A clock intelligent calibration method and system for an electric meter based on multi-modal fusion

Abstract

The present application relates to the technical field of electronic timer calibration, and particularly relates to a kind of electric meter clock intelligent calibration method and system based on multi-modal fusion, method includes: the present application obtains electric meter clock voltage and environmental data first, pre-processing generates transient disturbance sequence and steady voltage time sequence, then according to this respectively calculate voltage chronic influence factor, voltage acute influence factor and environmental acceleration factor;By the regression of the same kind electric meter historical data to determine the influence coefficient of each factor, the total life loss is fused and calculated, and the dynamic correction period of adaptive individual damage is iteratively generated;Finally, a reference group of the same batch and the same environment electric meter is constructed, similar clusters are screened by clustering, and the time consistency after correction is compared to determine the unconventional anomaly and output the targeted calibration strategy.The present application quantifies multi-dimensional error factors by multi-modal fusion, constructs a differentiated model and iteratively generates a dynamic correction period, solves the defects of traditional fixed cycle calibration, and improves the long-term timing accuracy and stability of electric meter.

Description

Technical Field

[0001] This invention relates to the field of electronic timer calibration technology. In particular, it relates to a smart calibration method and system for electricity meter clocks based on multimodal fusion. Background Technology

[0002] With the rapid development of industrial automation and intelligence, the accuracy of electricity meter clocks plays a crucial role in industrial metering and control. The accuracy of electricity meter clocks directly affects the operating efficiency of power systems, energy consumption management, and the safe operation of equipment. In complex industrial environments, errors in electricity meter clocks can lead to inaccurate metering, control signal delays, and even equipment malfunctions, thereby affecting the stability of the entire industrial system.

[0003] The error of an electricity meter clock is influenced by a combination of factors, including long-term environmental changes (such as temperature and humidity), equipment aging, and instantaneous power quality disturbances (such as voltage spikes, drops, and harmonics). These factors intertwine, making it difficult to accurately diagnose the root cause of meter clock deviations through single-dimensional analysis. Therefore, the introduction of multimodal fusion technology becomes particularly necessary. By simultaneously acquiring time signals, voltage data, and environmental status data, the complex environment in which the meter operates can be comprehensively perceived, thus providing a reliable basis for intelligent calibration. This multi-dimensional data acquisition and analysis method can more accurately identify the root cause of deviations, improving the accuracy and reliability of calibration.

[0004] However, most traditional meter clock calibration methods currently employ simplified clock drift models, correcting only through fixed compensation coefficients. This approach ignores the complexity of multi-dimensional errors, which can be categorized into long-term effects (such as environmental aging and equipment fatigue) and instantaneous effects (such as power quality disturbances and mechanical vibrations). Furthermore, electricity meters in industrial settings exhibit significant individual variations due to differences in installation location, service life, and microenvironment. Existing static calibration schemes cannot track these difficult-to-quantify, hidden environmental changes, leading to accelerated degradation of synchronization accuracy under long-term system operation, making it difficult to meet the high long-term stability requirements of industrial metering and control. Summary of the Invention

[0005] To address the problem that traditional electricity meter clock calibration methods, due to model simplification and fixed compensation, cannot cope with multi-dimensional errors and individual meter variations, leading to long-term operational accuracy degradation and difficulty in meeting the high stability requirements of industry, this invention provides solutions in the following aspects.

[0006] In the first aspect, a smart calibration method for an electricity meter clock based on multimodal fusion includes: acquiring electricity meter clock voltage data and environmental state data; preprocessing and extracting features from the voltage data and environmental state data to generate an instantaneous disturbance sequence and a steady-state voltage time series; based on the steady-state voltage time series, combining the statistical characteristics of total harmonic distortion rate, voltage load modulation effect, and the proportion of harmonic exceedance duration, calculating the voltage chronic impact factor of long-term harmonic pollution on the cumulative damage of the electricity meter clock circuit; based on the instantaneous disturbance sequence, quantifying and accumulating the impact stress of each type of transient event (voltage sudden change, instantaneous pulse, and high-frequency oscillation) to obtain the instantaneous damage of short-term voltage disturbance to the electricity meter clock circuit. The study identifies acute voltage influencing factors; based on environmental condition data, it analyzes the environmental acceleration factors that accelerate the aging process of the meter clock circuit, considering the accelerating effect of temperature on component aging and the promoting effect of humidity on failure mechanisms; it determines the corresponding influence coefficients of each influencing factor through regression analysis of historical data of similar meters, and obtains the total lifespan loss by integrating the lifespan loss calculated from each factor and its corresponding coefficient. The study then iteratively updates the dynamic correction cycle of the individual meter damage state using the baseline correction cycle as the initial value and according to the statistical cycle; it constructs a reference group of meters from the same batch and in the same environment, filters similar reference clusters through high-dimensional stress feature clustering, compares the time consistency after correction to determine whether the target meter has any abnormal anomalies, and outputs targeted calibration strategies.

[0007] Preferably, the step of obtaining the instantaneous perturbation sequence includes:

[0008] The extreme value of the root mean square value of the voltage data within a preset time period is calculated as the residual voltage, and the time from the first exceedance to recovery within the threshold is calculated as the duration.

