Method, system and equipment for predicting service life of relay through multi-working-condition simulation capacitance detection and medium

By combining multi-condition simulation and multi-dimensional parameter acquisition with physical mechanisms and machine learning algorithms, the problems of single-condition simulation and prediction lag in relay life testing are solved, enabling early fault warning and accurate life assessment.

CN121432167APending Publication Date: 2026-01-30GUIZHOU POWER GRID CO LTD
View PDF 0 Cites 5 Cited by

Patent Information

Application Number
CN202511757366.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-11-27
Publication Date
2026-01-30

AI Technical Summary

Technical Problem

Existing relay life testing methods suffer from limitations such as simplistic operating condition simulation, delayed life prediction, and limited detection methods. They are unable to accurately simulate complex operating modes and extreme power grid conditions, cannot provide early warning of potential faults, and the prediction models lack adaptive capabilities.

Method used

By applying various simulated operating conditions and collecting multi-dimensional parameters in real time, a hybrid prediction model is constructed. Combining physical mechanisms, statistical degradation, and machine learning algorithms, the model monitors changes in contact capacitance and enables early warning of performance degradation.

Benefits of technology

It significantly improves the accuracy and timeliness of relay life prediction, enabling early detection of potential faults and providing accurate remaining life and reliability assessments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121432167A_ABST
    Figure CN121432167A_ABST
Patent Text Reader

Abstract

The invention discloses a multi-condition simulation capacitance detection relay life prediction method, system, equipment and medium, and relates to the technical field of relays, and the method comprises the steps: applying a plurality of simulation conditions to a relay through a programmable driving power supply, simulating the complex conditions in actual operation, and collecting the multi-dimensional parameters of the relay in real time in the execution process of the simulation conditions, extracting characteristic parameters from the collected multi-dimensional parameters, carrying out data preprocessing on the characteristic parameters, constructing a life prediction model, carrying out prediction calculation and reliability grade evaluation on the residual life of the relay by utilizing the life prediction model, and carrying out reliability grade evaluation on the residual life of the relay by monitoring capacitance change between contacts of the relay. And early performance degradation early warning is carried out in combination with a prediction result. According to the invention, through fusion of multi-working-condition simulation, hybrid modeling and capacitance detection, full-chain optimization from data acquisition, accurate prediction to early warning is realized, and the accuracy, timeliness and engineering application value of relay life prediction are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of relay technology, specifically to a method, system, device, and medium for predicting the lifespan of a multi-condition simulated capacitance detection relay. Background Technology

[0002] Traditional relay life testing methods are mainly based on accelerated life tests under standard operating conditions, which assess the average life by statistically analyzing failure data from a large number of samples. In recent years, with the development of testing technology, researchers have begun to use electrical parameters such as contact resistance and operating time as characteristic quantities for life assessment, and have attempted to introduce machine learning algorithms to establish life prediction models. In terms of testing methods, current technologies generally rely on resistance measurement in the conducting state. Some advanced solutions have begun to explore non-invasive testing methods based on vibration analysis and thermal imaging. These methods have improved the accuracy of life assessment to some extent, but have not yet broken through the limitations of the traditional testing paradigm.

[0003] Despite the progress made in existing technologies, significant shortcomings remain in achieving accurate early prediction: First, the operating condition simulation capabilities of existing testing systems are relatively limited, making it difficult to reproduce the complex operating modes and extreme power grid conditions in actual operation. In particular, key failure mechanisms such as reduced contact pressure and increased arcing caused by insufficient coil attraction under undervoltage conditions cannot be effectively simulated, resulting in significant deviations between test results and actual applications. Second, the traditional determination of the end of life is usually based on complete contact failure, which is a destructive test and cannot effectively warn of early degradation of relay performance. Third, existing detection methods mostly rely on electrical parameters such as contact resistance, which only change significantly in the later stages of life and are not sensitive enough to potential faults such as early oxidation and fretting wear. In addition, existing prediction models mostly adopt a single modeling approach, either relying solely on data-driven approaches while ignoring physical mechanisms, or relying solely on empirical formulas and lacking adaptive capabilities. It is difficult to establish an accurate mapping relationship between parameters and lifespan, let alone achieve dynamic reliability assessment based on multi-dimensional feature fusion. Summary of the Invention

[0004] In view of the above-mentioned existing problems, the present invention provides a method, system, device and medium for predicting the life of a multi-condition simulated capacitor detection relay, in order to solve the problems of single-condition simulation, delayed life prediction and limited detection methods in the prior art.

[0005] To address the aforementioned technical issues, a multi-condition simulated capacitor detection relay life prediction method is proposed, including: By applying various simulated operating conditions to the relay using a programmable drive power supply, complex conditions in actual operation are simulated. During the execution of the simulated operating conditions, multi-dimensional parameters of the relay are collected in real time. Feature parameters are extracted from the collected multi-dimensional parameters to characterize the relay performance status. The feature parameters are preprocessed, and a life prediction model is constructed based on the preprocessed feature parameters to describe the mapping relationship between parameters and life. Using the life prediction model, the remaining life of the relay is predicted and the reliability level is assessed. Furthermore, by monitoring the capacitance changes between the relay contacts and combining the prediction results, early warning of performance degradation is provided.

