A Method and System for Anomaly Detection of Electrical Appliances Inside Cabinets Based on Multidimensional Environmental Parameters

By combining multi-gene genetic programming search and radio frequency identification technology with multi-dimensional environmental parameters, individualized life status detection of electrical tools has been realized, solving the problems of inaccurate detection and unstable abnormal thresholds in existing technologies, and improving the accuracy and completeness of detection.

CN123116064APending Publication Date: 2026-11-17STATE GRID HENAN ELECTRIC POWER COMPANY ZHENGZHOU POWER SUPPLY CO
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611271140.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-20
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies make it difficult to perform individualized lifespan status detection based on the actual usage history of electrical equipment, and fixed degradation models are difficult to reflect the interaction of multiple environmental parameters and the instability of abnormal thresholds under sparse test data.

Method used

A multi-gene genetic programming approach is used to search for environmental degradation rate functions. Combined with the interaction relationships of multi-dimensional environmental parameters, the status of the equipment is tracked by radio frequency identification. The environmental vector time series is collected and integrated to obtain the cumulative degradation amount. The measured values ​​of periodic insulation performance are used to fit the individual performance degradation mapping to generate a dynamic lower control limit for anomaly detection.

Benefits of technology

It enables stable anomaly detection of power tools under sparse test data conditions, improves the accuracy and completeness of detection, reduces the impact of historical test data of individual power tools, and can identify nonlinear interaction relationships and output missed collection reminders.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116064A_ABST
    Figure CN123116064A_ABST
Patent Text Reader

Abstract

This invention belongs to the field of electrical equipment cabinet management and condition monitoring, specifically involving a method and system for detecting anomalies in electrical equipment within cabinets based on multi-dimensional environmental parameters. It aims to solve the problems of fixed degradation models failing to reflect the interactions of multiple environmental parameters and the instability of anomaly thresholds under sparse test data. The invention includes: forming discrete degradation rate samples based on multi-stress accelerated degradation test data; generating an environmental degradation rate function using physically constrained multi-gene genetic programming; dividing the cabinet storage stage and the off-cabinet use stage based on RFID results and calculating the cumulative degradation amount; fitting individual performance degradation maps and calculating residuals; generating dynamic lower control limits through self-sampling of historical residuals from the same batch and test stage, and outputting anomaly judgment results. This invention can perform individualized anomaly detection by combining actual environmental history and periodic insulation performance data, and output missed item reminders based on work order requirements and actual off-cabinet results.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of electrical tool cabinet management and tool status detection technology, specifically to a method and system for detecting abnormalities in electrical tools inside the cabinet based on multi-dimensional environmental parameters. Background Technology

[0002] Electrical tools used in power system operations, such as insulated operating rods, voltage detectors, insulated boots, and insulated gloves, are typically stored in tool cabinets within the work area. Each cabinet is equipped with an RFID tag to record its status within the cabinet, as well as the time it is removed and placed back in. During storage, transportation, and operation, these tools are exposed to factors such as temperature, humidity, ultraviolet radiation, salt spray, and mechanical shock, causing their insulation performance to change with usage time and accumulated environmental stress. To ensure safe operation, insulation performance tests are usually conducted on these tools at prescribed intervals.

[0003] While existing tool cabinets can use RFID tags to record whether electrical tools are inside the cabinet, the corresponding identification results are usually mainly used for registration of issuance and return. They have not yet been correlated with the environmental history, mechanical shock conditions, and periodic insulation performance test results of the electrical tools during the storage and use phases inside and outside the cabinet. Therefore, it is difficult to conduct individualized life status detection based on the actual usage history of the tools.

[0004] Current testing methods primarily rely on periodic insulation performance tests to determine whether electrical appliances meet usage requirements. However, it's difficult to promptly grasp changes in the insulation performance of electrical appliances between adjacent tests. Some condition-based testing methods collect environmental parameters such as temperature and humidity and assess insulation status using preset thresholds, fixed degradation models, or linear weighting of multiple environmental parameters. However, different environmental parameters may exhibit multiplicative, power-law, or other nonlinear interactions, making it difficult for preset models or simple weighting methods to accurately reflect actual degradation patterns.

[0005] Furthermore, similar electrical appliances vary in material batches, initial insulation performance, structural dimensions, and usage history, making it difficult to form individualized judgments based solely on a uniform degradation model. The amount of data from periodic insulation performance tests is usually limited, and directly determining the anomaly threshold based on the historical residuals of individual electrical appliances may lead to unstable control limits due to insufficient residual samples.

[0006] Therefore, there is a need for an anomaly detection method for electrical tools inside the cabinet that can combine multi-dimensional environmental history and periodic insulation performance measurement data in the application scenario of tool cabinets in work areas, identify the nonlinear interaction relationship between environmental parameters, and form a stable anomaly judgment benchmark when there is limited experimental data. Summary of the Invention

[0007] To address the aforementioned problems in the prior art, namely, the difficulty of fixed degradation models in reflecting the interaction of multiple environmental parameters and the instability of abnormal thresholds under sparse experimental data, this invention provides a method and system for detecting abnormalities in cabinet electrical appliances based on multidimensional environmental parameters.

[0008] The first aspect of this invention provides a method for detecting abnormalities in electrical appliances inside cabinets based on multi-dimensional environmental parameters, comprising: Discrete degradation rate samples were formed based on multi-stress accelerated degradation test data of similar insulating materials; Multi-gene genetic programming is used to search for candidate functions that are linear combinations of multiple expression subtrees. Based on the monotonicity constraints of environmental parameters and the basic degradation rate constraints under the benchmark environment, environmental degradation rate functions containing environmental parameter interaction terms are obtained. The environmental vector time series of the electrical equipment to be tested in the cabinet is collected, and the environmental parameters that exceed the accelerated test range are truncated at the boundary. The truncated environmental parameters are input into the environmental degradation rate function and integrated to obtain the cumulative degradation amount. By fitting the individual performance degradation mapping with the measured values ​​of periodic insulation performance and the corresponding cumulative degradation, the residual between the predicted and measured values ​​of insulation performance is obtained. Self-sampling was performed on the historical residuals of the power equipment in the same batch and at the same test stage as the power equipment to be tested to obtain the compensated standard deviation. The dynamic lower control limit was generated by the larger of the historical residual mean and the historical residual standard deviation and the compensated standard deviation. Compare the residual with the dynamic lower control limit. If the residual is not lower than the dynamic lower control limit, update the dynamic baseline of the residual. If it is lower, output the anomaly judgment result.

[0009] Furthermore, the formation of the discrete degradation rate samples includes: In the multi-stress accelerated degradation test, temperature, humidity, ultraviolet irradiance intensity and equivalent salt spray deposition rate were collected for each test period. Logarithmic difference and duration normalization were performed on the measured insulation performance values ​​at the beginning and end of the same test period to obtain the degradation rate of the time period; The degradation rate over the specified time period is bound to the average environmental parameter within the same test period to form a discrete degradation rate sample that includes an environmental parameter field and a degradation rate field.

[0010] Furthermore, the screening of the environmental degradation rate function includes: Each candidate individual is encoded into multiple expression subtrees. Linear regression is performed on the output of each expression subtree to solve for the linear combination coefficients and intercept terms of each expression subtree, thereby generating candidate degradation rate prediction values. The prediction error is constructed based on the candidate degradation rate prediction value and the discrete degradation rate sample, and the fitness evaluation quantity is generated by combining the expression complexity, the sign of the partial derivative of the environmental parameter and the deviation of the basic degradation rate under the benchmark environment. A penalty is imposed on candidate individuals that violate the monotonicity constraint of environmental parameters or the basic degradation rate constraint under the benchmark environment, and candidate individuals whose fitness evaluation value meets the preset retention conditions are retained, thus obtaining the environmental degradation rate function.

[0011] Furthermore, the acquisition of the environmental vector time series includes: To bind radio frequency identification (RFID) tags to electrical tools to be tested stored in the tool cabinet in the work area, the radio frequency identification unit set in the tool cabinet in the work area is used to identify the out-of-cabinet and in-cabinet status of the electrical tools to be tested; The temperature, humidity, ultraviolet radiation intensity, salt spray deposition, and vibration and shock sensor outputs of the electrical equipment under test are collected during the storage stage inside the cabinet and the off-cabinet use stage from the time it is taken out of the cabinet to the time it is put back into the cabinet. The salt spray deposition sensor output is converted into an equivalent salt spray deposition rate according to laboratory calibration. According to the RFID tags and timestamps, the temperature, humidity, ultraviolet irradiation intensity, equivalent salt spray deposition rate, root mean square of vibration, peak acceleration, and impact duration are archived, and the storage area inside the cabinet and the usage area outside the cabinet are divided according to the time of exiting and entering the cabinet, forming the environmental vector time series.

