Life evaluation method based on historical fault data and operation data of electric energy meter

Abstract

This invention discloses a lifespan assessment method based on historical fault data and operational data of electricity meters, belonging to the field of electricity meter lifespan assessment technology. This invention calculates the basic lifespan of electricity meters in the same batch under standard operating conditions using a Weibull distribution model, constructs a multi-dimensional correlated dataset and a linear weighted multi-stress coupling model, calculates real-time lifespan loss by combining real-time operational data, and completes early warning assessment based on the proportion of loss and instantaneous growth rate after superimposing historical loss data. This invention models based on historical operational data and historical fault data, fitting single-stress fault loss functions under standard and non-standard operating conditions, comprehensively quantifying normal and abnormal lifespan losses caused by multi-stress coupling. This avoids the shortcomings of accelerated laboratory testing in simulating multi-stress synergy, and also compensates for the problems of traditional on-site statistical assessment not considering operating condition differences and incomplete coverage of single-stress assessment, accurately reflecting the aging state of smart electricity meters throughout their entire lifecycle.

Description

Technical Field

[0001] This invention belongs to the field of electricity meter life assessment technology, specifically a life assessment method based on historical fault data and operational data of electricity meters. Background Technology

[0002] As the core metering terminal of the smart grid, the operational reliability and service life of smart meters are directly related to the accuracy of electricity metering and the stability of grid operation. At present, the industry has formed a variety of technical routes for the life assessment of smart meters. The mainstream ones include assessment schemes based on accelerated laboratory tests, statistical assessment schemes based on historical field data, and loss assessment schemes based on single or partial stress.

[0003] Existing smart meter life assessment technologies still have many problems that need to be solved. First, some assessment schemes do not distinguish between standard and non-standard operating conditions of the meter, and directly use historical field data to calculate the average life. This is easily affected by extreme operating data and does not fully explore the implicit life loss information in the operating data, so it cannot accurately reflect the aging state of the smart meter throughout its entire life cycle. Second, laboratory accelerated testing schemes can only simulate idealized single extreme stress environments and cannot reproduce the actual multi-stress synergistic operating conditions in the field. This leads to the aging mechanism not matching reality and the life prediction results having a large deviation.

[0004] This invention provides a lifespan assessment method based on historical fault data and operational data of electricity meters to solve the above-mentioned technical problems. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a life assessment method based on historical fault data and operation data of electricity meters.

[0006] To achieve the above objectives, a first aspect of the present invention provides a lifespan assessment method based on historical fault data and operational data of an electricity meter, comprising:

[0007] The basic lifespan of the same batch of electricity meters under standard operating conditions was calculated using the Weibull distribution model.

[0008] A multi-dimensional associated dataset is constructed, and a multi-stress coupling model of the same batch of energy meters is built based on the associated dataset. The real-time life loss of the smart energy meter is calculated by combining the multi-stress coupling model and real-time operation data. The multi-stress coupling model is a linear weighted coupling model.

[0009] The real-time lifespan loss and historical lifespan loss of the smart meter are superimposed to obtain the lifespan loss of the smart meter; an early warning assessment of the smart meter is completed based on the basic lifespan and lifespan loss.

[0010] In one possible implementation, the baseline lifespan of electricity meters in the same batch under standard operating conditions is calculated using the Weibull distribution model, including:

[0011] Read historical operating data and historical fault data of the same batch of electricity meters under standard operating conditions; construct a basic sample set based on the historical operating data and historical fault data;

[0012] The basic lifespan of electricity meters in the same batch is calculated by combining the basic sample set and the Weibull distribution model.

[0013] In one possible implementation, a basic sample set is constructed based on historical operational data and historical failure data, including:

[0014] Valid operating data of smart meters in the same batch are selected from historical operating data as non-failed samples; core feature parameters are extracted from historical fault data as failed samples; among them, the core feature parameters include installation time, failure time, cumulative running time under standard operating conditions at the time of failure, and fault type.

[0015] The non-failed and failed samples are labeled and integrated to establish a basic sample set; the sample labeling includes runtime, runtime parameters and failure status.

[0016] In one possible implementation, the baseline lifetime of electricity meters in the same batch is calculated by combining a basic sample set and a Weibull distribution model, including:

[0017] The likelihood function is constructed based on the probability density function and reliability function of the Weibull distribution model; the likelihood function is solved using the Newton-Raphson method on the basis of the basic sample set to obtain the Weibull distribution model.

[0018] The basic lifespan of the same batch of electricity meters is calculated based on the reliability function and the preset reliability; the preset reliability is set to 0.95.

[0019] In one possible implementation, the Weibull distribution model is modified using a Bayesian correction method, including:

[0020] The prior distribution of the shape parameter is set to a gamma distribution, and the prior distribution of the scale parameter is set to a log-normal distribution; the posterior distribution function is constructed based on the probability density function and likelihood function of the gamma distribution and the log-normal distribution.

[0021] Based on the basic sample set, the posterior distribution function is solved by combining the Markov chain Monte Carlo algorithm to obtain the corrected shape parameters and scale parameters, thus completing the correction of the Weibull distribution model.

[0022] In one possible implementation, batch optimization coefficients are introduced to optimize the Weibull distribution model, including:

[0023] A batch optimization coefficient is set by randomly inspecting the quality of components in the same batch of electricity meters; the value range of the batch optimization coefficient is [0.95-1.05].