[0009] Wavelet decomposition analysis is used to distinguish three types of events: voltage surge, instantaneous pulse, and high-frequency oscillation. The type, start time, residual voltage, and duration of each type of event are recorded, as well as the dominant frequency and peak value of the high-frequency oscillation event, forming a list of instantaneous disturbance events.

[0010] Preferably, the step of obtaining the steady-state voltage time series includes:

[0011] The voltage data sequence is extracted according to the preset duration, the root mean square value of the voltage data sequence is calculated as the port voltage value, the fast Fourier transform is performed on the voltage data sequence to analyze the spectrum, and the total harmonic distortion rate and the amplitude of each harmonic are extracted.

[0012] By integrating port voltage values, total harmonic distortion rate, and amplitude of each harmonic according to timestamps, a steady-state voltage time series is constructed.

[0013] Preferably, the calculation method for the voltage chronic effect factor includes:

[0014] Extract the original value of total harmonic distortion, the synchronization port voltage value, and the rated voltage of the meter from the steady-state voltage time series within the statistical period;

[0015] The ratio of the port voltage to the rated voltage is used as a load weighting factor to weight the original total harmonic distortion (THD) value. The weighted THD values ​​are then sorted from smallest to largest, and the THD corresponding to the preset quantile is selected.

[0016] The percentage of time points in which the total harmonic distortion rate exceeds the standard limit after statistical weighting is used to obtain the harmonic time exceedance rate.

[0017] Based on the degree of exceeding the preset quantile and the standard limit, and the rate of exceeding the harmonic time limit, the voltage chronic influence factor is calculated, highlighting the nonlinear accelerated aging effect caused by severe exceedance.

[0018] Preferably, the calculation method for the acute voltage impact factor includes:

[0019] Based on the list of instantaneous disturbance events, voltage sudden change stress, instantaneous pulse stress, and high-frequency oscillation stress are calculated separately, and the voltage acute influence factor is obtained by summing the three types of stress.

[0020] Among them, voltage surge stress is calculated by combining the voltage excess, event duration, and fixed reference time obtained from the withstand curve; instantaneous pulse stress is calculated by combining the ratio of pulse peak value to equipment impulse withstand voltage, the ratio of reference rise time to pulse duration, and normalization constant obtained from historical data; high-frequency oscillation stress is calculated by combining the ratio of high-frequency oscillation peak value to high-frequency oscillation withstand voltage and the ratio of high-frequency oscillation duration to fixed reference time.

[0021] Preferably, the calculation method of the environmental acceleration factor includes:

[0022] The average ambient temperature and average ambient humidity within the statistical period are obtained, and the rated operating temperature and rated operating humidity of the meter are introduced as benchmarks. The accelerating effect of temperature on component aging and the accelerating effect of humidity on failure mechanism are obtained. The accelerating effects of temperature and humidity are combined to obtain the environmental acceleration factor that comprehensively reflects the aging acceleration effect of environmental stress on the meter clock circuit.

[0023] Preferably, the process of obtaining the influence coefficient includes:

[0024] The influence coefficients are divided into two groups: the first group corresponds to the chronic voltage influence factor and the environmental acceleration factor, and the second group corresponds to the voltage sudden change stress, instantaneous pulse stress and high-frequency oscillation stress.

[0025] Historical data of similar meters that have been in operation for a long time without severe acute shocks were collected. The first set of coefficients was obtained by using the correction cycle reduction time as the dependent variable and the historical average voltage chronic impact factor and environmental acceleration factor as independent variables through performance degradation regression analysis. Transient event records of similar meters with stable operating environment without major pulse or high-frequency oscillation events were collected. The difference in correction cycle before and after the event was used as the dependent variable and the corresponding stress value was used as the independent variable. The second set of coefficients was obtained by using linear regression.

[0026] Preferably, the iterative generation process of the dynamic correction cycle includes:

[0027] Using the baseline correction cycle as the initial value, the total lifespan loss within the baseline correction cycle is subtracted from the current correction cycle within each statistical cycle to obtain the updated correction cycle.

[0028] Starting from the time when the last meter calibration was completed, the calibration cycle is continuously iterated and updated until the time accumulated from the starting point reaches the current calibration cycle, triggering the meter clock calibration, and using this calibration cycle as the base cycle for the next iteration.

[0029] Secondly, a smart meter clock calibration system based on multimodal fusion includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned smart meter clock calibration method based on multimodal fusion is implemented.

[0030] The present invention has the following effects:

[0031] 1. This invention quantifies three core influencing factors—chronic voltage damage, acute impact, and environmental acceleration—through multimodal fusion, and calibrates the influence coefficients by combining historical data of similar meters. It generates a dynamic correction cycle that precisely matches the individual damage state of the meter in an iterative manner. This solves the problem of insufficient or over-calibration caused by neglecting multidimensional errors and individual differentiation in traditional fixed-cycle calibration, and significantly improves the stability of timing accuracy under long-term operation.

[0032] 2. This invention constructs a differentiated quantitative model by considering the multiple causes of meter clock errors, such as chronic accumulation, instantaneous impact, and environmental acceleration. It achieves accurate characterization of voltage stress through steady-state harmonic characteristics and transient event classification, and quantifies the environmental acceleration effect through a temperature and humidity coupling model. Compared with traditional simplified models, it comprehensively covers the sources of multimodal errors, making the calculation of life loss highly consistent with the actual aging mechanism, and providing reliable data support for calibration strategies. Attached Figure Description

[0033] Figure 1 This is a flowchart of steps S1-S4 in a smart calibration method for electricity meter clock based on multimodal fusion according to an embodiment of the present invention.

