Online monitoring method, system and equipment for mechanical characteristics of circuit breaker and medium

Abstract

The invention relates to the technical field of circuit breaker state monitoring, in particular to a circuit breaker mechanical characteristic on-line monitoring method, system and device and a medium, and the method comprises the steps: responding to an operation trigger signal of a circuit breaker, and synchronously collecting mechanical characteristic parameters and electrical characteristic parameters of the circuit breaker; the mechanical characteristic parameters and the electrical characteristic parameters are preprocessed through the intelligent monitoring IED, and time sequence characteristics and frequency domain characteristics used for representing dynamic changes in the operation process of the circuit breaker are extracted; carrying out multi-modal feature fusion on the data after preprocessing and feature extraction through a cloud analysis platform, and carrying out fault analysis by adopting a three-stage diagnosis model; and predicting the remaining service life of the circuit breaker through a degradation model based on a stochastic process based on the output of the diagnostic model and the historical operation data. The method has the advantages that the limitation of a traditional monitoring means is broken through, online iterative optimization of the diagnosis model is achieved through the edge-cloud collaborative architecture, and the fault recognition rate reaches 95% or above.

Description

Technical Field

[0001] This invention relates to the field of circuit breaker condition monitoring technology, and in particular to a method, system, device and medium for online monitoring of the mechanical characteristics of circuit breakers. Background Technology

[0002] High-voltage circuit breakers are critical protection and control devices in power systems, and the reliability of their mechanical characteristics directly affects the safe and stable operation of the power grid. Therefore, monitoring and diagnosing the mechanical condition of circuit breakers has always been a key focus of industry research. Traditional monitoring methods rely on periodic offline tests, which cannot reflect the equipment's condition under actual operating conditions. In recent years, with the development of sensing and communication technologies, online monitoring technologies have been gradually applied.

[0003] However, existing online monitoring technologies suffer from several drawbacks. First, they rely on a single monitoring dimension, making it difficult to comprehensively capture the multi-dimensional mechanical and electrical characteristics of circuit breakers during operation using a single sensor, resulting in incomplete state perception. Second, they lack forward-looking lifespan prediction capabilities, failing to effectively assess the degradation trend of mechanical characteristics or provide early fault warnings. These issues collectively hinder further improvements in the effectiveness of online monitoring technologies for circuit breaker mechanical characteristics. Summary of the Invention

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for online monitoring of the mechanical characteristics of a circuit breaker, including synchronously acquiring the mechanical characteristic parameters and electrical characteristic parameters of the circuit breaker in response to the operation trigger signal of the circuit breaker; The mechanical and electrical characteristic parameters are preprocessed by the intelligent monitoring IED, and the time-series and frequency-domain characteristics used to characterize the dynamic changes in the circuit breaker operation process are extracted. The preprocessed and feature-extracted data are fused using a cloud-based analytics platform, and a three-level diagnostic model is employed for fault analysis. Based on the output of the diagnostic model and historical operating data, the remaining service life of the circuit breaker is predicted by a degradation model based on stochastic processes.

[0005] As a preferred embodiment of the online monitoring method for the mechanical characteristics of circuit breakers of the present invention, the three-level diagnostic model includes: First-level diagnosis based on preset threshold rules for rapid screening and alarm of feature parameters; Second-level diagnosis based on pattern recognition of temporal features using a long short-term memory network model; The third level of diagnosis is based on graph neural network model to reason about and analyze the associated faults between different components.

[0006] In a preferred embodiment of the online monitoring method for the mechanical characteristics of circuit breakers of the present invention, the degradation model based on the stochastic process is the Wiener process model. Build and update the Wiener process model, including, Based on historical monitoring data, the drift coefficient and diffusion coefficient of the model are updated in real time using the maximum likelihood estimation algorithm; Based on the contact material properties and factory parameters of the circuit breaker, set the failure threshold of mechanical characteristics; Based on the updated model parameters and failure threshold, the probability distribution of remaining useful life is obtained analytically using the inverse Gaussian distribution.

[0007] As a preferred embodiment of the online monitoring method for the mechanical characteristics of circuit breakers of the present invention, the mechanical characteristic parameters include the moving contact stroke and vibration signal; Electrical characteristic parameters include the trip coil current, the closing coil current, and the energy storage motor current.

