Method for detecting abnormal metering performance of intelligent electric energy meter

Abstract

The invention discloses a method for detecting abnormal metering performance of an intelligent electric energy meter, which belongs to the technical field of electric energy metering equipment and comprises the following steps of: 1, establishing a reference response curved surface of parasitic parameters of a voltage sampling resistor and a current transformer in the electric energy meter relative to temperature and humidity; step 2, during the operation period of the electric energy meter, acquiring the real-time measurement value of the parasitic parameter in situ; and step 3, matching the real-time measurement value with the reference response curved surface. According to the method, an active matching model library can be constructed by fusing an underlying physical model of a component and a mapping rule set for a complex environment with sudden change of plateau outdoor temperature and humidity, the internal mechanism that parasitic parameters change along with temperature and humidity is explained from the physical essence level, and the limitation of a pure data statistical method is made up; the dynamic error judgment threshold value can adapt to environmental stress changes in real time, and the accuracy of metering performance anomaly detection and the complex environment adaptability are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of electrical energy metering equipment technology, and more specifically, to a method for detecting abnormal metering performance of a smart energy meter. Background Technology

[0002] In high-altitude outdoor settings, distribution boxes often face extreme temperature and humidity fluctuations, with significant day-night temperature differences and high and volatile air humidity. The metering accuracy of smart meters directly depends on the performance stability of core components such as voltage sampling resistors and current transformers. However, the extreme temperature and humidity environment at high altitudes can easily cause performance degradation and parameter drift of these components, leading to metering deviations and affecting the accuracy and fairness of electricity metering.

[0003] Current mainstream methods for detecting anomalies in smart meter metering performance mostly use fixed metering error thresholds as the basis for anomaly judgment. These methods lack adaptability to the special outdoor environments of high-altitude areas. Such detection methods do not establish a dynamic correlation mechanism between temperature and humidity parameters and component performance degradation. They cannot accurately identify metering deviations caused by environmental factors and real anomalies caused by component aging. In scenarios with sudden changes in temperature and humidity, metering deviations caused by component parameter drift are often misjudged as normal fluctuations, while substantial metering anomalies caused by long-term component aging are missed because they do not reach the fixed threshold. This results in a significant decrease in detection sensitivity and makes it difficult to meet the accuracy requirements for metering performance testing in high-altitude outdoor scenarios. Summary of the Invention

[0004] To address the problems existing in the prior art, the purpose of this invention is to provide a method for detecting abnormal metering performance of smart energy meters. This method can be used in complex environments with sudden changes in temperature and humidity, such as high-altitude outdoor environments. By integrating the underlying physical model of the components with the mapping rule set to construct an active matching model library, it explains the intrinsic mechanism of parasitic parameters changing with temperature and humidity from a physical perspective, thus overcoming the limitations of pure data statistical methods. The dynamic error judgment threshold can be adapted to changes in environmental stress in real time, effectively improving the accuracy of metering performance anomaly detection and adaptability to complex environments.

[0005] To solve the above problems, the present invention adopts the following technical solution:

[0006] A method for detecting abnormal metering performance of a smart energy meter includes the following steps:

[0007] Step 1: Establish the reference response surface of the parasitic parameters of the voltage sampling resistor and current transformer in the electricity meter with respect to temperature and humidity.

[0008] Step 2: During the operation of the electricity meter, the real-time measurement value of the parasitic parameter is acquired in situ.

[0009] Step 3: Match the real-time measured values ​​with the reference response surface to determine the theoretical performance state under the current environmental stress; based on the theoretical performance state and the preset component stress life correlation model, calculate and generate the expected performance degradation trajectory for a predetermined future period.

[0010] Step 4: Based on the expected performance degradation trajectory, determine the theoretical measurement error component caused by environmental stress; combine the theoretical measurement error component with the legal maximum permissible error to obtain a dynamic error judgment threshold.

[0011] Step 5: If the real-time metering error of the electricity meter exceeds the dynamic error judgment threshold, an environmental stress alarm is triggered; if the deviation of the real-time measurement value from the expected performance degradation trajectory exceeds a predetermined number of times and exceeds the preset process deviation threshold, a component aging alarm is triggered.

[0012] Step 6: When the component aging alarm is triggered, obtain the actual performance trajectory of the parasitic parameters under multiple historical environmental cycles, and analyze the irreversible offset characteristics of the actual performance trajectory relative to the reference response surface.

[0013] Step 7: Periodically calibrate the energy meter on-site using an external standard, and use the calibration results to verify and calibrate the benchmark of the real-time measurement value obtained in situ.

[0014] Furthermore, a baseline response surface for its parasitic parameters with respect to temperature and humidity is established, including:

[0015] Step 11: Design and execute a time-series change path for temperature and humidity that simulates the sudden changes in outdoor conditions at high altitudes. Simultaneously capture the transient fluctuation sequence of the resistance value of the voltage sampling resistor inside the electricity meter and the relaxation curve of the loss tangent of the current transformer to obtain the dynamic response sequence of the path.

[0016] Step 12: Based on the path dynamic response sequence, identify and extract the cooperative response mode determined by the temperature change rate, humidity change rate and their phase difference, and establish a mapping rule set from the temperature and humidity change path feature vector to the parasitic parameter dynamic response feature vector.

[0017] Step 13: Combine the mapping rule set with the underlying physical model of the components to construct an active matching model library. The active matching model library is used to generate the corresponding dynamic evolution trajectory of parasitic parameters based on temperature and humidity time series data.

[0018] Step 14: Run the active matching model library, input virtual environmental time series data covering the full temperature and humidity range and having randomly changing path characteristics, batch calculate the dynamic trajectory of parasitic parameters, extract the measured values ​​and characteristic peak points after reaching steady state under a specific temperature and humidity combination, and form a dependency model library.

[0019] Furthermore, real-time measurements of parasitic parameters are obtained, including:

[0020] Step 21: Analyze the metering cycle, determine the idle time window for analog-to-digital conversion and digital processing, plan the microsecond-level deterministic detection time slots, and allocate exclusive and periodically repeating specific time slots for the voltage sampling resistor and current transformer in the energy meter, and generate a time slot mapping table.

[0021] Step 22: For the exclusive time slot, design and inject an orthogonal characteristic detection signal that has a predetermined correspondence with the predetermined parasitic parameter changes of the component. The signal is orthogonal to the main metering signal in the time domain and frequency domain.

[0022] Step 23: In the same time slot of injecting the orthogonal feature detection signal, synchronously acquire the response waveform of the detected component, and perform coherent demodulation processing with the injected signal as a reference to extract the complex number of the baseband response.

[0023] Step 24: Based on the complex baseband response, compensation is performed using a pre-stored parasitic effect model of the detection loop, and the real-time measured value of the parasitic parameter is calculated according to the excitation response transfer function of the component.

[0024] Furthermore, the theoretical performance state under the current environmental stress is determined, including:

[0025] Step 31: Based on the dependent model library, the feature vector of temperature and humidity change path is discretized into a set of actions, and the feature vector of parasitic parameter dynamic response is discretized into a set of states. The state transition probability matrix and observation likelihood model are statistically generated.

[0026] Step 32: Obtain the real-time measurement sequence of historical parasitic parameters and the synchronized temperature and humidity data sequence. Convert the temperature and humidity data sequence into the action sequence. Based on the state transition probability matrix and the observation likelihood model, decode the action sequence and the measurement sequence to obtain the hidden state sequence and its terminal state as the current theoretical performance state.

[0027] Furthermore, the expected performance degradation trajectory for future predetermined periods is calculated, including:

[0028] Step 33: Starting from the terminal state, combining the action distribution of future temperature and humidity prediction data transformation and the state transition probability matrix, perform Monte Carlo simulation, randomly generate multiple state transition paths, and call the active matching model library to calculate the performance degradation of each path to obtain a set of random trajectories.

[0029] Step 34: Perform statistical analysis on the random trajectory set, calculate the mean value of performance degradation at each future time point and the preset confidence interval, and form the expected performance degradation trajectory.

[0030] Furthermore, a dynamic error determination threshold is obtained, including:

[0031] Step 41: For the expected performance degradation trajectory, establish an inverse mapping function to inversely analyze the trajectory into a time-varying fault mode parameter sequence;

[0032] Step 42: Based on the fault parameter sequence, run the pre-stored metering link error transfer function to obtain the expected error trend line and the error risk envelope.

[0033] Step 43: Analyze the time-domain characteristics of the error risk envelope, and apply an asymmetric compression function to process it based on the correlation between its volatility and trend, thereby generating a dynamic error component baseline.

[0034] Step 44: Perform asymmetric synthesis of the dynamic error component baseline and the statutory maximum permissible error under hard limiting to determine the dynamic error judgment threshold.

[0035] Furthermore, the analysis of the irreversible offset characteristics of the actual performance trajectory relative to the reference response surface includes:

[0036] Step 61: When the component aging alarm is triggered, obtain the time series of actual measured values ​​of parasitic parameters within the historical environmental cycle, query the dependency model library based on the temperature and humidity values ​​at each time point in the series to obtain the ideal response value series, and vector subtract the actual measured value series from the ideal response value series to obtain the aging and noise residual series.