[0034] Figure 2 This is a structural block diagram of a smart meter clock calibration system based on multimodal fusion according to an embodiment of the present invention. Detailed Implementation

[0035] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0036] Specific application scenarios: When a large number of smart meters deployed in the smart grid are running in precision component manufacturing workshops under complex industrial environments, frequent power quality disturbances, and long-term non-uniform aging conditions, the defects of traditional clock calibration systems, such as misjudgment of individual health status, inaccurate calibration timing, and sudden increase in cumulative error caused by sudden disturbances, lead to the risk of metering inaccuracy.

[0037] Reference Figure 1 A smart calibration method for electricity meter clocks based on multimodal fusion includes steps S1-S4, as detailed below:

[0038] S1: Acquire the meter clock voltage data and environmental status data, preprocess and extract features from the voltage data and environmental status data, and generate instantaneous disturbance sequences and steady-state voltage time series.

[0039] In this invention, two modal data are acquired for analysis, wherein, mode one: voltage data;

[0040] Specifically, voltage data is preprocessed, and pulse-type outliers are removed using the nearest neighbor comparison method to obtain cleanroom quality data. Based on the cleanroom quality data, the root mean square (RMS) value is calculated every 20ms (one complete voltage cycle), and thresholds are set (0.9 pu for droop and 1.1 pu for swell). When the RMS value exceeds the threshold, secondary analysis is triggered. Next, the residual voltage (minimum for droop and maximum for swell) and event duration (time from exceeding the limit to recovery within the threshold) are calculated for the triggered transient events. Using Daubechies 4 wavelet 5-level decomposition, high-frequency coefficients (levels 1-3) are analyzed to distinguish between high-frequency oscillating transients and instantaneous pulses. Finally, various transient events are recorded. The transient disturbance event list is formed by taking the type, start time, residual voltage, duration, main frequency of high-frequency oscillation (only for high-frequency oscillation events), and peak value of the event (voltage sudden change: sag or swell, instantaneous pulse, high-frequency oscillation) as data. Based on the transient disturbance event list, labels are set for each type of transient event, such as 0 for voltage sudden change, 1 for instantaneous pulse, and 2 for high-frequency oscillation. Feature vectors are constructed in the order of label, start time, residual voltage, duration, main frequency of high-frequency oscillation, and peak value of high-frequency oscillation, and arranged in chronological order to obtain the transient disturbance sequence.

[0041] Voltage data is captured for 200ms (10 consecutive cycles) every minute. The port voltage value for each time period is calculated. The spectrum is analyzed by Fast Fourier Transform to extract the total harmonic distortion rate and harmonic amplitude. The harmonic amplitudes are the amplitudes of the 3rd, 5th, and 7th harmonics. The steady-state voltage time series is obtained by using the timestamp, total harmonic distortion rate, 3rd harmonic amplitude, 5th harmonic amplitude, and 7th harmonic amplitude as the data structure.

[0042] Modal 2: Environmental state data includes temperature data and humidity data.

[0043] The steps for processing environmental status data include: validating the data, removing outliers that significantly exceed the sensor's range, and retaining environmental status data that accurately reflects the crystal oscillator's operating environment.

[0044] S2: Based on steady-state voltage time series, combined with the statistical characteristics of total harmonic distortion rate, voltage load modulation effect, and the proportion of harmonic exceedance duration, calculate the voltage chronic impact factor of long-term harmonic pollution on the cumulative damage of the meter clock circuit; based on transient disturbance series, quantify and accumulate the impact stress of each transient event according to three types of events: voltage sudden change, instantaneous pulse, and high-frequency oscillation, to obtain the voltage acute impact factor of short-term voltage disturbance on the instantaneous damage of the meter clock circuit; based on environmental state data, combined with the accelerating effect of temperature on component aging and the promoting effect of humidity on failure mechanism, analyze the environmental acceleration factor of the aging process acceleration effect of the meter clock circuit.

[0045] The steps for analyzing the effects of chronic fatigue based on steady-state voltage time series analysis are as follows:

[0046] Harmonics increase heat loss and electromagnetic interference in equipment, leading to accelerated aging of components. This aging effect can be quantified by analyzing THD and critical harmonics. The impact of harmonics on equipment varies under different load conditions; the heavier the load, the more significant the harmonic effect. By introducing a load weighting factor, the harmonic stress under actual operating conditions can be more accurately reflected.

[0047] Starting from the moment a meter calibration is completed, and with a preset statistical cycle of 1 day, the following calculations are performed repeatedly until the current accumulated time reaches the calculated calibration cycle, triggering the next calibration operation. The specific operation steps are as follows:

[0048] Based on steady-state voltage time series, the ratio of port voltage value to rated voltage value is calculated as a load weighting factor to characterize the modulation effect of actual voltage level on harmonic damage. When the port voltage value is higher than the rated voltage value, the aging effect of harmonics on the clock circuit will be amplified; conversely, the effect will be weakened, ensuring that the quantization results are consistent with the actual working load scenario.