[0006] As a preferred embodiment of the multi-condition simulated capacitor detection relay life prediction method of the present invention, the data preprocessing of the feature parameters includes: cleaning the feature parameters, identifying and processing outliers and missing values ​​in the data; normalizing the cleaned feature parameters, unifying the data scale and eliminating dimensional differences; and reducing the dimensionality of the normalized feature parameters to extract core features.

[0007] As a preferred embodiment of the multi-condition simulated capacitor detection relay life prediction method of the present invention, the construction of the life prediction model includes: selecting key feature parameters from the core features after dimensionality reduction, training the prediction model based on the key feature parameters, learning the correlation between parameters and life through historical data, optimizing the parameters of the prediction model, and adjusting the model structure.

[0008] As a preferred embodiment of the multi-condition simulated capacitance detection relay life prediction method of the present invention, the remaining life prediction calculation of the relay includes calculating the cumulative damage index based on the prediction model and quantifying the degree of performance degradation of the relay during operation. The remaining lifespan is estimated based on the cumulative damage index, the remaining lifespan is assessed, the health status is evaluated in combination with parameters, and a comprehensive prediction result is output.

[0009] As a preferred embodiment of the multi-condition simulated capacitor detection relay life prediction method of the present invention, the multi-condition simulated operating conditions include undervoltage condition, overvoltage condition, noise interference condition and high-speed pulse condition. Among them, the overvoltage condition applies a drive signal higher than the rated voltage to test the life under overexcitation shock; the high-speed pulse condition outputs a high-frequency pulse sequence to test the relay's tolerance under frequent switching conditions. The real-time acquisition relay's multi-dimensional parameters include drive voltage, coil current, contact voltage, contact current, operating time, bounce time, contact open-circuit capacitance, temperature rise, noise figure, and over / under voltage characteristics. Among them, the open-circuit capacitance of the contacts is measured by a high-precision capacitance measuring instrument when the relay is open, and is specifically determined by the dielectric constant of the dielectric, the effective contact area and the contact spacing based on the capacitance formula; the temperature rise is monitored by an infrared thermal imager to monitor the contact temperature; the action time and bounce time are recorded by a data acquisition card and the time parameters are extracted.

[0010] As a preferred embodiment of the multi-condition simulated capacitor detection relay life prediction method of the present invention, the life prediction model is constructed by combining a physical mechanism model, a statistical degradation model and an intelligent prediction model. The physical mechanism model is based on the cumulative damage equation to calculate the cumulative damage degree; the statistical degradation model is based on historical data distribution and updates the posterior degradation rate through Bayesian analysis; the intelligent prediction model is based on machine learning algorithms, including a grey prediction model, a BP neural network and an LSTM time series model, to process multi-dimensional feature parameters. The cumulative damage equation is expressed as: in, The cumulative damage at the current moment. For time, Let t be the total number of actions up to time t. The material erosion coefficient, Let be the arc energy of the i-th action. The thermal degradation coefficient, To activate energy, Let be the ideal gas constant. Let be the absolute temperature at time τ. For time indexing, This is the weighting factor for the increase in contact resistance. This is the increase in contact resistance. The contact resistance at the current moment. This represents the initial contact resistance. The method for predicting the remaining life of the relay also includes using a grey prediction model to make predictions based on time-series trends, obtaining the final remaining life value by weighted fusion of prediction results, and optimizing the prediction accuracy using a deviation correction term. The formula for obtaining the final remaining lifetime value is expressed as: in, This is a predicted remaining useful life based on a physical model. The cumulative damage at the current moment. For the posterior degradation rate, The coefficient of variation of the load. The failure threshold contact resistance, This is the contact resistance value at the previous measurement moment. This is the predicted remaining useful life based on the grey prediction model. and The development coefficient and grey action quantity of the grey prediction model. Let be the predicted value of a cumulative generated sequence obtained at time t using the grey model. The sampling interval is... This represents the final predicted remaining lifetime after fusion. These are the weighting coefficients. This is a deviation correction term; The reliability level assessment includes single-parameter assessment by calculating the degree to which each monitoring parameter deviates from the health threshold, weighting and integrating the assessment results into a comprehensive health index, using the comprehensive health index to correct the baseline failure rate to obtain the equivalent failure rate, classifying the equipment status into four levels according to the value of the comprehensive health index, and automatically matching the remaining life prediction range with maintenance recommendations. The formula for calculating the comprehensive health index is as follows: in, For health index, For monitoring parameter index, Let j be the current actual measured value of the j-th monitoring parameter. The attention threshold for the j-th parameter. Let the limit threshold be the j-th parameter. The weight coefficient for the j-th parameter; The formula for obtaining the equivalent failure rate is expressed as: in, Let be the equivalent failure rate at time t. Let be the baseline failure rate at time t. For shape parameters, For scale parameters, As a health influencing factor, For time, It is an exponential function.