[0037] Step 62: Extract the aging trend component from the aging and noise residual sequence, identify the pattern features of the trend component, and obtain the aging pattern feature vector.

[0038] Furthermore, it also includes:

[0039] Step 63: Select common temperature and humidity state points in multiple historical environmental cycles as anchor points. For each anchor point, calculate its aging residual in different historical cycles and the current cycle to form a historical evolution dataset of aging residual for each anchor point.

[0040] Step 64: Analyze the aging residual historical evolution dataset, examine its monotonicity or cumulativeity, calculate the offset vector of the current aging residual relative to the historical value, and combine it with the aging mode feature vector to confirm the aging type and quantization offset.

[0041] Furthermore, the energy meter is periodically calibrated on-site using an external standard, and the calibration results are used to verify and calibrate the benchmark of the real-time measurement values ​​acquired in situ, including:

[0042] Step 71: During the normal power supply and metering of the electricity meter, apply a differential test signal with an amplitude lower than a predetermined proportion of the rated electrical quantity corresponding to the current operating condition of the electricity meter, and record the difference between the standard value and the sampling response to obtain a micro-differential standard value-sampling response difference dataset.

[0043] Step 72: Based on the dataset, combined with the temperature and humidity data during calibration and the real-time measurement values ​​of the parasitic parameters, a hierarchical error recursive decomposition model is constructed. The overall error is decomposed and the calibration chain error and the expected component error component calculated by the real-time measurement values ​​of the parasitic parameters through the metering link error transfer function are deducted in sequence to obtain the non-component error component.

[0044] Step 73: Correct the internal compensation parameters of the energy meter using the non-component error components;

[0045] The parasitic parameter measurement values ​​used to calculate the expected component error components are compared with the equivalent parasitic parameters derived by error decomposition to calculate the deviation vector of the parasitic parameter measurement.

[0046] The deviation vector is used to perform feedback correction on the parameter reconstruction model and the signal control logic.

[0047] Furthermore, it also includes:

[0048] Step 74: Apply the corrected parasitic parameter measurement, re-execute the steps of matching the real-time measurement value with the dependent model library and generating the dynamic error judgment threshold, and analyze the over-limit situation of real-time measurement error in the new period. At the same time, calculate the residual sequence between the parasitic parameter measurement value and the predicted value of the dependent model library in this period and analyze its statistical characteristics.

[0049] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0050] (1) This solution is designed for the complex environment of sudden temperature and humidity changes in the plateau. It integrates the underlying physical model of the components with the mapping rule set to build an active matching model library. It explains the intrinsic mechanism of parasitic parameters changing with temperature and humidity from the physical essence level, makes up for the limitations of pure data statistical methods, and the dynamic error judgment threshold can adapt to environmental stress changes in real time, effectively improving the accuracy of metrological performance anomaly detection and the adaptability to complex environments.

[0051] (2) This solution realizes the classification and triggering of environmental stress alarm and component aging alarm. By using anchor point analysis and aging residual evolution dataset to verify the irreversible aging characteristics, it accurately distinguishes the causes of abnormalities, avoids the problem location deviation caused by a single alarm, provides maintenance personnel with a clear direction for maintenance, significantly improves maintenance efficiency, and reduces the time and cost losses caused by blind maintenance.

[0052] (3) This solution uses non-component error components to accurately correct internal compensation parameters, optimizes the model and control logic through deviation vector feedback, continuously calibrates the metering benchmark, ensures the metering accuracy of the electricity meter in long-term complex environment operation, and extends the stable working cycle of the equipment. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0054] Figure 1 This is a flowchart of the method for detecting abnormal metering performance of smart energy meters according to the present invention.

[0055] Figure 2 This is a flowchart of the dynamic threshold generation process in step 4 of the present invention;

[0056] Figure 3 This is a flowchart of the aging analysis in step 6 of the present invention;

[0057] Figure 4 This is a flowchart of the calibration and correction process in step 7 of the present invention. Detailed Implementation

[0058] 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 only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0059] Please see Figures 1 to 4 A method for detecting abnormal metering performance of a smart energy meter includes the following steps:

[0060] Step 1: Establish the reference response surface of the parasitic parameters of the voltage sampling resistor and current transformer in the electricity meter with respect to temperature and humidity.

[0061] The process of establishing the reference response surface of its parasitic parameters with respect to temperature and humidity involves the following steps: Step 11, designing and executing a time-series change path for temperature and humidity that simulates the sudden changes in outdoor conditions at high altitudes, simultaneously capturing the transient fluctuation sequence of the resistance value of the voltage sampling resistor inside the energy meter and the relaxation curve of the loss tangent of the current transformer, to obtain the path dynamic response sequence. The specific operations are as follows:

[0062] To replicate the extreme environmental characteristics of rapid temperature and humidity changes in high-altitude outdoor environments, a time-series temperature and humidity variation path must first be designed to fit the scenario. This path must encompass significant diurnal temperature differences, drastic humidity fluctuations, and abrupt changes in temperature and humidity, while also considering environmental variation patterns across different seasons and time periods to ensure consistency between the simulated environment and real-world conditions. During the execution of this time-series variation path, a data acquisition mechanism is simultaneously activated. For the voltage sampling resistor within the electricity meter, the transient fluctuation sequence of its resistance value with changes in temperature and humidity is collected in real time. This sequence accurately reflects the dynamic response characteristics of the resistance value under rapid environmental stress changes. For the current transformer, the relaxation curve of its loss tangent is captured. This curve effectively characterizes the hysteretic response of the transformer core loss and winding parameters with changes in temperature and humidity. Through the synchronous acquisition and integration of these two types of data, a path dynamic response sequence is ultimately formed. This sequence fully records the dynamic evolution of parasitic parameters of the voltage sampling resistor and the current transformer under simulated extreme temperature and humidity conditions in high-altitude environments.

[0063] Step 12: Based on the path dynamic response sequence, identify and extract the cooperative response pattern determined by the rate of temperature change, the rate of humidity change, and the phase difference between the two, and establish a mapping rule set from the path feature vector of temperature and humidity change to the feature vector of the dynamic response of parasitic parameters. The specific operations are as follows:

[0064] The temperature and humidity sensitive physical model is constructed based on the "Influence of Environmental Humidity on Resistance" in Baidu Wenku and the error assessment approach for high-altitude environments in patent CN121208742A. It is established by combining the material characteristics of the voltage sampling resistor, such as the temperature sensitivity of the metal film resistor and the humidity adsorption characteristics of the ceramic substrate. The model expression is as follows: ;

[0065] In the formula, R( H) represents the current temperature The real-time resistance value of the voltage sampling resistor under the current humidity H; This is the reference resistance value under standard temperature and humidity conditions; The standard reference temperature is 25℃. The standard reference humidity is 60%RH. The temperature coefficient of resistance is determined by the composition of the resistive material, and its typical value is [value missing]. / ℃ level; The resistance humidity coefficient is determined by the humidity adsorption characteristics of the resistive ceramic matrix, with a typical value of [value missing]. The magnitude is in the range of / %RH; The temperature and humidity cross-coupling coefficient is used to adapt the influence of the synergistic effect of temperature and humidity on the resistance value under the extreme environment of plateau. It is obtained by fitting multiple sets of plateau field experimental data. The formula first establishes a linear temperature response relationship based on the temperature effect of resistance, then introduces the influence of humidity on the adsorption loss of the resistive matrix to construct an exponential humidity response term, and finally adds a temperature and humidity cross term to adapt to the synergistic effect under the sudden change environment of plateau. The coefficient is determined by fitting multiple sets of plateau temperature and humidity experimental data to ensure that the model can accurately characterize the dynamic changes of the parasitic parameters (resistance value) of the resistor with temperature and humidity.

[0066] The electromagnetic loss physical model was constructed with reference to the loss separation theory in "Research on the Loss Characteristics of Non-Sinusoidal Excitation High-Frequency Transformer Core Based on Bertotti Model" and the electromagnetic-temperature field coupling approach in patent CN117421914A. It was optimized by combining the structural characteristics of the current transformer, such as core material, number of winding turns, and winding resistance and capacitance. The model expression is as follows: ;

[0067] In the formula, Current temperature The total electromagnetic loss of the current transformer under the current humidity H, magnetic flux density B, and operating frequency f; This is the hysteresis loss component; This represents the eddy current loss component. The remaining loss components are calculated using the following formulas: ,in is the hysteresis loss coefficient, and n is the Steinmetz exponent. Both are determined by the magnetic properties of the current transformer core material. ,in It is the eddy current loss coefficient, which is related to the core thickness, the resistivity of the core material, and the winding structure of the current transformer. ,in The residual loss factor is determined by the microscopic magnetic domain structure of the core material, while a temperature and humidity correction factor is also introduced. ;

[0068] in, This is the loss correction factor under the current temperature and humidity conditions; This is a temperature loss correction factor, used to characterize the effect of temperature changes on the resistivity and permeability of the iron core; This is a humidity loss correction factor, used to characterize the effect of humidity changes on the insulation performance of the iron core surface. and All losses were obtained through accelerated aging experiments of current transformers in high-altitude environments, and the final loss model is as follows: ;

[0069] Its derivation is based on the Bertotti model, which is divided into three types of loss components. The loss coefficient is adjusted in combination with the power frequency operating scenario of the current transformer (f=50Hz). Then, referring to the electromagnetic-temperature field coupling idea, a correction factor for the influence of temperature and humidity on the resistivity and permeability of the iron core is introduced, so that the model can accurately characterize the relationship between the parasitic parameters of the transformer (loss tangent) and temperature, humidity and magnetic flux density.