[0008] As a preferred embodiment of the online monitoring method for the mechanical characteristics of circuit breakers of the present invention, the preprocessing includes differential calculation of the moving contact stroke-time curve to obtain the opening and closing speed characteristics; Preprocessing also includes performing fast Fourier transform analysis on the trip coil current, closing coil current, and energy storage motor current to extract the harmonic components of the current signal.

[0009] As a preferred embodiment of the online monitoring method for the mechanical characteristics of circuit breakers of the present invention, the time synchronization accuracy of synchronous acquisition is at the microsecond level; The multimodal sensing unit includes a high-precision angular displacement sensor, a closed-loop Hall current sensor group, and a vibration sensor.

[0010] As a preferred embodiment of the online monitoring method for the mechanical characteristics of circuit breakers of the present invention, the high-precision angular displacement sensor adopts a magnetic grating ruler structure to measure the output spindle rotation angle of the circuit breaker's operating mechanism in a non-contact manner.

[0011] Secondly, the present invention provides an online monitoring system for the mechanical characteristics of a circuit breaker, comprising: a data acquisition module, used to synchronously acquire the mechanical characteristic parameters and electrical characteristic parameters of the circuit breaker in response to the operation trigger signal of the circuit breaker; The extraction module is used to preprocess mechanical and electrical characteristic parameters through intelligent monitoring IED, and extract time-series and frequency-domain features to characterize the dynamic changes in the circuit breaker operation process. The analysis module is used to perform multimodal feature fusion on preprocessed and feature-extracted data through a cloud-based analysis platform, and to perform fault analysis using a three-level diagnostic model. The prediction module is used to predict the remaining service life of circuit breakers based on the output of the diagnostic model and historical operating data, using a degradation model based on stochastic processes.

[0012] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0013] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method described above.

[0014] Compared with existing technologies, the beneficial effects of this invention are as follows: Multimodal synchronous acquisition technology achieves precise μs-level alignment of mechanical and electrical parameters, fundamentally solving the problem of misjudgment of status caused by timing mismatch in traditional monitoring. Furthermore, the use of non-contact magnetic grating measurement can completely eliminate mechanical wear, ensuring stable output from the sensor throughout the circuit breaker's entire lifespan. An edge-cloud collaborative computing architecture organically integrates rapid edge response with deep cloud analysis (supporting GNN correlation inference and Wiener process life prediction), meeting both real-time alarm requirements and predictive maintenance needs. The adaptive diagnostic model continuously absorbs new data and iteratively optimizes through an online learning mechanism, increasing the fault identification rate from 80% to 95% compared to traditional methods, significantly reducing the risk of sudden failures. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the online monitoring method for the mechanical characteristics of circuit breakers.

[0017] Figure 2 This is a schematic diagram of the system topology.

[0018] Figure 3 This is a diagram of the LSTM network structure. Detailed Implementation

[0019] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0020] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a method for online monitoring of the mechanical characteristics of a circuit breaker, including: S100: In response to the circuit breaker's operation trigger signal, synchronously acquire the circuit breaker's mechanical and electrical characteristic parameters; S200: The mechanical and electrical characteristic parameters are preprocessed by the intelligent monitoring IED, and the time-series and frequency-domain characteristics used to characterize the dynamic changes in the circuit breaker operation process are extracted. S300: Performs multimodal feature fusion on preprocessed and feature-extracted data through a cloud-based analytics platform, and uses a three-level diagnostic model for fault analysis; S400: Based on the output of the diagnostic model and historical operating data, the remaining service life of the circuit breaker is predicted by a degradation model based on a stochastic process.

[0021] It should be noted that traditional mechanical performance monitoring of high-voltage circuit breakers often adopts periodic maintenance or offline testing methods. However, this method can only obtain static parameters and cannot capture the dynamic characteristics under actual load operation. Moreover, the test cycle is long and the cost is high, and the safety warning has a significant lag. Existing online monitoring technologies mostly use single sensors, which are difficult to fully reflect the coupling relationship of displacement-force-vibration during the mechanical condition degradation process. At the same time, the accuracy of multi-source data synchronous acquisition is insufficient, which cannot meet the precise alignment requirements of mechanical transient processes. Furthermore, the independent analysis of various monitoring data lacks effective spatiotemporal correlation fusion methods, making it difficult to achieve early mechanical fault identification and predictive maintenance.