[0024] The optimized scale parameters are obtained by multiplying the batch optimization coefficients by the scale parameters of the Weibull distribution model, thus completing the optimization of the Weibull distribution model.

[0025] In one possible implementation, a multi-dimensional related dataset is constructed, including:

[0026] Based on the preset stress factor, the historical operating data and historical fault data are divided into several subsets, and then further divided into subsets two according to standard and non-standard operating conditions.

[0027] The fault type and life loss of each historical fault data in Subset 2 are labeled, and a multi-dimensional associated dataset is generated based on Subset 1, Subset 2 and the labeling results; among them, the life loss is calculated in combination with historical operation data.

[0028] In one possible implementation, a multi-stress coupling model and real-time operational data are combined to calculate the real-time lifetime loss of the smart energy meter, including:

[0029] By using regression analysis to fit the multi-dimensional correlated dataset, a fault loss function for a single stress is obtained.

[0030] A multi-stress coupling model is constructed based on the fault loss function of a single stress. Real-time operating data is then input into the multi-stress coupling model to calculate the real-time life loss of the smart energy meter.

[0031] In one possible implementation, the fault loss function for a single stress is obtained, including:

[0032] Extract sub-dataset two corresponding to sub-dataset one from a multi-dimensional associated dataset;

[0033] The regression analysis method was used to fit the subset data to obtain the fault loss function under a single stress; the fault loss function includes fault loss functions under standard and non-standard operating conditions.

[0034] In one possible implementation, the early warning assessment of the smart energy meter is completed based on the base lifespan and the amount of lifespan depletion, including:

[0035] When the instantaneous growth rate of the real-time life loss exceeds the preset growth rate threshold, or when the life loss reaches a preset proportion of the basic life, an early warning signal is generated; the preset growth rate threshold is set according to the fault loss pattern, and the preset proportion is 80%.

[0036] Compared with the prior art, the beneficial effects of the present invention are:

[0037] 1. This invention calculates the basic lifespan of a batch of electricity meters under standard operating conditions using the Weibull distribution model, constructs a multi-dimensional correlated dataset and a linear weighted multi-stress coupling model, calculates real-time lifespan loss based on real-time operating data, and completes early warning assessment based on the proportion of loss and instantaneous growth rate after superimposing historical loss data. This invention strictly distinguishes between standard and non-standard operating conditions, models based on historical operating data and historical fault data, fits the single-stress fault loss function under standard and non-standard operating conditions, and comprehensively quantifies the normal and abnormal lifespan loss caused by multi-stress coupling. It avoids the shortcomings of laboratory accelerated testing in simulating multi-stress synergy, and makes up for the problems of traditional on-site statistical assessment not considering operating condition differences and single-stress assessment not being comprehensive. It accurately reflects the aging status of smart electricity meters throughout their entire life cycle, and the early warning mechanism takes into account both cumulative loss and instantaneous changes, providing a clear basis for operation and maintenance and improving the accuracy of lifespan assessment.

[0038] 2. This invention addresses the problem of insufficient accuracy in fitting the Weibull distribution model using maximum likelihood estimation with small samples. It employs a Bayesian correction method, setting the prior shape parameter as a gamma distribution and the prior scale parameter as a log-normal distribution. The posterior distribution is then solved using a Markov chain Monte Carlo algorithm to correct the shape and scale parameters of the Weibull distribution model. This invention introduces a prior distribution adapted to the reliability characteristics of smart meters, combined with iterative solving using a basic sample set. This effectively compensates for the modeling bias caused by insufficient early failure samples, improves the fitting accuracy of the Weibull distribution model, and makes the basic lifetime calculation more consistent with actual field conditions. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a schematic diagram of the method steps for assessing the lifespan of a smart energy meter in Embodiment 1 of the present invention;

[0041] Figure 2 This is a schematic diagram illustrating the steps of correcting the Weibull distribution model using the Bayesian correction method in Embodiment 2 of the present invention.

[0042] Figure 3 This is a schematic diagram illustrating the steps of the method for optimizing the Weibull distribution model by introducing batch optimization coefficients in Embodiment 3 of the present invention. Detailed Implementation

[0043] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] As the core metering terminal of the smart grid, the operational reliability and lifespan of smart meters directly affect the accuracy of electricity metering and the stability of grid operation. In the field of smart meter lifespan assessment technology, a development pattern has emerged with multiple technical approaches coexisting. All these approaches revolve around the core process of "data acquisition - model building - lifespan assessment," and the main technical approaches are as follows:

[0045] 1. Life assessment scheme based on laboratory accelerated testing: By applying extreme stress (such as high temperature, high humidity, overvoltage, etc.) higher than the normal operating level of smart energy meters in a laboratory environment, the aging process of the energy meter is accelerated. Combined with accelerated life models (such as the Arrhenius model and the inverse power law model), the aging data under laboratory accelerated conditions is extrapolated to the actual operating conditions, thereby obtaining the life prediction results of the energy meter. This technical route is mainly applied to the development of new smart energy meters and the verification of factory quality.