[0011] As a preferred embodiment of the multi-condition simulated capacitance detection relay life prediction method of the present invention, the early performance degradation warning includes: applying a high-frequency test signal using a high-precision capacitance measuring instrument to measure the capacitance value between relay contacts, analyzing the changing trend of the capacitance value, determining whether it exceeds the normal fluctuation range, triggering a warning signal when the changing trend indicates performance degradation, and inputting the capacitance change data into the life prediction model to update the prediction result; The formula for calculating capacitance is as follows: in, The dielectric constant of the medium between the contacts is... This is the capacitance value. For effective contact area, This refers to the contact spacing; The early performance degradation warning also includes combining capacitance change data with contact resistance and temperature rise parameters to construct a multi-dimensional feature vector. By analyzing the capacitance change trend through a hybrid model, potential faults such as early oxidation and fretting wear can be identified. When the capacitance value shows a continuous monotonic change, even if the contact resistance has not deteriorated significantly, an early warning is triggered and the health status assessment is updated.

[0012] The beneficial effects of this preferred technical solution are as follows: by monitoring the capacitance change trend between contacts and combining it with hybrid model analysis, sensitive detection of early performance degradation of relays is achieved. By utilizing the high sensitivity of capacitance measurement to minute physical changes, potential faults can be detected before the contact resistance changes significantly, thus providing early fault warning and laying the foundation for predictive maintenance.

[0013] As a preferred embodiment of the multi-condition simulated capacitor detection relay life prediction system of the present invention, it is characterized by including a multi-condition simulation driving module, a multi-dimensional data acquisition and feature extraction module, an intelligent life prediction module, and a reliability assessment and decision support module.

[0014] The multi-condition simulation drive module is used to simulate rated operating conditions and extreme operating conditions through a programmable drive power supply.

[0015] The multi-dimensional data acquisition and feature extraction module is used to collect multi-dimensional electrical and mechanical parameters such as driving voltage, coil current, contact voltage, contact current, action time, bounce time, contact open-circuit capacitance, and temperature rise in real time, and to extract key features characterizing performance degradation.

[0016] The intelligent lifespan prediction module is used to fuse a physical model based on the cumulative damage equation, a statistical degradation model based on Bayesian updates, and an intelligent prediction model based on machine learning. By weighted fusion of the physical damage prediction results and the time-series trend prediction results, it outputs the remaining lifespan prediction value and calculates the comprehensive health index.

[0017] The reliability assessment and decision support module is used to classify the relay status into four levels based on the comprehensive health index, and automatically associate the remaining lifespan range with maintenance recommendations. It calculates the equivalent failure rate by substituting the health index into the failure rate model.

[0018] A computer device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of a method for predicting the lifespan of a multi-condition simulated capacitor detection relay.

[0019] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a method for predicting the lifespan of a multi-condition simulated capacitor detection relay.

[0020] The beneficial effects of this invention are as follows: This invention applies various simulated operating conditions to a programmable drive power supply and collects multi-dimensional parameters in real time, making the test results closer to actual engineering conditions. Furthermore, by establishing a preprocessing workflow including data cleaning, normalization, and dimensionality reduction, a high-quality data foundation is provided for subsequent modeling, significantly improving model stability. By constructing a hybrid prediction architecture combining physical mechanism models, statistical degradation models, and intelligent prediction models, the physical meaning of the model is clear, and it possesses adaptive learning capabilities. By weighted fusion of physical damage prediction and time-series trend prediction results, high-precision prediction is maintained at different life stages. By establishing a comprehensive health index and equivalent failure rate calculation model, quantitative assessment and fine-grained classification of relay health status are achieved. By monitoring the capacitance change trend between contacts and combining it with hybrid model analysis, potential faults are detected before significant changes in contact resistance, significantly advancing fault warnings. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0022] Figure 1 This is a flowchart illustrating the overall process of a multi-condition simulated capacitor detection relay life prediction method provided in one embodiment of the present invention.

[0023] Figure 2This is a schematic diagram of a test panel for a multi-condition simulated capacitance detection relay life prediction method provided in one embodiment of the present invention.

[0024] Figure 3 This is a schematic diagram of the framework of a multi-condition simulated capacitance detection relay life prediction method provided in an embodiment of the present invention.

[0025] Figure 4 This is a schematic diagram illustrating the construction and evaluation process of a big data model for relay reliability in a multi-condition simulated capacitance detection relay life prediction method provided in one embodiment of the present invention.

[0026] Figure 5 The flowchart shows a system scheme for a multi-condition simulated capacitance detection relay life prediction system provided in one embodiment of the present invention. Detailed Implementation

[0027] 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.

[0028] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for predicting the lifespan of a multi-condition simulated capacitance detection relay is provided, comprising: S100: Apply various simulated operating conditions to the relay through a programmable drive power supply to simulate complex conditions in actual operation, and collect multi-dimensional parameters of the relay in real time during the execution of the simulated operating conditions.

