Circuit breaker electric contact life state evaluation system and method based on multi-parameter fusion

Abstract

The present application belongs to the technical field of power system high-voltage electrical equipment state monitoring, and particularly relates to a circuit breaker electric contact life state evaluation system and method based on multi-parameter fusion. A cloud-edge collaborative architecture is adopted to directly complete signal preprocessing, feature extraction and operation life state evaluation model at the edge computing node, to realize state evaluation within 100 ms after tripping, and to meet the strong real-time requirement of power equipment state monitoring. A weight distribution strategy based on the analytic hierarchy process (AHP) is proposed, a unified quantitative index comprehensive degradation index (CDI) is constructed, and a unified, physically interpretable decision basis is provided for life prediction. Through a multi-dimensional real-time feature extraction engine, deep-level features such as arc voltage ripple coefficient, reignition frequency and micro-arc segment proportion are extracted, and a model is established in combination with I2t cumulative value, which can accurately reflect the micro-ablation and material transfer on the contact surface and improve the accuracy of electrical life evaluation.

Description

Technical Field

[0001] This invention belongs to the field of condition monitoring technology for high-voltage electrical equipment in power systems, specifically relating to a system and method for assessing the lifespan of circuit breaker electrical contacts based on multi-parameter fusion. Background Technology

[0002] High-voltage circuit breakers are crucial equipment in power systems, responsible for breaking and closing high-voltage, high-current circuits. Their operational reliability directly impacts the safety and stability of the power system. Among various types of circuit breakers, spring-operated circuit breakers are widely used in high-voltage transmission and distribution systems due to their simple structure, reliable operation, and convenient maintenance. The electrical contacts, as the core components of spring-operated circuit breakers for opening and closing operations, are subjected to electric arcing, electrodynamic impact, and thermal stress during each breaking process, making them prone to ablation, welding, and wear. As the contacts gradually degrade, their contact resistance increases, breaking capacity decreases, and may eventually lead to circuit breaker failure or malfunction, posing a serious threat to power grid operation. Therefore, assessing the lifespan of circuit breakers has a significant impact on power grid operation. There are two main methods for assessing the lifespan of traditional high-voltage circuit breaker electrical contacts: The first is the mechanical count method or the single electrical parameter method, which relies solely on the number of breaking operations (mechanical lifespan) or monitoring indicators such as single breaking current and arcing time, using empirical formulas to evaluate the circuit breaker's condition. This method is one-sided and cannot reflect the actual wear of the contacts under complex operating conditions (such as arc erosion and mechanical bounce). The second method is the manual experience-based judgment method, which depends on regular manual inspections and visual observation, lacking quantitative standards and being highly subjective. The one-sidedness and subjectivity of traditional high-voltage circuit breaker electrical contact lifespan assessments lead to inaccurate results, easily resulting in untimely maintenance or over-maintenance.