[0070] The public release of the mapping rule set is based on the path dynamic response sequence constructed in step 11, and includes three parts: feature vector definition, mapping relationship modeling, and rule verification. First, two types of feature vector dimensions are defined: temperature and humidity change path feature vectors. The meanings of each symbol are as follows: For the rate of temperature change, The rate of change of humidity. The phase difference is the change in temperature and humidity. The average temperature over the period is... Average humidity over a period of time; dynamic response feature vector of parasitic parameters ,in, This represents the peak offset of the voltage sampling resistor value. This represents the steady-state offset of the voltage sampling resistor value. The relaxation time of the voltage sampling resistor response. This represents the peak offset of the loss tangent of the current transformer. This represents the steady-state offset of the loss tangent of the current transformer. Let be the relaxation time of the current transformer's loss tangent response. Then, based on the random forest algorithm, using the path dynamic response sequence as training samples, construct... The mapping relationship model outputs a mapping rule set, including feature importance ranking, decision tree branching rules, and prediction error threshold. Finally, the rule set is verified through 10 independent high-altitude temperature and humidity change experiments to ensure that the mapping accuracy is ≥95% and the error is ≤3%.

[0071] After completing the above model and rule set construction, the mapping rule set is deeply integrated with the underlying physical model of the components to build an active matching model library. During the integration process, the statistical regularity of the mapping rule set is used as a constraint condition for the core parameters of the physical model, such as... , , , The model is optimized and calibrated by minimizing the error between the parasitic parameter trajectory output by the physical model and the trajectory predicted by the mapping rule set using the least squares method. This ensures that the model conforms to physical principles and accurately matches the parasitic parameter response characteristics in the actual environment. The final active matching model library has the following capabilities: when inputting temperature and humidity time series data, the model library can determine the feature vector X of the temperature and humidity change path based on the mapping rule set, and then calculate the corresponding dynamic evolution trajectory of parasitic parameters through the optimized underlying physical model of the components, thereby realizing the active prediction and matching of the changing trend of parasitic parameters under different temperature and humidity environments.

[0072] Step 14: Run the active matching model library. Input virtual environmental time-series data covering the entire temperature and humidity range and featuring randomly changing paths. Calculate the dynamic trajectory of parasitic parameters in batches, extract the measured values ​​and characteristic peak points after reaching steady state under specific temperature and humidity combinations, and form a dependency model library. The specific operations are as follows:

[0073] To construct a parasitic parameter benchmark data support system covering all operating conditions, the active matching model library built in step 13 needs to be run, and customized virtual environment time series data needs to be input. This virtual environment time series data needs to meet two main characteristics: first, it needs to cover the full temperature and humidity range that may occur in plateau outdoor scenarios, including extreme high temperature, extreme low temperature, high humidity, and boundary conditions such as sudden temperature and humidity changes; second, it needs to have random change path characteristics, simulating the temperature and humidity change process under different random disturbances to ensure the diversity and representativeness of the data, so as to adapt to the uncertainty of temperature and humidity changes in the actual environment. The model library receives the virtual environment time series data. After the data is collected, a batch calculation process is initiated. For each set of input data, a corresponding dynamic evolution trajectory of parasitic parameters is generated, forming a massive trajectory dataset. Based on this dataset, further feature extraction is carried out: for a specific temperature and humidity combination, the measured values ​​of parasitic parameters after reaching a steady state are extracted from the trajectory. These values ​​can be used as the benchmark reference values ​​of parasitic parameters under that temperature and humidity condition. At the same time, feature peak points in the trajectory are extracted. These points can characterize the extreme value response features of parasitic parameters when temperature and humidity change abruptly. All extracted steady-state measured values ​​and feature peak points are associated and integrated with the corresponding temperature and humidity combinations to finally form a dependency model library.

[0074] In a preferred embodiment of the present invention, step 2 is further included: during the operation of the electricity meter, real-time measurement values ​​of parasitic parameters are acquired in situ.

[0075] In obtaining real-time measurements of parasitic parameters, the following steps are also performed:

[0076] Step 21: Analyze the metering cycle, determine the idle time window for analog-to-digital conversion and digital processing, plan microsecond-level deterministic detection time slots, and allocate exclusive and periodically repeating specific time slots for the voltage sampling resistor and current transformer within the energy meter, generating a time slot mapping table. The specific operations are as follows:

[0077] First, the metering cycle of the electricity meter needs to be broken down into time sequences to clarify the core links and timing parameters of the main metering process. The metering cycle of a smart electricity meter is usually 20ms, corresponding to a 50Hz power frequency signal. Within this cycle, the analog-to-digital converter (ADC) module needs to complete the sampling of voltage and current signals, while the digital signal processing (DSP) module needs to perform operations such as filtering, Fourier transform, and energy calculation. The typical operation time is 500 to 800μs. Through timing simulation and actual measurement analysis, two idle time windows are located: one after the ADC sampling ends and before the DSP operation starts, and the other after the DSP operation ends and before the next round of sampling starts. These windows need to meet the conditions of no main metering data interaction and no operation resource occupation, with a typical duration of 10 to 30μs.

[0078] Based on the timing characteristics of the idle window, microsecond-level deterministic detection time slots are planned. Considering the parasitic parameters of voltage sampling resistors and current transformers, such as the transient fluctuations in resistance values ​​and the fast response speed of loss tangent relaxation, the time slot accuracy needs to reach the 1μs level to ensure that transient changes in parasitic parameters can be captured. At the same time, to avoid mutual interference between the detection signals of the two types of components, exclusive time slots need to be allocated to them: for voltage sampling resistors, due to their fast parasitic parameter response speed, a 3μs exclusive time slot is allocated; for current transformers, due to their longer loss tangent relaxation response time, a 5μs exclusive time slot is allocated. Both types of time slots adopt a periodic reproduction mechanism, with the reproduction period synchronized with the metering period, to ensure that parasitic parameter detection can be completed once in each metering cycle.

[0079] Finally, a time slot mapping table is generated. This table needs to specify the core parameters: unique identifiers of components, such as voltage sampling resistor R1 and current transformer CT1; the start time of the corresponding detection time slot and the offset relative to the start of the metering cycle, such as 10.002ms for R1 and 10.007ms for CT1; the duration of the time slot, the recurrence cycle, and the priority of the time slot, which is higher than non-core auxiliary functions but lower than the main metering function. The time slot mapping table will be stored in the main control unit (MCU) of the energy meter as the timing reference for subsequent detection signal injection and response acquisition, ensuring that the detection operation and the main metering operation are completely isolated in timing, and avoiding the decrease in the accuracy of the main metering or the distortion of the detection data due to time slot conflicts.

[0080] Step 22: For the exclusive time slot, design and inject an orthogonal characteristic detection signal that has a predetermined correspondence with the changes in predetermined parasitic parameters of the components. The signal is orthogonal to the main metering signal in both the time and frequency domains. The specific operation is as follows:

[0081] First, the characteristics of the main metering signal are defined: a 50Hz power frequency voltage and current signal with a frequency range of 0-2kHz. In the time domain, it is a continuous sinusoidal signal. Based on the orthogonality requirement, the detection signal needs to achieve dual isolation from the main metering signal in both the time and frequency domains. In the time domain, the injection of the detection signal is strictly limited to the exclusive time slot planned in step 21, completely staggered from the sampling and processing period of the main metering signal, and the signal injection duration does not exceed the exclusive time slot duration, i.e., 3μs / 5μs, to avoid time domain superposition interference. In the frequency domain, the frequency of the detection signal needs to avoid the frequency band of the main metering signal, selecting a high-frequency band of 500kHz-1MHz. This frequency band is far from power frequency harmonics, and the low-pass filter of the main metering link of the energy meter can effectively suppress the signal in this frequency band, preventing it from entering the main metering link. At the same time, by adjusting the signal phase, the cross-correlation coefficient between the detection signal and the main metering signal is ≤0.01, achieving frequency domain orthogonality.

[0082] Secondly, a predetermined correspondence between the detection signal and the parasitic parameter changes is designed. Through component characteristic testing, a correlation model between the parasitic parameter changes and signal characteristic parameters is established: for the voltage sampling resistor, its resistance change... Amplitude attenuation coefficient of the detected signal They are linearly correlated, that is , Since the resistance is the reference value, the detection signal is a sinusoidal amplitude-modulated signal, and the amplitude fluctuates linearly with the resistance value; for the current transformer, its loss tangent is... Phase offset from the detection signal They are positively correlated, that is , The proportional coefficient is determined by the transformer core material and winding structure. Therefore, the detection signal adopts a sinusoidal phase-modulated signal, and the phase shifts with the loss tangent. The specific parameters of the detection signal need to be optimized and determined: the carrier frequency is 800kHz, the amplitude is 50mV, which is 1% lower than the amplitude of the main metering signal to avoid affecting the normal operation of the components, and the modulation coefficient is 0.5 to ensure that the signal has sufficient anti-interference capability and recognition.