[0022] Therefore, to address the aforementioned issues, this method, through steps S100-S400, achieves μs-level multi-channel synchronous acquisition via auxiliary switch signal triggering, enabling precise timing alignment of mechanical and electrical characteristics; it utilizes edge computing resources to perform preprocessing locally and extracts multi-dimensional feature vectors, reducing cloud load and enabling rapid response; it integrates heterogeneous data in the cloud and establishes a three-level diagnostic model, achieving progressive analysis from anomaly detection to fault tracing; finally, it establishes a degradation model based on stochastic processes, transforming diagnostic output into lifespan prediction, enabling early warning of mechanical faults 3-6 months in advance.

[0023] Example 2, refer to Figures 1-3 As an embodiment of the present invention, based on the above embodiment, an online monitoring method for the mechanical characteristics of a circuit breaker is provided.

[0024] In this embodiment of the application, in step S100, in response to the operation trigger signal of the circuit breaker, the mechanical characteristic parameters and electrical characteristic parameters of the circuit breaker are synchronously acquired. These parameters are acquired synchronously by a multi-modal sensing unit triggered by the contact signal of an auxiliary switch, and include the following steps A1-A2: It should be noted that the time synchronization accuracy of the synchronous acquisition is at the microsecond level, and the multimodal sensing unit includes a high-precision angular displacement sensor, a closed-loop Hall current sensor group and a vibration sensor. Among them, the high-precision angular displacement sensor adopts a magnetic grating ruler structure to measure the output spindle rotation angle of the circuit breaker's operating mechanism in a non-contact manner.

[0025] Understandably, due to the strict causal and temporal relationship between the mechanical actions of a circuit breaker (such as contact movement) and electrical signals (such as coil current), this step determines the time synchronization accuracy to the microsecond level to ensure that data points of different physical quantities such as displacement, current, and vibration are precisely aligned on the time axis. This further enables subsequent analysis to accurately calculate key mechanical characteristic parameters. Specifically, μs-level synchronous acquisition is achieved by directly transmitting sensor data via shielded twisted-pair cables to the edge computing device deployed in the operating mechanism box, with data transmission time on the μs level.

[0026] It should be further noted that the high-precision angular displacement sensor has a measurement error of ≤±0.01°, and is used to directly and accurately monitor the position and trajectory of the moving contact (converted to stroke through rotation angle), which is the direct basis for calculating core mechanical characteristics such as opening / closing speed and overtravel; the closed-loop Hall current sensor group: synchronously collects the opening / closing coil current (range 0-50A) and the energy storage motor current (range 0-30A); the vibration sensor is used to capture the high-frequency stress waves generated by the collision and friction of the mechanism components (such as cam, connecting rod, and latch) during operation.

[0027] Preferably, traditional contact displacement sensors (such as resistance rulers) suffer from mechanical wear and installation difficulties, while this method uses a magnetic grating ruler to measure the spindle rotation angle by reading the magnetic signal on the annular magnetic grating. Since there is no physical contact with the spindle, mechanical wear can be avoided, and the sensor life and measurement stability can be improved. Furthermore, non-contact installation is more convenient and can reduce power outage construction time.

[0028] A1: Mechanical characteristic parameters include the stroke of the moving contact and vibration signal.

[0029] Understandably, the contact stroke is measured by the aforementioned high-precision angular displacement sensor. Furthermore, the vibration signal is measured by a vibration sensor installed on key mechanical components such as the camshaft bearing housing, and is used to extract time-frequency domain features for subsequent fault diagnosis. The main parameters of the vibration sensor are: frequency response range 0.5Hz-20kHz, sensitivity 50-500mV / g, range ±50g, and resolution ≤0.001g (RMS).

[0030] A2: Electrical characteristic parameters include the opening coil current, closing coil current, and energy storage motor current.

[0031] It should be noted that the trip coil current is measured by a Hall current sensor installed on the KM1 terminal block side, the closing coil current is measured by a Hall current sensor installed on the KM2 terminal block side, and the energy storage motor current is measured by a Hall current sensor installed on the M terminal block side.

[0032] Understandably, these current waveform characteristics (such as start-up time, peak value, and duration) are key to determining the electrical health of the operating mechanism.

[0033] Reference Figure 2 As shown in the system topology diagram, the acquisition process uses the contact signal of the auxiliary switch (a switch signal that indicates the opening and closing position) as a unified trigger source. Once this signal changes, the data acquisition card for multiple channels such as angular displacement, current, and vibration is activated simultaneously, thus ensuring microsecond-level synchronization accuracy from the source.