[0003] To address the aforementioned issues, existing technologies have proposed various solutions. Chinese patent CN121769771A proposes a dynamic closed-loop management method for bridge-type contact performance. This method integrates multiple parameters such as contact resistance, closing coil current, and temperature rise, calculates a health index using fuzzy logic or machine learning models, and extrapolates trends based on historical health index data to predict remaining lifespan and generate maintenance recommendations. Chinese patent CN121834390A discloses an online diagnostic method for high-voltage switchgear contact wear. Based on time-aligned vibration and current signals, it extracts multi-dimensional time-frequency domain features and diagnoses contact wear through wear state degree matching and similarity calculation. Chinese patent CN109164382B provides a fault diagnosis method for high-voltage circuit breaker contact electrical erosion. It collects dynamic contact resistance and travel curves, optimizes support vector machine (SVM) parameters using the bat algorithm, and establishes a nonlinear regression model to evaluate contact erosion status. The aforementioned existing technical solutions for assessing the condition of circuit breaker electrical contacts abandon single, one-dimensional indicators and shift to multi-source information fusion, more realistically reflecting the health status of the circuit breaker. Furthermore, by introducing intelligent algorithms and data-driven models, they transform ambiguous fault phenomena into precise numerical values ​​(such as health indices and remaining lifespan), achieving objectivity in the assessment. However, the aforementioned existing technical solutions still have three shortcomings in practical applications: (1) Real-time performance and transmission bottleneck: The above-mentioned existing technical solutions require uploading massive amounts of raw waveform data to a host computer or cloud for processing, which may lead to difficulties in data transmission and high latency in cloud processing, making it difficult to meet the strong real-time requirements of power equipment status monitoring; (2) The decision-making basis is inconsistent, and there is a lack of interpretable unified weighted evaluation index after the fusion of multi-dimensional features: CN121769771A is based on the logical combination of threshold judgment, which makes it difficult to quantify the specific weight of different physical quantities on lifespan; CN121834390A focuses on "similarity matching" and only pays attention to geometric morphological differences, which makes it difficult to effectively characterize intangible losses (such as the accumulation of arc heat effect); CN109164382B adopts "regression prediction", which lacks an explicit weighting mechanism, makes it difficult to distinguish the contribution of each factor, and has poor physical interpretability. (3) The characterization of electrical wear mechanism is inaccurate and neglects the assessment of cumulative damage: the temperature rise and resistance of CN121769771A are hysteresis parameters after wear; the vibration of CN121834390A is an indirect mechanical parameter; although CN109164382B uses dynamic resistance to reflect morphology, it is essentially an offline or periodic quasi-static detection. They all lack real-time capture and cumulative calculation of the core damage parameter of arc energy during the breaking process, are not sensitive to early electrical wear, cannot accurately detect microscopic ablation and material transfer on the contact surface, and have insufficient accuracy in electrical life assessment. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the technical problem to be solved by this invention is to provide a system and method for evaluating the life status of circuit breaker electrical contacts based on multi-parameter fusion, which solves the problems of data transmission and processing delays, realizes the unified quantification of multiple physical dimensions to establish an interpretable unified weighted evaluation index, deeply mines the microscopic characteristics of electric arc energy and combines them with cumulative values ​​to establish a model, and achieves accurate and real-time online evaluation.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The first aspect of the present invention provides a circuit breaker electrical contact life status assessment system based on multi-parameter fusion, which aims to realize real-time online monitoring and assessment of the life status of high-voltage spring circuit breaker electrical contacts. The assessment system adopts a cloud-edge collaborative real-time processing architecture, including a cloud and an edge processing end, wherein the edge processing end includes a perception layer and an edge computing layer. The sensing layer is used for non-intrusive acquisition of key physical signals during the circuit breaker tripping process, including an electrical parameter acquisition module, an auxiliary status monitoring module, and an environmental temperature control module. The electrical parameter acquisition module includes an open-loop Rogowski coil and a high-voltage differential probe. The open-loop Rogowski coil is used to acquire the discharge circuit current, and the high-voltage differential probe is used to acquire the arc voltage across the circuit breaker. The auxiliary status monitoring module includes an auxiliary contact status monitoring unit and an operating coil monitoring unit. The auxiliary contact status monitoring unit provides a time reference for tripping, used to synchronize all acquisition channels. The operating coil monitoring unit is used to synchronously record the voltage and current waveforms of the tripping coil, assisting in determining the operating status of the operating mechanism. The environmental temperature control module includes an NTC temperature probe, used to monitor the temperature of the supercapacitor energy storage unit and the test environment in real time, ensuring that all data are acquired under standard environmental conditions (10–40 ℃), eliminating the influence of temperature drift on the measurement results.

[0006] The edge computing layer is used for real-time data processing and intelligent analysis, including a synchronous acquisition and preprocessing module, a multi-dimensional real-time feature extraction engine, a lifetime assessment and prediction module, and a data management and communication module. Preferably, the edge computing layer includes edge computing nodes composed of x86 / ARM architecture processors, wherein the processors are configured with 4 or more cores and ≥8 GB of RAM.

[0007] The synchronous acquisition and preprocessing module is used to realize multi-channel synchronous sampling, hardware timestamp marking of signals, detrending, digital filtering and phase correction; to ensure the fidelity of the original signal.