[0083] Finally, the detection signal is injected: the dedicated signal generator integrated inside the energy meter, implemented by the DDS chip, has a frequency accuracy of 1ppm. According to the timing requirements of the time slot mapping table, the detection signal is accurately injected into both ends of the detected component within the corresponding exclusive time slot. During the injection process, an impedance matching circuit with a matching impedance of 50Ω is required to ensure that the signal is not reflected. At the same time, an isolation circuit is used to prevent the detection signal from being inserted into the main metering link, thus ensuring the purity of the main metering signal.

[0084] Step 23: Within the same time slot of the injected orthogonal feature detection signal, synchronously acquire the response waveform of the detected component, and perform coherent demodulation processing with the injected signal as a reference to extract the complex number of the baseband response. The specific operation is as follows:

[0085] First, synchronous acquisition of the response waveform is achieved: a clock signal from the same source as the probe signal injection is used as the acquisition trigger signal to ensure that the timing synchronization error between the acquisition module and the signal injection module is ≤10ns. The acquisition module uses a high-speed ADC chip, starting acquisition within the same time slot as the probe signal injection. The acquisition duration is consistent with the probe signal injection duration, and the number of data points acquired is 30-50 to ensure complete capture of the transient characteristics of the response waveform. During acquisition, common-mode interference is suppressed using differential acquisition, with a common-mode rejection ratio ≥80dB. A low-noise amplifier is used, with a noise figure ≤1dB, to amplify the weak response signal and avoid feature loss due to signal attenuation.

[0086] Subsequently, the acquired response waveform is coherently demodulated. The advantage of coherent demodulation is that it can effectively suppress random noise and environmental interference, highlighting the effective signal components related to parasitic parameters. The demodulation process consists of three steps: First, the acquired response waveform and the injected probe signal are used as reference signals for mixing. After mixing, a mixed signal containing low-frequency baseband signals and high-frequency carrier signals is obtained, in which the low-frequency baseband signal carries parasitic parameter information. Second, the high-frequency carrier component is filtered out by a low-pass filter, retaining the low-frequency baseband signal. An elliptic filter is used to ensure flat amplitude-frequency characteristics and linear phase-frequency characteristics, avoiding baseband signal distortion. Third, the filtered baseband signal is phase-synchronized and calibrated. Using the phase of the reference signal as a benchmark, the phase offset of the baseband signal is corrected, which is caused by circuit delay and environmental interference, to ensure the accuracy of the phase information.

[0087] Finally, the complex number of the baseband response is extracted. The baseband signal after coherent demodulation is a complex signal, and its complex form can be expressed as follows: ,in This represents the amplitude of the baseband signal. For the phase of the baseband signal, this complex number fully characterizes the response of the component to the probe signal: amplitude It reflects the magnitude of changes in parasitic parameters, such as the fluctuation range of resistance value and the magnitude of transformer loss; Represents the imaginary unit; phase It reflects the changing phase of parasitic parameters, such as the relaxation phase of the loss tangent. The complex baseband response will be used as the input for subsequent parasitic parameter calculation. Its extraction accuracy directly determines the accuracy of parasitic parameter measurement. The extraction error needs to be controlled within 0.1%.

[0088] Step 24: Based on the complex baseband response, compensation is performed using a pre-stored parasitic effect model of the detection loop. Then, based on the excitation response transfer function of the component, the real-time measured values ​​of the parasitic parameters are calculated. The specific operations are as follows:

[0089] First, parasitic effect compensation is performed on the detection loop. The detection loop, including the signal generator, acquisition module, connecting wires, and component pins, has inherent parasitic parameters, such as the distributed capacitance of the wires, contact resistance, and pin inductance. These parasitic parameters interfere with the response of the detection signal, causing deviations in the complex baseband response. The pre-stored parasitic effect model of the detection loop is an equivalent circuit model based on the loop hardware structure. This model includes parameters of various parasitic components in the loop and temperature and humidity correction coefficients. Through simulation calculation, the interference components caused by parasitic effects under different operating conditions, such as different temperatures, humidity levels, and detection frequencies, can be determined. .

[0090] During the compensation process, the complex baseband response extracted in step 23 is... With interference components Perform vector subtraction to obtain the complex baseband response after compensation. The complex number has eliminated the interference of parasitic effects in the detection loop and truly reflects the parasitic parameter response characteristics of the detected component. Subsequently, the excitation response transfer function of the detected component is called. This transfer function is based on the underlying physical characteristics of the component and is a quantitative relationship between input and output obtained by fitting a large amount of experimental data. The establishment process is as follows: by changing the temperature and humidity environment, detection signals with different parameters are applied to the component, the corresponding baseband response complex number is recorded, and the coefficients of the transfer function are obtained by fitting with the least squares method, thus forming a personalized transfer function for a specific component.

[0091] For the voltage sampling resistor, its excitation response transfer function is: ,in To compensate for the amplitude of the complex baseband response, the function is in the form of a linear function. , This is the amplitude resistance conversion coefficient. The real-time resistance value can be directly calculated using this function, which is the core parasitic parameter. For a current transformer, its excitation response transfer function is... ,in To compensate for the phase of the complex baseband response, the function is in the form of a quadratic function. , , This is the phase loss angle conversion coefficient. The reference loss tangent value is used to calculate the real-time loss tangent value of the current transformer, which is the core parasitic parameter. After the calculation is completed, the validity of the parasitic parameter measurement value is verified. If the measurement value exceeds the normal parameter range of the component, such as the resistance value deviation exceeding ±5% or the loss tangent value deviation exceeding ±0.01, a data anomaly prompt is triggered to ensure the reliability of the measurement results.

[0092] In a preferred embodiment of the present invention, step 3 is further included: matching the real-time measured value with the reference response surface to determine the theoretical performance state under the current environmental stress; and calculating and generating the expected performance degradation trajectory for a predetermined future period based on the theoretical performance state and the preset component stress life correlation model.

[0093] Determining the theoretical performance state under the current environmental stress also involves the following steps:

[0094] Step 31: Based on the dependency model library, discretize the feature vector of temperature and humidity change path into a set of actions, and discretize the feature vector of parasitic parameter dynamic response into a set of states. Statistically generate the state transition probability matrix and the observation likelihood model. The specific operations are as follows:

[0095] First, the dependency model library stores the dynamic trajectories of parasitic parameters corresponding to time-series data of virtual environments covering the entire temperature and humidity range, as well as the mapping relationship between temperature and humidity change path feature vectors and parasitic parameter dynamic response feature vectors. This provides rich basic data for constructing action sets and state sets. For the temperature and humidity change path feature vectors, its constituent dimensions include the temperature change rate, humidity change rate, and the phase difference between the two. Through statistical analysis of all temperature and humidity path data in the dependency model library, an adaptive interval partitioning method is used to discretize each dimension. The partitioning is based on the clustering characteristics of the data distribution to avoid boundary distortion caused by equal interval partitioning. Each discretized interval combination constitutes an independent action. The set of all actions constitutes the action set. Each action uniquely corresponds to a type of temperature and humidity change path with similar characteristics. Secondly, for the feature vector of the dynamic response of parasitic parameters, its core constituent dimensions revolve around the core indicators of parasitic parameters. For voltage sampling resistors, it is the peak value, steady-state value, and attenuation coefficient of the transient fluctuation of the resistance value. For current transformers, it is the relaxation time, steady-state value, and fluctuation frequency of the loss tangent. The same adaptive interval partitioning method as the feature vector of temperature and humidity path is used to discretize each core indicator dimension. Each discretized interval combination constitutes an independent state. The set of all states constitutes the state set. Each state uniquely corresponds to a type of dynamic response mode of parasitic parameters with similar characteristics.