[0034] In an optional implementation, the synchronous acquisition of the mechanical and electrical characteristic parameters of the circuit breaker in step S100 can adopt a multi-channel parallel acquisition method based on an FPGA hardware synchronous bus. This employs a centralized parallel architecture, using a programmable logic device to generate a nanosecond-level synchronous sampling clock, driving multiple independent ADC chips to perform parallel analog-to-digital conversion on the travel, current, and vibration signals. When a hardware interrupt is triggered by a change in the auxiliary switch contact state, the FPGA latches all channel data on the same clock edge and automatically embeds a hardware timestamp. The synchronization error between channels is less than 50 nanoseconds, thereby ensuring precise timing alignment of the mechanical and electrical characteristics during transient processes.

[0035] In another optional implementation, the synchronous acquisition of the mechanical and electrical characteristic parameters of the circuit breaker in step S100 can adopt a distributed acquisition and synchronization method based on the IEEE 1588 PTP precise time protocol. This involves using distributed intelligent sensing nodes, each with a built-in PTP timing module interconnected via industrial Ethernet. The system initiates sampling via self-triggering due to current surges or a master station command. Each node achieves sub-microsecond clock synchronization based on the precise time protocol and embeds a PTP timestamp into the sampling data frame. The cloud platform performs cross-node data alignment based on the timestamp, not the receiving timing.

[0036] In this embodiment of the application, step S200 preprocesses the mechanical and electrical characteristic parameters using an intelligent monitoring IED and extracts time-series and frequency-domain features to characterize the dynamic changes during circuit breaker operation, including the following steps B1-B2: Understandably, this step is to transform the raw high-precision data collected in step S100 into more refined and representative structured feature information. This not only alleviates the pressure on network transmission, but also prepares high-quality data for subsequent in-depth and complex intelligent analysis in the cloud.

[0037] It should be noted that the intelligent monitoring IED mainly includes the following functions: Equipped with the IEC61850 communication protocol, it can collect measurement data from multi-modal sensors, including mechanical characteristics, opening and closing time, opening and closing speed, full-break time of ABC three phases and arcing time calculation (including three-phase synchronous estimation), as well as SF6 meter data. It can also upload SF6 gas monitoring data, including density, pressure, temperature, and trace moisture, through the IED via the IEC61850 protocol.

[0038] It should be further noted that the installation of the intelligent monitoring IED is fixed by the guide rail inside the circuit breaker mechanism box (DIN35 standard), the power supply is taken from the lighting circuit (protected by a 1A circuit breaker), and the cross-sectional area of ​​the grounding wire is ≥4mm².

[0039] B1: Preprocessing includes differential calculation of the moving contact stroke-time curve to obtain the opening and closing speed characteristics.

[0040] It should be noted that the intelligent monitoring IED receives the moving contact stroke data transmitted by the angular displacement sensor, which is precisely time-corresponding, and forms a stroke-time curve. Furthermore, through differential calculation (i.e., calculating the rate of change of stroke with respect to time), the speed-time curve or speed-stroke curve is obtained in real time. Even further, based on the speed curve, key parameters such as average speed, maximum speed, initial separation speed, and initial engagement speed are extracted.

[0041] The specific form of extracting the maximum speed is as follows: In the formula: This indicates the maximum speed captured during the entire motion. Indicates the travel of the moving contact; The instantaneous speed is the derivative of the distance traveled with respect to time.

[0042] Ideally, a continuous displacement curve can be refined into a few key velocity characteristic values, which can greatly reduce the amount of data that needs to be uploaded while retaining the core information.

[0043] B2: Preprocessing also includes fast Fourier transform analysis of the trip coil current, closing coil current and energy storage motor current to extract the harmonic components of the current signal.

[0044] Understandably, this step is to discover fault features hidden in the current time-domain waveform that are difficult for the human eye to detect.

[0045] It should be noted that by using an intelligent monitoring IED to synchronously acquire the time-domain signals of coil current and motor current, a Fast Fourier Transform (FFT) algorithm is applied, and the time-domain current signal is decomposed into sinusoidal components of different frequencies through FFT, thus obtaining the signal spectrum, which contains the amplitude and phase information of the fundamental wave and each harmonic. Finally, specific harmonic components (such as the rate of change of the amplitude of the second and third harmonics) and other features are extracted from the spectrum.

[0046] The formula for calculating the current integral is as follows: In the formula: It is the coil current.

