetic interference index ( ),in, 、 、 is the weight coefficient, represents the standard deviation operation, is the standard frequency; when When , the FPGA hardware filtering is triggered and / or the sensor sampling frequency is adjusted, where is the preset interference threshold.

3. The system according to claim 1, wherein: The edge computing node has a built-in data preprocessing module for: Time synchronization of collected raw data; Calculate the data collection compliance index: in, For the The actual collection parameters of class data, To preset standard parameters, is the number of parameter categories; right The data is resampled or marked as suspicious data, where The default compliance threshold.

4. The system according to claim 1, wherein: The edge computing node has a built-in lightweight risk assessment module for: Extract real-time deviation array ,in, For the Group data The deviation value of the parameter, is the measured value, is the standard value; By formula Calculate real-time risk value ,in, is the parameter weight coefficient; when When the local sound and light alarm is triggered and an early warning log is generated, The preset risk threshold.

5. The system according to claim 1, wherein: The cloud server has a built-in global analysis module for: Receive key feature values uploaded by edge computing nodes ,in, is the real-time risk value, is the rate of change of risk value, is the electromagnetic interference index, Collect compliance indices for data; A device cluster association model is constructed based on a graph neural network, where the node feature is the key feature value, and the edge weight is determined according to the electrical connection strength and space between devices. Analyze the state propagation effect between devices and predict system-level failure risks, including: For nodes Features, For the edge Attributes, For the message function: Index of the risk of system-level failure ,in, Risk contribution value for each node; Model optimization parameters are regularly pushed to edge computing nodes, and parameter increments are transmitted using differential update technology.

6. A method for monitoring the operating status of electrical equipment, based on the system according to any one of claims 1 to 5, characterized in that: The following steps are involved: S1. Device initialization and edge computing node configuration: Build a lightweight digital twin model and load it to the edge computing node, configure sensor acquisition parameters and edge task priorities; S2. Multi-source data edge acquisition and preprocessing: The sensor group collects operating and environmental parameters, and performs electromagnetic interference suppression and data compliance verification at the edge computing node. S3. Initial screening of edge layer risks and local early warning: Extract real-time characteristic parameters, including temperature growth rate and vibration signal main frequency components, and combine them with real-time risk values to generate edge layer early warning signals; S4. Cloud-based collaborative analysis and global warning: Upload key feature values to the cloud, generate a comprehensive warning report through the global model, and send it to the edge computing node for execution.

7. The method according to claim 6, characterized in that The device initialization and edge computing node configuration steps include: Load the lightweight digital twin model, which includes the physical parameters and simplified mathematical models of the equipment components; Configure the edge computing task priority and set the real-time data processing ratio to ≥ 90%.

8. The method according to claim 6, characterized in that The multi-source data edge collection and preprocessing steps include: Synchronously collect voltage, current, temperature, vibration acceleration, and electromagnetic signals of electrical equipment to generate raw data sets with microsecond-level time stamps; Electromagnetic interference index The data is filtered and the FPGA real-time filtering algorithm is enabled to suppress high-frequency noise.

9. The method according to claim 6, characterized in that The steps of initial screening and local early warning of edge layer risks include: Calculate the temperature growth rate within a 5-minute sliding time window: in, 、 is the extreme temperature in the window, minute; Perform fast Fourier transform on the vibration signal to extract the main frequency component , calculate the mean of the historical main frequency distribution Deviation rate , identify periodic shock characteristics; Comprehensive real-time risk value , temperature growth rate and main frequency deviation rate Generate edge layer warning signals, including: when and hour, The weight is increased by 20% and a Level II warning is triggered; when When the alarm is triggered, the local sound and light alarm will be triggered and an early warning log including temperature field cloud map and vibration spectrum waterfall map will be generated.

10. The method according to claim 6, characterized in that The cloud-based collaborative analysis and global warning steps include: When the edge computing node detects or When , upload the feature vector to the cloud ,in, is the maximum allowable rate of change of risk value per unit time; The cloud simulates the status of the equipment cluster through a global twin model and graph neural network, generating a comprehensive early warning report that includes fault level, potential hazard area, and maintenance priority. After receiving the instructions, the edge computing node links the on-site equipment to adjust the operating parameters or trigger the maintenance process, and updates the local model parameters.

Images