[0046] The extreme stresses applied in the laboratory (single high temperature, high humidity, or overpressure) of this scheme represent an idealized single-stress environment. This cannot simulate the multi-stress synergistic effects (such as the synergistic effects of temperature, humidity, harmonics, and overload) encountered in actual smart energy meter operation. Consequently, the laboratory aging mechanism differs from the actual field aging mechanism. The lifespan predictions extrapolated from this aging data deviate significantly from the actual lifespan of the energy meter in the field, failing to accurately reflect the lifespan under field operating conditions. Furthermore, laboratory tests can only simulate rapid aging failure under extreme stress, failing to simulate the implicit lifespan losses caused by mild environments and slight overloads in actual field operation. This results in predicted lifespans that only reflect failure under extreme conditions, deviating from the actual lifespan losses of the energy meter throughout its entire lifespan. The predictions are overly idealized and cannot provide reliable support for field operation and maintenance.

[0047] 2. Statistical evaluation scheme based on historical field data: Relying on the historical fault data and field operation data of smart meters accumulated during the operation and maintenance process of power companies, a reliability statistical model (such as the Weibull distribution model) or a data-driven model (such as the LSTM model) is used to statistically analyze the life of the electricity meter and output the life assessment results. This scheme is mainly applied to the operation and maintenance evaluation scenario of batch smart meters of power companies.

[0048] This scheme fails to clearly distinguish between the "standard operating conditions" and "actual operating conditions" of smart meters, directly using historical fault data to calculate the average lifespan. This makes it susceptible to interference from extreme on-site operating data (such as short-term overloads or extreme temperature and humidity), and it does not consider individual differences in components among meters of the same model and batch. This causes the baseline lifespan to deviate from the actual design lifespan, affecting overall prediction accuracy. Furthermore, it focuses only on the explicit lifespan loss corresponding to historical fault data, failing to fully explore the implicit lifespan loss information contained within historical operating data (such as slow losses caused by long-term mild overloads or harmonic interference). This results in incomplete lifespan loss calculations, and the prediction results cannot reflect the full-cycle aging state of the smart meter.

[0049] 3. Loss assessment scheme based on single or partial stress: For single key stress (such as high temperature, overload) or partial stress that affects the life of smart energy meters, establish a correlation model between stress and life loss, quantify the impact of single or partial stress on the life of energy meters, and thus achieve life prediction. This technical approach is mostly used for energy meter life assessment under specific working conditions.

[0050] This solution only considers the impact of a single stress (such as high temperature) or a subset of stresses on lifespan, failing to account for the cumulative effect of multiple stresses working together in actual operation (such as the combined acceleration of component aging by high temperature and high humidity). It cannot comprehensively quantify the lifespan loss of the entire electricity meter lifecycle, resulting in significant deviations in lifespan predictions and failing to reflect the actual aging state of the electricity meter. Furthermore, this solution is designed only for specific stress conditions and cannot adapt to stress differences in different application scenarios (such as harmonic stress in industrial scenarios and load fluctuation stress in residential scenarios). It cannot meet the lifespan prediction needs of multi-scenario, batch-produced smart electricity meters, lacking versatility and hindering large-scale engineering deployment.

[0051] Implementation 1: Please refer to Figure 1 The first aspect of this invention provides a lifespan assessment method based on historical fault data and operational data of electricity meters, comprising: calculating the basic lifespan of electricity meters in the same batch under standard operating conditions using a Weibull distribution model; constructing a multi-dimensional associated dataset, and constructing a multi-stress coupling model of electricity meters in the same batch based on the associated dataset; calculating the real-time lifespan loss of smart electricity meters by combining the multi-stress coupling model and real-time operational data; superimposing the real-time lifespan loss of smart electricity meters with the historical lifespan loss to obtain the lifespan loss of smart electricity meters; and completing an early warning assessment of smart electricity meters based on the basic lifespan and lifespan loss.

[0052] The basic lifespan calculation targets smart meters from the same production batch, with the same component models, the same manufacturing process, and the same rated parameters. A group of smart meters that meet this requirement is called a batch of smart meters to ensure the batch adaptability of the calculated basic lifespan.

[0053] The baseline lifespan is calculated based on historical operating and fault data under standard operating conditions for the same batch of energy meters. It is the theoretical reliable lifespan under standard operating conditions, calculated using the Weibull distribution model. Essentially, the baseline lifespan is the theoretical lifespan benchmark of a smart energy meter from commissioning to failure under standard operating conditions. It only reflects the ideal aging trend under standard operating conditions and does not quantify the cumulative lifespan loss under standard operating conditions.

[0054] The standard operating condition of the smart meter disclosed in this invention refers to the smart meter's operating parameters and operating environment meeting the rated operating range, and without extreme stress interference. Under this standard operating condition, the smart meter's aging process is a natural aging process without accelerated wear factors. Historical operating data under the standard operating condition includes voltage, current, load rate, ambient temperature and humidity, harmonic content, voltage fluctuation amplitude, etc.

[0055] It should be noted that while smart meters operate within their rated operating range, the conditions under extreme stress interference are not considered standard operating conditions. Under these conditions, it means that individual operating parameters do not exceed the rated range, but there are situations such as "long-term critical fluctuations of parameters," "multiple parameters experiencing extreme stress," and "short-term high-frequency impacts." These situations can lead to accelerated aging of smart meters and hidden lifespan losses.

[0056] It is worth noting that if the historical operating data or historical fault data required by this invention cannot be fully extracted from the working records of the same batch of electricity meters, it can be supplemented by laboratory simulation as needed.

[0057] In one example, for the "long-term critical fluctuation of parameters" working condition: the voltage of the electricity meter is in the critical range of ±8% to ±10% of the rated voltage for a long time, and the fluctuation frequency is high. Although it does not exceed the rated range of ±10%, the continuous critical voltage impact will accelerate the aging of components such as metering chips and capacitors.