[0008] The multi-dimensional real-time feature extraction engine is used to calculate the key features of the signal processed by the synchronous acquisition and preprocessing module. The key features include peak current, peak voltage, arcing time, I²t cumulative value, arc voltage ripple coefficient, re-ignition times, and micro-arc segment ratio. The multi-dimensional real-time feature extraction engine completes the automatic calculation of all the key features within 100ms after each tripping event. The life assessment and prediction module has a built-in support vector machine (SVM) model and a multi-parameter fusion model, which are used to receive standardized feature vectors, calculate the comprehensive degradation index (CDI) in real time, and output the health status classification and remaining life prediction results of the contacts. The data management and communication module is used to cache compressed raw waveform data, extracted features, and evaluation results, and transmit the data to a host computer or cloud platform via fiber optic isolation. This enables hierarchical data archiving, remote monitoring, model optimization, and historical data analysis.

[0009] Preferably, the edge computing layer is deployed at the monitoring site. Signal preprocessing and feature extraction are performed locally, and the lifetime status assessment model and remaining lifetime prediction algorithm are run locally. Only the compressed feature data and assessment results are uploaded to the cloud.

[0010] A second aspect of the present invention provides a method for assessing the life status of circuit breaker electrical contacts based on multi-parameter fusion, implemented based on the aforementioned assessment system, comprising the following steps: Step S1, Data Acquisition; Preferably, in step S1, when the circuit breaker tripping command is issued, the system uses the auxiliary contact disconnection time t0 provided by the auxiliary contact status monitoring unit as a unified time reference to synchronously trigger multi-channel data acquisition; it fully captures and records the original signals, including current i(t) and arc voltage u. a (t), coil parameters and auxiliary contact status, covering the entire decay process from the rising edge of the current to the end of the arc.

[0011] Step S2, standardize feature vector generation; Preferably, step S2 includes: S2.1, preprocess the raw signal acquired in step S1, including detrending, filtering, and integral correction, to eliminate noise and baseline drift; S2.2, based on the preprocessed signal in S2.1, extract multi-dimensional feature parameters, including peak current, peak voltage, arcing time, cumulative I²t value, arc voltage ripple coefficient, re-ignition frequency, and micro-arc segment ratio; standardize the feature parameters to generate a standardized feature vector with unified dimensions. .

[0012] Preferably, in step S2.2, the feature parameters are standardized using Min-Max normalization or Z-Score normalization methods, mapping each feature parameter to a unified interval [0, 1] or converting it into a distribution with a mean of 0 and a standard deviation of 1, thus eliminating the influence of dimensions; where I²t is a feature parameter that has a cumulative effect on contact erosion, a historical cumulative scaling factor is introduced for the cumulative value of I²t to reflect the historical total loss of the contact, and the calculation formula is as follows: I²t_cumulative(n) = I²t_cumulative(n-1) + I²t_current, Wherein, I²t_cumulative(n) represents the cumulative I²t value after the nth opening operation, which is used to characterize the cumulative arc heat energy loss borne by the contact up to the current moment; I²t_cumulative(n-1) represents the historical cumulative I²t value after the (n-1)th opening operation; and I²t_current represents the instantaneous I²t value generated during the current opening operation.

[0013] Step S3: Calculate the Comprehensive Degradation Index (CDI); Preferably, in step S3, the standardized feature vector generated in S2.2 is... The input is fed into the multi-parameter fusion model built into the life assessment and prediction module. The multi-parameter fusion model is based on the analytic hierarchy process (AHP) and uses standardized feature vectors. The weights of each characteristic quantity are assigned, and the Comprehensive Degradation Index (CDI) is calculated to quantify the overall wear and tear on the contact caused by a single operation.

[0014] Preferably, in step S3, calculating the Comprehensive Degradation Index (CDI) based on the Analytic Hierarchy Process (AHP) includes the following steps: (1) Constructing a hierarchical model: The target layer is the life state of the electrical contact, and the criterion layer is the standardized feature vector. The characteristic quantities in; (2) Constructing the judgment matrix: Based on the relative importance of each feature quantity to the contact life, compare each feature quantity in the criterion layer pairwise to construct the judgment matrix; (3) Calculate the weight vector and perform consistency test: By calculating the largest eigenvalue of the judgment matrix and its corresponding eigenvector, the AHP weight allocation of each feature is obtained and the consistency ratio CR test is performed to ensure the logical consistency of the judgment. (4) Determine the Comprehensive Degradation Index (CDI): Standardize the eigenvectors The overall degradation index (CDI) after each tripping operation is calculated by linearly weighting and fusing the corresponding AHP weights. CDI is a value between 0 and 1. The larger the value, the greater the wear on the contact in a single operation, or the worse the current health status of the contact.