[0096] After constructing the action set and state set, all data samples in the dependency model library are traversed and statistically analyzed to count the frequency of transition from the current state to the next state under each action trigger. The frequencies are then normalized to obtain the state transition probability matrix. Each element in this matrix quantitatively represents the probability of transitioning from one state to another under a specific action. Simultaneously, a key focus is on constructing an observation likelihood model. This model quantitatively represents the probability distribution of observed values ​​of corresponding parasitic parameters under specific states. The core of this model is to fit the probability density function of the observed values ​​under each state using sample data from the dependency model library, ensuring that the model can... To accurately match the statistical characteristics of actual observation data, the construction process involves four key steps: sample selection, distribution hypothesis, parameter estimation, and fit verification. The first step, sample selection, involves extracting all corresponding data samples from the dependent model library for each state, removing outliers caused by model simulation errors and data acquisition noise, and using the 3σ criterion (removing data deviating more than three standard deviations from the sample mean) to ensure the validity and representativeness of the samples. The effective sample size for each state is no less than 500 sets to guarantee subsequent fitting accuracy. The second step, distribution hypothesis, combines the statistical characteristics of parasitic parameter observations with the influence of environmental factors. Due to environmental noise and detection errors, the observed values ​​fluctuate randomly around the state characteristic values. It is assumed that the observed values ​​in each state follow a Gaussian distribution, i.e., a normal distribution. This distribution can effectively fit the fluctuation pattern of the observed data dominated by random noise, and its parameters are simple and easy to implement in engineering. The third step is parameter estimation. Based on the selected effective samples, the maximum likelihood estimation method is used to estimate the core parameters of the Gaussian distribution in each state: the mean μ and the variance σ². The mean μ is the statistical average of the parasitic parameter observed values ​​in that state, corresponding to the typical response characteristic value of that state; the variance σ² is the degree of dispersion of the observed values ​​relative to the mean, reflecting the degree of dispersion of the observed values ​​in that state. The fourth step is to verify the fit of the observed values ​​with the assumed Gaussian distribution using a chi-square goodness-of-fit test. The significance level is set at 0.05. If the p-value of the test result is greater than 0.05, the observed values ​​under this condition are considered to conform to the Gaussian distribution and the model is considered to be fit effectively. If the p-value is less than 0.05, the distribution hypothesis is adjusted, such as adopting a mixture Gaussian distribution, and the parameters are re-estimated and verified until the fit requirement is met.

[0097] Ultimately, each state corresponds to a validated probability density function, and the probability density functions of all states together constitute a complete observation likelihood model. The function of this model is to quickly calculate the likelihood value of any input parasitic parameter observation in each state, that is, the probability that the state produces this observation.

[0098] Step 32: Obtain the real-time measurement sequence of historical parasitic parameters and the synchronized temperature and humidity data sequence. Convert the temperature and humidity data sequence into an action sequence. Based on the state transition probability matrix and the observation likelihood model, decode the action sequence and the measurement sequence to obtain the hidden state sequence and its terminal state as the current theoretical performance state. The specific operations are as follows:

[0099] First, obtain the historical real-time measurement sequence of parasitic parameters and the synchronized temperature and humidity data sequence. The time span of the historical sequence needs to cover at least three complete plateau outdoor temperature and humidity cycles to ensure sufficient statistical representativeness of the data. The real-time measurement sequence of parasitic parameters is obtained by solving step 24, including the real-time resistance value of the voltage sampling resistor and the real-time loss tangent value of the current transformer. The temperature and humidity data sequence is synchronously collected by the temperature and humidity sensor built into the energy meter, ensuring complete temporal alignment with the parasitic parameter measurements. Second, convert the temperature and humidity data sequence into an action sequence. For each time point in the sequence, extract the temperature change rate, humidity change rate, and phase difference between them to construct a temperature and humidity change path feature vector. By matching it with the action set constructed in step 31, find the discretized action interval to which the feature vector belongs, and convert it into an action sequence. After traversing the entire temperature and humidity data sequence, the corresponding action identifier is obtained, and the action sequence synchronized with the time axis is obtained. Then, based on the state transition probability matrix and observation likelihood model generated in step 31, the action sequence and the real-time measurement value sequence of parasitic parameters are decoded. The decoding process adopts the optimal path search logic. Its principle is to find the hidden state sequence that conforms to the transition law of the state transition probability matrix and has the highest likelihood with the parasitic parameter observation value sequence among all possible state sequences. This process can effectively filter out random noise and detection loop interference in the real-time measurement values ​​and restore the true state evolution process of parasitic parameters dominated by environmental stress. Finally, the terminal state is extracted from the decoded hidden state sequence. This state is the theoretical performance state of the component under the current environmental stress, and its corresponding parasitic parameter response mode is highly matched with the current temperature and humidity change path characteristics.

[0100] Step 33: Starting from the terminal state, and combining the action distribution and state transition probability matrix of the future temperature and humidity prediction data, perform Monte Carlo simulation to randomly generate multiple state transition paths. Then, call the active matching model library to calculate the performance degradation of each path and obtain a set of random trajectories. The specific operations are as follows:

[0101] First, the final state obtained in step 32 is used as the simulation starting point. This state accurately represents the theoretical performance level of the components under the current environmental stress and is the common starting point for all simulation trajectories. Second, future temperature and humidity prediction data is obtained. This data is obtained through time series analysis of historical temperature and humidity data from high-altitude outdoor environments. It can predict the temperature and humidity change trends within a predetermined period, including the baseline values ​​of temperature and humidity, the rate of change, and the dynamic evolution of phase difference. The future temperature and humidity prediction data is processed according to the action discretization rules in step 31 and transformed into a corresponding action distribution. This action distribution not only includes the optimal matching action at each time point but also the probability of occurrence of each action, fully representing the uncertainty of future environmental stress changes. Next, combining the action distribution with the state transition probability matrix generated in step 31, the Monte Carlo simulation process is started. The simulation logic generates multiple state transition paths through random sampling: at each simulation time point, one is randomly selected based on the current action distribution. The action is then performed, and the state transition probability matrix is ​​used to transition from the current state to the next state. This process is repeated until all time points of the predetermined future period are covered, thus generating a complete state transition path. For each generated state transition path, the active matching model library constructed in step 13 is called. Based on the parasitic parameter response mode corresponding to each state in the path, and combined with the underlying physical model of the components in the active matching model library, the specific values ​​of the core parasitic parameter indicators in that state are calculated. By comparing the differences in parasitic parameter values ​​between the initial state and the states at each time point, the performance degradation amount at each time point under that path is obtained. The core characterization indicators of the performance degradation amount are the resistance drift rate of the voltage sampling resistor and the increment of the loss tangent of the current transformer. By repeating the above simulation process multiple times, hundreds of statistically representative state transition paths and corresponding performance degradation amount sequences are generated. All sequences together constitute a random trajectory set, which fully covers the performance degradation trajectory differences caused by the uncertainty of future environmental stress changes.

[0102] Step 34: Perform statistical analysis on the random trajectory set, calculate the mean of performance degradation values ​​and the preset confidence interval at each future time point, and form the expected performance degradation trajectory. The specific operation is as follows:

[0103] First, the random trajectory set generated in step 33 is time-axis aligned to ensure that all trajectories are fully synchronized at each time point for point-by-point statistical analysis. For each time point within a predetermined future period, the performance degradation data of all trajectories at that time point is extracted. First, the arithmetic mean is calculated. This mean reflects the most likely trend of component parasitic parameter performance degradation under the current environmental stress prediction and serves as the baseline for the expected performance degradation trajectory. Second, a pre-set confidence interval for the performance degradation data at that time point is calculated. The selection of the confidence interval must balance statistical validity and engineering practicality; a 95% confidence interval is typically used. This is based on calculating the standard deviation and standard error of the data. The system determines a numerical range that covers 95% of the sample data. This range characterizes the uncertainty boundary of future performance degradation and can effectively reflect the impact of the randomness of environmental stress changes on component performance. After calculating the mean and confidence interval for all time points, the mean data for each time point are smoothed and fitted to obtain the trend line of expected performance degradation. At the same time, the upper and lower limits of the confidence interval for each time point are fitted to obtain the upper and lower boundary lines of expected performance degradation. The three curves together constitute a complete trajectory of expected performance degradation. This trajectory can not only accurately predict the degradation trend of component parasitic parameters within a predetermined period in the future, but also clearly show the uncertainty range of the degradation process.

[0104] In a preferred embodiment of the present invention, step 4 is further included: determining the theoretical measurement error component caused by environmental stress based on the expected performance degradation trajectory; and combining the theoretical measurement error component with the statutory maximum permissible error to obtain a dynamic error judgment threshold.

[0105] In obtaining the dynamic error determination threshold, the following steps are also performed:

[0106] Step 41: For the expected performance degradation trajectory, establish an inverse mapping function to inversely analyze the trajectory into a time-varying fault mode parameter sequence. The specific operations are as follows:

[0107] The expected performance degradation trajectory generated in step 34 is the degradation law of parasitic parameters of key components such as voltage sampling resistors and current transformers over time, such as the resistance drift of resistors and the increment of loss tangent of transformers. These parameters directly reflect the degree of performance degradation of components under environmental stress. However, they are physical parameters of the components themselves and cannot be directly input into the metering error calculation model. They must be transformed into fault mode parameters that are directly related to metering error through inverse mapping. These parameters specifically characterize the contribution of component performance degradation to metering link error and are key intermediate variables connecting component physical degradation and metering error.

[0108] The inverse mapping function is based on the inherent physical relationship between component parasitic parameters and fault modes, and is fitted by a large amount of experimental data and simulation results from high-altitude environments. For voltage sampling resistors, the core of their parasitic parameter decay is the change of resistance drift ΔR(t) with time t, and the corresponding fault mode parameter is the resistance drift coefficient. This coefficient quantitatively characterizes the weight of the influence of resistance drift on the power metering error per unit time; the larger the resistance drift, the greater the impact. The larger the value, the more significant its contribution to measurement error; for current transformers, the attenuation of their parasitic parameters is the change of the increment of the loss tangent Δtanδ(t) with time t, and the corresponding fault mode parameter is the loss tangent offset coefficient. This coefficient quantitatively characterizes the degree of influence of the loss angle change on the measurement error; the more obvious the loss angle shift, the lower the error. The larger the value, the more significant the error contribution. The contribution of the resistance drift coefficient is calculated through actual measurement and ratio: First, under standard temperature and humidity conditions, the reference resistance value of the sampling resistor and the reference measurement error of the energy meter are recorded. Then, in the target environment, such as high-altitude temperature and humidity, the real-time total resistance drift is continuously monitored, and the change in measurement error of the energy meter relative to the reference state is recorded within the same time period. Finally, the change in measurement error during that time period is divided by the total resistance drift during that time period to obtain the measurement error change corresponding to a unit resistance drift, which is the resistance drift coefficient. The degree of contribution represented.