[0058] For example, in the case of "multi-parameter synergistic extreme" operating conditions: the voltage fluctuation is within the rated range (e.g., ±5%-±8%), but the harmonic content of the power grid is close to the rated upper limit (total harmonic distortion of voltage THD≈5%, current THD≈10%). The combined effect of the two leads to an increase in harmonic losses in the internal circuit of the energy meter and an increased computational load on the metering chip.

[0059] Historical fault data refers to the failure-related data collected through maintenance records and fault reporting systems after a batch of electricity meters failed due to natural aging (under standard operating conditions) during operation. Historical fault data includes basic information and failure characteristic data. Basic information includes the smart meter's asset number, production batch information, and rated parameter information. Failure characteristic data includes installation time, fault reporting time, cumulative operating time under standard operating conditions at the time of failure, and snapshots of operating parameters at the time of failure.

[0060] In a preferred embodiment, a basic sample set is constructed based on historical operating data and historical fault data.

[0061] When calculating the basic lifespan of electricity meters in the same batch, historical operating data and historical fault data under standard operating conditions of the same batch of electricity meters are first read from the database. The historical operating data and historical fault data of the same batch of electricity meters are combined with the Weibull distribution model to calculate the basic lifespan under standard operating conditions. This basic lifespan serves as the lifespan assessment benchmark for all smart electricity meters in the same batch, providing a unified baseline for subsequent rated lifespan loss calculations and remaining lifespan assessments.

[0062] Valid operating data from all smart meters in the same batch were selected from the historical operating data and used as the non-failure sample. Valid operating data corresponds to the operating data of the smart meters under standard operating conditions within the rated operating parameter range. Abnormal data caused by sensor failures, communication interruptions, etc., were removed during the selection process. The non-failure sample should ideally cover different operating periods of the same batch of smart meters, such as different seasons and different time periods with varying loads, to reflect the fluctuations in operating conditions within the standard operating range and ensure the comprehensiveness of the sample.

[0063] Historical fault data was standardized, and core characteristic parameters of each failed energy meter were extracted as fault samples. Standardization mainly included unifying data format and dimensions, normalizing data units and units, removing invalid data, and correcting anomalies. Core characteristic parameters included installation time, failure time, cumulative operating time under standard conditions at the time of failure, and fault type. It is worth noting that when acquiring fault samples, failure data caused by non-standard operating conditions was strictly excluded; only failure data resulting from natural aging under standard operating conditions was retained, ultimately forming a standardized fault sample set to provide fault data support for subsequent model parameter fitting.

[0064] It should be noted that when collecting historical fault data from the same batch of electricity meters, the focus should be on supplementing early failure samples. If the number of early failure samples from the same batch of electricity meters is insufficient, early failure data from similar batches of the same model and with the same components can be selected as supplementary samples.

[0065] Data filtering, feature extraction, and field standardization were performed on both unfailed and failed samples. The runtime, operating parameters (such as voltage, load, temperature, and humidity), and failure status (unfailed / failed) of each sample were labeled to establish a basic sample set, providing data support for the subsequent fitting of the Weibull distribution model.

[0066] In one example, each sample in the base sample set (either a non-failed sample or a failed sample) includes data items such as asset number, production batch, cumulative runtime, average voltage, average load rate, average humidity, failure status, and fault type. It should be noted that the cumulative runtime of a non-failed sample is the truncated duration, indicating that the smart meter is not yet failed; the cumulative runtime of a failed sample is the failure duration, indicating that the smart meter has already failed.

[0067] In a preferred embodiment, the basic lifespan of the same batch of electricity meters is calculated by combining the basic sample set and the Weibull distribution model.

[0068] This invention selects a two-parameter Weibull distribution model to calculate the basic lifespan of electricity meters in the same batch. Based on the Weibull distribution model, maximum likelihood estimation is used to solve for the shape and scale parameters in the Weibull distribution model. During the solution process, a likelihood function is first constructed based on the probability density function and reliability function of the Weibull distribution model. The runtime and failure states of the basic sample set are used as inputs to solve for the shape and scale parameters corresponding to the maximum likelihood function, thus obtaining the fitted Weibull distribution model.

[0069] Based on the fitted Weibull distribution model, the basic lifespan of the same batch of energy meters under standard operating conditions was calculated, and the reliability was set. This reliability level is set based on the reliability requirements of power companies for the operation of smart meters. The reliability function of the Weibull distribution model is... ,Will Substituting the values ​​into the solution yields... Value, that This value represents the basic lifespan of electricity meters in the same batch, calculated using the following formula: .

[0070] In one example, the probability density function and reliability function of the Weibull distribution model are fundamental to solving the Weibull distribution model. The probability density function is: Used to describe different runtimes The failure probability density is given by the following; the reliability function is: Used to describe runtime of The probability that it has not yet failed. Among them, To accumulate runtime, For shape parameters, Reflecting the failure distribution characteristics of smart energy meters, At that time, the probability of failure increases with time, which is consistent with the aging failure pattern of electricity meters. This is a scale parameter, representing the lifetime when the failure probability is 1-e⁻¹≈36.8%.

[0071] The failed and non-failed samples required for maximum likelihood estimation are extracted from the standardized base sample set. (Failed samples) It refers to the first The cumulative operating time under standard operating conditions of a failed energy meter is given by its likelihood contribution as a probability density function. Unexpired samples It refers to the first The truncated operating time of the non-failed energy meters is represented by their likelihood contribution as a reliability function. .