[0012] Furthermore, the calculation of the cumulative degradation includes: Read the environmental parameters at adjacent sampling times in the environmental vector time series, and truncate the environmental parameters that are outside the coverage of the multi-stress accelerated degradation test. The environmental parameters truncated at the boundary between adjacent sampling times are respectively input into the environmental degradation rate function, and trapezoidal integration is performed according to the adjacent sampling interval to obtain the cumulative environmental degradation. Mechanical impact events that meet the impact criteria are identified from the vibration and impact sensor output. The number of mechanical impact events is multiplied by the single impact degradation increment of the same type of insulating material to obtain the cumulative impact degradation. The cumulative environmental degradation and the cumulative impact degradation are combined into the cumulative degradation.

[0013] Furthermore, the fitting of the individual performance degradation map includes: After each periodic insulation performance test is completed, the cumulative degradation amount within the preset time window before the test time is read and the test degradation input is generated. The measured insulation performance value of the same test is read and the test performance output is generated. Query the material batch prior parameter table. When there is a batch-level prior or a material-level prior, input the test degradation input and the test performance output into Bayesian exponential regression. When there is no available prior, input the test degradation input and the test performance output into least squares regression. Solve for the initial insulation performance parameters and degradation ratio parameters to generate the individual performance degradation mapping of the electrical equipment to be tested.

[0014] Furthermore, the updating of the residual dynamic benchmark includes: After a new periodic insulation performance test is completed, the current cumulative degradation amount is input into the individual performance degradation mapping to obtain the insulation performance prediction value, and the residual between the measured insulation performance value and the insulation performance prediction value is calculated. The dynamic lower control limit is generated based on the larger of the historical residual mean, the historical residual standard deviation, and the compensated standard deviation. When the residual is not lower than the dynamic lower control limit, the mean of the historical residual and the standard deviation of the historical residual are updated using the forgetting factor; When the residual is lower than the dynamic lower control limit, the mean of the historical residual and the standard deviation of the historical residual remain unchanged.

[0015] Furthermore, the generation of the compensation standard deviation includes: Historical residuals that are in the same test cycle stage as the electrical equipment to be tested are selected from the same batch of electrical equipment to form a set of historical residuals; The historical residual set is subjected to auto-sampling to form a synthetic residual sample set; Quantile statistics are performed on the synthetic residual sample set, and the lower quantile is extracted as the compensation warning lower limit; The compensation standard deviation is calculated based on the difference between the historical residual mean and the compensation warning lower limit, as well as the control limit coefficient. When generating the dynamic lower control limit, the larger value between the historical residual standard deviation and the compensated standard deviation is selected to participate in the calculation of the dynamic lower control limit, and residuals that are judged to be abnormal are not used to update the historical residual set.

[0016] Furthermore, the anomaly determination includes: When the residual is lower than the dynamic lower control limit, output the residual abnormality judgment result; The critical degradation amount is calculated by back-calculating the lower limit of safe insulation performance and the individual performance degradation mapping. When the current cumulative degradation exceeds the critical degradation, an abnormal degradation threshold determination result is output. When the residual is not lower than the dynamic lower control limit and the current cumulative degradation amount does not exceed the critical degradation amount, a normal judgment result is output.

[0017] The second aspect of this invention provides an anomaly detection system for electrical appliances inside cabinets based on multidimensional environmental parameters, including a discrete sample generation module, a degradation function generation module, an environmental degradation processing module, a performance mapping module, a control limit generation module, and an anomaly determination module; The discrete sample generation module is used to generate discrete degradation rate samples based on multi-stress accelerated degradation test data of similar insulating materials; The degradation function generation module is used to search for candidate functions that are linear combinations of multiple expression subtrees using multi-gene genetic programming, and to filter them based on the monotonicity constraints of environmental parameters and the basic degradation rate constraints under the benchmark environment, so as to obtain an environmental degradation rate function containing environmental parameter interaction terms. The environmental degradation processing module is used to identify the out-of-cabinet and in-cabinet states of the electrical tools to be tested through the radio frequency identification unit of the tool cabinet in the work area, collect the environmental vector time series of the electrical tools to be tested during the storage stage in the cabinet and the off-cabinet use stage from out of the cabinet to in the cabinet, truncate the environmental parameters that exceed the accelerated test range, input the truncated environmental parameters into the environmental degradation rate function and integrate to obtain the cumulative degradation amount; The performance mapping module is used to fit an individual performance degradation mapping using the measured values ​​of periodic insulation performance and the corresponding cumulative degradation amount, so as to obtain the residual between the predicted insulation performance value and the measured value. The control limit generation module is used to select historical residuals of power equipment from the same batch and in the same test cycle as the power equipment to be tested for self-sampling, obtain the compensated standard deviation, and generate a dynamic lower control limit by using the larger value between the historical residual mean and the historical residual standard deviation and the compensated standard deviation. The anomaly detection module is used to compare the residual with the dynamic lower control limit. When the residual is not lower than the dynamic lower control limit, the dynamic baseline of the residual is updated; when it is lower, the anomaly detection result is output.

[0018] The beneficial effects of this invention are: This invention utilizes multi-stress accelerated degradation test data to search for environmental degradation rate functions through multi-gene genetic programming. During the search process, it introduces environmental parameter monotonicity constraints and baseline degradation rate constraints under the benchmark environment. This enables the obtained environmental degradation rate function to represent the nonlinear interaction between temperature, humidity, ultraviolet irradiation intensity, and equivalent salt spray deposition rate, while reducing the possibility that candidate functions violate actual degradation laws.

[0019] This invention utilizes the identification results of RFID tags in the work area tool cabinet to determine the out-of-cabinet and in-cabinet states of electrical tools. It further divides the storage stage and the usage stage outside the cabinet based on the out-of-cabinet and in-cabinet times, enabling the data acquired by environmental acquisition terminals at different locations to be continuously archived according to the specific electrical tools and their actual usage process. The invention calculates the cumulative environmental degradation based on the formed environmental vector time series and combines it with the cumulative degradation corresponding to mechanical impact events to form a cumulative degradation value. This allows anomaly detection to utilize the actual environmental history of the electrical tools, rather than being limited to single measurement results at periodic test times.

[0020] When linked with the work order management system, this invention can also obtain the set of required tool identifiers corresponding to the current work order, and identify the situation of missing items based on the actual set of tool identifiers generated during the cabinet door opening, and output corresponding reminders to the recipients.

[0021] This invention utilizes the measured values ​​of periodic insulation performance and the corresponding cumulative degradation to fit an individual performance degradation mapping. It characterizes the performance differences between different power tools through initial insulation performance parameters and degradation ratio parameters, thereby forming a predicted insulation performance value corresponding to the state of the power tool under test.

[0022] This invention performs self-sampling on historical residuals from the same batch and at the same test number stage, generates a compensated standard deviation based on the lower quantile statistics, and calculates the dynamic lower control limit using the larger of the historical residual standard deviation and the compensated standard deviation. This can reduce the impact of limited historical test data on the stability of the anomaly judgment benchmark for individual power tools.

[0023] This invention uses only residuals that are not below the dynamic lower control limit to update the dynamic baseline of the residuals, thus avoiding abnormal residuals from entering the normal historical residual set and causing the dynamic lower control limit to shift. It also combines residual abnormality judgment and cumulative degradation threshold judgment to improve the completeness of abnormal state identification of power tools. Attached Figure Description

[0024] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart of the cabinet-based electrical appliance anomaly detection method based on multi-dimensional environmental parameters provided in the first embodiment of the present invention; Figure 2 This is a flowchart of the environmental degradation rate function generation step in the first embodiment of the present invention; Figure 3 This is a flowchart of the steps for determining the radio frequency identification status, forming the environmental vector time series, and calculating the cumulative degradation in the first embodiment of the present invention; Figure 4This is a flowchart of the steps for generating the compensation standard deviation and dynamic lower control limit in the first embodiment of the present invention; Figure 5 This is a block diagram of the architecture of the cabinet-based electrical appliance anomaly detection system based on multi-dimensional environmental parameters provided in the second embodiment of the present invention. Detailed Implementation

[0025] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0026] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0027] The first embodiment of this invention proposes a method for detecting abnormalities in electrical appliances inside cabinets based on multi-dimensional environmental parameters, comprising the following steps: S10: Based on multi-stress accelerated degradation test data of similar insulating materials, a discrete degradation rate sample is formed; S20: Multi-gene genetic programming is used to search for candidate functions that are linear combinations of multiple expression subtrees. Based on the monotonicity constraints of environmental parameters and the basic degradation rate constraints under the benchmark environment, environmental degradation rate functions containing environmental parameter interaction terms are obtained. S30: Collect the environmental vector time series of the electrical equipment to be tested in the cabinet, truncate the environmental parameters that exceed the accelerated test range, input the truncated environmental parameters into the environmental degradation rate function and integrate to obtain the cumulative degradation amount; S40: By fitting the individual performance degradation mapping with the measured values ​​of periodic insulation performance and the corresponding cumulative degradation, the residual between the predicted and measured values ​​of insulation performance is obtained; S50: Select historical residuals of electrical equipment from the same batch and in the same test cycle as the electrical equipment to be tested for self-sampling, obtain the compensated standard deviation, and generate the dynamic lower control limit by the larger of the historical residual mean and the historical residual standard deviation and the compensated standard deviation. S60: Compare the residual with the dynamic lower control limit. If the residual is not lower than the dynamic lower control limit, update the dynamic benchmark of the residual. If it is lower, output the anomaly judgment result.