[0109] Based on the above correlation, the expression for the inverse mapping function is: In the formula, This is the time-varying fault mode parameter sequence, the output of this step. Essentially, it is a set of parameters that dynamically change over time, including the resistance drift coefficient. and transformer loss angle offset coefficient ,Right now The superscript T indicates matrix transpose, which converts a row vector into a column vector; The parameter sequence representing the expected performance degradation trajectory is the input to the function, containing the resistance drift ΔR(t) and the increment of the transformer loss tangent Δtanδ(t), i.e. ; This represents the inverse mapping function. The inverse aspect is reflected in its function of reversing the parasitic parameter attenuation into an error contribution coefficient, rather than calculating the parameter attenuation in the forward direction.

[0110] The derivation logic of this function is clear and practical: First, through multiple sets of accelerated aging experiments in high-altitude environments, the performance degradation of components under different temperature and humidity stresses is artificially simulated, and the degradation of parasitic parameters, such as ΔR(t) and Δtanδ(t), and their corresponding contribution to measurement error are recorded simultaneously to establish a sample mapping relationship between the two; Second, based on the underlying physical models of the components, such as the temperature and humidity sensitive model of resistors and the electromagnetic loss model of transformers, the intrinsic mechanism of measurement error caused by parasitic parameter degradation is analyzed, and the basic mathematical form of the mapping function is determined; Third, the least squares method is used to fit the experimental sample data, optimize the core parameters of the function, and ensure that the error of the mapping result is ≤2%, so as to ensure that the analyzed fault mode parameters can accurately reflect the actual impact of parasitic degradation on measurement error.

[0111] Finally, through this inverse mapping function, the parasitic parameter decay in the expected performance degradation trajectory can be analyzed time-by-time, generating a continuous, dynamic time-varying fault mode parameter sequence. This sequence fully preserves the characteristics of component performance degradation caused by environmental stress and is directly related to metrological errors.

[0112] Step 42: Based on the fault parameter sequence, run the pre-stored metering link error transfer function to obtain the expected error trend line and error risk envelope. The specific operation is as follows:

[0113] First, the time-varying fault mode parameter sequence generated in step 41 is... The metering link error transfer function is input at each time step. The function calculates the theoretical metering error value E(t) caused by environmental stress at each time step by analyzing the error contribution of each fault mode parameter to each link of the metering link. The theoretical metering error values ​​at all time steps are connected along the time axis to form the expected error trend line. This trend line reflects the average change trend of the metering error caused by environmental stress over time. The core features include the error growth rate, steady-state error value and fluctuation period, which are strongly correlated with the change law of the expected performance degradation trajectory.

[0114] Furthermore, from the perspective of existing technology, the principle of the error transfer function in the metering link of an electricity meter has been widely disclosed in industry standards, academic papers, and patent documents. For example, GB / T 17215.321-2021 "Special Requirements for Electrical Measuring Equipment (AC) Part 21: Static Active Energy Meters (Class A, B, C, D and E)" clearly defines the error sources and basic transfer laws of each link in the metering link, such as the sampling link, signal processing link, and energy calculation link. Patent CN121208742A (Metering Performance Evaluation Method for Smart Electricity Meters) discloses an error evaluation model for the sampling circuit in a high-altitude environment, which includes the transfer logic of voltage sampling resistor temperature and humidity errors on the metering results.

[0115] Then, considering the random fluctuations in the fault mode parameter sequence caused by the uncertainty of future temperature and humidity predictions, and the small modeling errors in the metering link error transfer function, it is necessary to calculate the uncertainty range of the error to generate the error risk envelope. By performing Monte Carlo random sampling on the fault mode parameter sequence with a sampling number ≥1000, multiple sets of fault mode parameter sequences after random perturbation are obtained. These are then input into the error transfer function to obtain multiple sets of random error trajectories. Statistical analysis is performed on the random error values ​​at each time point to calculate the upper and lower limits of the 95% confidence interval. The upper and lower limits are fitted along the time axis to form the error risk envelope. The upper and lower boundaries of this envelope represent the maximum and minimum possible errors caused by environmental stress, respectively.

[0116] Step 43: Analyze the time-domain characteristics of the error risk envelope. Based on its volatility and trend correlation, apply an asymmetric compression function to generate a dynamic error component baseline. The specific operation is as follows:

[0117] First, a time-domain characteristic analysis is performed on the error risk envelope. The analysis indicators include volatility and trend correlation: volatility represents the rate of change of the upper and lower boundaries of the envelope over time, reflecting the severity of error fluctuations. Volatility increases significantly when temperature and humidity change abruptly in a high-altitude environment. Trend correlation represents the correlation coefficient between the envelope and the expected error trend line, reflecting the consistency between the changing trend of the envelope and the core error trend. The correlation coefficient must be ≥0.9 to ensure that the envelope does not deviate from the core error pattern.

[0118] Based on the time-domain characteristic analysis results, an asymmetric compression function is designed to process the error risk envelope. The asymmetric design is adopted because the risk weights for positive error fluctuations (leading to over-counting of electrical energy) and negative fluctuations (leading to under-counting of electrical energy) in the high-altitude environment differ, requiring targeted adjustment of the compression intensity. The expression for the asymmetric compression function is: ,

[0119] in For the dynamic error component baseline, , These represent the upper and lower boundaries of the error risk envelope. This is the expected error trend line. , It is an asymmetric compressibility coefficient. , The compression coefficient is dynamically adjusted based on volatility; the higher the volatility, the smaller the compression coefficient, thus enhancing the noise removal effect.

[0120] This formula first determines the piecewise structure of the function based on the asymmetry of error fluctuations, and designs compression formulas for positive and negative fluctuations respectively. Then, combined with statistical data on errors in the plateau environment, it fits and obtains the range of values ​​for the compression coefficient, ensuring that the compressed baseline retains the core error trend while eliminating extreme fluctuations. Finally, through error verification experiments, the compression coefficient is adjusted to ensure that the baseline has a good fit of ≥96% with the actual environmental error. After asymmetric compression processing, the generated dynamic error component baseline eliminates the interference of extreme fluctuations, accurately reflects the change law of measurement error caused by environmental stress over time, and has dynamic adaptability to changes in environmental stress.

[0121] Step 44: Perform asymmetric synthesis of the dynamic error component baseline and the legally permissible maximum error under hard limiting to determine the dynamic error judgment threshold. The specific operation is as follows:

[0122] First, clarify the standard for determining the legally permissible maximum error, such as the legal document GB / T 17215.321-2021. Then, combine this with the accuracy class of the smart meter, such as 0.5 class, to determine its legally permissible maximum error. The active power is ±0.5%, which is a legal constraint on metering accuracy and cannot be exceeded.

[0123] Dynamic error component baseline This is the theoretical error component caused by environmental stress. Its positive and negative values ​​correspond to positive error (over-counted) and negative error (under-counted), respectively. The synthesis with the legally mandated maximum permissible error requires an asymmetric method because the measurement impartiality impact of positive and negative errors differs, and the fluctuation amplitude of errors caused by environmental stress differs in the positive and negative directions. The logic of asymmetric synthesis is as follows: the positive dynamic threshold corresponds to the over-counted error being the legally mandated maximum permissible positive error minus the positive dynamic error component, i.e. The underreporting error corresponding to the negative dynamic threshold is the statutory maximum permissible negative error minus the negative dynamic error component, i.e. ,in , , These are the positive and negative absolute values ​​of the dynamic error component baseline, respectively.

[0124] A hard threshold constraint needs to be introduced during the synthesis process to prevent the synthesized dynamic threshold from exceeding the legal error range and to ensure the legality of the threshold: when the synthesized positive threshold... When the threshold is zero, the hard limit is 0; when the negative threshold is... When the calculated threshold is within the legally defined range, the hard limit is 0; when the composite threshold is within the legally defined range, the calculated value is retained. The core purpose of the hard limit is to prevent the dynamic threshold from exceeding the legal limit when the error caused by environmental stress is too large, thus ensuring the legality and impartiality of measurement. The final dynamic error judgment threshold is... This threshold changes dynamically over time, accurately adapting to error fluctuations caused by sudden changes in temperature and humidity in high-altitude outdoor environments. It avoids misjudgments caused by environmental errors due to fixed thresholds, and ensures measurement accuracy through legal error constraints, thus meeting the precise detection needs in high-altitude outdoor scenarios.