[0072] The likelihood contribution of all sample data is calculated based on the basic sample set. The likelihood function is obtained by multiplying the likelihood contributions of all sample data. The likelihood function is: After substituting the function expansion, we get: .

[0073] The expanded likelihood function is difficult to solve, so we take its natural logarithm to transform it into a log-likelihood function: Based on the log-likelihood function, we obtain information about... and The first-order partial derivatives are obtained by setting them to zero, resulting in a system of partial derivative equations. These equations are then solved using a numerical iteration method. and The parameter estimates.

[0074] Scale parameters The partial derivative equation is After simplification, we get .

[0075] Shape parameters The partial derivative equation is Since the partial derivative equation has no analytical solution, the Newton-Raphson method is used for iterative data analysis to find the solution, which is the shape parameter. Set initial values Based on the aging characteristics of smart energy meters, The value is usually 2.

[0076] During numerical iterative calculations, the initial values ​​of the parameters are... Substitute In the calculation formula, the obtained Input shape parameters The partial derivative equations yield the shape parameters. Then Substitute In the calculation formula, ..., if the preset convergence condition is not met, the process continues to iterate according to this process.

[0077] The preset convergence condition is: until the parameter change in two consecutive iterations is less than a preset change threshold (usually taken as...). The estimated values ​​of the shape and scale parameters are obtained by solving the problem.

[0078] In a preferred embodiment, a multi-dimensional associated dataset is constructed, and a multi-stress coupling model of the same batch of energy meters is constructed based on the associated dataset; the real-time life loss of the smart energy meter is calculated by combining the multi-stress coupling model and real-time operating data.

[0079] The lifespan degradation of smart meters is primarily based on historical fault data and baseline lifespan of meters from the same batch, as well as real-time operating data for each smart meter. Real-time operating data is collected according to a preset collection cycle, such as every 1-5 minutes. This data includes: real-time voltage, real-time current, real-time load rate, real-time ambient temperature and humidity, real-time harmonic content, and real-time voltage fluctuation amplitude, covering key stress parameters that affect the lifespan of smart meters.

[0080] By filtering historical operating and fault data from the same batch of electricity meters, failure data caused by preset stress factors under standard and non-standard operating conditions were identified. Failure data caused by non-stress factors (such as human-caused damage) were removed, resulting in a multi-dimensional correlation dataset of fault-stress-loss. For real-time operating data, abnormal real-time data caused by sensor failures and communication interruptions were removed. A moving average method was used to correct data fluctuations, data units were standardized, and the acquisition time and stress type corresponding to each set of real-time data were labeled. Preset stress factors include: voltage stress, load stress, temperature and humidity stress, and harmonic stress, covering fluctuating stress under standard operating conditions and extreme stress under non-standard operating conditions.

[0081] Real-time operational data and historical fault data are normalized according to a unified standard to ensure data consistency, providing data support for subsequent fault loss function identification and multi-stress coupling model fitting.

[0082] In practice, historical operating data and historical fault data are first divided into several subsets according to a preset stress factor. Each subset contains only the dataset under the corresponding preset stress factor. Within subset one, the datasets are further divided into subset two according to standard and non-standard operating conditions.

[0083] For the historical fault data in subset two, the corresponding fault types are labeled; combining the stress duration of historical operating data and the failure time in historical fault data, the lifespan loss of the smart energy meter is calculated. After standardizing subset two and the standard content, a fault loss function for a single stress can be fitted, including fault loss functions under standard and non-standard operating conditions.

[0084] A regression analysis algorithm is used to fit the multi-dimensional correlation dataset of fault-stress-loss to establish the correlation function between various single stresses and lifetime loss, and to determine the loss coefficient of a single stress. The fitted single stress loss function is compared with the actual loss in historical fault samples to ensure that the fitting deviation does not exceed 5%. If the deviation exceeds the range, the parameters of the regression analysis algorithm are adjusted and refitted to finally obtain a stable and reliable fault loss function.

[0085] The fault loss function for a single stress is used to describe the mapping relationship between the stress parameter value and stress duration of a single stress and the life loss of a smart energy meter. Its core function is to quantify the life loss value of a smart energy meter corresponding to a unit stress and a unit stress duration.

[0086] The preset stress factors affecting smart meters include voltage stress, load stress, temperature and humidity stress, and harmonic stress. Voltage stress refers to the degree to which the actual voltage deviates from the rated voltage and the amplitude of voltage fluctuations during the operation of the smart meter. Load stress refers to the ratio of the actual load to the rated load and the frequency of load fluctuations. Temperature and humidity stress refers to the temperature and relative humidity of the smart meter's operating environment, as well as the duration of these temperature and humidity effects. Harmonic stress refers to the harmonic content, harmonic order, and duration of harmonic effects in the power grid.

[0087] The single failure sample is extracted from the standardized historical failure data, which is dominated by a single stress. The extracted failure sample excludes failure data caused by non-stress factors and multiple stress coupling, ensuring that the single failure sample is caused by only a single stress.

[0088] Stress characteristics, loss characteristics, and operating condition characteristics are extracted from a single failure sample. Stress characteristics refer to the stress parameters and duration of a single stress, loss characteristics refer to the actual lifespan loss of the smart meter under a single stress, and operating condition characteristics include standard or non-standard operating conditions.