[0028] To more clearly illustrate the technical solution of the present invention, the following is combined with... Figure 1 Each step is explained in detail.

[0029] S10: Discrete degradation rate samples are formed based on multi-stress accelerated degradation test data of similar insulating materials.

[0030] In this embodiment, step S10 includes: Step S11: In the multi-stress accelerated degradation test, collect the temperature, humidity, ultraviolet irradiation intensity and equivalent salt spray deposition rate for each test period. Step S12: Logarithmic difference and duration normalization are performed on the measured insulation performance values ​​at the beginning and end of the same test period to obtain the degradation rate of the time period; Step S13: Bind the degradation rate of the time period to the average environmental parameter within the same test period to form a discrete degradation rate sample containing an environmental parameter field and a degradation rate field.

[0031] The standard test specimens are made of the same type of insulating material as the electrical equipment to be tested. The insulating material includes natural rubber, silicone rubber, epoxy resin, or glass fiber reinforced insulating material. Multiple sets of standard test specimens are prepared for the same type and batch of insulating material, with multiple parallel specimens in each set.

[0032] Standard samples were placed in a multi-factor aging test chamber and subjected to temperature, humidity, ultraviolet radiation, and salt spray deposition according to different pre-set stress combinations. Each stress combination covered the environmental range that the electrical equipment under test might experience during storage, transportation, and operation.

[0033] In this embodiment, the temperature test range is from room temperature to 85°C, the relative humidity test range is from 30% to 98%, the ultraviolet irradiation intensity test range is from 0 to 1200 W / m², and the equivalent salt spray deposition rate test range is from 0 to 2.5 μg / (cm²). 2 •h). The test range is also used to determine the upper and lower limits of subsequent environmental parameter boundary cutoff.

[0034] During each test period, temperature, humidity, ultraviolet radiation intensity, and equivalent salt spray deposition rate were continuously or periodically collected, and the average values ​​of each environmental parameter during that test period were calculated.

[0035] The insulation performance of the standard sample was measured at the beginning and end of the test period. The measured insulation performance values ​​can be surface insulation resistance, volume insulation resistance, leakage current, or withstand voltage test results. In this embodiment, surface insulation resistance is used as the measured insulation performance value.

[0036] Before measuring the surface insulation resistance, the standard sample is placed in the same recovery environment to allow the temperature and surface humidity of the standard sample to stabilize, thereby reducing the impact of differences in the measurement environment on the test results.

[0037] Let the insulation resistance at the beginning of a certain test period be... The insulation resistance at the end of the test period is The test period length is Then the time-period degradation rate of this test period Represented as: ; in, Represents the natural logarithm. and These are the measured insulation performance values ​​of the same standard sample at the beginning and end of the same test period. This refers to the duration of the trial period.

[0038] When the insulation resistance at the end of the test period is lower than the insulation resistance at the beginning of the test period, the degradation rate of the test period is... The result is positive. Through logarithmic difference and duration normalization, standard samples with different initial insulation properties and different test durations can participate in subsequent function searches under a unified degradation rate scale.

[0039] The degradation rate over a given time period was linked to the average temperature, average humidity, average UV irradiance, and average equivalent salt spray deposition rate within the same experimental time period to form a discrete degradation rate sample: ; in, Indicates the first The average temperature during each test period Indicates the first Average relative humidity during each test period Indicates the first The average ultraviolet radiation intensity during each test period Indicates the first The average equivalent salt spray deposition rate over the test period Indicates the first The degradation rate corresponding to each test period.

[0040] Salt spray deposition rates were obtained using the standard copper sheet weight loss method or a calibrated atmospheric corrosion monitoring unit.

[0041] When using an atmospheric corrosion monitoring unit, multiple salt spray deposition conditions are set up in the laboratory. The sensor output values ​​of the atmospheric corrosion monitoring unit and the salt spray deposition rate determined by the standard copper sheet weight loss method are acquired simultaneously, and a calibration function is established based on the correspondence between the two. ; in, This indicates the output of the salt spray deposition sensor. This represents the transformation function obtained by fitting laboratory calibration data. This represents the equivalent salt spray deposition rate.

[0042] The salt spray deposition sensor outputs subsequently collected on-site were converted according to the same calibration function, so that the equivalent salt spray deposition rate in the on-site environmental vector and the equivalent salt spray deposition rate in the multi-stress accelerated degradation test adopted the same data definition.

[0043] S20: A multi-gene genetic programming search is used to search for candidate functions that are linear combinations of multiple expression subtrees. Based on the monotonicity constraints of environmental parameters and the basic degradation rate constraints under the benchmark environment, an environmental degradation rate function containing environmental parameter interaction terms is obtained.

[0044] See Figure 2 In this embodiment, step S20 includes: Step S21: Encode each candidate individual into multiple expression subtrees, perform linear regression on the output of each expression subtree, solve for the linear combination coefficients and intercept terms of each expression subtree, and generate candidate degradation rate prediction values. Step S22: Construct a prediction error based on the candidate degradation rate prediction value and the discrete degradation rate sample, and generate a fitness evaluation quantity by combining the expression complexity, the partial derivative sign of the environmental parameter, and the deviation of the basic degradation rate under the benchmark environment. Step S23: Penalize candidate individuals that violate the monotonicity constraint of environmental parameters or the basic degradation rate constraint under the benchmark environment, retain candidate individuals whose fitness evaluation value meets the preset retention conditions, and obtain the environmental degradation rate function.

[0045] In step S21, each candidate individual in multi-gene genetic programming includes multiple genes, and each gene corresponds to an expression subtree.

[0046] The terminal input of the expression subtree includes temperature. relative humidity Ultraviolet radiation intensity Equivalent salt spray deposition rate And constant terms. The function operators of the expression subtree include addition, subtraction, multiplication, protected division, power functions, exponential functions, protected logarithms, and protected square roots.

[0047] Protected division is used to return a finite output value when the denominator is zero or close to zero; protected logarithm is used to restrict the input when the input does not satisfy the domain of the logarithm; protected square root is used to prevent negative input from causing the operation to be interrupted.

[0048] Multiple expression subtrees can each form a single environmental parameter action term and an interaction term between different environmental parameters. Environmental parameter interaction terms include combinations of temperature and humidity, temperature and ultraviolet radiation intensity, humidity and salt spray deposition rate, and ultraviolet radiation intensity and salt spray deposition rate.

[0049] For a candidate individual, the outputs of each expression subtree in the candidate individual to all discrete degradation rate samples are used to form a regression input matrix, and the time-period degradation rates in each discrete degradation rate sample are used to form a regression output vector. Then, linear regression is used to solve for the linear combination coefficients and intercept terms of each expression subtree.

[0050] Let the first The output of each expression subtree is The linear combination coefficients corresponding to the expression subtree are The intercept term is Then the predicted candidate degradation rate Represented as: ; in, This represents summing all expression subtrees within a candidate individual.

[0051] Expression subtrees are used to determine the nonlinear combination of environmental parameters, and linear combination coefficients are used to determine the contribution of each expression subtree to the candidate degradation rate prediction.

[0052] In step S22, the prediction error is calculated based on the predicted candidate degradation rate and the time-period degradation rate in the discrete degradation rate sample. The prediction error is expressed as mean square error: ; in, The number of samples representing the discrete degradation rate. Indicates the first The time-period degradation rate in a discrete degradation rate sample. Indicates the candidate individual's relationship to the first... The candidate degradation rate prediction values ​​are output from the discrete degradation rate samples.

[0053] The complexity of an expression is determined by the number of expression subtrees contained in a candidate individual, the total number of expression nodes, the maximum tree depth, and the number of nested operation levels. The more expression nodes a candidate individual has and the greater the tree depth, the higher the expression complexity.

[0054] Multiple evaluation points were selected within the coverage area of ​​the multi-stress accelerated degradation test, and the partial derivative signs of the candidate function with respect to temperature, humidity, ultraviolet irradiation intensity, and equivalent salt spray deposition rate were calculated respectively.

[0055] When the partial derivative of a candidate function with respect to any environmental parameter at any evaluation point is less than zero, it indicates that the candidate function is experiencing an increase in environmental stress and a decrease in the predicted value of the candidate degradation rate at the evaluation point. A monotonic penalty is then generated for the candidate function.

[0056] For candidate functions that involve protective operations or are inconvenient to directly analyze and differentiate, the finite difference method is used to determine the changing trend of the candidate function relative to the target environment parameters.