[0029] S200: Extract feature parameters from the collected multi-dimensional parameters to characterize the relay performance status, perform data preprocessing on the feature parameters, and construct a life prediction model based on the preprocessed feature parameters to describe the mapping relationship between parameters and life.

[0030] S300: Utilizes a life prediction model to predict and calculate the remaining life of relays and assess their reliability level. It also monitors capacitance changes between relay contacts and uses the prediction results to provide early warnings of performance degradation.

[0031] It should be noted that through multi-condition simulation and multi-dimensional parameter acquisition, as well as data preprocessing and hybrid model construction, a deep integration of physical mechanisms, statistical laws and intelligent algorithms has been achieved, improving the accuracy and robustness of the prediction model. By comprehensively evaluating the health index and reliability level, and introducing high-sensitivity monitoring of contact capacitance, early warning of early micro-degradation has been achieved, improving the accuracy, timeliness and engineering application value of relay life prediction.

[0032] Example 2, refer to Figures 1-4 This is a second embodiment of the present invention, which provides a method for predicting the lifespan of a multi-condition simulated capacitance detection relay, including: In step S100, the programmable drive power supply includes a programmable relay drive power supply, the output frequency, voltage / current amplitude and output power are all adjustable, and various working conditions can be simulated by changing the power supply output frequency, capacity, current limiting point and duty cycle.

[0033] Furthermore, in step S100, the various simulated operating conditions include undervoltage operating conditions, overvoltage operating conditions, noise interference operating conditions, and high-speed pulse operating conditions, specifically including steps S101~S104: S101: The undervoltage condition is set to be 70%~85% below the rated voltage to simulate insufficient contact pressure caused by power grid fluctuations.

[0034] S102: Apply a drive signal higher than the rated voltage under overvoltage conditions to test the lifespan under overexcitation shock.

[0035] S103: Noise interference condition where more than 30% noise is superimposed on the drive signal to simulate a harsh industrial electromagnetic environment.

[0036] S104: High-speed pulse mode outputs a high-frequency pulse sequence to test the relay's tolerance under frequent switching conditions.

[0037] Furthermore, in step S100, the multi-dimensional parameters of the real-time acquired relay include driving voltage, coil current, contact voltage, contact current, operating time, bounce time, contact open-circuit capacitance, temperature rise, noise figure, and over / under voltage characteristics. Among them, the open-circuit capacitance of the contacts is measured by a high-precision capacitance meter when the relay is open, and is specifically determined by the dielectric constant of the dielectric, the effective contact area and the contact spacing based on the capacitance formula; the temperature rise is monitored by an infrared thermal imager to ensure that the temperature rise is less than 65K under continuous operation; the action time and bounce time are recorded by a data acquisition card and the time parameters are extracted.

[0038] In step S200, the extraction of feature parameters includes extracting key features from the collected raw data, including extracting the dynamic curve of contact resistance, action time drift, rebound energy, contact open circuit capacitance, temperature rise, noise figure, and over / under voltage characteristics.

[0039] Furthermore, in this embodiment of the application, in step S200, the data preprocessing includes steps S201 to S203: S201: Data cleaning includes calculating the historical mean and standard deviation of each feature parameter, and marking data points that exceed the mean ± 3 times the standard deviation as outliers; for missing values, linear interpolation is used to calculate and fill in the missing values ​​using the valid data points adjacent to the current data point. The formula for linear interpolation is expressed as: in, For timestamp At point q, the estimated value after imputing the missing values ​​of the q-th feature. The time points corresponding to the missing values. For the previous valid time point At point q, the measured value of the q-th feature. For the later valid time point The measured value of the q-th feature at that location.

[0040] S202: Data normalization includes using the Min-Max normalization method to identify the minimum and maximum values ​​in the entire test dataset for each feature parameter (driving voltage and contact open-circuit capacitance), and mapping each data point to the [0,1] interval through a linear transformation; The Min-Max normalization formula is expressed as: in, For the q-th feature, the normalized value of the p-th data point is... This represents the original value of the p-th data point within the q-th feature. The minimum value among all data points for the q-th feature. The maximum value among all data points for the q-th feature. The upper limit of the target interval is 1. The lower limit of the target interval is 0.

[0041] S203: Data dimensionality reduction includes using principal component analysis to center all data (i.e., subtract their respective averages), calculate the covariance matrix of the dataset, and perform mathematical decomposition to identify the directions with the largest variance in the data, i.e., principal components. These principal components are then ranked according to their proportion of the original data variance explained (i.e., cumulative contribution rate), and the top n principal components that can explain more than 95% of the information are selected. The original high-dimensional data is then projected into a new space composed of n principal components, thereby achieving data dimensionality reduction, eliminating redundant information, and extracting the most representative core features. The covariance matrix is ​​expressed as: in, Let covariance matrix be the variance matrix. The original data matrix has already undergone centralized processing. Let X be the transpose of matrix X. The number of samples; The mathematical decomposition formula is expressed as: in, Let Σ be the eigenvalues ​​of the covariance matrix. For the corresponding eigenvalues eigenvectors; The projection matrix and dimensionality reduction formula are expressed as follows: in, Let be the projection matrix. The new data matrix after dimensionality reduction; The cumulative contribution rate is expressed as: in, The sum of the first n eigenvalues ​​represents the total variance captured by the first n principal components. The sum of all eigenvalues ​​represents the total variance of the original data. The total number of all eigenvalues.