[0047] Preferably, after processing in step S100, the data is no longer the massive original waveform data, but rather a significantly simplified set of velocity parameters, time parameters, harmonic features, and 12-dimensional time-frequency domain features extracted from the vibration signal (such as wavelet packet energy entropy, high-frequency wavelet packet energy ratio, time-domain kurtosis, centroid frequency, 1 / 3 octave band energy ratio, Hilbert marginal spectral entropy, short-time energy zero-crossing rate, and time-domain synchronous average residual energy).

[0048] In an optional implementation, step S200 extracts the time-series and frequency-domain features characterizing the dynamic changes during circuit breaker operation. This extraction can be performed using variational mode decomposition (VMD) and time-frequency image texture. Specifically, VMD adaptively decomposes the original vibration signal into K bandwidth-limited intrinsic mode functions (IMFs). Synchronous compressed short-time Fourier transform (SST) is calculated for each IMF component to generate a high-resolution time-frequency spectrum, transforming the time-frequency joint distribution into a two-dimensional grayscale image. Then, texture feature vectors such as contrast, correlation, energy, and homogeneity are extracted using the gray-level co-occurrence matrix (GLCM). For the current signal, Hilbert-Huang transform (HHT) is used instead of FFT to obtain instantaneous frequency ridge features, accurately capturing the frequency modulation phenomenon caused by coil core jamming. For the travel signal, piecewise polynomial fitting is used to identify inflection points and curvature change rates, characterizing trajectory distortion caused by buffer performance degradation.

[0049] In another optional implementation, the extraction of time-series and frequency-domain features characterizing the dynamic changes in the circuit breaker operation process in step S200 can also be achieved through feature extraction based on sparse representation and online dictionary learning. Specifically, an overcomplete dictionary trained from historical normal operation data and typical fault data is constructed, where each atom corresponds to a signal primitive under a standard operating mode. The edge computing unit performs frame-based processing on the real-time acquired vibration, current, and travel signals (frame length covering the complete operating cycle), and uses the Orthogonal Matching Pursuit (OMP) algorithm to solve for the sparse coefficient vector. The position and amplitude of its non-zero elements characterize the deviation features between the current operation and the standard mode. The L1 norm of the sparse coefficients characterizes the overall anomaly degree, and the non-zero distribution pattern characterizes the fault type. An incremental dictionary update module is deployed; when a new anomaly pattern that cannot be reconstructed by existing atoms is detected, it is automatically expanded into new dictionary atoms, achieving continuous evolution of the feature library.

[0050] In this embodiment of the application, step S300 involves multimodal feature fusion of the preprocessed and feature-extracted data using a cloud-based analysis platform, and fault analysis using a three-level diagnostic model, including the following steps C1-C3: It should be noted that the cloud-based analytics platform uses a heterogeneous data access layer (supporting MQTT / Modbus protocols) to achieve asynchronous access for multiple terminals via 4G / LoRaWAN, and processes tens of thousands of concurrent data streams through a streaming computing framework (Apache Flink). The core analytics engine includes a multimodal feature fusion module (constructing a three-dimensional feature space of travel-current-vibration and dynamically weighting it based on an attention mechanism), a three-level diagnostic model (threshold alarm / LSTM time series analysis / GNN correlation inference), and a Wiener process lifetime prediction module (error <15%).

[0051] C1: First-level diagnosis that quickly screens and alarms feature parameters based on preset threshold rules.

[0052] It should be noted that the preset threshold is set based on historical data and engineering experience, and is not limited here.

[0053] Understandably, this step involves comparing characteristic parameters received from the intelligent monitoring IED, such as opening and closing time, maximum speed, and peak coil current, with preset threshold (safe operation threshold) ranges stored in the database through a cloud-based analysis platform. Once any characteristic value exceeds the threshold (for example, the opening time exceeds the manufacturer's specified upper limit, or the peak closing current is abnormally low), the system immediately triggers an alarm and indicates the abnormal parameter.

[0054] C2: A second-level diagnostic method for pattern recognition of temporal features based on a long short-term memory network model.

[0055] It should be noted that the temporal features extracted in step S200 (such as speed-stroke curves of multiple consecutive operations, current waveform sequences, etc.) are input into the Long Short-Term Memory network model (whose structure is attached). Figure 3 As shown, the system processes sequential data through a sophisticated internal gating mechanism (forget gate, input gate, and output gate). Specifically, the forget gate determines which historical information to discard from previous states (e.g., ignoring early normal data irrelevant to the current fault mode); the input gate determines which new temporal feature information to store in the current cell state (e.g., focusing on minor dips in the velocity curve during recent operations); and the output gate, based on the current input and the updated cell state, determines which information to output to the next layer, ultimately classifying the current operating state (e.g., normal, early mechanism jamming, slight coil aging). Furthermore, a dynamic alarm threshold is set using the 3σ criterion, i.e., the alarm line is dynamically and adaptively set based on the statistical distribution (standard deviation σ) of the deviation (residual) between the LSTM model's prediction and the actual value. It's worth noting that using a dynamic alarm threshold not only better adapts to the slow aging process of the equipment itself, reducing false alarms, but also makes it more sensitive to sudden anomalies or accelerated degradation, achieving early and accurate warnings.