[0057] Taking temperature as an example, while keeping humidity, ultraviolet irradiation intensity, and equivalent salt spray deposition rate constant, a preset perturbation amount is added to the temperature, and the predicted values ​​of candidate degradation rates before and after the temperature perturbation are calculated respectively. When the predicted value of the candidate degradation rate decreases after the temperature increases, it is determined that the evaluation point violates the monotonicity constraint of environmental parameters.

[0058] The baseline environment includes baseline temperature, baseline humidity, baseline ultraviolet irradiance, and baseline equivalent salt spray deposition rate. In this embodiment, the baseline temperature is 25°C, the baseline relative humidity is 50%, the baseline ultraviolet irradiance is 0, and the baseline equivalent salt spray deposition rate is 0.

[0059] Standard samples from the same batch were subjected to room temperature storage tests, and the basic degradation rate was calculated based on the changes in insulation performance during room temperature storage. The output value of the candidate function under the baseline environment is compared with the basic degradation rate. By comparison, the deviation of the basic degradation rate under the reference environment is obtained.

[0060] Fitness Evaluation Quantity Represented as: ; in, Indicates the prediction error. Indicates the complexity of the expression. This represents the monotonicity penalty for environmental parameters. This represents the deviation between the output value of the candidate function in the baseline environment and the base degradation rate. , and These represent the weights of the corresponding evaluation items.

[0061] The weights of each evaluation item are determined based on the prediction error on the validation samples, the stability of the candidate function, and the changing trends of environmental parameters. The smaller the fitness evaluation value, the better the overall evaluation result of the candidate individual.

[0062] In step S23, candidate individuals are arranged in ascending order of fitness evaluation scores, and candidate individuals ranked within a preset retention ratio are retained. The retained candidate individuals are then allowed to participate in the next round of genetic selection, crossover, and mutation.

[0063] The preset retention conditions include that the fitness evaluation score of the candidate individual is lower than the set evaluation threshold, or that the candidate individual is within the set retention ratio after being sorted according to the fitness evaluation score.

[0064] The search stops when the preset number of iterations is reached, or when the change in the optimal fitness evaluation value obtained from multiple consecutive rounds of search is less than the set convergence threshold. The expression subtree, linear combination coefficients, and intercept term of the final retained candidate individuals are written into the environment degradation rate function.

[0065] The environmental degradation rate function is expressed as: ; in, The function representing the rate of environmental degradation. Represents the intercept term. Indicates the first The linear combination coefficients of the expression subtrees, Indicates the first An expression subtree.

[0066] One expression subtree can represent the power-law or logarithmic interaction term between temperature and humidity, while the other expression subtree can represent the product interaction term between ultraviolet irradiation intensity and equivalent salt spray deposition rate. Specific expressions and parameters are determined from discrete degradation rate samples and multigene genetic programming search results.

[0067] In this embodiment, step S23 further includes: Step S231: After outputting the environment parameter interaction term in the candidate expression subtree, generate penalty amounts for the partial derivative sign, baseline environment output value, and expression complexity of the environment parameter interaction term at multiple sampling points in the sample space. Step S232: Combine the penalty amount and the prediction error into a fitness evaluation amount, remove candidate individuals that show a negative degradation trend in environmental parameters or whose baseline environmental output deviates from the basic degradation rate, and write the expression subtree, linear combination coefficients and intercept term of the retained candidate individuals into the environmental degradation rate function.

[0068] In step S231, multiple environmental parameter combination points are selected from the coverage area of ​​the multi-stress accelerated degradation test, and the partial derivative sign of the candidate function with respect to each environmental parameter is calculated for each environmental parameter combination point.

[0069] When the partial derivative of a candidate function with respect to a certain environmental parameter is less than zero, a monotonicity penalty is generated based on the absolute value of the partial derivative. The larger the negative change, the larger the corresponding monotonicity penalty.

[0070] The penalty for the baseline environment output value is determined based on the absolute or relative deviation between the candidate function's output value in the baseline environment and the base degradation rate.

[0071] The expression complexity penalty is determined based on the number of nodes in the expression subtree, the tree depth, the number of genes, and the number of protection operation calls.

[0072] In step S232, the prediction error, monotonicity penalty, baseline environment output value penalty, and expression complexity penalty are combined into a fitness evaluation quantity.

[0073] Candidate individuals exhibiting a negative degradation trend in environmental parameters and whose monotonicity penalty exceeds the set upper limit are eliminated; candidate individuals whose deviation between the baseline environmental output value and the basic degradation rate exceeds the set range are also eliminated.

[0074] The remaining candidate individuals are sorted according to their fitness evaluation, and the expression subtrees, linear combination coefficients, and intercept terms corresponding to the candidate individuals that meet the preset retention conditions are written into the environmental degradation rate function.

[0075] S30: Collect the environmental vector time series of the electrical equipment to be tested in the cabinet, truncate the environmental parameters that exceed the accelerated test range, input the truncated environmental parameters into the environmental degradation rate function and integrate to obtain the cumulative degradation amount.

[0076] See Figure 3 In this embodiment, step S30 includes: Step S31: Bind radio frequency identification (RFID) tags to the electrical tools to be tested stored in the tool cabinet in the work area. Identify the out-of-cabinet and in-cabinet status of the electrical tools to be tested by the RFID unit set in the tool cabinet in the work area, and collect temperature, humidity, ultraviolet radiation intensity sensor output, salt spray deposition sensor output and vibration and shock sensor output during the storage stage in the cabinet and the off-cabinet use stage between the out-of-cabinet and in-cabinet. Step S32: Convert the salt spray deposition sensor output into an equivalent salt spray deposition rate according to the laboratory calibration relationship; Step S33: Archive temperature, humidity, ultraviolet irradiation intensity, equivalent salt spray deposition rate, root mean square of vibration, peak acceleration and impact duration according to the RFID tag and timestamp, and divide the storage area inside the cabinet and the usage area outside the cabinet according to the time of exiting the cabinet and the time of entering the cabinet to form the environmental vector time series. Step S34: Read the environmental parameters at adjacent sampling times in the environmental vector time series, and truncate the environmental parameters that exceed the coverage of the multi-stress accelerated degradation test. Step S35: Input the environmental parameters truncated at the boundary between adjacent sampling times into the environmental degradation rate function, and perform trapezoidal integration according to the adjacent sampling interval to obtain the cumulative environmental degradation. Step S36: Identify mechanical impact events that meet the impact criteria from the vibration and impact sensor output, and multiply the number of mechanical impact events by the single impact degradation increment of the same type of insulating material to obtain the cumulative impact degradation amount. Step S37: Combine the cumulative environmental degradation and the cumulative impact degradation into the cumulative degradation amount.

[0077] In step S31, the electrical tools to be tested include insulated operating rods, voltage detectors, insulated boots, insulated gloves, insulated ladders, insulated barriers, or other electrical tools that require insulation performance management.

[0078] Radio frequency identification (RFID) uses radio frequency tags set on the electrical tools to be tested. The RFID tag code is bound to the tool type, tool number, insulation material type, material batch, initial insulation performance, commissioning time, and periodic insulation performance test records of the electrical tools to be tested.

[0079] The tool cabinet in the work area is equipped with an RFID unit and a cabinet door status detection unit. During the cabinet door opening period, when the RFID tag changes from an identifiable state inside the cabinet to an out-of-cabinet state, the time of exiting the corresponding electrical tool to be tested is recorded; when the RFID tag re-enters the identification range inside the cabinet, the time of entering the corresponding electrical tool to be tested is recorded, and the time period between the exit time and the corresponding entry time is defined as one out-of-cabinet usage phase.

[0080] Environmental data acquisition terminals are installed in tool cabinets, transport vehicles, work sites, or temporary storage areas within the work area. These terminals include temperature and humidity sensors, ultraviolet radiation sensors, salt spray deposition sensors, and triaxial accelerometers.

[0081] Temperature and humidity sensors collect temperature and relative humidity, ultraviolet radiation sensors collect ultraviolet radiation intensity, salt spray deposition sensors output sensing signals related to salt spray deposition rate, and triaxial accelerometers collect vibration and shock sensor outputs.

[0082] Temperature, humidity, ultraviolet radiation intensity, and salt spray deposition sensor outputs are collected according to the set environmental acquisition cycle. The triaxial accelerometer collects raw acceleration data at a sampling rate higher than the environmental acquisition frequency, and calculates the root mean square of vibration, peak value of triaxial absolute acceleration, and continuous duration exceeding the impact threshold in each environmental acquisition cycle.

[0083] In step S32, the salt spray deposition sensor output is input into the laboratory calibration function established in step S10 to obtain the equivalent salt spray deposition rate.

[0084] When the output of the salt spray deposition sensor exceeds the range of laboratory calibration data, an out-of-range flag is set for the corresponding data, and the boundary is truncated in subsequent steps according to the range of multi-stress accelerated degradation test.