[0015] Step S4: Pattern recognition and status output.

[0016] Preferably, in step S4, the standardized feature vector is... Together with the CDI value obtained in step S3, they are used as independent variables and input into the Support Vector Machine (SVM) model built into the life assessment and prediction module. Based on the decision boundary learned during its training phase, the SVM model performs pattern recognition on the input features and outputs the health status classification result of the contact and the remaining effective number of operations (RUL) of the contact. The health status classification includes: "healthy", "attention", "abnormal", and "critical".

[0017] If the status is "healthy", the data is recorded without warning. If the status reaches the "Attention" level, the system will issue a primary warning to remind maintenance personnel to pay closer attention. If the status enters the "abnormal" or "critical" level, or if the predicted RUL is lower than the preset safety threshold (e.g., 100 operations remaining), the system will immediately issue an advanced alarm and push maintenance suggestions to the mobile terminals of maintenance personnel and the back-end monitoring center.

[0018] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention adopts a real-time processing architecture of “cloud-edge collaboration”, deploys high-performance processors on edge computing nodes (field monitoring devices), and performs preprocessing such as detrending, filtering, integral correction and feature extraction locally. It also runs the life status assessment model and remaining life prediction algorithm locally, and only uploads the compressed feature data and assessment results to the cloud, which solves the problem of difficult transmission of massive waveform data. At the same time, it realizes the completion of event feature extraction and status assessment within 100ms after the circuit breaker is opened, which meets the strong real-time requirements of power equipment status monitoring.

[0019] (2) This invention proposes a weight allocation strategy based on the Analytic Hierarchy Process (AHP). For different dimensional parameters such as breaking current (instantaneous stress), arcing time (time effect), and I²t cumulative value (cumulative effect), a judgment matrix is ​​constructed to calculate the weights, and the comprehensive degradation index (CDI) is calculated by linear weighted fusion. Through the constructed unified quantitative index (CDI) between 0 and 1, the wear and current health status of the contact after a single operation can be characterized, providing a unified and physically interpretable decision basis for life prediction.

[0020] (3) Based on the arc ablation mechanism, this invention designs a multi-dimensional real-time feature extraction engine, which specifically extracts deep features such as arc voltage ripple coefficient, re-ignition times, and micro-arc segment ratio, and establishes a model in combination with I²t cumulative value; by directly monitoring the micro state of arc combustion, it can accurately reflect the micro ablation and material transfer on the contact surface, and improve the accuracy of electrical life assessment. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on the structures shown in these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the multi-dimensional electrical parameter acquisition and feature database construction process of the present invention.