[0042] In an optional implementation, in step S201, data cleaning further includes using clustering algorithms such as K-means to cluster all the collected multi-dimensional parameters (contact resistance, action time, temperature rise), dividing the data into several clusters, and identifying the clusters with very few samples or the furthest from the cluster center as outlier clusters and removing them as a whole. For missing values, the mean of the remaining samples in the same cluster is used to fill them.

[0043] In another optional implementation, in step S201, data cleaning may further include using a fixed number of action cycles as a sliding window, calculating the local mean and variance of a specific parameter (rebound energy) within each window, and determining that a data point in the window is abnormal and removed when its deviation from the local mean exceeds a preset dynamic threshold (the threshold is related to the local variance). For missing values, the local mean in the current window is used to fill them.

[0044] In an optional implementation, step S202, data normalization further includes calculating the mean and standard deviation of each feature parameter across the entire test dataset, and converting each data point into a Z-score obtained by dividing the difference from the mean by the standard deviation.

[0045] In another alternative implementation, in step S202, data normalization may further include reducing the maximum absolute value of each feature parameter to below 1 by finding a suitable power of 10.

[0046] Furthermore, in step S200, the construction of the lifetime prediction model includes steps S211-S213: S211: Key characteristic parameters to be selected include, and the mandatory parameters include, contact resistance: measured using the four-wire method, with a sampling frequency ≥ 1 time / 100 actions; dynamic time parameters: pull-in time, overtravel time, bounce time, and arcing time; temperature rise: the contact temperature is monitored by an infrared thermal imager, and the temperature rise should be < 65K under continuous operation; and arc energy.

[0047] Parameter priority includes, first-level features: contact resistance, overtravel time (Fisher discrimination value is the highest 8.2); Secondary characteristics: arcing time, absorption time, and temperature rise; Level 3 characteristics: load current and operating frequency.

[0048] S212: Constructing a hybrid model architecture includes adopting a hybrid architecture based on the DB-G model, establishing a physics-based cumulative damage equation, and quantifying the cumulative damage level of the relay at any time. The cumulative damage equation is expressed as: in, The cumulative damage at the current moment. For time, Let t be the total number of actions up to time t. The material erosion coefficient, Let be the arc energy of the i-th action. The thermal degradation coefficient, To activate energy, Let be the ideal gas constant. Let be the absolute temperature at time τ. For time indexing, This is the weighting factor for the increase in contact resistance. This is the increase in contact resistance. The contact resistance at the current moment. The initial contact resistance; the cumulative damage calculated by the physical model. or When the time comes, the lifespan is determined to have ended.

[0049] S213: In order to capture the complex nonlinear relationships and time-series dependencies in the data, the model integrates advanced machine learning algorithms, including a grey prediction model suitable for small sample trend prediction to train the prediction model, a BP neural network for fitting complex mapping relationships, and an LSTM model that is good at processing time series data. It should be noted that the grey prediction model (GM(1,1)) is suitable for small samples (M<20) and has a prediction error of <5%; the BP neural network includes an input layer = number of features, a hidden layer = 10-15 nodes, and an output layer = 1 (remaining lifetime); the LSTM time series model captures the contact resistance degradation trend.

[0050] In this embodiment of the application, step S211, the selection of key feature parameters includes steps A1~A2: A1: The Fisher score is calculated by using the feature scoring method based on Fisher discrimination to calculate the ratio of inter-class dispersion to intra-class dispersion of each core feature parameter (overrun time, arc time) in different life stages (healthy period, decline period). The formula for calculating the Fisher score is expressed as follows: in, The Fisher score for the q-th feature. Let be the inter-class scatter of the q-th feature. Let be the intra-class scatter of the q-th feature. The total number of categories, This is a categorical index, where g=1 represents the healthy period and g=2 represents the declining period. Let g be the number of samples belonging to the g-th category. Let q be the mean of the q-th feature in the g-th category (life stage). Let q be the global mean of the q-th feature across all samples. Let be the specific value of the q-th feature of the c-th sample. Let g be the set of sample indices belonging to the g-th category.

[0051] A2: Sort all features from highest to lowest score and select the highest-ranking feature as the key feature parameter.

[0052] In an optional implementation, step S211, the selection of key feature parameters further includes calculating the mutual information between each core feature parameter and the relay remaining life tag. The larger the mutual information, the stronger the correlation between the current feature and the life, and the more information about the life it contains. The top K features with the largest mutual information with the life are selected as key feature parameters.