[0056] Understandably, the data features used to train the Long Short-Term Memory (LSTM) network model include: Phase A travel, Phase B travel, Phase C travel, Phase A trip 1 current, Phase B trip 1 current, Phase C trip 1 current, Phase A closing current, Phase B closing current, Phase C closing current, Phase A trip 2 current, Phase B trip 2 current, Phase C trip 2 current, Phase A main circuit current, Phase B main circuit current, Phase C main circuit current, Phase A 101 circuit breaker position, Phase B 101 circuit breaker position, Phase C 101 circuit breaker position, Phase A trip voltage (voltage 1), Phase B trip voltage (voltage 1), Phase C trip voltage (voltage 1), Phase A closing voltage (voltage 2), Phase B closing voltage (voltage 2), Phase C closing voltage (voltage 2), Phase A vibration, Phase B vibration, and Phase C vibration.

[0057] Ideally, LSTM can capture patterns of feature parameters that slowly degrade or fluctuate over time, enabling early identification of faults before they reach the threshold alarm level.

[0058] C3: The third level of diagnosis, which uses a graph neural network model to reason about and analyze the associated faults between different components.

[0059] Understandably, this step treats the circuit breaker's operating mechanism as a graph structure, where nodes represent various components (such as opening coils, closing coils, energy storage motors, contact systems, and transmission links), and edges represent the physical connections or functional relationships between components. Furthermore, domain knowledge (such as the equipment's operating principles and common failure modes) is embedded into the graph structure, enabling the model to reason based on prior knowledge.

[0060] It should be noted that by inputting the features and results of the first two levels of diagnosis, along with the raw observation data from the sensors, as attributes of nodes and edges into the graph neural network model, the GNN allows adjacent nodes in the graph to exchange information through a message passing mechanism. For example, a node feature of slow contact travel will interact and iteratively update with node features of low trip coil current and transmission link. Through multi-layer information propagation and aggregation, the GNN can synthesize the correlation information of the entire system and infer the most likely root cause node (for example, ultimately determining that the root cause of the fault is transmission link jamming, rather than trip coil failure).

[0061] Preferably, step S300 enables comprehensive fault analysis, ranging from simple to complex and from local to systemic.

[0062] In an optional implementation, the multimodal feature fusion of the preprocessed and feature-extracted data in step S300 can be performed using a multimodal high-order correlation feature fusion method based on tensor decomposition. Specifically, the preprocessed time-series signals of travel, current, and vibration are constructed as a third-order tensor X∈R^(T×F×M), where T is the time dimension of the operation cycle, F is the frequency / time domain feature dimension of each mode, and M is the number of modes. Tucker decomposition is used to decompose the tensor into a product of a core tensor G∈R^(r1×r2×r3) and the three mode matrices. The core tensor G implicitly represents the high-order nonlinear interaction relationship between travel, current, and vibration, rather than the linear weighting of the non-attention mechanism. The decomposition process is implemented through high-order singular value decomposition (HOSVD), retaining principal components with a cumulative contribution rate >95% as the fusion feature vector.

[0063] In another optional implementation, step S300, which performs multimodal feature fusion on the preprocessed and feature-extracted data, can also be based on multimodal physical association fusion using a heterogeneous graph neural network. This involves constructing a heterogeneous feature graph G=(V,E): the node set V contains three types of heterogeneous nodes: travel nodes, current nodes, and vibration nodes. Each node carries a temporal feature vector for its own mode. The edge set E contains two types: 1) intramodal edges (temporal edges), connecting adjacent time-step nodes within the same mode, using an LSTM update function to pass temporal dependencies; 2) intermodal edges (physical association edges), predefined based on circuit breaker mechanism knowledge (e.g., rigid connection between travel nodes and current nodes at the moment of opening / closing, and impact response association between vibration nodes and travel nodes), using an attention mechanism to calculate the intermodal information transfer weights. Through the message passing mechanism of the heterogeneous graph neural network, each node aggregates neighbor information under graph topology constraints, ultimately generating a fused node representation.