[0023] Figure 2 This is a schematic diagram of the process for evaluating the life status of circuit breaker electrical contacts based on multi-parameter fusion, as described in this invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described 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. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0025] This invention relates to a circuit breaker electrical contact life status assessment system based on multi-parameter fusion, employing a cloud-edge collaborative real-time processing architecture, including a cloud platform and an edge processing terminal. The edge processing terminal comprises a perception layer and an edge computing layer. The circuit breaker electrical contact life status assessment method of this invention is implemented based on the aforementioned assessment system. This embodiment is divided into four parts: construction of a high-voltage spring circuit breaker electrical contact detection platform, multi-dimensional electrical parameter acquisition and feature database construction (offline stage), model training and deployment, and real-time assessment (online stage), as detailed below: I. Construction of a High-Voltage Spring Circuit Breaker Electrical Contact Testing Platform The high-voltage spring circuit breaker electrical contact testing platform includes a charging power supply, a protective resistor, a supercapacitor energy storage unit, a current measurement unit, a voltage probe, and a circuit breaker testing device. The charging power supply output range is 0-200 V DC, suitable for 1–2 kA discharge current conditions. Rated power is 3–5 kW, with a ripple factor ≤0.5%. Voltage / current regulation accuracy is ±0.5%. Protection functions include overvoltage / overcurrent / overtemperature / short circuit protection. The protection resistor structure is a metal film or water-cooled non-inductive series-parallel network, with a nominal resistance of 10–50 mΩ (set according to the target peak current). Transient power consumption is ≥100 kJ, with a 50 ms pulse tolerance. Temperature drift is ≤100 ppm / ℃, and inductance is ≤100 nH. The supercapacitor energy storage unit has a rated capacity of 50–200 F (modular stacking) and a rated voltage of ≤150 V. The equivalent series resistance (ESR) is ≤1 mΩ / module. The charging / discharging circuit uses a relay / IGBT controlled connection, with the circuit breaker under test connected in series in the discharge circuit. Safety features include passive voltage equalization. Active equalizing plate with built-in temperature probe (NTC); Circuit breaker under test and operating trigger: High-voltage spring-operated circuit breaker (rated voltage level adapted), trigger: closing / opening coil driven by isolation, with auxiliary contact lead-out as time reference trigger (jitter ≤100 μs), environment: test room temperature 10–40 ℃, relative humidity 20–80%, optional low-voltage enclosure inert gas protection; Current measurement unit: open-loop Rogowski coil + integrator, bandwidth: DC–5 MHz (integrator lower limit 100 Hz), rise time response ≤200 ns, range: ±5 kA pk; accuracy: ±1%, insulation and common mode: ≥10 kV (impulse); High-voltage differential probe: range 300 V / 1 kV, bandwidth ≥20MHz, common mode rejection ratio (CMRR): ≥80 dB@100 kHz, input capacitance: ≤5 pF; accuracy: ±1%; Data Acquisition (DAQ) and Time Synchronization: DAQ Channels: ≥8 channels of synchronous sampling (current, arc voltage, auxiliary contacts, coil voltage / current, temperature, etc.), Sampling Rate: ≥2 MS / s / channel (5 MS / s recommended), Quantization Bit Width: 16 bit, Front End: 50 Ω / 1 MΩ optional, Time Base: Hardware in-phase sampling, Absolute Timestamp Error ≤1 μs, Isolation: ≥600V CAT II between channels; Fiber optic isolation from host computer; Edge computing and storage: Processor: x86 / ARM (4 cores or more, ≥8 GB RAM), Real-time performance: Complete event feature extraction within 100 ms after the circuit breaker is triggered, Storage: Local ≥2 TB (NVMe), Event ring cache; RAID1 archiving to NAS, Software: Real-time acquisition service + Feature extraction module + Remaining lifetime (RUL) prediction module (containerized).

[0026] II. Multi-dimensional Electrical Parameter Acquisition and Feature Database Construction (Offline Stage)

[0027] This section completes the offline data collection and model training library construction, such as... Figure 1 As shown, it includes the following steps: (1) Experimental preparation Verify the nameplate and rated parameters of the circuit breaker under test, and confirm the test level and safety clearance. Complete the wiring: close the circuit according to the sequence "Power Supply → Protective Resistor → Supercapacitor → Circuit Breaker Under Test → Return Current"; connect the sensor according to the measurement polarity and ground reference; check the shielding and grounding. Perform interlock self-test and sensor self-test (zero point / noise / saturation margin). Load the latest calibration coefficients and operating condition configuration file (voltage setpoint, target peak current, sampling rate, trigger threshold).

[0028] (2) Charging and Readiness

[0029] The charging power supply is started, and constant voltage and current limiting charging is performed to the target voltage U_ch. The target voltage U_ch is set according to the parameters of the target peak current I_pk, the equivalent circuit resistance R_eq, and the equivalent circuit inductance L_eq to ensure that the preset peak discharge current and arc energy output can be obtained during the opening and discharge process. The DAQ enters the pre-trigger buffer (≥50 ms), waiting for the auxiliary contact and the current rising edge to trigger under dual conditions.

[0030] (3) Single tripping event acquisition

[0031] The controller issues a tripping command; records coil voltage / current and auxiliary contact status. Current i(t) and arc voltage u_a(t) are simultaneously acquired, fully covering the arcing stage and the subsequent 50–100 ms decay period. The system automatically completes the first round of feature extraction and quality assessment (signal-to-noise ratio, saturation, shearing, synchronization drift).