[0049] The total harmonic distortion rate is multiplied by the load weighting factor to obtain the weighted harmonic distortion rate, which is then sorted to obtain the harmonic distortion rate corresponding to the preset percentile. In this embodiment, the 95th percentile harmonic distortion rate is used. Within the statistical period, the number of data points whose weighted harmonic distortion rate of the environment where the meter is located exceeds the standard limit is marked as the number of exceeding the standard. The ratio of the number of exceeding the standard to the total number of data points is used as the harmonic time exceedance rate. The preset percentile uses the 95th percentile to represent the harmonic level of the worst 5% operating conditions. The core is to focus on the nonlinear accelerated aging caused by the meter's clock circuit. The root cause of extreme harmonic scenarios is directly related to the characteristics of the industrial environment and the damage mechanism of devices: First, the distribution of harmonics in industrial sites is characterized by meeting the standard for most periods and severely exceeding the standard for a few periods. Conventional mean or median values ​​will mask the damage effect of extreme working conditions, while the 95th percentile can accurately identify low-probability, high-intensity exceeding events, i.e., the worst 5% of working conditions; Second, the aging of meter crystal oscillators and timing circuits follows a non-linear cumulative law. The accelerating effect of severe harmonic pollution on aging is far greater than that of slight exceeding the standard. It is necessary to highlight such key influencing factors through quantiles to avoid damage quantification distortion caused by data averaging.

[0050] The difference between the harmonic distortion rate corresponding to the preset percentile and the standard limit is squared and normalized. The product of the normalized result and the harmonic time exceedance rate is used as the chronic fatigue influencing factor.

[0051] Specifically, the influencing factors of chronic fatigue satisfy the following relationship:

[0052] ;

[0053] In the formula, Indicating factors affecting chronic fatigue, This represents the harmonic distortion rate corresponding to the 95th percentile. This indicates the standard limit (5.0%). This indicates the rate of harmonic time exceeding the standard. This indicates the severity of exceeding the standard.

[0054] By analyzing the harmonic levels and harmonic time exceedance rate (the percentage of time exceeding the national standard limit) under the worst 5% operating conditions, combined with the nonlinear aging effect of severely excessive square term amplification, the degree of chronic damage to the meter clock caused by long-term harmonic distortion is finally quantified.

[0055] Specifically, acute shock effects include: voltage surge stress, instantaneous pulse stress, and high-frequency oscillation stress.

[0056] Based on a list of transient disturbance events, various stress values ​​are calculated according to event type, and finally accumulated to obtain a comprehensive voltage acute impact factor, quantifying the impact damage of short-term transient events on the correction cycle. The specific steps are as follows:

[0057] The calculation steps for voltage surge stress include:

[0058] Extract records with the event type "voltage change" (including voltage sags and voltage swells) from the list of transient disturbance events. The core input parameters include:

[0059] Residual voltage: The lowest point (voltage sag scenario) or the highest point (voltage swell scenario) of the effective voltage value during the event; Event duration: The time from the first time the voltage RMS value exceeds the limit to the first time it recovers to within the threshold (unit: ms); Preset parameters: ITIC (Information Technology Industry Council) tolerance curve (as a benchmark for device voltage surge tolerance), reference time (example value is 100ms, used for time normalization).

[0060] Based on the ITIC tolerance curve, the device tolerance voltage threshold for the corresponding scenario is obtained by querying according to the duration of the event; where voltage sag is the lower limit of the tolerance voltage and voltage swell is the upper limit of the tolerance voltage.

[0061] Calculate voltage excess: in sag scenarios Temporary upgrade scenario ;in, The voltage sag corresponding to the duration of the event is taken as the lower limit of the withstand voltage. Indicates the duration of the event. This indicates the residual voltage.

[0062] Events with voltage exceedance greater than 0 are filtered out, indicating that such events fall within the equipment damage zone, while events in the safe zone with voltage exceedance less than or equal to 0 are not counted as damage.

[0063] Extract the residual voltage and duration of the target voltage surge event from the list of transient disturbance events, call the preset ITIC withstand curve and reference time, obtain the corresponding preset ITIC withstand curve based on the residual voltage, obtain the lower limit of the withstand voltage based on the voltage sag based on the time duration, and take the difference between the residual voltage and the lower limit of the withstand voltage sag as the voltage excess.

[0064] The ratio of the event duration to the reference time is used for normalization. The product of the normalized result and the absolute value of the voltage excess is used to obtain the voltage surge stress.

[0065] Specifically, the voltage surge stress satisfies the following relationship:

[0066] ;

[0067] In the formula, Indicates voltage sudden change stress. This indicates that the voltage exceeds the limit. Indicates the duration of the event. Indicates a reference time (example, 100ms) used for normalization.

[0068] The calculation steps for instantaneous pulse stress include:

[0069] Extract records with the event type "transient pulse" from the list of transient disturbance events. The core input parameters include: Pulse peak value: the peak voltage of the transient pulse event (unit: kV); Pulse duration: the duration of the pulse from start to end (unit: μs); Preset parameters: equipment impulse withstand voltage (8kV~12kV, which can be configured according to the meter model and IEC61000-4-5 standard), and reference rise time (example value 1.2μs).