[0085] In step S33, the data platform determines the electrical equipment to be tested corresponding to the environmental data based on the RFID tag, arranges the collected data according to the timestamp, and marks the storage segment or usage segment outside the cabinet for each collected data according to the time of leaving the cabinet and the time of entering the cabinet, forming an environmental vector time series: ; in, Indicates the first Each sampling time, Indicates temperature. Indicates relative humidity. Indicates ultraviolet radiation intensity. Indicates the equivalent salt spray deposition rate. This represents the root mean square of the vibration within the corresponding acquisition period. Indicates the peak absolute acceleration of the three axes. This indicates the duration of continuous impact exceeding the impact threshold.

[0086] When the electrical tools to be tested are transferred from the tool cabinet in the work area to the transport vehicle or the work site, the data platform continuously classifies the data collected by the environmental acquisition terminals at different locations into the same environmental vector time series based on the radio frequency identification (RFID) identifier; when the electrical tools to be tested are put back into the cabinet, the subsequent data collected will be marked as a new storage segment in the cabinet.

[0087] In a specific application scenario, the data platform also acquires the set of required tool identifiers corresponding to the current work order. Upon detecting that the tool cabinet is closed or receiving a completion signal, the data platform determines the actual set of tool identifiers retrieved based on the exit records generated during the cabinet opening period, and compares this actual set with the required tool identifier set. If any tool identifier in the required tool identifier set is not included in the actual retrieved tool identifier set, a missed item reminder is output to the current user. This missed item reminder is not included in the calculation of the environmental degradation rate function, cumulative degradation amount, individual performance degradation mapping, and dynamic lower control limit.

[0088] In step S34, adjacent sampling times are read. and The corresponding environmental parameters, including temperature, humidity, ultraviolet radiation intensity, and equivalent salt spray deposition rate, are subject to boundary truncation in the calculation of the environmental degradation rate function.

[0089] Temperature boundary truncation is represented as: ; Humidity boundary truncation is represented as: ; The ultraviolet radiation intensity boundary cutoff is expressed as: ; The equivalent salt spray deposition rate boundary truncation is expressed as: ; in, and These represent the lower and upper limits of the temperature test range, respectively. and These represent the lower and upper limits of the humidity test range, respectively. and These represent the lower and upper limits of the ultraviolet irradiation intensity test range, respectively. and These represent the lower and upper limits of the equivalent salt spray deposition rate test range, respectively.

[0090] Temperature, humidity, ultraviolet radiation intensity, and equivalent salt spray deposition rate, after boundary truncation, constitute the truncation environment vector. .

[0091] In step S35, and Input the environmental degradation rate function obtained in step S20 to obtain the degradation rate at adjacent sampling times. and .

[0092] The cumulative environmental degradation is calculated using trapezoidal integrals. : ; in, This indicates the number of sampling moments involved in the integration. This indicates the time interval between adjacent sampling times.

[0093] When the output value of the environmental degradation rate function exceeds the upper limit of the degradation rate observed in the multi-stress accelerated degradation test, the output value is limited to within the set upper limit of the degradation rate. The upper limit of the degradation rate is determined based on the highest observed degradation rate in the multi-stress accelerated degradation test.

[0094] In step S36, a mechanical impact event is identified based on the peak acceleration and the impact duration. The impact criteria include the peak acceleration condition and the impact duration condition.

[0095] When the peak value of the triaxial absolute acceleration reaches the impact threshold corresponding to the electrical tool to be tested, and the duration of the continuous time exceeding the impact threshold is within the set impact duration range, the vibration record is determined as a mechanical impact event.

[0096] In this embodiment, the peak acceleration condition is 5.0g, and the impact duration ranges from 2ms to 50ms. For different types of electrical appliances, the corresponding impact criteria are adjusted according to the material type, structural strength, and on-site impact waveform test results.

[0097] Single impact degradation increment This was obtained through impact testing of similar insulating material samples.

[0098] Under the same recovery environment and measurement conditions, the insulation properties of the samples were measured before and after being subjected to an impact of a specified magnitude, and the degradation increment of a single impact was determined based on the change in insulation properties. .

[0099] When identified within the statistical period During a secondary mechanical impact event, the cumulative degradation due to impact is... Represented as: ; In step S37, the cumulative environmental degradation and the cumulative impact degradation are added together to obtain the cumulative degradation amount. : ; Environmental cumulative degradation represents the degradation caused by the continuous effects of temperature, humidity, ultraviolet radiation, and salt spray deposition, while shock cumulative degradation represents the additional degradation caused by mechanical shock events.

[0100] After accumulating sufficient data on the combined effects of impact events and environmental parameters, the degradation increment of a single impact is corrected based on the comprehensive environmental stress level at the time of the impact. The corrected cumulative degradation is expressed as: ; in, This indicates the overall environmental stress level at the time of the impact. This represents the interaction coefficient between the impact and environmental parameters. The values ​​were determined based on the normalized values ​​of temperature, humidity, ultraviolet radiation intensity, and equivalent salt spray deposition rate at the time of the impact. Calibrate based on additional tests or historical data.

[0101] S40: By fitting the individual performance degradation mapping with the measured values ​​of periodic insulation performance and the corresponding cumulative degradation, the residual between the predicted and measured values ​​of insulation performance is obtained.

[0102] In this embodiment, step S40 includes: Step S41: After each periodic insulation performance test is completed, the cumulative degradation amount within the preset time window before the test time is read and the test degradation input is generated; the measured insulation performance value of the same test is read and the test performance output is generated. Step S42: Query the material batch prior parameter table. When there is a batch-level prior or a material-level prior, input the test degradation input and the test performance output into Bayesian exponential regression. When there is no available prior, input the test degradation input and the test performance output into least squares regression. Step S43: Solve for the initial insulation performance parameters and degradation ratio parameters to generate the individual performance degradation mapping of the electrical equipment to be tested; Step S44: Input the current cumulative degradation amount into the individual performance degradation mapping to obtain the insulation performance prediction value, and calculate the residual between the measured insulation performance value and the insulation performance prediction value.

[0103] In step S41, after each periodic insulation performance test is completed, the measured value of the insulation performance for that test is obtained. .

[0104] The cumulative degradation value is read within a preset time window prior to the test time. The preset time window is used to reduce the impact of pre-test handling, environmental changes, and test preparation operations on the cumulative degradation value.

[0105] In this embodiment, the preset time window is 24 hours before the test, and the average value of the cumulative degradation within this time window is used as the test degradation input.

[0106] The test degradation input is linked to the measured insulation performance value of the same periodic insulation performance test to form a mapped data point: ; in, Indicates the first The test degradation input corresponding to the sub-periodic insulation performance test, Indicates the first The measured insulation performance values ​​obtained from the periodic insulation performance test.

[0107] In step S42, the material batch prior parameter table stores statistical information on initial insulation performance parameters and degradation ratio parameters according to the type of electrical equipment, the type of insulation material, and the material batch.

[0108] When multiple electrical appliances have accumulated effective periodic insulation performance test data from the same batch of materials, batch-level priors are formed based on the parameter statistics of the electrical appliances in the same batch.

[0109] When data for the same material batch is insufficient, but other batches of the same insulating material have accumulated valid data, a material-level prior is formed based on the statistical results of the parameters of the same insulating material.

[0110] When there are no batch-level or material-level priors, least squares regression is used to generate the individual performance degradation mapping for the cold start phase.

[0111] Individual performance degradation mapping is represented as: ; in, This represents the predicted insulation performance value. Indicates the initial insulation performance parameters. This represents the degradation ratio parameter. This indicates the cumulative amount of degradation.

[0112] When using least squares regression, a logarithmic transformation is performed on the exponential relationship: ; Degraded input of each experiment As the independent variable in the regression, the natural logarithm of the corresponding measured insulation performance value will be used. As the dependent variable in the regression, the intercept is solved using least squares regression. and slope Thus, the initial insulation performance parameters are obtained. and degradation ratio parameter .

[0113] When using Bayesian exponential regression, the prior parameters are read from the material batch prior parameter table. and The corresponding prior statistical parameters are used, and the test degradation input and test performance output of the electrical equipment under test are used as observation data to calculate... and The posterior distribution.

[0114] Determine the individual performance degradation mapping based on the mean, median, or maximum a posteriori estimate of the posterior distribution. and .

[0115] When there are no batch-level or material-level priors, least squares regression is used as the first solution. and When the accumulated effective periodic insulation performance test data of the electrical equipment to be tested reaches the conditions for Bayesian exponential regression modeling, the initial prior is constructed using the parameters obtained from least squares regression and the fluctuation of the regression residuals, and Bayesian updates are performed using the accumulated test degradation inputs and test performance outputs.

[0116] The parameters obtained from the Bayesian update are saved as the initial prior parameters of the corresponding insulation material, which can be used for individual performance degradation mapping modeling of similar power tools in the future.