[0089] The single failure sample is divided into several subsets according to a preset stress factor. Each subset contains only failure samples with a single preset stress factor and extracted features. Regression analysis is used to fit mathematical relationships for each subset to obtain the life loss function for a single stress.

[0090] It should be noted that the life loss function is fitted separately for standard and non-standard operating conditions in the same subset of data, that is, the life loss function corresponds to both standard and non-standard operating conditions under a single stress.

[0091] In one example, a linear function form can be used under standard operating conditions. Nonlinear forms can be used under non-standard operating conditions. ;in, The loss coefficient is... In These are non-linear exponents, all obtained through regression analysis.

[0092] The multi-stress coupling model is a linear weighted coupling model, and its specific form is as follows: .

[0093] In the formula:

[0094] This represents the lifespan loss caused by real-time stress.

[0095] This represents the number of coupling stress factors;

[0096] For the first The loss weights for stress-like parameters are obtained by fitting the failure loss pattern; the greater the impact of stress on lifespan, the greater the weight.

[0097] For the first A single loss function similar to stress. For the first Real-time values ​​of stress-like structures. For the first The real-time duration of stress-like effects;

[0098] The stress coupling coefficient ranges from 0.05 to 0.15. It is obtained by fitting multi-stress failure samples from historical fault data and reflects the amplification effect of multi-stress synergy.

[0099] For the first A single loss function similar to stress. , Class and Stress-like coupling effects are formed.

[0100] The real-time stress parameters (including parameter values ​​and duration) of the preset stress factor are calculated based on real-time operating data. By substituting these real-time stress parameters into the multi-stress coupling model, the lifespan loss of the smart energy meter under the current real-time operating conditions can be calculated. It should be noted that the losses corresponding to fluctuating stress under standard operating conditions are included in the normal life loss, while the losses corresponding to extreme stress under non-standard operating conditions are included in the abnormal life loss.

[0101] Life loss includes normal life loss under standard operating conditions and abnormal life loss under non-standard operating conditions.

[0102] Normal lifespan loss under standard operating conditions refers to the cumulative lifespan loss of a smart meter during long-term normal operation under standard operating conditions (meeting rated operating parameters and without extreme stress interference). This loss is caused by unavoidable natural aging processes such as natural aging of components, continuous application of normal electrical stress (rated voltage and rated load), and slight environmental wear. This loss is a gradual accumulation of losses from commissioning to reaching the basic lifespan of the meter under ideal standard operating conditions. It is an inherent loss during normal operation of the meter and cannot be completely avoided.

[0103] Abnormal lifespan loss under non-standard operating conditions refers to the additional lifespan loss of smart meters that occurs when they operate under non-standard conditions (exceeding rated operating parameters, experiencing extreme stress), such as overload, overvoltage, extreme temperature and humidity, harmonic interference, etc. This results in accelerated aging and increased wear of components, exceeding the normal aging range. This loss is not inherent and is directly related to fluctuations in actual operating conditions; the abnormal loss varies significantly between different individual meters.

[0104] It should be noted that normal life loss under standard operating conditions is not considered in the basic life calculation. The core reason is that the essence of basic life is the "theoretical reliable life benchmark under standard operating conditions," which is obtained by fitting a Weibull distribution model and represents the "theoretical lifespan" of an electricity meter from commissioning to failure under standard operating conditions. Its core function is to determine the life benchmark for electricity meters in the same batch. It only reflects the ideal aging trend under standard operating conditions and does not quantify "the actual accumulated life loss of the electricity meter from commissioning to the present moment." Normal life loss, on the other hand, is a correction to the actual operating loss of the "basic life benchmark." It quantifies the share of life that the electricity meter has consumed after operating under standard operating conditions for a period of time. The two have clear divisions of labor and no overlap, together constituting the accurate assessment logic of "theoretical benchmark - actual loss."

[0105] In a preferred embodiment, the smart energy meter is assessed for early warning based on the base lifespan and the amount of lifespan loss. When the instantaneous growth rate of the real-time lifespan loss exceeds a preset growth rate threshold, or when the lifespan loss reaches a preset proportion of the base lifespan, an early warning signal is generated.

[0106] The real-time lifespan loss calculated each time is superimposed with the historical lifespan loss to obtain the lifespan loss of a single smart energy meter. At the same time, the lifespan loss is repeatedly calculated for each new set of real-time operating data, and the real-time lifespan loss and cumulative loss value are dynamically updated.

[0107] When the instantaneous growth rate of real-time life loss exceeds the preset threshold (set based on historical failure loss patterns), or when the cumulative real-time life loss reaches 80% of the basic life, an early warning signal is triggered to remind maintenance personnel to promptly investigate the operating conditions.

[0108] Example 2: Based on Example 1, the Weibull distribution model in Example 1 is modified.

[0109] In a preferred embodiment, please refer to Figure 2 The Weibull distribution model is modified using the Bayesian correction method.

[0110] When fitting a Weibull distribution model using the maximum likelihood estimation method, insufficient fitting accuracy may occur with small samples (e.g., few early failure samples). Therefore, a Bayesian correction method is introduced to correct the solved shape and scale parameters. Specifically, the prior distribution of the shape parameter is chosen as the gamma distribution, and the prior distribution of the scale function is chosen as the log-normal distribution. The mean of the posterior distribution is then calculated using the base sample set, and these mean values ​​are used as the corrected shape and scale parameters.