[0070] Based on the specific model of the meter (such as a residential single-phase meter or an industrial three-phase meter), port type (such as a power input port or a signal acquisition port), and the international standards followed (such as IEC 61000-4-5 "Electromagnetic compatibility - Part 4-5: Test and measurement techniques - Surge (impulse) immunity test"), configure the equipment's impulse withstand voltage, with a range of 8kV to 12kV, to ensure consistency with the meter's actual pulse impulse resistance capability.

[0071] Specifically, the instantaneous pulse stress satisfies the following relationship:

[0072] ;

[0073] In the formula, Indicates instantaneous pulse stress. Represents residual voltage. Indicates the device's impulse withstand voltage. Indicates duration, This indicates the reference rise time (for example, 1.2 μs). This represents the median severity score.

[0074] The calculation steps for high-frequency oscillation stress include:

[0075] Extract the discrete frequency points and corresponding test voltages for the high-frequency oscillation withstand test of the clock circuit of the meter in the international standard; select the highest test voltage under the common mode scenario from all the test voltages of the discrete frequency points as the weight normalization benchmark.

[0076] For example, the weight corresponding to the dominant frequency is obtained through the IEC61000-4-18 standard "Damped Oscillating Wave Immunity Test" (existing technology), where the common-mode test includes:

[0077] frequency:

[0078] Standard test voltage (example):

[0079] The weighting factor is proportional to the test voltage at the corresponding frequency point; that is, the expression for the weighting is:

[0080] ;

[0081] In the formula, This represents the high-frequency oscillation frequency weighting factor. Represents frequency The corresponding standard test voltage for humans, This represents the highest test voltage at all frequency points under common mode.

[0082] Records of high-frequency oscillation events are extracted from the transient disturbance sequence. The core input parameters include: high-frequency oscillation peak value: the peak value of the high-frequency oscillation component superimposed on the fundamental wave (unit: V); high-frequency oscillation duration: the duration of the high-frequency oscillation event (unit: ms); high-frequency oscillation dominant frequency: the core frequency of the high-frequency oscillation (unit: kHz, used to assist in verifying the characteristics of the high-frequency oscillation).

[0083] Based on the specific model of the meter, port type, and the international standards followed (such as IEC61000-4-5), configure the high-frequency oscillation withstand voltage, with a value range of 2.5kV~4.0kV; construct a weight-frequency discrete curve based on the discrete frequency point weighting factor; and use linear interpolation to determine the corresponding frequency weighting factor on the discrete curve according to the monitored high-frequency oscillation main frequency.

[0084] Specifically, the high-frequency oscillation stress satisfies the following relationship:

[0085] ;

[0086] in, Indicates high-frequency oscillation stress. This represents the peak value of the high-frequency oscillation voltage. Indicates the high-frequency oscillation withstand voltage. The duration of high-frequency oscillation. For reference time (example, 1ms). This represents the high-frequency oscillation weighting factor.

[0087] The voltage acute influence factor is obtained by summing the voltage sudden change stress, instantaneous pulse stress, and high-frequency oscillation stress.

[0088] The aging process of core components in the clock circuit of an electricity meter (such as crystal oscillators and capacitors) is closely related to ambient temperature and humidity. Increased ambient temperature accelerates atomic diffusion and chemical reactions within these components, while increased humidity exacerbates failure mechanisms such as metal corrosion and insulation degradation. The combined effect of these factors leads to a significantly faster decline in the timing accuracy of the clock circuit. To accurately quantify this environmental acceleration effect, normalized calculations must be performed based on environmental condition data and the meter's rated operating conditions. The specific logic is as follows:

[0089] First, based on real-time ambient temperature and humidity data within the cycle, the average ambient temperature (thermodynamic temperature, unit K) and average relative humidity within that cycle are calculated through averaging to ensure that the data accurately reflects the actual environmental stress level of the clock circuit. Second, the rated operating temperature (unit K) and rated operating relative humidity (unit %RH) specified during the meter design phase are introduced as benchmark conditions. These benchmark parameters are standard operating environment indicators determined by component life tests during the meter design phase, ensuring that the quantitative results of environmental acceleration effects have a unified reference dimension. Finally, based on the above actual environmental parameters and rated benchmark parameters, the temperature acceleration model (Arrhenius equation) and humidity acceleration model (Peck model) are integrated to achieve accurate quantification of the aging acceleration effect of the clock circuit caused by environmental stress, providing a scientific environmental impact input for subsequent total lifespan reduction calculations.

[0090] Specifically, the environmental acceleration factor satisfies the following relationship:

[0091] ;

[0092] In the formula, Indicates environmental acceleration factor, This indicates the activation energy, with an example value of 0.7 eV, which falls within the typical activation energy range for the aging of electronic components. Represents the Boltzmann constant. Indicates the rated operating temperature. This indicates the average temperature during actual use. This indicates the average relative humidity during actual use. Indicates the rated operating relative humidity. This represents the humidity acceleration index, with an example value of 2.7, reflecting the degree to which humidity promotes the failure mechanism. Represented by natural constant An exponential function with base 0.

[0093] It should be noted that the steady-state voltage time series records long-term harmonic distortion, voltage fluctuations and other steady-state anomalies, corresponding to the chronic fatigue effect of voltage on the meter clock; the instantaneous disturbance series records short-term sudden events such as voltage dips, dips, instantaneous pulses and high-frequency oscillations, corresponding to the acute impact of voltage on the meter clock; the environmental temperature and humidity data reflect the stress state of the meter's working environment, and its anomalies can accelerate the cumulative effect of chronic fatigue, which needs to be quantified in conjunction with power quality stress.