[0117] Standard samples differ from actual electrical appliances in terms of geometry, insulation thickness, manufacturing condition, and usage. Degradation ratio parameter. This is used to map the cumulative degradation amount obtained by integrating the environmental degradation rate of the standard sample to the actual insulation performance change of the electrical equipment under test.

[0118] In step S43, the initial insulation performance parameters and degradation ratio parameters obtained by solving are bound to the radio frequency identification mark of the electrical equipment to be tested, and the mapping generation time, the range of test data used and the modeling method are recorded to form an individual performance degradation mapping of the electrical equipment to be tested.

[0119] In step S44, the current cumulative degradation amount up to the current periodic insulation performance test time is read. ,Will Input individual performance degradation mapping to obtain predicted insulation performance values. .

[0120] Based on the current measured insulation performance values and predicted insulation performance values Calculate residuals : ; When the residual is negative, it indicates that the measured insulation performance is lower than the predicted insulation performance obtained by mapping environmental history and individual performance degradation. The degree of negative deviation of the residual is used for subsequent dynamic lower control limit determination.

[0121] If a periodic insulation performance test has abnormal test equipment calibration, missing test records, or operation procedures that do not meet the test requirements, the test data will be marked as invalid data and will not be used to update the individual performance degradation mapping and residual dynamic benchmark.

[0122] S50: Select historical residuals of electrical equipment from the same batch and in the same test cycle as the electrical equipment to be tested for self-sampling, obtain the compensated standard deviation, and generate the dynamic lower control limit by the larger of the historical residual mean, the historical residual standard deviation and the compensated standard deviation.

[0123] See Figure 4 In this embodiment, step S50 includes: Step S51: Select historical residuals from the same batch of electrical tools that are in the same test cycle stage as the electrical tools to be tested, and form a set of historical residuals; Step S52: Perform self-sampling on the historical residual set to form a synthetic residual sample set; Step S53: Perform quantile statistics on the synthetic residual sample set and extract the lower quantile as the compensation warning lower limit; Step S54: Calculate the compensation standard deviation based on the difference between the historical residual mean and the compensation warning lower limit, and the control limit coefficient; Step S55: When generating the dynamic lower control limit, the larger value between the historical residual standard deviation and the compensated standard deviation is selected to participate in the calculation of the dynamic lower control limit, and residuals that are judged to be abnormal are not used to update the historical residual set.

[0124] In step S51, historical residuals are screened according to material batch and number of tests.

[0125] Material batches are used to ensure that the selected electrical appliances have similar insulation materials, manufacturing processes, and initial performance. The number of trials phase is used to ensure that historical residuals are at a similar data accumulation stage.

[0126] For example, when the electrical equipment under test completes its third periodic insulation performance test, the normal historical residuals corresponding to the third periodic insulation performance test are selected from other electrical equipment of the same material batch to form a set of historical residuals: ; in, This represents the number of residuals in the historical residual set.

[0127] In step S52, sampling with replacement is performed on the historical residual set. Each sampling... Each historical residual is used to form a synthetic residual sample set. This sampling with replacement is repeated to form multiple synthetic residual sample sets.

[0128] The number of self-sampled iterations is determined based on the number of historical residuals and quantile statistical stability. When the change in the compensation warning lower limit obtained after increasing the number of self-sampled iterations is less than the set range, it is determined that the sampling results have reached a steady state.

[0129] In step S53, the residuals in each synthetic residual sample set are sorted from smallest to largest, and the residual values ​​corresponding to the set lower quantile positions are extracted.

[0130] The lower quantile values ​​of multiple synthetic residual sample sets are summarized, and the average or median of these lower quantile values ​​is used as the lower limit of the compensation warning. .

[0131] The lower quantile position is determined based on the normal negative fluctuation probability allowed by anomaly detection. In this embodiment, the residual value corresponding to the 5th quantile position is selected as the lower quantile value.

[0132] In step S54, let the mean of the historical residuals be... The control limit coefficient is The lower limit of the compensation warning is Then compensate for the standard deviation Represented as: ; in, Below This calculation relationship is obtained by inverse calculation from the dynamic lower control limit expression, so that the control limit calculated using the compensated standard deviation corresponds to the compensated warning lower limit obtained by autopilot sampling.

[0133] In step S55, let the standard deviation of the historical residuals of the electrical equipment to be tested be . Then the dynamic lower control limit Represented as: ; in, Indicates the selection of historical residual standard deviation and compensation standard deviation The larger value in the range.

[0134] When the number of historical residuals of the electrical equipment to be tested is small, resulting in Less than At that time, adopt Generate dynamic lower control limits to prevent them from becoming too close to the historical residual mean due to a small number of individual residual samples.

[0135] When the electrical equipment to be tested has accumulated a large number of normal historical residuals, and Not less than At that time, adopt Generate a dynamic lower control limit so that the dynamic lower control limit reflects the normal residual fluctuation of the electrical equipment under test.

[0136] In this embodiment, step S55 further includes: Step S551: Screen the historical residual set according to the material batch and test number stage of the electrical equipment to be tested, sort the sampled synthetic residual sample set by lower quantile, and convert the lower quantile sorting result into the equivalent residual fluctuation lower limit; Step S552: When the standard deviation of the historical residuals is less than the compensation standard deviation converted from the lower limit of the equivalent residual fluctuation, the compensation standard deviation is used in the calculation of the dynamic lower control limit; when the standard deviation of the historical residuals is not less than the compensation standard deviation, the standard deviation of the historical residuals is used in the calculation of the dynamic lower control limit, and the use of residuals judged as abnormal to update the historical residual set is prohibited.

[0137] In step S551, the lower quantiles of each synthetic residual sample set are sorted to obtain multiple lower quantile residual values. The median or average value of the multiple lower quantile residual values ​​is calculated to obtain the lower limit of the equivalent residual fluctuation.

[0138] The lower limit of equivalent residual fluctuation is used as the lower limit of compensation warning. And based on the historical residual mean Compensation warning lower limit and control limit coefficient Calculate the compensation standard deviation .

[0139] In step S552, when Less than At that time, adopt Generate dynamic lower control limits; when Not less than At that time, adopt Generate dynamic lower control limits.

[0140] Residuals deemed abnormal are not included in the historical residual mean and standard deviation of the electrical equipment under test, nor are they included in the normal historical residual set of the same batch and the same number of tests in the subsequent period, in order to avoid abnormal deterioration data from changing the statistical characteristics of normal residuals.

[0141] S60: Compare the residual with the dynamic lower control limit. If the residual is not lower than the dynamic lower control limit, update the dynamic benchmark of the residual. If it is lower, output the anomaly judgment result.

[0142] In this embodiment, step S60 includes: Step S61: After a new periodic insulation performance test is completed, the current cumulative degradation amount is input into the individual performance degradation mapping to obtain the insulation performance prediction value, and the residual between the measured insulation performance value and the insulation performance prediction value is calculated. Step S62: Generate a dynamic lower control limit based on the larger of the historical residual mean, the historical residual standard deviation, and the compensated standard deviation; Step S63: When the residual is not lower than the dynamic lower control limit, update the historical residual mean and the historical residual standard deviation using the forgetting factor; Step S64: When the residual is lower than the dynamic lower control limit, keep the historical residual mean and the historical residual standard deviation unchanged; Step S65: When the residual is lower than the dynamic lower control limit, output the residual abnormality judgment result; Step S66: Calculate the critical degradation amount based on the lower limit of safe insulation performance and the individual performance degradation mapping; Step S67: When the current cumulative degradation exceeds the critical degradation, output the degradation threshold anomaly determination result; Step S68: When the residual is not lower than the dynamic lower control limit and the current cumulative degradation amount does not exceed the critical degradation amount, output a normal judgment result.

[0143] In step S61, the current cumulative degradation amount is... Input individual performance degradation mapping and calculate the predicted insulation performance value. Based on the measured insulation performance values ​​obtained from the current periodic insulation performance tests. Calculate the current residual : ; In step S62, based on the historical residual mean obtained in step S50... Historical residual standard deviation and compensation standard deviation Calculate the dynamic lower control limit : ; The historical residual mean and historical residual standard deviation are calculated based on historical residuals that have been judged to be normal.

[0144] When establishing the dynamic baseline for residuals for the first time, the initial historical residual mean and historical residual standard deviation are calculated using the obtained normal residuals. When the number of normal residuals of the power equipment to be tested is insufficient, the compensation standard deviation is used to limit the residual fluctuation range corresponding to the dynamic lower control limit.

[0145] In step S63, when When the current residual is within the normal residual fluctuation range, the historical residual mean and historical residual standard deviation are updated using the forgetting factor.

[0146] The historical residual mean is updated as follows: ; The historical residual variance is updated as follows: ; in, Indicates the forgetting factor, and These represent the mean and standard deviation of the historical residuals before the update, respectively. and represents the updated historical residual mean and historical residual standard deviation, respectively.