[0053] In another optional implementation, in step S211, the selection of key feature parameters may further include adding an L1 regularization term to the loss function when constructing the regression prediction model. By adjusting the regularization strength, some feature coefficients of the model are compressed to zero, thereby achieving feature selection, and retaining features with non-zero coefficients as key feature parameters.

[0054] It should be noted that, in the embodiments of this application, in step S213, the training prediction model includes training with a gray prediction model GM(1,1), generating a key feature parameter sequence (time series of contact resistance) by accumulating it once to weaken randomness and highlight trend, establishing a first-order linear differential equation, i.e., the GM(1,1) model, fitting it, solving the development coefficient and gray action of the model, and using the current model to extrapolate the development of the feature parameters to calculate the remaining life.

[0055] In an optional implementation, in step S213, the training prediction model further includes constructing a fully connected neural network containing an input layer (the number of nodes equals the number of key feature parameters), a hidden layer, and an output layer (single node, outputting remaining lifetime), training the network using a historical dataset (feature parameters as input, actual remaining lifetime as target output), and continuously adjusting the weights and biases of neurons in each layer through an error backpropagation algorithm until the network output error converges to a predetermined range.

[0056] In another optional implementation, in step S213, the training prediction model may further include mapping key feature parameters to a high-dimensional feature space, finding the optimal regression hyperplane in the current space such that the deviation between all training sample data points and the current hyperplane is minimized, handling nonlinear relationships by using a kernel function (radial basis kernel function), and introducing slack variables to tolerate some outliers, and constructing a regression model for lifetime prediction.

[0057] In step S300, the prediction calculation of the remaining lifespan of the relay includes steps S301 to S303: S301: In the early stage of damage, the life is estimated based on the difference between the current damage value and the average degradation rate, and will be corrected in combination with the fluctuation of the load conditions. When the damage enters the later stage, the model switching strategy focuses on the accelerated deterioration trend of the contact resistance and calculates the time to reach the failure threshold by comparing the current value with the rate of change of the previous measurement point. The physical prediction formula based on damage degree is expressed as: in, This is a predicted remaining useful life based on a physical model. The cumulative damage at the current moment. For the posterior degradation rate, The coefficient of variation of the load. The failure threshold contact resistance, This is the contact resistance value at the previous measurement moment.

[0058] S302: Using a grey prediction model, analyze the time series data of contact resistance parameters and calculate the time required for the contact resistance to reach the preset failure threshold (three times the initial value); The time series prediction formula based on the grey model is expressed as: in, This is the predicted remaining useful life based on the grey prediction model. and The development coefficient and grey action quantity of the grey prediction model. Let be the predicted value of a cumulative generated sequence obtained at time t using the grey model. The sampling interval is denoted as .

[0059] S303: Introduce a bias correction term to fine-tune the results based on the accuracy of historical predictions, and obtain the final remaining lifetime value through weighted fusion of prediction results; The formula for obtaining the final remaining lifetime value is expressed as: in, This represents the final predicted remaining lifetime after fusion. These are the weighting coefficients. This is a deviation correction term.

[0060] Furthermore, in step S300, the reliability level assessment includes steps S311 to S313: S311: Single-parameter assessment is performed by calculating the degree to which each monitoring parameter deviates from the health threshold, and the assessment results are weighted and integrated into a comprehensive health index; The formula for calculating the comprehensive health index is as follows: in, For health index, For monitoring parameter index, Let j be the current actual measured value of the j-th monitoring parameter. The attention threshold for the j-th parameter. Let the limit threshold be the j-th parameter. The weight coefficient for the j-th parameter.

[0061] S312: The baseline failure rate is corrected using a comprehensive health index, and the equivalent failure rate formula is expressed as follows: in, Let be the equivalent failure rate at time t. Let be the baseline failure rate at time t. For shape parameters, For scale parameters, As a health influencing factor, For time, It is an exponential function.

[0062] S313: Based on the value of the comprehensive health index, the equipment status is divided into four levels, and the remaining life prediction range and maintenance recommendations are automatically matched. The specific equipment levels include: Level A (Health): HIS score of 85-100. Rated lifespan Normal operation, routine monitoring; Grade B (Note): HIS value 70~84. For 30% to 60% of the rated life, The value is 0.1~0.5×10. -6 / h, enhanced monitoring (sampling frequency ×2); Level C (early warning): HIS is 50~69. For 10% to 30% of the rated life, The value is 0.5~2.0×10. -6 / h, planned maintenance, prepare spare parts; Class D (hazardous): , Rated lifespan Replace immediately; do not continue using.

[0063] Furthermore, in step S300, the early performance degradation warning includes steps S321-S322: S321: Using a high-precision capacitance meter to apply a high-frequency test signal, the capacitance value between relay contacts is measured, the trend of capacitance value change is analyzed, and it is determined whether it exceeds the normal fluctuation range. When the trend of change indicates performance degradation, an early warning signal is triggered, and the capacitance change data is input into the life prediction model to update the prediction results. The formula for calculating capacitance is as follows: in, The dielectric constant of the medium between the contacts is... This is the capacitance value. For effective contact area, This refers to the contact spacing.