[0125] In a preferred embodiment of the present invention, step 5 is further included: if the real-time metering error of the energy meter exceeds a dynamic error judgment threshold, an environmental stress alarm is triggered; if the deviation of the real-time measured value from the expected performance degradation trajectory exceeds a predetermined number of times and exceeds a preset process deviation threshold, a component aging alarm is triggered. The specific operation is as follows:

[0126] Firstly, regarding the triggering logic of environmental stress alarm, the energy meter will collect the metering data under the current operating conditions in real time, and obtain the benchmark metering value under the same operating conditions through the built-in standard calibration module. The deviation between the two is the real-time metering error, which comprehensively reflects the metering accuracy under the combined effect of the current environment and the state of the components. The real-time measurement error is compared in real-time with the dynamic error judgment threshold determined in step 44. The dynamic error judgment threshold is the allowable error range adapted to the current environmental stress. If the real-time measurement error exceeds the upper and lower boundaries of the threshold, it indicates that the measurement error caused by the current environmental stress has exceeded the allowable range after adaptation. At this time, it is judged as a measurement anomaly caused by environmental stress, triggering an environmental stress alarm. This alarm can prompt maintenance personnel to promptly investigate abnormal fluctuations in environmental factors such as temperature and humidity. Subsequently, for the triggering logic of the component aging alarm, the real-time measurement value here is the real-time value of the parasitic parameters calculated in step 2, including the real-time resistance value of the voltage sampling resistor and the real-time loss tangent value of the current transformer. First, the difference between the real-time measurement value and the predicted value at the corresponding moment in the expected performance degradation trajectory generated in step 34 is calculated to obtain the degree of deviation of the real-time measurement value from the trajectory. The process deviation threshold is the parasitic value determined based on the component manufacturing process standard. The normal fluctuation range of parameters represents the reasonable performance deviation of components under no aging conditions. The predetermined number of times is the number of consecutive judgments set to avoid misjudgment caused by a single random fluctuation. It is usually set to 3 to 5 times based on the parasitic parameter fluctuation characteristics in high-altitude environments. When the deviation of the real-time measurement value from the expected performance degradation trajectory meets the two conditions of continuously exceeding the predetermined number of times and each deviation exceeding the preset process deviation threshold, it indicates that the deviation is not a temporary environmental fluctuation or measurement noise, but a continuous deterioration of the component's own performance beyond the normal process fluctuation range, that is, the component has aged. At this time, a component aging alarm is triggered. This alarm can prompt maintenance personnel to check or replace the corresponding component. Through the classification and triggering of these two types of alarms, step 5 achieves accurate differentiation of the cause of the energy meter abnormality, avoids the defect that a single alarm cannot locate the root cause of the problem, and effectively improves the maintenance efficiency and the operational reliability of the energy meter.

[0127] In a preferred embodiment of the present invention, step 6 is further included: when the component aging alarm is triggered, the actual performance trajectory of the parasitic parameters under multiple historical environmental cycles is obtained, and the irreversible offset characteristics of the actual performance trajectory relative to the reference response surface are analyzed.

[0128] The analysis of the irreversible offset characteristics of the actual performance trajectory relative to the reference response surface also includes the following steps:

[0129] Step 61: After triggering the component aging alarm, obtain the time series of actual measured values ​​of parasitic parameters within the historical environmental cycle. Based on the temperature and humidity values ​​at each time point in the series, query the dependency model library to obtain the ideal response value series. Subtract the actual measured value series from the ideal response value series vectorively to obtain the aging and noise residual series. The specific operation is as follows:

[0130] First, the typical temperature and humidity change cycle of the plateau region, such as day and night or season, is determined as the historical environmental cycle. The actual measured values ​​of parasitic parameters within this cycle are obtained as a time series. These sequences are historical data of the real-time resistance value of the voltage sampling resistor and the real-time loss tangent value of the current transformer, which are continuously collected and stored in step 2. Each data point is synchronously associated with the temperature and humidity measurement value at the corresponding time. For each time point in the sequence, the normal response value of the parasitic parameters under the temperature and humidity conditions without aging is queried in the dependency model library according to the corresponding temperature and humidity value. That is, the ideal response value sequence. The ideal response values ​​in the dependency model library have been tested and calibrated by the component performance under standard environment and can characterize the parameter response under normal working conditions without aging or abnormalities. Then, the vector subtraction operation of the actual measured value sequence and the ideal response value sequence is performed at the corresponding time points. That is, the actual value of each time point is subtracted from the corresponding ideal value. The resulting difference sequence is the aging and noise residual sequence. This sequence includes the performance deviation caused by component aging.

[0131] Step 62: Extract the aging trend component from the aging and noise residual sequences, identify the pattern features of the trend component, and obtain the aging pattern feature vector. The specific operations are as follows:

[0132] First, for the aging and noise residual sequences obtained in step 61, the differences in component characteristics are clarified: the aging-induced shift is a low-frequency trend component that evolves slowly over time, with a change period typically measured in days; random measurement noise is a high-frequency, small-amplitude fluctuation component, with a fluctuation period of less than 1 hour; and temporary environmental disturbances are occasional, sudden fluctuation components, with no fixed period, large amplitude, but short duration. Based on this, a two-step processing method of first denoising and then extracting the trend is adopted: First, sudden disturbances are removed by adaptive median filtering. This filtering method can dynamically adjust the filtering window according to the fluctuation amplitude of the sequence, effectively filtering out instantaneous sudden disturbances while retaining the trend component and small-amplitude noise; Second, Butterworth low-pass filtering is used to further separate the trend and noise. The filter cutoff frequency is set based on the typical rate of change of the aging trend in a high-altitude environment. Through statistical analysis of the aging data of components in high-altitude areas, the highest change frequency of the aging trend is determined to be 0.001Hz, corresponding to a change period of approximately 11.5 hours. Therefore, the cutoff frequency is set to 0.001Hz to ensure that all high-frequency noise is filtered out, retaining only the low-frequency aging trend component.

[0133] Subsequently, pattern feature recognition was performed on the extracted aging trend components. The recognition process consisted of three stages: trend fitting, feature extraction, and vector construction. In the trend fitting stage, a combination of linear and exponential fitting was used to fit the aging trend components. The best-fit model was selected by calculating the goodness-of-fit R², i.e., the coefficient of determination. If the goodness-of-fit R² for linear fitting was ≥0.85, it was determined to be a linear aging trend; if the goodness-of-fit R² for exponential fitting was ≥0.85, it was determined to be an exponential aging trend; if both goodness-of-fit were below 0.85, it was determined to be a step-like aging trend, requiring further extraction of the number of inflection points and the inflection point offset. In the feature extraction stage, core feature indicators were extracted for different trend types, including the trend change rate, i.e., the average increment of aging offset per unit time; the cumulative offset, i.e., the difference in aging offset between the current time and the initial time; the goodness-of-fit, representing the regularity of the trend; the number of inflection points, only for step-like trends, representing the number of aging mutations; and the inflection point offset, only for step-like trends, representing the offset magnitude of each mutation. Vector construction stage: The extracted feature indicators are standardized according to preset dimensions, trend type identifier, rate of change, cumulative offset, goodness of fit, number of inflection points, and inflection point offset. Each indicator is normalized to the [0,1] interval to eliminate dimensional differences. Finally, they are integrated to form an aging mode feature vector. This vector can comprehensively and quantitatively characterize the evolution law and core features of the current aging.

[0134] Step 63: Select common temperature and humidity state points from multiple historical environmental cycles as anchor points. For each anchor point, calculate its aging residuals in different historical cycles and the current cycle to form an aging residual historical evolution dataset for each anchor point. The specific operations are as follows:

[0135] First, common temperature and humidity state points in multiple historical environmental cycles are selected as anchor points. These anchor points are state points where the temperature and humidity are completely consistent or within a preset allowable range in different cycles. For example, state points of 20℃ and 50%RH that appear in different seasonal cycles. The reason for selecting such anchor points is that the influence of environmental stress is consistent under the same temperature and humidity conditions, which can eliminate the interference of environmental factors and focus only on the aging changes of the components themselves. For each determined anchor point, the aging residual corresponding to the anchor point is found in different historical environmental cycles and the current cycle. That is, the residual value of the anchor point time point in the residual sequence obtained in step 61. These residual values ​​are arranged in cyclical order to form the aging residual historical evolution dataset corresponding to each anchor point. This dataset can intuitively reflect the changes of component aging residual with time cycle under the same environmental conditions.