[0032] (4) Post-event processing and warehousing

[0033] Signal preprocessing: detrending, distortion point repair, bandpass / lowpass filtering (1 Hz–500 kHz), numerical integration and phase correction of the Rogowski coil link.

[0034] Based on the auxiliary contact disconnection time t0, the measurements of i(t), u_a(t), and coil are uniformly aligned.

[0035] Key feature calculations: Peak values ​​and statistics: I_pk, U_pk, arcing time t_arc (criterion for arc extinction from current zero crossing) and cumulative value of I²t, arc voltage ripple coefficient, number of reignitions, and proportion of micro-arc segments; Quality Labels: Data is labeled based on its integrity or abnormal events (saturation, missing samples, overshoot); qualified data, along with its feature values, is stored in the local feature database to form a standard sample set for subsequent SVM model training and optimization, while unqualified data is entered into the retest queue.

[0036] III. Model Training and Deployment

[0037] Based on a feature database constructed from historical accelerated aging test and field operation data, an SVM classifier and regression model were trained. According to the contact resistance, temperature rise, and disassembly inspection results, the contact status at different time points was labeled, such as "healthy," "caution," "abnormal," and "critical." An SVM classifier was trained using the standardized feature sequence and the corresponding CDI as input features and the status label as output. An SVM regression model was trained using the feature sequence and CDI sequence as input, and the remaining number of operations before the contact reaches its end-of-life (e.g., contact resistance increases to 200% of its initial value or temperature rise exceeds 70°C) as output. The classifier was used to identify the contact health status ("healthy," "caution," "abnormal," "critical"), and the regression model was used to predict the remaining life (RUL).

[0038] The trained SVM classifier and regression model are containerized and deployed to the lifetime assessment and prediction module of the edge computing layer.

[0039] IV. Real-time Assessment Process (Online Phase)

[0040] like Figure 2 As shown, based on the deployed SVM classifier and regression model, the system performs the following real-time inference steps: 1. Real-time acquisition and preprocessing: The system continuously monitors the tripping trigger signal, captures the event waveform based on the trigger signal, and performs real-time filtering and correction.

[0041] 2. Feature extraction and fusion: (1) Using the auxiliary contact disconnection time t0 as the reference, align the signal, calculate the key features and standardize them to generate feature vectors. .

[0042] The original features extracted from each tripping event vary greatly in scale and numerical range, and direct fusion would cause the model to favor features with larger numerical values. Therefore, standardization preprocessing is necessary. Normalization: Using Min-Max normalization or Z-Score standardization methods, each feature quantity is mapped to a uniform interval (such as [0, 1]) or converted into a distribution with a mean of 0 and a standard deviation of 1, thus eliminating the influence of dimensions.

[0043] Introduction of scaling factor: For parameters such as the cumulative value of I²t, which have a cumulative effect on contact erosion, a historical cumulative scaling factor is introduced to reflect the total historical loss of the contact. I²t_cumulative(n) = I²t_cumulative(n-1) + I²t_current.

[0044] Using the same algorithm logic as the second part of the database construction stage, feature calculation is completed within 100ms, but complex quality screening is no longer performed; instead, the features are directly vectorized.

[0045] (2) Calculate the Comprehensive Degradation Index (CDI) of the current event based on the AHP weights.

[0046] The lifespan of contacts is affected to varying degrees by different electrical parameters. This invention employs the Analytic Hierarchy Process (AHP) to assign weights to each characteristic quantity, constructing a scientific multi-parameter comprehensive evaluation system.

[0047] Construct a hierarchical model: the target layer is the "life state of electrical contacts"; the criterion layer consists of key feature quantities, such as: breaking current I_pk, arcing time t_arc, cumulative value of I²t, arc voltage U_pk, etc.

[0048] Constructing the judgment matrix: Based on the relative importance of each parameter to the contact life, the feature quantities in the criterion layer are compared pairwise to construct the judgment matrix.

[0049] Calculate the weight vector and perform consistency check: By calculating the largest eigenvalue of the judgment matrix and its corresponding eigenvector, the weight allocation of each feature is obtained and a consistency ratio (CR) check is performed to ensure the logical consistency of the judgment.