[0111] The shape parameter is based on the reliability characteristics of smart energy meters. and scale parameters Set the appropriate prior distribution, and assume that the two are independent of each other.

[0112] Shape parameters The prior distribution is set to a gamma distribution, and the gamma distribution is adapted... The value characteristics are consistent with the aging and failure patterns of electricity meters, that is... The probability of failure increases over time, shape parameters The probability density function is: .

[0113] Hyperparameter values:

[0114] Rate parameter, with a value range of [0.5, 2], and a default value of 1;

[0115] : Shape parameter of gamma distribution, with a value range of [2, 5], and a default value of 3;

[0116] The gamma function is used for distribution normalization and can be solved using numerical computation tools.

[0117] Scale parameters The prior distribution is set to a log-normal distribution, and the scale parameter is... Let be the characteristic lifespan of the smart energy meter, with a positive value that conforms to the distribution law of lifespan parameters of industrial products. Let... ,but Scale parameters The probability density function is: .

[0118] Hyperparameter values:

[0119] The mean of the normal distribution is derived from the scale parameters of the same batch of smart energy meters. Typical values ​​are obtained by logarithmic transformation, for example, the scale parameter. When the typical value is 12 years, ;

[0120] : Standard deviation of normal distribution, with a value range of [0.2, 0.5], and a default value of 0.3.

[0121] Assuming shape parameters and scale parameters They are independent of each other, and their joint prior distribution is: According to Bayes' theorem, the parameters The posterior distribution is proportional to the product of the likelihood function and the joint prior distribution. The likelihood function is exactly the same as that used in maximum likelihood estimation, specifically: ;in, It refers to the basic sample set.

[0122] The posterior distribution is: Substituting the likelihood function and the joint prior distribution, we get: .

[0123] Since the Weibull distribution and the pre-defined prior distribution (Gamma distribution + log-normal distribution) are not conjugate, the Markov chain Monte Carlo algorithm (MCMC) is chosen to numerically iteratively solve for the posterior distribution. The specific solution process is as follows:

[0124] Initial parameter values ​​can be set using either the maximum likelihood estimation result or the prior distribution mean; if the prior distribution mean is chosen, then... ;

[0125] The Metropolis-Hastings algorithm is used to generate candidate parameters through random perturbation, calculate the acceptance probability, and determine whether to update the parameters. The number of iterations is typically set to 10,000-50,000 until the Markov chain converges (convergence criterion: ); This is a classic statistic for determining the convergence of the MCMC algorithm;

[0126] To avoid the influence of initial values ​​on the results, the first 20%-30% of the combustion period samples were removed. The mean of the remaining valid samples was then calculated as the corrected parameter value. , Where N is the number of valid samples, and These are the converged parameter sample values.

[0127] Example 3: Based on Example 1 or Example 2, optimize the Weibull distribution model.

[0128] In a preferred embodiment, please refer to Figure 3 We introduce batch optimization coefficients to optimize the Weibull distribution model.

[0129] After fitting the Weibull distribution model, batch optimization coefficients for the same batch of energy meters are introduced to adjust the scaling parameters of the fit. Optimization is performed. The batch optimization coefficient is used to quantify the lifespan differences in energy meters within the same batch due to individual component quality variations and minor differences in manufacturing processes. The specific correction formula is as follows: Among them, batch optimization coefficient The value range is 0.95-1.05, which is determined based on the pass rate of component sampling and the production process deviation data of the same batch of electricity meters. That is, the higher the pass rate and the smaller the process deviation, the closer k is to 1.

[0130] Batch optimization coefficient The acquisition process can be found here:

[0131] Samples were randomly selected from the same batch of electricity meters for inspection. Component performance testing was conducted on the smart meters in the sampled samples, and the pass rate was calculated based on the test results. Core process parameters, including welding temperature, calibration error, and assembly gap, were extracted from the production process traceability system for the same batch of electricity meters. The process deviation rate for each core process parameter was calculated, and the average of all process deviation rates was taken as the average deviation rate for the same batch of electricity meters.

[0132] A linear normalization method is used, with a pass rate of 100% corresponding to a standardized value of 1 and a pass rate of 90% corresponding to a standardized value of 0, to normalize the sampling pass rate and obtain the standard pass rate value. A reverse normalization method is used, with a process deviation of 0% corresponding to a standardized value of 1 and a process deviation of 5% corresponding to a standardized value of 0, to normalize the average deviation rate and obtain the standard deviation rate value.

[0133] Weighting coefficients of 0.6 and 0.4 were set for the pass rate standard value and the deviation rate standard value, respectively. The weighted result was obtained after weighting. The weighted results Constrain to 0.95-1.05, the constraint formula is: .

[0134] It should be noted that the modification and optimization of the Weibull distribution model can be carried out as independent schemes, or the modification and optimization of the Weibull distribution model can be carried out sequentially.

[0135] After calculating the baseline lifespan of the same batch of electricity meters based on the Weibull distribution model, the baseline lifespan is verified using the aforementioned failure samples. During verification, the operating time of the failed electricity meters under standard operating conditions in the failure samples is compared with the baseline lifespan to ensure that the operating time of over 95% of the non-failed electricity meters does not exceed the baseline lifespan, and the deviation between the operating time of the failed electricity meters and the baseline lifespan does not exceed 8%. If the verification fails, the shape and scale parameters of the Weibull distribution model need to be refitted. During refitting, adjustments include, but are not limited to, the weights of the likelihood function input parameters, adjustments to the parameters of the prior distribution function, and adjustments to the batch correction parameters.