[0147] Forgetting factor The forgetting factor is greater than 0 and less than 1. A larger forgetting factor indicates a greater impact of recent normal residuals on the dynamic baseline of the residuals; a smaller forgetting factor indicates a smoother change in the dynamic baseline of the residuals. In this embodiment, the forgetting factor... Take 0.6.

[0148] In step S64, when When the current residual exceeds the normal negative fluctuation range, the mean and standard deviation of the historical residuals remain unchanged, and the current residual is prohibited from entering the normal historical residual set.

[0149] In step S65, when When the measured insulation performance deviates negatively from the insulation performance predicted by the environmental history, exceeding the normal fluctuation range, the residual abnormality judgment result is output.

[0150] In step S66, the lower limit of safe insulation performance is set to... The initial insulation performance parameters in the individual performance degradation mapping are: The degradation ratio parameter is Then, the critical degradation amount can be calculated back based on the following relationship. : ; ; Lower limit of safety insulation performance Determined according to the insulation performance test requirements of the corresponding electrical equipment.

[0151] In step S67, the current cumulative degradation amount is... With critical degradation Comparison. When When the time is right, output the abnormal determination result of the degradation threshold.

[0152] Degradation threshold anomaly determination and residual anomaly determination are performed separately. Even if the current residual is not lower than the dynamic lower control limit, as long as the current cumulative degradation exceeds the critical degradation, the degradation threshold anomaly determination result will still be output.

[0153] When the individual performance degradation map is generated by least squares regression, the individual performance degradation map is utilized. and Calculate the critical degradation amount and perform degradation threshold judgment during the cold start phase.

[0154] Once the accumulated periodic insulation performance measured data of the electrical equipment to be tested meets the conditions for Bayesian exponential regression modeling, the individual performance degradation mapping and corresponding critical degradation amount are updated, and the residual dynamic benchmark judgment and degradation threshold judgment are performed simultaneously.

[0155] In step S68, when and When the condition is met, output the normal judgment result and update the dynamic benchmark of the residual using the current residual.

[0156] When both the residual abnormality condition and the degradation threshold abnormality condition are met, the residual abnormality and degradation threshold abnormality are recorded simultaneously in the abnormality judgment result.

[0157] The anomaly determination results include the radio frequency identification mark of the electrical equipment under test, the determination time, the anomaly type, the current cumulative degradation amount, the predicted value of insulation performance, the measured value of insulation performance, the current residual, the dynamic lower control limit, and the critical degradation amount.

[0158] The anomaly determination results are written into the testing file of the electrical equipment to be tested according to the radio frequency identification mark and timestamp, and then sent to the safety management terminal.

[0159] The test file also records the version of the environmental degradation rate function used for each judgment, the version of the individual performance degradation mapping, the historical residual mean, the historical residual standard deviation, the compensated standard deviation, and the judgment result.

[0160] Taking a batch of identical insulated operating rods centrally managed in a tool cabinet in a power supply unit's work area as an example. These insulated operating rods are manufactured using the same type and batch of insulating materials and are each equipped with an RFID tag. The tool cabinet in the work area records the time each insulated operating rod leaves and enters the cabinet via an RFID unit. Environmental data acquisition terminals are installed in the tool cabinet, transport vehicles, and work sites to collect temperature, humidity, ultraviolet radiation intensity, salt spray deposition sensor output, and vibration and shock sensor output for each insulated operating rod during its storage in the cabinet and its use after leaving the cabinet.

[0161] First, standard samples using the same insulation material as the batch of insulating operating rods were subjected to multi-stress accelerated degradation tests. Discrete degradation rate samples were generated based on the average environmental parameters, test duration, and insulation resistance before and after the test for each test period. These discrete degradation rate samples were then input into a multi-gene genetic programming algorithm for searching. Candidate functions were selected based on prediction error, expression complexity, monotonicity constraints of environmental parameters, and baseline degradation rate constraints under the benchmark environment to obtain an environmental degradation rate function suitable for this batch of insulation materials.

[0162] Insulating operating rod numbered JG-07 was selected as the insulating operating rod to be tested. During a certain testing cycle, JG-07 experienced an average temperature of 32℃, an average relative humidity of 76%, and an average ultraviolet radiation intensity of 420 W / m². 2 The average equivalent salt spray deposition rate was 0.48 μg / (cm³). 2 •h), and detected two mechanical impact events that met the impact criterion. Integrating the environmental parameters within the detection period according to the environmental degradation rate function, and combining this with the cumulative degradation amount corresponding to the mechanical impact events, the cumulative degradation amount of JG-07 up to the current test time is obtained as follows: ; The measured insulation resistance values ​​and corresponding cumulative degradation values ​​from all periodic insulation performance tests of JG-07 were retrieved. The prior parameters for this batch of insulating operating rods are already stored in the material batch prior parameter table. Therefore, Bayesian exponential regression was used to fit the individual performance degradation mapping of JG-07, yielding the initial insulation performance parameters and degradation ratio parameters as follows: ;

[0163] This leads to the individual performance degradation mapping of JG-07: ; Inputting the current cumulative degradation amount into the individual performance degradation mapping yields the predicted insulation performance value. The measured insulation performance values ​​obtained from this periodic insulation performance test are... Therefore, the current residual is: ; From other insulating operating rods in this batch, normal historical residuals that were in the same test cycle as JG-07 were selected, and self-sampling and lower quantile statistics were performed to obtain the compensated standard deviation. The historical residual mean of JG-07 is The standard deviation of historical residuals is The control limit coefficient is Therefore, the dynamic lower control limit is calculated using the compensated standard deviation: ; Since the current residual is below the dynamic lower control limit, that is: ; Therefore, it is determined that the measured insulation performance of JG-07 deviates abnormally negatively from the insulation performance predicted by the environmental history, and the residual abnormality judgment result is output, while keeping the historical residual mean and historical residual standard deviation of JG-07 unchanged.

[0164] Furthermore, based on the lower limit of safety insulation performance in JG-07 and the individual performance degradation mapping, the critical degradation amount is calculated as follows: Due to the current cumulative degradation Since the threshold degradation amount has not been exceeded, no degradation threshold abnormality judgment result will be output this time.

[0165] Finally, the radio frequency identification mark of JG-07, the current cumulative degradation amount, the predicted value of insulation performance, the measured value of insulation performance, the current residual, the dynamic lower control limit, and the anomaly judgment result are written into the corresponding test file.

[0166] See Figure 5 The second embodiment of the present invention proposes an abnormal detection system for electrical appliances in cabinets based on multi-dimensional environmental parameters, including a discrete sample generation module, a degradation function generation module, an environmental degradation processing module, a performance mapping module, a control limit generation module, and an abnormal judgment module; The discrete sample generation module is used to generate discrete degradation rate samples based on multi-stress accelerated degradation test data of similar insulating materials; The degradation function generation module is used to search for candidate functions that are linear combinations of multiple expression subtrees using multi-gene genetic programming, and to filter them based on the monotonicity constraints of environmental parameters and the basic degradation rate constraints under the benchmark environment, so as to obtain an environmental degradation rate function containing environmental parameter interaction terms. The environmental degradation processing module is used to identify the out-of-cabinet and in-cabinet states of the electrical tools to be tested through the radio frequency identification unit of the tool cabinet in the work area, collect the environmental vector time series of the electrical tools to be tested during the storage stage in the cabinet and the off-cabinet use stage from out of the cabinet to in the cabinet, truncate the environmental parameters that exceed the accelerated test range, input the truncated environmental parameters into the environmental degradation rate function and integrate to obtain the cumulative degradation amount; The performance mapping module is used to fit an individual performance degradation mapping using the measured values ​​of periodic insulation performance and the corresponding cumulative degradation amount, so as to obtain the residual between the predicted insulation performance value and the measured value. The control limit generation module is used to select historical residuals of power equipment from the same batch and in the same test cycle as the power equipment to be tested for self-sampling, obtain the compensated standard deviation, and generate a dynamic lower control limit by using the larger value between the historical residual mean and the historical residual standard deviation and the compensated standard deviation. The anomaly detection module is used to compare the residual with the dynamic lower control limit. When the residual is not lower than the dynamic lower control limit, the dynamic baseline of the residual is updated; when it is lower, the anomaly detection result is output.

[0167] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process and related explanations of the methods described above can be found in the corresponding processes in the foregoing system embodiments, and will not be repeated here.

[0168] It should be noted that the cabinet-based electrical appliance anomaly detection system based on multi-dimensional environmental parameters provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the modules or steps in the embodiments of the present invention can be further decomposed or combined. For example, the modules in the above embodiments can be merged into one module, or further divided into multiple sub-modules to complete all or part of the functions described above. The names of the modules and steps involved in the embodiments of the present invention are only for distinguishing the various modules or steps and are not considered as an improper limitation of the present invention.

[0169] The terms “first”, “second”, etc., are used to distinguish similar objects, not to describe or indicate a specific order or sequence.