[0050] Determine the Comprehensive Degradation Index (CDI): The standardized feature values ​​are linearly weighted and fused with their corresponding AHP weights to calculate the comprehensive degradation index after each tripping operation. The CDI is a value between 0 and 1; a higher value indicates greater contact wear from a single operation, or a worse current contact health condition.

[0051] 3. Marginal Reasoning and Decision Making: The CDI input is fed into the SVM classifier and regression model that have been trained and deployed in the third part, and the contact health status label and RUL prediction value are output in real time.

[0052] The workflow is as follows: Feature input: After each circuit breaker trip, the edge computing node inputs the preprocessed and standardized latest feature values ​​and the calculated CDI into the trained SVM model.

[0053] State assessment: The SVM classifier outputs the current state category of the contact ("healthy", "attention", "abnormal", "critical") in real time based on the input features.

[0054] RUL Prediction: The SVM regression model predicts the remaining effective number of operations (RUL) of the contact based on the current and historical feature sequence and CDI sequence.

[0055] Result: The edge computing node encapsulates the state assessment results and RUL prediction values.

[0056] 4. Early warning and operation and maintenance suggestions: Implement tiered early warning based on the output status.

[0057] If the status is "healthy", the data is recorded and no warning is issued.

[0058] If the status reaches the "Attention" level, the system will issue a primary warning to remind maintenance personnel to pay closer attention.

[0059] If the status enters the "abnormal" or "critical" level, or if the predicted RUL is lower than the preset safety threshold (e.g., 100 operations remaining), the system will immediately issue an advanced alarm and push maintenance suggestions to the mobile terminals of maintenance personnel and the back-end monitoring center through the communication module.

[0060] This system interacts with the monitoring platform of the circuit breaker operating mechanism via standard communication protocols (such as IEC 61850 MMS or Modbus TCP). It performs collaborative diagnosis with the contact life assessment results and the monitoring results of the mechanism's mechanical characteristics (opening and closing speeds, stroke curves, mechanical vibration, etc.) to comprehensively assess the overall health status of the circuit breaker and avoid false alarms and missed alarms.

[0061] The system has a multi-level early warning mechanism and provides differentiated operation and maintenance strategies from "observation" to "immediate repair" based on the predicted RUL and status assessment results, so as to achieve predictive maintenance and effectively avoid the problems of "over-maintenance" and "faults not being detected in time".

Claims

1. A multi-parameter fusion based circuit breaker electrical contact life condition assessment system, characterized in that, The system adopts a cloud-edge collaborative real-time processing architecture, which includes a cloud and an edge processing terminal, wherein the edge processing terminal includes a perception layer and an edge computing layer. The sensing layer includes an electrical parameter acquisition module, an auxiliary status monitoring module, and an environmental temperature control module. The electrical parameter acquisition module includes an open-loop Rogowski coil and a high-voltage differential probe. The auxiliary status monitoring module includes an auxiliary contact status monitoring unit and an operating coil monitoring unit. The auxiliary contact status monitoring unit provides a time reference for the tripping moment. The operating coil monitoring unit is used to synchronously record the voltage and current waveforms of the tripping coil. The environmental temperature control module includes an NTC temperature probe. The edge computing layer includes a synchronous acquisition and preprocessing module, a multi-dimensional real-time feature extraction engine, a lifetime assessment and prediction module, and a data management and communication module. The synchronous acquisition and preprocessing module is used to realize multi-channel synchronous sampling, hardware timestamp marking of signals, detrending, digital filtering and phase correction; The multi-dimensional real-time feature extraction engine is used to calculate the key features of the signal processed by the synchronous acquisition and preprocessing module. The key features include peak current, peak voltage, arcing time, I²t cumulative value, arc voltage ripple coefficient, re-ignition times, and micro-arc segment ratio. The multi-dimensional real-time feature extraction engine completes the automatic calculation of all the key features within 100ms after each tripping event. The life assessment and prediction module incorporates a support vector machine (SVM) model and a multi-parameter fusion model to calculate the comprehensive degradation index (CDI) in real time and output the health status classification and remaining life prediction results of the contacts. The data management and communication module is used to cache compressed feature data and evaluation results, and transmit the data to the host computer or cloud platform through fiber optic isolation.

2. The life condition assessment system according to claim 1, characterized in that, The edge computing layer includes edge computing nodes composed of x86 / ARM architecture processors, wherein the processors are configured with 4 or more cores and ≥8 GB of RAM.