[0094] S3: By using regression analysis of historical data of similar meters, the influence coefficients of each influencing factor are determined. The life loss calculated by integrating each factor and its corresponding coefficient is obtained to obtain the total life loss. The dynamic correction cycle of the individual meter damage status is generated by iteratively updating the baseline correction cycle as the initial value according to the statistical cycle.

[0095] After quantifying the severity of voltage chronic fatigue factors, environmental acceleration factors, and voltage acute impact factors (including voltage sudden change stress, instantaneous pulse stress, and high-frequency oscillation stress), it is necessary to calibrate the coefficients corresponding to each influencing factor using historical operating data of similar meters in order to accurately quantify the degree of influence of different influencing factors on the meter clock correction cycle.

[0096] Considering that the damage mechanism of acute impact events is weakly correlated with environmental stress, and that the indirect influence of the environment on acute impact is negligible, this invention calibrates the influence coefficients in two independent groups to ensure the accuracy of the coefficient calibration:

[0097] Group 1: Coefficients corresponding to the voltage chronic fatigue factor The coefficients corresponding to environmental acceleration factors Both contribute to the long-term chronic degradation of the main conductor clock; the second group: the coefficient corresponding to voltage surge stress. The coefficient corresponding to instantaneous pulse stress The coefficient corresponding to high-frequency oscillation stress These three correspond to the instantaneous injuries caused by different types of acute impacts.

[0098] Historical data of similar meters of the same model as the target meter that have been running stably for 3-5 years were collected. Samples without severe acute shock records were screened, and their clock degradation was mainly caused by chronic fatigue and environmental acceleration.

[0099] Independent variables: The cumulative reduction time of the correction cycle for each sample meter throughout its entire lifespan, directly reflecting the degree of clock degradation; Dependent variables: Calculated historical average voltage chronic fatigue factor and historical average environmental acceleration factor for each sample meter; A performance degradation regression analysis model is used to establish the mapping relationship between the dependent and independent variables, and the results are obtained through fitting. and .

[0100] Collect operating data of similar meters of the same model as the target meter that meet the following conditions: no major pulse / high-frequency oscillation events were recorded between the two calibrations, and the operating environment was relatively stable.

[0101] Independent variables: Voltage surge event records extracted from sample meters, calculating the voltage excess and corresponding voltage surge stress for each event; Dependent variable: The difference between the current correction cycle and the previous correction cycle after a voltage surge event occurs in the sample meters, reflecting the direct impact of the surge event on the correction cycle; A linear regression model is used for fitting, yielding... .

[0102] Independent variables: instantaneous pulse stress, high-frequency oscillation stress; dependent variable: the change in the correction period after the corresponding event occurs. These are successively fitted using a linear regression model to obtain... and .

[0103] Multiplying each influencing factor by its corresponding calibration coefficient yields the specific reduction in the lifespan of the meter's clock due to various types of damage:

[0104] Since environmental acceleration factors amplify the damage effects of chronic fatigue, a coupled calculation method of "chronic fatigue loss × (environmental acceleration loss + 1)" is adopted, and various acute impact losses are superimposed to obtain the total life loss within the statistical period. The formula logic is as follows:

[0105] ;

[0106] in, This indicates the total lifespan reduction of the meter caused by power quality issues within the statistical period. This indicates the chronic voltage fatigue life reduction of the meter caused by power quality issues within the statistical period. This indicates the environmental acceleration factor of the electricity meter during the statistical period. This indicates the reduction in the lifespan of the meter caused by sudden voltage changes due to sudden events within the statistical period. This indicates the reduction in the instantaneous pulse lifespan of the meter caused by pulse events within the statistical period. This indicates the reduction in the lifespan of the meter caused by high-frequency oscillation events within the statistical period.

[0107] The initial calibration cycle is set at the factory of the electricity meter, and the calibration cycle is updated iteratively according to the following logic:

[0108] At the end of each statistical period, the total life loss within that period is calculated.

[0109] Specifically, the total lifetime loss satisfies the following relationship:

[0110] ;

[0111] In the formula, Indicates the first Next correction cycle Indicates the first Next correction cycle This represents the sum of the total lifetime loss for each statistical period before correction was performed.

[0112] The total lifespan loss is calculated from the last calibration completion time to the end of the current statistical cycle. The cumulative running time since the last calibration completion time is monitored in real time. When the cumulative running time equals the currently updated calibration cycle, the meter clock calibration operation is triggered, and the calibration cycle is used as the initial reference cycle for the next iteration. The above steps are repeated to achieve dynamic adaptive updating of the calibration cycle, ensuring that the calibration cycle accurately matches the actual damage state of the meter clock and avoiding undercalibration or overcalibration caused by fixed-cycle calibration.

[0113] Through the above coefficient calibration and iterative calculation, the dynamic correction cycle of the electricity meter clock is generated, so that the electricity meter avoids a 1-second time anomaly at the end of the correction cycle, ensuring the timekeeping accuracy and operational stability of the electricity meter.