[0136] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments.

[0137] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any other combination thereof. When implemented using a software program, it can be implemented entirely or partially in the form of a computer program product. This computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device containing one or more servers, data centers, etc., that can be integrated with the medium. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0138] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A life assessment method based on historical failure data and operation data of an electric energy meter, characterized in that, include: The basic lifespan of the same batch of electricity meters under standard operating conditions was calculated using the Weibull distribution model. Construct a multi-dimensional associated dataset, and build a multi-stress coupling model of the same batch of electricity meters based on the associated dataset; By combining a multi-stress coupling model and real-time operating data, the real-time lifespan loss of smart energy meters is calculated; where the multi-stress coupling model is a linear weighted coupling model. The real-time lifespan loss of the smart meter is superimposed with the historical lifespan loss to obtain the lifespan loss of the smart meter; an early warning assessment of the smart meter is completed based on the basic lifespan and the lifespan loss. Construct a multi-dimensional related dataset, including: Based on the preset stress factor, the historical operating data and historical fault data are divided into several subsets, and then further divided into subsets two according to standard and non-standard operating conditions. The fault type and life loss of each historical fault data in Subset 2 are labeled, and a multi-dimensional associated dataset is generated based on Subset 1, Subset 2 and the labeling results; among them, the life loss is calculated in combination with historical operation data.

2. The method for life assessment based on historical failure data and operational data of an electric energy meter according to claim 1, characterized in that, The baseline lifespan of electricity meters in the same batch under standard operating conditions was calculated using the Weibull distribution model, including: Read historical operating data and historical fault data of the same batch of electricity meters under standard operating conditions; construct a basic sample set based on the historical operating data and historical fault data; The basic lifespan of electricity meters in the same batch is calculated by combining the basic sample set and the Weibull distribution model. 3.The life assessment method based on the historical fault data and operation data of the electric energy meter according to claim 2, characterized in that, A basic sample set is constructed based on historical operational data and historical failure data, including: Valid operating data of smart meters in the same batch are selected from historical operating data as non-failed samples; core feature parameters are extracted from historical fault data as failed samples; among them, the core feature parameters include installation time, failure time, cumulative running time under standard operating conditions at the time of failure, and fault type. The non-failed and failed samples are labeled and integrated to establish a basic sample set; the sample labeling includes runtime, runtime parameters and failure status.

4. The lifespan assessment method based on historical fault data and operational data of electricity meters according to claim 2, characterized in that, The baseline lifespan of electricity meters in the same batch was calculated by combining the basic sample set and the Weibull distribution model, including: The likelihood function is constructed based on the probability density function and reliability function of the Weibull distribution model; the likelihood function is solved using the Newton-Raphson method on the basis of the basic sample set to obtain the Weibull distribution model. The basic lifespan of the same batch of electricity meters is calculated based on the reliability function and the preset reliability; the preset reliability is set to 0.

95.

5. The lifespan assessment method based on historical fault data and operational data of electricity meters according to claim 4, characterized in that, The Weibull distribution model is modified using Bayesian correction methods, including: The prior distribution of the shape parameter is set to a gamma distribution, and the prior distribution of the scale parameter is set to a log-normal distribution; the posterior distribution function is constructed based on the probability density function and likelihood function of the gamma distribution and the log-normal distribution. Based on the basic sample set, the posterior distribution function is solved by combining the Markov chain Monte Carlo algorithm to obtain the corrected shape and scale parameters, thus completing the correction of the Weibull distribution model.

6. The lifespan assessment method based on historical fault data and operational data of an electricity meter according to claim 4 or 5, characterized in that, The Weibull distribution model is optimized by introducing batch optimization coefficients, including: A batch optimization coefficient is set by randomly inspecting the quality of components in the same batch of electricity meters; the value range of the batch optimization coefficient is [0.95-1.05]. The optimized scale parameters are obtained by multiplying the batch optimization coefficients by the scale parameters of the Weibull distribution model, thus completing the optimization of the Weibull distribution model.

7. The lifespan assessment method based on historical fault data and operational data of electricity meters according to claim 1, characterized in that, Combining a multi-stress coupling model and real-time operational data, the real-time lifespan loss of smart energy meters is calculated, including: By using regression analysis to fit the multi-dimensional correlated dataset, a fault loss function for a single stress is obtained. A multi-stress coupling model is constructed based on the fault loss function of a single stress. Real-time operating data is then input into the multi-stress coupling model to calculate the real-time life loss of the smart energy meter.

8. The lifespan assessment method based on historical fault data and operational data of electricity meters according to claim 7, characterized in that, The fault loss function for a single stress is obtained, including: Extract sub-dataset two corresponding to sub-dataset one from a multi-dimensional associated dataset; The regression analysis method was used to fit the subset data to obtain the fault loss function under a single stress; the fault loss function includes fault loss functions under standard and non-standard operating conditions.

9. The lifespan assessment method based on historical fault data and operational data of electricity meters according to claim 1, characterized in that, Based on the baseline lifespan and lifespan depletion, a warning assessment of the smart energy meter is completed, including: When the instantaneous growth rate of the real-time life loss exceeds the preset growth rate threshold, or when the life loss reaches a preset proportion of the basic life, an early warning signal is generated; the preset growth rate threshold is set according to the fault loss pattern, and the preset proportion is 80%.

Images