[0170] The term "comprising" or any other similar term is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus / device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent in such process, method, article, or apparatus / device.

[0171] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for detecting anomalies in electrical appliances inside a cabinet based on multi-dimensional environmental parameters, characterized in that, include: Discrete degradation rate samples were formed based on multi-stress accelerated degradation test data of similar insulating materials; Multi-gene genetic programming is used to search for candidate functions that are linear combinations of multiple expression subtrees. Based on the monotonicity constraints of environmental parameters and the basic degradation rate constraints under the benchmark environment, environmental degradation rate functions containing environmental parameter interaction terms are obtained. The environmental vector time series of the electrical equipment to be tested in the cabinet is collected, and the environmental parameters that exceed the accelerated test range are truncated at the boundary. The truncated environmental parameters are input into the environmental degradation rate function and integrated to obtain the cumulative degradation amount. By fitting the individual performance degradation mapping with the measured values ​​of periodic insulation performance and the corresponding cumulative degradation, the residual between the predicted and measured values ​​of insulation performance is obtained. Self-sampling was performed on the historical residuals of the power equipment in the same batch and at the same test stage as the power equipment to be tested to obtain the compensated standard deviation. The dynamic lower control limit was generated by the larger of the historical residual mean and the historical residual standard deviation and the compensated standard deviation. Compare the residual with the dynamic lower control limit. If the residual is not lower than the dynamic lower control limit, update the dynamic baseline of the residual. If it is lower, output the anomaly judgment result.

2. The method according to claim 1, characterized in that, The formation of the discrete degradation rate sample includes: In the multi-stress accelerated degradation test, temperature, humidity, ultraviolet irradiance intensity and equivalent salt spray deposition rate were collected for each test period. Logarithmic difference and duration normalization were performed on the measured insulation performance values ​​at the beginning and end of the same test period to obtain the degradation rate of the time period; The degradation rate over the specified time period is bound to the average environmental parameter within the same test period to form a discrete degradation rate sample that includes an environmental parameter field and a degradation rate field.

3. The method according to claim 1, characterized in that, The screening of the environmental degradation rate function includes: Each candidate individual is encoded into multiple expression subtrees. Linear regression is performed on the output of each expression subtree to solve for the linear combination coefficients and intercept terms of each expression subtree, thereby generating candidate degradation rate prediction values. The prediction error is constructed based on the candidate degradation rate prediction value and the discrete degradation rate sample, and the fitness evaluation quantity is generated by combining the expression complexity, the sign of the partial derivative of the environmental parameter and the deviation of the basic degradation rate under the benchmark environment. A penalty is imposed on candidate individuals that violate the monotonicity constraint of environmental parameters or the basic degradation rate constraint under the benchmark environment, and candidate individuals whose fitness evaluation value meets the preset retention conditions are retained, thus obtaining the environmental degradation rate function.

4. The method according to claim 1, characterized in that, The acquisition of the environmental vector time series includes: To bind radio frequency identification (RFID) tags to electrical tools to be tested stored in the tool cabinet in the work area, the radio frequency identification unit set in the tool cabinet in the work area is used to identify the out-of-cabinet and in-cabinet status of the electrical tools to be tested; The temperature, humidity, ultraviolet radiation intensity, salt spray deposition, and vibration and shock sensor outputs of the electrical equipment under test are collected during the storage stage inside the cabinet and the off-cabinet use stage from the time it is taken out of the cabinet to the time it is put back into the cabinet. The salt spray deposition sensor output is converted into an equivalent salt spray deposition rate according to laboratory calibration. According to the RFID tags and timestamps, the temperature, humidity, ultraviolet irradiation intensity, equivalent salt spray deposition rate, root mean square of vibration, peak acceleration, and impact duration are archived, and the storage area inside the cabinet and the usage area outside the cabinet are divided according to the time of exiting and entering the cabinet, forming the environmental vector time series.

5. The method according to claim 4, characterized in that, The calculation of the cumulative degradation includes: Read the environmental parameters at adjacent sampling times in the environmental vector time series, and truncate the environmental parameters that are outside the coverage of the multi-stress accelerated degradation test. The environmental parameters truncated at the boundary between adjacent sampling times are respectively input into the environmental degradation rate function, and trapezoidal integration is performed according to the adjacent sampling interval to obtain the cumulative environmental degradation. Mechanical impact events that meet the impact criteria are identified from the vibration and impact sensor output. The number of mechanical impact events is multiplied by the single impact degradation increment of the same type of insulating material to obtain the cumulative impact degradation. The cumulative environmental degradation and the cumulative impact degradation are combined into the cumulative degradation.

6. The method according to claim 1, characterized in that, The fitting of the individual performance degradation map includes: After each periodic insulation performance test is completed, the cumulative degradation amount within the preset time window before the test time is read and the test degradation input is generated. The measured insulation performance value of the same test is read and the test performance output is generated. Query the material batch prior parameter table. When there is a batch-level prior or a material-level prior, input the test degradation input and the test performance output into Bayesian exponential regression. When there is no available prior, input the test degradation input and the test performance output into least squares regression. Solve for the initial insulation performance parameters and degradation ratio parameters to generate the individual performance degradation mapping of the electrical equipment to be tested.

7. The method according to claim 1, characterized in that, The updating of the residual dynamic benchmark includes: After a new periodic insulation performance test is completed, the current cumulative degradation amount is input into the individual performance degradation mapping to obtain the insulation performance prediction value, and the residual between the measured insulation performance value and the insulation performance prediction value is calculated. The dynamic lower control limit is generated based on the larger of the historical residual mean, the historical residual standard deviation, and the compensated standard deviation. When the residual is not lower than the dynamic lower control limit, the mean of the historical residual and the standard deviation of the historical residual are updated using the forgetting factor; When the residual is lower than the dynamic lower control limit, the mean of the historical residual and the standard deviation of the historical residual remain unchanged.

8. The method according to claim 7, characterized in that, The generation of the compensated standard deviation includes: Historical residuals that are in the same test cycle stage as the electrical equipment to be tested are selected from the same batch of electrical equipment to form a set of historical residuals; The historical residual set is subjected to auto-sampling to form a synthetic residual sample set; Quantile statistics are performed on the synthetic residual sample set, and the lower quantile is extracted as the compensation warning lower limit; The compensation standard deviation is calculated based on the difference between the historical residual mean and the compensation warning lower limit, as well as the control limit coefficient. When generating the dynamic lower control limit, the larger value between the historical residual standard deviation and the compensated standard deviation is selected to participate in the calculation of the dynamic lower control limit, and residuals that are judged to be abnormal are not used to update the historical residual set.

9. The method according to claim 7, characterized in that, The anomaly determination includes: When the residual is lower than the dynamic lower control limit, output the residual abnormality judgment result; The critical degradation amount is calculated by back-calculating the lower limit of safe insulation performance and the individual performance degradation mapping. When the current cumulative degradation exceeds the critical degradation, an abnormal degradation threshold determination result is output. When the residual is not lower than the dynamic lower control limit and the current cumulative degradation amount does not exceed the critical degradation amount, a normal judgment result is output.

10. A cabinet-based electrical appliance anomaly detection system based on multi-dimensional environmental parameters, characterized in that, It includes a discrete sample generation module, a degradation function generation module, an environment degradation processing module, a performance mapping module, a control limit generation module, and an anomaly detection module; The discrete sample generation module is used to generate discrete degradation rate samples based on multi-stress accelerated degradation test data of similar insulating materials; The degradation function generation module is used to search for candidate functions that are linear combinations of multiple expression subtrees using multi-gene genetic programming, and to filter them based on the monotonicity constraints of environmental parameters and the basic degradation rate constraints under the benchmark environment, so as to obtain an environmental degradation rate function containing environmental parameter interaction terms. The environmental degradation processing module is used to identify the out-of-cabinet and in-cabinet states of the electrical tools to be tested through the radio frequency identification unit of the tool cabinet in the work area, collect the environmental vector time series of the electrical tools to be tested during the storage stage in the cabinet and the off-cabinet use stage from out of the cabinet to in the cabinet, truncate the environmental parameters that exceed the accelerated test range, input the truncated environmental parameters into the environmental degradation rate function and integrate to obtain the cumulative degradation amount; The performance mapping module is used to fit an individual performance degradation mapping using the measured values ​​of periodic insulation performance and the corresponding cumulative degradation amount, so as to obtain the residual between the predicted insulation performance value and the measured value. The control limit generation module is used to select historical residuals of power equipment from the same batch and in the same test cycle as the power equipment to be tested for self-sampling, obtain the compensated standard deviation, and generate a dynamic lower control limit by using the larger value between the historical residual mean and the historical residual standard deviation and the compensated standard deviation. The anomaly detection module is used to compare the residual with the dynamic lower control limit. When the residual is not lower than the dynamic lower control limit, the dynamic baseline of the residual is updated; when it is lower, the anomaly detection result is output.