[0064] In this embodiment of the application, step S400, based on the output of the diagnostic model and historical operating data, predicts the remaining service life of the circuit breaker using a degradation model based on a stochastic process, including the following steps D1-D2: D1: The degradation model based on stochastic processes is the Wiener process model.

[0065] Specifically, the Wiener process model is represented as follows: X(t) = μt + σW(t) In the formula: X(t) represents the cumulative degradation at time t; μt represents the deterministic drift component of degradation (μ is the drift coefficient), which is used to describe the average degradation trend of performance; σW(t) represents the random fluctuation component of degradation (where σ is the diffusion coefficient and W(t) is the standard Brownian motion), which is used to describe individual differences and random fluctuations caused by factors such as load, environment, and operating frequency.

[0066] It should be noted that the data sources used for life prediction include vibration acceleration, velocity, displacement and derived variables of these characteristics, as well as calculated contact travel distance, opening and closing current, number of interruptions, and monitored temperature, etc. In other words, by integrating mechanical vibration, electrical drive, cumulative load and environmental factors, a three-dimensional and comprehensive equipment health profile is constructed to avoid the one-sidedness of prediction based on a single parameter.

[0067] Better yet, the Wiener process model can reflect both the overall trend of degradation and its random fluctuations. Compared with simple linear models or fixed threshold models, it provides a more accurate description of the real situation and more reliable prediction results.

[0068] D2: Build and update the Wiener process model, including, Based on historical monitoring data, the drift coefficient and diffusion coefficient of the model are updated in real time using the maximum likelihood estimation algorithm; It should be noted that the cloud platform will continuously collect the current circuit breaker's historical and latest multimodal characteristics (such as contact stroke, speed, vibration characteristics, number of operations, etc., as listed in the handover document), and every certain period (such as one month or 100 operations), the system will use all this data to recalculate and update the core parameters of the model (drift coefficient μ and diffusion coefficient σ) through the maximum likelihood estimation method.

[0069] Based on the contact material properties and factory parameters of the circuit breaker, set the failure threshold of mechanical characteristics; It should be noted that the failure threshold is not a fixed value, but is set based on engineering principles and physical knowledge. For example, it is based on the electrical life wear standard of circuit breaker contacts (usually copper-tungsten alloy).

[0070] Based on the updated model parameters and failure threshold, the probability distribution of remaining useful life is obtained analytically using the inverse Gaussian distribution.

[0071] It should be noted that the cloud-based analysis platform substitutes the updated μ, σ, and the set failure threshold into the analytical formula of the inverse Gaussian distribution. The result obtained is not a single point in time, but a probability distribution of the remaining service life.

[0072] Ideally, maintenance personnel can scientifically arrange maintenance plans based on probability distribution.

[0073] In summary, this method achieves μs-level precise alignment of mechanical and electrical parameters through multimodal synchronous acquisition technology, fundamentally solving the problem of misjudgment of status caused by timing mismatch in traditional monitoring. Furthermore, the use of non-contact magnetic grating measurement can completely eliminate mechanical wear, ensuring stable output of the sensor throughout the entire life cycle of the circuit breaker. The edge-cloud collaborative computing architecture organically integrates the rapid response at the edge with the deep analysis in the cloud (supporting GNN correlation inference and Wiener process life prediction), meeting both real-time alarm requirements and predictive maintenance. The adaptive diagnostic model continuously absorbs new data and iteratively optimizes it periodically through an online learning mechanism, increasing the fault identification rate from 80% in traditional methods to 95%, significantly reducing the risk of sudden failures.

[0074] Example 3 illustrates a schematic scheme for an online monitoring method of circuit breaker mechanical characteristics. It should be noted that the technical solution of this online monitoring system for circuit breaker mechanical characteristics is based on the same concept as the aforementioned online monitoring method for circuit breaker mechanical characteristics. Details not described in detail in this embodiment can be found in the description of the aforementioned online monitoring method for circuit breaker mechanical characteristics.