[0114] S4: Construct a reference group of electricity meters in the same batch and environment, screen similar reference clusters through high-dimensional stress feature clustering, compare the time consistency after correction to determine whether there are any abnormal anomalies in the target electricity meter and output targeted calibration strategies.

[0115] To verify the rationality of the dynamic correction cycle and investigate potential influencing factors not covered by the algorithm (such as manufacturing quality differences, lack of microenvironment monitoring, etc.), cluster analysis is needed to screen reference groups with similar stress trajectories to the target meter. This avoids the drawback of directly comparing correction cycles while ignoring historical stress differences in the equipment. The core logic is as follows: based on multi-dimensional stress data from meters in the same batch and operating environment, a high-dimensional feature matrix is ​​constructed according to a unified time window. The reference meter cluster with the highest similarity is screened through distance quantization, and then the consistency of time after correction is compared to determine whether the target meter has any abnormal anomalies.

[0116] Select meters from the same batch and operating environment as the target meters to construct a reference group (exemplary quantity: 50 meters). Process the historical data of the target meters and each meter in the reference group after commissioning in segments according to a fixed time window (exemplary: 1 month) to ensure that the feature extraction dimensions of each meter are consistent with the time benchmark.

[0117] For each fixed time window, calculate the following characteristic parameters: the cumulative value of environmental accelerated loss corresponding to all statistical periods (24 hours) within the time window; the cumulative value of voltage chronic fatigue life loss corresponding to all statistical periods within the time window; the cumulative value of total life loss for all statistical periods within the time window; the average temperature and humidity of the environment where the meter is located within the time window; and construct a high-dimensional stress feature vector for that time window.

[0118] High-dimensional feature matrix construction for extracting the target meter and continuous features of each meter in the reference group High-dimensional stress feature vectors for each time window (exemplarily 6 months) are used to construct a high-dimensional feature matrix for each electricity meter, where at least the vectors are continuous. A fixed time window, with the matrix dimension consistent with the number of time windows and the number of feature parameters.

[0119] Calculate the Euclidean distance between the feature matrix of the target meter and the feature matrix of each meter in the reference group to quantify the similarity of their historical stress trajectories. The smaller the distance, the higher the similarity.

[0120] Sort all Euclidean distances from largest to smallest to obtain a distance sequence. Calculate the first quartile of the sequence (corresponding to the upper boundary of the 25% of meters with smaller distances) and the interquartile range (reflecting the distance dispersion of the middle 50% of meters).

[0121] The screening threshold is set as the sum of the distance value corresponding to the first quartile and the distance value corresponding to the interquartile range (a common boundary for identifying moderate outliers in statistics, taking into account both similarity and inclusiveness). Reference meters with Euclidean distances less than this threshold are grouped into a similar reference cluster of the target meter.

[0122] Extract the dynamic correction cycle of all meters in the similar reference cluster, calibrate the clock according to their respective dynamic correction cycles, and record the calibrated time data; count the mode of the calibrated time in the similar reference cluster (i.e., the time value that appears most frequently), and compare the consistency between the time of the target meter after calibration according to its own dynamic correction cycle and this mode; if the time of the target meter after calibration is inconsistent with the mode of the reference cluster time, it is determined that the target meter has an unusual anomaly (the anomaly may be due to manufacturing quality differences, unmonitored micro-environmental influences, or other factors not covered by the algorithm); if they are consistent, it is determined to be a regular anomaly, indicating that the dynamic correction cycle algorithm has fully covered the main influencing factors, and after calibration according to the currently calculated correction cycle, the meter will only have an error of 1 second, which meets the design expectations.

[0123] By classifying high-dimensional stress features, a reference group most similar to the aging trajectory of the target meter is accurately selected, avoiding the limitations of comparing a single parameter. Based on the consistency judgment of the mode after calibration, regular anomalies within the algorithm's coverage area can be effectively distinguished from non-regular anomalies caused by potential factors. For non-regular anomalies, users can formulate targeted handling strategies based on actual scenarios (such as replacing meters with manufacturing defects, or addressing local microenvironmental issues), providing a clear basis for accurate meter maintenance and further ensuring the stability and accuracy of the smart grid metering system.

[0124] This invention also provides a smart calibration system for electricity meter clocks based on multimodal fusion. For example... Figure 2 As shown, the system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a smart clock calibration method for electricity meters based on multimodal fusion according to the first aspect of the present invention. The system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, the settings and functions of which are known in the art and will not be described in detail here.