3. The life condition assessment system according to claim 1, characterized in that, The edge computing layer is deployed at the monitoring site.

4. A method for assessing the lifespan status of circuit breaker electrical contacts based on multi-parameter fusion, implemented based on the lifespan status assessment system according to any one of claims 1-3, characterized in that, Includes the following steps: Step S1, Data Acquisition; Step S2, standardize feature vector generation; Step S3: Calculate the Comprehensive Degradation Index (CDI); Step S4: Pattern recognition and status output.

5. The lifespan condition assessment method according to claim 4, characterized in that, In step S1, when the circuit breaker tripping command is issued, the system uses the auxiliary contact disconnection time t0 provided by the auxiliary contact status monitoring unit as a unified time reference to synchronously trigger multi-channel data acquisition. The complete signal is captured and recorded, including the current i(t), the arc voltage u a (t), the coil parameters and the auxiliary contact status, covering the complete process from the current rise to the end of the arc decay.

6. The lifespan condition assessment method according to claim 5, characterized in that, Step S2 includes: S2.1, preprocess the raw signal acquired in step S1, including detrending, filtering, and integral correction, to eliminate noise and baseline drift; S2.2, based on the preprocessed signal in S2.1, extract multi-dimensional feature parameters, including peak current, peak voltage, arcing time, cumulative I²t value, arc voltage ripple coefficient, re-ignition frequency, and micro-arc segment ratio; standardize the feature parameters to generate a standardized feature vector with unified dimensions. .

7. The lifespan status assessment method according to claim 6, characterized in that, In step S2.2, the feature parameters are standardized using Min-Max normalization or Z-Score standardization methods, mapping each feature parameter to a unified interval [0, 1] or converting it into a distribution with a mean of 0 and a standard deviation of 1, thus eliminating the influence of dimensions; for the cumulative value of I²t, a historical cumulative proportion factor is introduced, calculated using the following formula: I²t_cumulative(n) = I²t_cumulative(n-1) + I²t_current, Wherein, I²t_cumulative(n) represents the cumulative I²t value after the nth opening operation, which is used to characterize the cumulative arc heat energy loss borne by the contact up to the current moment; I²t_cumulative(n-1) represents the historical cumulative I²t value after the (n-1)th opening operation; and I²t_current represents the instantaneous I²t value generated during the current opening operation.

8. The lifespan condition assessment method according to claim 7, characterized in that, In step S3, the standardized feature vector generated in S2.2 is... The input is fed into the multi-parameter fusion model built into the life assessment and prediction module. The multi-parameter fusion model is based on the analytic hierarchy process (AHP) and uses standardized feature vectors. The weights of each characteristic quantity are assigned, and the comprehensive degradation index (CDI) is calculated.

9. The lifespan condition assessment method according to claim 8, characterized in that, In step S3, the Comprehensive Degradation Index (CDI) is calculated based on the Analytic Hierarchy Process (AHP), including the following steps: (1) Constructing a hierarchical model: The target layer is the life state of the electrical contact, and the criterion layer is the standardized feature vector. The characteristic quantities in; (2) Constructing the judgment matrix: Based on the relative importance of each feature quantity to the contact life, compare each feature quantity in the criterion layer pairwise to construct the judgment matrix; (3) Calculate the weight vector and perform consistency test: By calculating the largest eigenvalue of the judgment matrix and its corresponding eigenvector, the AHP weight allocation of each feature is obtained and the consistency ratio CR test is performed to ensure the logical consistency of the judgment. (4) Determine the Comprehensive Degradation Index (CDI): Standardize the eigenvectors The comprehensive degradation index (CDI) after each circuit breaker operation is calculated by linearly weighting and fusing the corresponding AHP weights.

10. The lifespan status assessment method according to claim 9, characterized in that, In step S4, the standardized feature vectors are... Together with the CDI value obtained in step S3, they are used as independent variables and input into the support vector machine (SVM) model built into the life assessment and prediction module. The SVM model performs pattern recognition on the input features and outputs the health status classification result of the contact and the remaining effective number of operations (RUL) of the contact. The health status classification includes: "healthy", "attention", "abnormal" and "critical".

Images