[0075] This embodiment also provides an online monitoring system for the mechanical characteristics of circuit breakers, including: The data acquisition module is used to synchronously acquire the mechanical and electrical characteristic parameters of the circuit breaker in response to the circuit breaker's operation trigger signal. The extraction module is used to preprocess mechanical and electrical characteristic parameters through intelligent monitoring IED, and extract time-series and frequency-domain features to characterize the dynamic changes in the circuit breaker operation process. The analysis module is used to perform multimodal feature fusion on preprocessed and feature-extracted data through a cloud-based analysis platform, and to perform fault analysis using a three-level diagnostic model. The prediction module is used to predict the remaining service life of circuit breakers based on the output of the diagnostic model and historical operating data, using a degradation model based on stochastic processes.

[0076] This embodiment also provides an electronic device suitable for online monitoring of the mechanical characteristics of circuit breakers, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the online monitoring method for the mechanical characteristics of circuit breakers as proposed in the above embodiment.

[0077] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the method for online monitoring of the mechanical characteristics of a circuit breaker as proposed in the above embodiments.

[0078] The storage medium proposed in this embodiment and the method for online monitoring of the mechanical characteristics of circuit breakers proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0079] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0080] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for online monitoring of the mechanical characteristics of a circuit breaker, characterized in that: include, In response to the circuit breaker's operation trigger signal, the mechanical and electrical characteristic parameters of the circuit breaker are collected synchronously; The mechanical and electrical characteristic parameters are preprocessed by intelligent monitoring IED, and time-series and frequency-domain features are extracted to characterize the dynamic changes in the circuit breaker operation process. The preprocessed and feature-extracted data are fused using a cloud-based analytics platform, and a three-level diagnostic model is employed for fault analysis. Based on the output of the diagnostic model and historical operating data, the remaining service life of the circuit breaker is predicted by a degradation model based on stochastic processes.

2. The method for online monitoring of the mechanical characteristics of a circuit breaker as described in claim 1, characterized in that: The three-level diagnostic model includes, First-level diagnosis based on preset threshold rules for rapid screening and alarm of feature parameters; The second-level diagnosis of the temporal features is based on a long short-term memory network model for pattern recognition. The third level of diagnosis is based on graph neural network model to reason about and analyze the associated faults between different components.

3. The method for online monitoring of the mechanical characteristics of a circuit breaker as described in claim 2, characterized in that: The degradation model based on stochastic processes is the Wiener process model; Build and update the Wiener process model, including, Based on historical monitoring data, the drift coefficient and diffusion coefficient of the model are updated in real time using the maximum likelihood estimation algorithm; Based on the contact material properties and factory parameters of the circuit breaker, set the failure threshold of mechanical characteristics; Based on the updated model parameters and the failure threshold, the probability distribution of the remaining service life is obtained analytically according to the inverse Gaussian distribution.

4. The method for online monitoring of the mechanical characteristics of a circuit breaker as described in claim 3, characterized in that: The mechanical characteristic parameters include the moving contact stroke and vibration signal; The electrical characteristic parameters include the opening coil current, the closing coil current, and the energy storage motor current.

5. The method for online monitoring of the mechanical characteristics of a circuit breaker as described in claim 4, characterized in that: The preprocessing includes performing differential calculations on the moving contact stroke-time curve to obtain the opening and closing speed characteristics; The preprocessing also includes performing fast Fourier transform analysis on the trip coil current, closing coil current and energy storage motor current to extract the harmonic components of the current signal.

6. The method for online monitoring of the mechanical characteristics of a circuit breaker as described in any one of claims 1-5, characterized in that: The time synchronization accuracy of the synchronous acquisition is at the microsecond level; The multimodal sensing unit includes a high-precision angular displacement sensor, a closed-loop Hall current sensor group, and a vibration sensor.

7. The method for online monitoring of the mechanical characteristics of a circuit breaker as described in claim 6, characterized in that: The high-precision angular displacement sensor adopts a magnetic grating ruler structure to measure the rotation angle of the main shaft output of the circuit breaker's operating mechanism in a non-contact manner.

8. An online monitoring system for the mechanical characteristics of a circuit breaker, using the method described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to synchronously acquire the mechanical and electrical characteristic parameters of the circuit breaker in response to the circuit breaker's operation trigger signal. The extraction module is used to preprocess the mechanical and electrical characteristic parameters through intelligent monitoring IED, and extract the time-series and frequency-domain features that characterize the dynamic changes in the circuit breaker operation process. The analysis module is used to perform multimodal feature fusion on preprocessed and feature-extracted data through a cloud-based analysis platform, and to perform fault analysis using a three-level diagnostic model. The prediction module is used to predict the remaining service life of circuit breakers based on the output of the diagnostic model and historical operating data, using a degradation model based on stochastic processes.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images