[0125] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A smart calibration method for electricity meter clocks based on multimodal fusion, characterized in that, include: Acquire meter clock voltage data and environmental status data, preprocess and extract features from the voltage data and environmental status data, and generate instantaneous disturbance sequence and steady-state voltage time series; Based on steady-state voltage time series, combined with the statistical characteristics of total harmonic distortion rate, voltage load modulation effect, and the proportion of harmonic exceedance duration, the chronic voltage impact factor of long-term harmonic pollution on the cumulative damage of the meter clock circuit is calculated; based on transient disturbance series, the impact stress of each type of transient event is quantified and accumulated according to three types of events: voltage sudden change, instantaneous pulse, and high-frequency oscillation, to obtain the acute voltage impact factor of short-term voltage disturbance on the instantaneous damage of the meter clock circuit. Based on environmental condition data, and combining the accelerating effect of temperature on component aging and the promoting effect of humidity on failure mechanism, we analyze the environmental acceleration factors that accelerate the aging process of the meter clock circuit. By using regression analysis of historical data of similar electricity meters, the influence coefficients of each influencing factor are determined. The life loss calculated by integrating each factor and its corresponding coefficient is obtained as the total life loss. The dynamic correction cycle of the individual damage status of the electricity meter is generated by iteratively updating the baseline correction cycle as the initial value according to the statistical cycle. Construct a reference group of electricity meters from the same batch and in the same environment. Select similar reference clusters through high-dimensional stress feature clustering. Compare the time consistency after correction to determine whether there are any abnormal anomalies in the target electricity meter and output a targeted calibration strategy. The calculation method for the voltage chronic influence factor includes: Extract the original value of total harmonic distortion, the synchronization port voltage value, and the rated voltage of the meter from the steady-state voltage time series within the statistical period; The ratio of the port voltage to the rated voltage is used as a load weighting factor to weight the original total harmonic distortion (THD) value. The weighted THD values ​​are then sorted from smallest to largest, and the THD corresponding to the preset quantile is selected. The percentage of time points in which the total harmonic distortion rate exceeds the standard limit after statistical weighting is used to obtain the harmonic time exceedance rate. Based on the degree of exceedance of preset quantiles and standard limits, and the rate of exceedance of harmonic time, the voltage chronic influence factor is calculated, highlighting the nonlinear accelerated aging effect caused by severe exceedance. The calculation method for the acute impact factor of voltage includes: Based on the list of instantaneous disturbance events, voltage sudden change stress, instantaneous pulse stress, and high-frequency oscillation stress are calculated separately, and the voltage acute influence factor is obtained by summing the three types of stress. Among them, voltage surge stress is calculated by combining the voltage excess, event duration, and fixed reference time obtained from the withstand curve; instantaneous pulse stress is calculated by combining the ratio of pulse peak value to equipment impulse withstand voltage, the ratio of reference rise time to pulse duration, and normalization constant obtained from historical data; high-frequency oscillation stress is calculated by combining the ratio of high-frequency oscillation peak value to high-frequency oscillation withstand voltage and the ratio of high-frequency oscillation duration to fixed reference time. The calculation method for the environmental acceleration factor includes: The average ambient temperature and average ambient humidity within the statistical period are obtained, and the rated operating temperature and rated operating humidity of the meter are introduced as benchmarks. The effects of temperature on component aging and humidity on failure mechanism are obtained. The effects of temperature acceleration and humidity acceleration are combined to obtain an environmental acceleration factor that comprehensively reflects the environmental stress on the aging acceleration effect of the meter clock circuit. The process of obtaining the influence coefficient includes: The influence coefficients are divided into two groups: the first group corresponds to the chronic voltage influence factor and the environmental acceleration factor, and the second group corresponds to the voltage sudden change stress, instantaneous pulse stress and high-frequency oscillation stress. Historical data of similar meters that have been in operation for a long time without severe acute shocks were collected. The first set of coefficients was obtained by using the correction cycle reduction time as the dependent variable and the historical average voltage chronic impact factor and environmental acceleration factor as independent variables through performance degradation regression analysis. Transient event records of similar meters with stable operating environment without major pulse or high-frequency oscillation events were collected. The difference in correction cycle before and after the event was used as the dependent variable and the corresponding stress value was used as the independent variable. The second set of coefficients was obtained by using linear regression.

2. The method for intelligent calibration of a meter clock based on multimodal fusion according to claim 1, characterized in that, The steps for obtaining the instantaneous perturbation sequence include: The extreme value of the root mean square value of the voltage data within a preset time period is calculated as the residual voltage, and the time from the first exceedance to recovery within the threshold is calculated as the duration. Wavelet decomposition analysis is used to distinguish three types of events: voltage surge, instantaneous pulse, and high-frequency oscillation. The type, start time, residual voltage, and duration of each type of event are recorded, as well as the dominant frequency and peak value of the high-frequency oscillation event, forming a list of instantaneous disturbance events.

3. The method for intelligent calibration of a meter clock based on multimodal fusion according to claim 1, characterized in that, The steps for obtaining the steady-state voltage time series include: The voltage data sequence is extracted according to the preset duration, the root mean square value of the voltage data sequence is calculated as the port voltage value, the fast Fourier transform is performed on the voltage data sequence to analyze the spectrum, and the total harmonic distortion rate and the amplitude of each harmonic are extracted. By integrating port voltage values, total harmonic distortion rate, and amplitude of each harmonic according to timestamps, a steady-state voltage time series is constructed.

4. The intelligent calibration method for electricity meter clocks based on multimodal fusion according to claim 1, characterized in that, The iterative generation process of the dynamic correction cycle includes: Using the baseline correction cycle as the initial value, the total lifespan loss within the baseline correction cycle is subtracted from the current correction cycle within each statistical cycle to obtain the updated correction cycle; Starting from the time when the last meter calibration was completed, the calibration cycle is continuously iterated and updated until the time accumulated from the starting point reaches the current calibration cycle, triggering the meter clock calibration, and using this calibration cycle as the base cycle for the next iteration.

5. A smart calibration system for electricity meter clocks based on multimodal fusion, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the smart calibration method for electricity meter clocks based on multimodal fusion according to any one of claims 1-4.

Images