[0064] S322: Combines capacitance change data with contact resistance and temperature rise parameters to construct a multi-dimensional feature vector. Analyzes capacitance change trends through a hybrid model to identify potential faults such as early oxidation and fretting wear. When the capacitance value shows a continuous monotonic change, even if the contact resistance has not deteriorated significantly, it triggers an early warning and updates the health status assessment.

[0065] 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.

[0066] Example 3, referring to Figure 5 This is the third embodiment of the present invention, which provides a multi-condition simulated capacitance detection relay life prediction system, including a multi-condition simulated drive module, a multi-dimensional data acquisition and feature extraction module, an intelligent life prediction module, and a reliability assessment and decision support module.

[0067] The multi-condition simulation drive module is used to simulate rated operating conditions and extreme operating conditions through a programmable drive power supply.

[0068] The multi-dimensional data acquisition and feature extraction module is used to collect multi-dimensional electrical and mechanical parameters such as driving voltage, coil current, contact voltage, contact current, action time, bounce time, contact open-circuit capacitance, and temperature rise in real time, and to extract key features characterizing performance degradation.

[0069] The intelligent lifespan prediction module is used to fuse a physical model based on the cumulative damage equation, a statistical degradation model based on Bayesian updates, and an intelligent prediction model based on machine learning. By weighted fusion of the physical damage prediction results and the time-series trend prediction results, it outputs the remaining lifespan prediction value and calculates the comprehensive health index.

[0070] The reliability assessment and decision support module is used to classify the relay status into four levels based on the comprehensive health index, and automatically associate the remaining lifespan range with maintenance recommendations. It calculates the equivalent failure rate by substituting the health index into the failure rate model.

[0071] 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.

[0072] Example 4, the fourth embodiment of the present invention, differs from the previous three embodiments in that: If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0073] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0074] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0075] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

Claims

1. A method for predicting the lifespan of a multi-condition simulated capacitance detection relay, characterized in that: The application relates to a relay performance state prediction method and device. The method comprises the following steps: extracting feature parameters from the collected multi-dimensional parameters to represent the relay performance state, and performing data preprocessing on the feature parameters, constructing a life prediction model based on the preprocessed feature parameters, and describing the mapping relationship between the parameters and the life; using the life prediction model to predict the remaining life of the relay and evaluate the reliability level, and combining the prediction result to perform early performance degradation warning by monitoring the capacitance change between the relay contacts.

2. The multi-condition analog capacitance detection relay life prediction method of claim 1, wherein: The data preprocessing on the feature parameters comprises the following steps: data cleaning is performed on the feature parameters to identify and process abnormal values and missing values in the data, data normalization is performed on the cleaned feature parameters to unify the data scale and eliminate the dimension difference, and data dimension reduction is performed on the normalized feature parameters to extract core features.

3. The multi-condition analog capacitance detection relay life prediction method of claim 2, wherein: The construction of the life prediction model comprises the following steps:

4. The multi-condition analog capacitance detection relay life prediction method of claim 3, wherein: key feature parameters are selected from the reduced core features, a prediction model is trained based on the key feature parameters, the correlation between the parameters and the life is learned and established through historical data, and the prediction model is optimized in parameters to adjust the model structure. The prediction calculation of the remaining life of the relay comprises the following steps:

5. The multi-condition analog capacitance detection relay life prediction method of claim 4, wherein: a cumulative damage index is calculated based on the prediction model to quantify the performance degradation degree of the relay in operation; a remaining life estimate value is derived from the cumulative damage index to evaluate the remaining service life, the health state is evaluated in combination with the parameters, and a comprehensive prediction result is output. The multiple simulation working conditions comprise an under-voltage working condition, an over-voltage working condition, a noise interference working condition and a high-speed pulse working condition; wherein the over-voltage working condition applies a driving signal higher than the rated voltage to test the life under over-excitation impact, and the high-speed pulse working condition outputs a high-frequency pulse sequence to test the resistance of the relay in the frequent switching state; 6. The multi-condition analog capacitance detection relay life prediction method of claim 5, wherein: The multi-dimensional parameters of the relay are collected in real time, including driving voltage, coil current, contact voltage, contact current, action time, bounce time, contact open-circuit capacitance, temperature rise, noise coefficient and over-voltage and under-voltage characteristics. The contact open-circuit capacitance is measured by a high-precision capacitance measuring instrument in the relay off state, and is determined by the dielectric constant, effective contact area and contact spacing based on the capacitance formula; the temperature rise is monitored by an infrared thermal imager; the action time and bounce time are recorded by a data acquisition card and the time parameters are extracted. The construction of the life prediction model comprises the following steps: combining a physical mechanism model, a statistical degradation model and an intelligent prediction model, wherein the physical mechanism model is based on a cumulative damage equation to calculate the cumulative damage degree; the statistical degradation model is based on the historical data distribution to update the posteriori degradation rate through Bayes; and the intelligent prediction model is based on a machine learning algorithm, including a gray prediction model, a BP neural network and an LSTM time series model, and processes multi-dimensional feature parameters; The cumulative damage equation is expressed as: wherein, is the accumulated damage degree at the current time, is the time, is the total number of actions up to time t, is the material erosion coefficient, is the arc energy of the i-th action, is the thermal degradation coefficient, is the activation energy, is the ideal gas constant, is the absolute temperature at time t, is the time index, is the weight coefficient of the contact resistance increment, is the contact resistance increment, is the contact resistance at the current time, is the initial contact resistance; The prediction calculation of the remaining life of the relay further comprises: performing time series trend-based prediction by using a grey prediction model, obtaining a final remaining life value by weightedly fusing prediction results, and optimizing prediction accuracy by using a deviation correction term; The final remaining life value formula is: wherein, is a remaining useful life prediction value based on a physical model, is a cumulative damage degree at a current time, is a posteriori degradation rate, is a load variation coefficient, is a failure threshold contact resistance, is a contact resistance value at a previous measurement time, is a remaining useful life prediction value based on a grey prediction model, and are a development coefficient and a grey action amount of the grey prediction model, is a prediction value of a first cumulative generation sequence obtained by the grey model at time t, is a sampling interval, is a final remaining life prediction value after fusion, is a weight coefficient, is a bias correction term; The reliability level evaluation comprises: performing single-parameter evaluation by calculating the degree of deviation of each monitoring parameter from a health threshold, fusing evaluation results by weighting into a comprehensive health index, correcting a benchmark failure rate by using the comprehensive health index to obtain an equivalent failure rate, dividing the device state into four levels according to the numerical value of the comprehensive health index, and automatically matching the remaining life prediction interval and the maintenance suggestion; The comprehensive health index calculation formula is: wherein, is a health index, is an index of the monitoring parameter, is a current actual measurement value of the jth monitoring parameter, is an attention threshold value for the jth parameter, is a limit threshold value for the jth parameter, is a weight coefficient for the jth parameter; The equivalent failure rate formula is: wherein, is the equivalent failure rate at time t, is the reference failure rate at time t, is the shape parameter, is the scale parameter, is the health impact factor, is time, is the exponential function.

7. The multi-condition analog capacitance detection relay life prediction method of claim 6, wherein: The early performance degradation early warning comprises: applying a high-frequency test signal by using a high-precision capacitance measuring instrument, measuring the capacitance value between relay contacts, analyzing the change trend of the capacitance value, determining whether it exceeds the normal fluctuation range, triggering a warning signal when the change trend indicates performance degradation, and inputting the capacitance change data into the life prediction model to update the prediction result; The capacitance value calculation formula is: wherein, is the dielectric constant of the inter-contact medium, is the capacitance value, is the effective contact area, is the contact pitch; The early performance degradation early warning further comprises: combining the capacitance change data with the contact resistance and temperature rise parameters to construct a multi-dimensional feature vector, analyzing the capacitance change trend by using a hybrid model, identifying potential faults of early oxidation and fretting wear, and triggering a warning and updating the health state evaluation when the capacitance value appears a sustained monotonic change even if the contact resistance is not significantly degraded.

8. A multi-condition simulation capacitor detection relay life prediction system, applying the multi-condition simulation capacitor detection relay life prediction method according to any one of claims 1 to 7, characterized in that, It comprises a multi-working-condition simulation driving module, a multi-dimensional data acquisition and feature extraction module, an intelligent life prediction module, and a reliability evaluation and decision support module. The multi-working-condition simulation driving module is configured to simulate the rated working condition and extreme working condition by using a programmable driving power supply. The multi-dimensional data acquisition and feature extraction module is configured to acquire multi-dimensional electrical parameters and mechanical parameters in real time, including driving voltage, coil current, contact voltage, contact current, action time, bounce time, contact open-circuit capacitance, and temperature rise, and extract key features representing performance degradation. The intelligent life prediction module is configured to fuse a physical model based on a cumulative damage equation, a statistical degradation model based on Bayesian update, and an intelligent prediction model based on machine learning, output a remaining life prediction value by weightedly fusing physical damage degree prediction results and time series trend prediction results, and calculate a comprehensive health index. The reliability evaluation and decision support module is configured to divide the relay state into four levels according to the comprehensive health index, automatically associate the remaining life interval with the maintenance suggestion, and calculate an equivalent failure rate by substituting the health index into a failure rate model. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor executes the computer program to implement the steps of the multi-working-condition simulation capacitance detection relay life prediction method of any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the multi-working-condition simulation capacitance detection relay life prediction method of any one of claims 1 to 7.

Citation Information

Cited By

  • Intelligent circuit breaker multi-parameter state monitoring and fault diagnosis system

    CN121656827A

  • Intelligent circuit breaker multi-parameter state monitoring and fault diagnosis system

    CN121656827B

  • Monitoring and life prediction method of motor for electric hoist based on multi-source data

    CN121933926A

  • Method for monitoring and life prediction of motor for electric hoist based on multi-source data

    CN121933926B

  • A relay life detection method, system, device, and storage medium

    CN122330675A