[0136] Step 64: Analyze the historical evolution dataset of aging residuals, examine its monotonicity or cumulativeity, calculate the offset vector of the current aging residuals relative to historical values, and combine it with the aging mode feature vector to confirm the aging type and quantization offset. The specific operations are as follows:

[0137] First, the irreversibility of the aging residual historical evolution dataset is tested. The test indicators are the monotonicity and cumulativeity of the dataset. Monotonicity indicates whether the aging residual continuously increases with the progress of the cycle. Component aging is irreversible, and the residual should show a continuous upward trend without any decline or rebound. Cumulativeity indicates whether the total amount of aging residual gradually accumulates with the progress of the cycle. The aging offset is irreversible, and the total amount of residual should continuously increase. The specific test procedure is as follows: The Kendall rank correlation test is used to verify monotonicity. The significance level is set at 0.05. If the correlation coefficient τ of the test results is ≥0.7 and the p-value is <0.05, then the dataset is determined to have significant monotonicity. Cumulative growth rate analysis is used to verify cumulativeity. The residual growth rate of each cycle relative to the previous cycle is calculated. If the residual growth rate of all cycles is ≥0, with no negative growth, and the cumulative growth rate is ≥10%, ensuring that the aging offset is significant, then the dataset is determined to have cumulativeity. Only when the dataset simultaneously satisfies both monotonicity and cumulativeity can it be confirmed that the offset is an irreversible change caused by component aging, excluding spurious offsets caused by temporary environmental disturbances or measurement errors.

[0138] Subsequently, the offset vector of the current aging residual relative to historical values ​​is calculated. Taking each anchor point as the dimension, the difference between the representative value of the aging residual in the current period and the representative value of the anchor point residual in each historical period is calculated to obtain the offset of each anchor point relative to different historical periods. Then, the offsets of each anchor point relative to the most recent historical period are integrated in the order of the anchor points to form an offset vector. Each element of this vector corresponds to the current aging offset of an anchor point, which can quantitatively characterize the degree of offset of the current aging relative to the recent historical state. To further improve the accuracy of offset quantification, it is also necessary to calculate the statistical characteristic values ​​of the offset vector, including the maximum offset, i.e., the offset corresponding to the anchor point with the largest aging offset; the average offset, i.e., the arithmetic mean of the offsets of all anchor points, which characterizes the overall aging offset level; and the standard deviation of the offset, which characterizes the uniformity of the aging offset of each anchor point, comprehensively quantifying the overall characteristics and local differences of irreversible offset.

[0139] Finally, combining the aging mode feature vector obtained in step 62, the aging type is confirmed and the offset is quantified. The logic for confirming the aging type is as follows: if the trend type in the feature vector is linear and the correlation coefficient τ of the monotonicity test is ≥0.8, it is determined to be progressive aging, with a stable aging rate and no accelerating trend; if the trend type is exponential and the cumulative growth rate is ≥20%, it is determined to be accelerated aging, with the aging rate gradually increasing with the cycle; if the trend type is stepped and the number of inflection points is ≥2, it is determined to be abrupt aging, with multiple performance abrupt changes during the aging process. Offset quantification is achieved by combining the statistical feature values ​​of the offset vector with the aging mode features: for progressive aging, the product of the average offset and the rate of change represents the cumulative irreversible offset; for accelerated aging, the product of the maximum offset and the exponential fitting coefficient represents the potential maximum offset; for abrupt aging, the sum of the inflection point offsets represents the cumulative irreversible offset. Finally, through the above analysis, the accurate confirmation of the aging type and the quantitative characterization of irreversible offset are completed, providing data support for maintenance personnel to formulate targeted component repair or replacement strategies.

[0140] In a preferred embodiment of the present invention, step 7 is further included: periodically calibrating the energy meter on-site using an external standard, and using the calibration results to verify and calibrate the benchmark of the real-time measurement value obtained in situ.

[0141] Step 7 also includes the following steps:

[0142] Step 71: During normal power supply and metering of the electricity meter, apply a differential test signal with an amplitude lower than a predetermined proportion of the rated electrical quantity corresponding to the current operating condition of the electricity meter, and record the difference between the standard value and the sampling response to obtain a dataset of micro-differential standard value - sampling response difference. The specific operation is as follows:

[0143] First, the selection and configuration of the external standard should be clarified. A 0.01-level high-precision standard power source conforming to GB / T 17215.321-2021 standard should be selected as the external standard. This standard has the ability to accurately output voltage, current and power signals with an error controlled within ±0.01%, which can meet the high-precision requirements of on-site calibration. At the same time, a dedicated calibration terminal block should be used to connect to the metering port of the energy meter to avoid introducing additional errors due to wiring contact resistance. After the wiring is completed, an insulation resistance test should be performed to ensure the reliability of the wiring.

[0144] Subsequently, the parameter settings for the differential test signal were determined. Considering the normal power supply and metering requirements of the electricity meter, the amplitude of the applied differential test signal was set to 5% to 10% of the rated electrical quantity corresponding to the current operating condition of the electricity meter. For example, when the current rated voltage is 220V, the amplitude of the differential voltage signal is 11V to 22V. This amplitude range can ensure that the sampling module generates a recognizable response signal without interfering with the normal metering data of the electricity meter. The frequency of the test signal is consistent with the current grid frequency, the waveform is a standard sine wave, the signal is applied continuously for 30 minutes, and the sampling frequency is set to 1 time / second to ensure that a sufficient number of sample data are obtained.

[0145] During signal application, the standard values ​​output by the standard device are recorded synchronously, including standard voltage, standard current, standard power, and the real-time sampling response value of the energy meter sampling module. The difference corresponding to each set of data is calculated, i.e., the sampling response value minus the standard value, forming the original difference dataset. To improve the quality of the dataset, the original data needs to be preprocessed: Grubbs' criterion is used to remove abnormal differences, which are mostly caused by poor wiring contact or transient electromagnetic interference. Then, a small amount of missing data is filled in using linear interpolation, with the missing rate controlled within 0.5%. Finally, the preprocessed data is normalized to normalize the differences to the [-1,1] interval to eliminate dimensional differences, ultimately forming a micro-difference standard value sampling response difference dataset. This dataset can accurately reflect the error characteristics of the current sampling stage of the energy meter.

[0146] Step 72: Based on the dataset, and combining the temperature and humidity data during calibration with the real-time measurements of parasitic parameters, a hierarchical error recursive decomposition model is constructed. The overall error is decomposed and the calibration chain error and the expected component error component calculated from the real-time parasitic parameter measurements through the metrology link error transfer function are deducted sequentially to obtain the non-component error component. The specific operations are as follows:

[0147] First, the composition of the overall error is clarified. The overall error in the field calibration of the power meter is the comprehensive error corresponding to the sampled response difference dataset, which mainly includes three types of components: calibration chain error, which is introduced by standard, wiring and environmental factors; expected component error, which is caused by parasitic parameter attenuation; and non-component error, which is introduced by signal processing module drift, calculation algorithm deviation, etc. The logic of the hierarchical error recursive decomposition model is to successively subtract the known error components from the overall error to finally obtain the unknown non-component error components.

[0148] The model is constructed based on the physical mechanism and statistical laws of error propagation, and adopts a top-down, layer-by-layer recursive structure: The first layer deducts the calibration chain error. The calibration chain error is calculated based on the allowable error of the standard, the wiring error correction formula, and the temperature and humidity environmental data during calibration. The allowable error of the standard is determined by its accuracy level, the wiring error is corrected by a dedicated wiring correction coefficient, and the environmental error is adjusted by a temperature and humidity correction factor. The temperature and humidity correction factor is fitted based on the environmental adaptability parameters of the standard. The three are superimposed to obtain the calibration chain error component. The second layer deducts the expected component error. The real-time measured values ​​of parasitic parameters obtained in step 2, namely the real-time resistance value and the real-time loss tangent value of the current transformer, are substituted into the metering link error propagation function in step 42 to calculate the expected component error component caused by the attenuation of component parasitic parameters under the current operating conditions. This component has a strong correlation with the expected performance degradation trajectory in step 34, and its rationality can be verified by trajectory data.

[0149] To ensure the accuracy of error decomposition, the model also introduces an error residual detection mechanism. After deducting each type of error component, the goodness of fit of the remaining error is calculated, i.e., the determination coefficient R². If R² ≥ 0.9, it indicates that the error deduction of that layer is effective; if R² < 0.9, the calculation parameters of that type of error component are recalibrated until the goodness of fit requirement is met. Finally, by deducting the calibration chain error and the expected component error layer by layer, the remaining error component is the non-component error component, which reflects the error characteristics of the non-component links inside the energy meter.

[0150] Step 73: Correct the internal compensation parameters of the energy meter using non-component error components;

[0151] The parasitic parameter measurements used to calculate the expected component error components are compared with the equivalent parasitic parameters derived through error decomposition to calculate the deviation vector of the parasitic parameter measurements.

[0152] The deviation vector is used to perform feedback correction on the parameter reconstruction model and signal control logic. The specific operation is as follows:

[0153] First, the internal compensation parameters of the energy meter are corrected based on the non-component error components. The non-component error components mainly correspond to circuit drift of the signal processing module, quantization deviation of the A / D converter, and coefficient deviation of the energy calculation algorithm. These errors can be offset by adjusting the internal compensation parameters. The correction of the compensation parameters is optimized using the least squares method: with the non-component error components as the objective function and the adjustable compensation parameters inside the energy meter, such as A / D conversion gain, phase compensation coefficient, and algorithm correction factor, as optimization variables, an error minimization optimization model is constructed. The optimal compensation parameter value is obtained by solving the model and then written into the parameter configuration module of the energy meter to replace the original parameters.

[0154] After the correction is completed, the compensation effect needs to be verified. The differential test signal is reapplied, and the non-component error components after correction are calculated. If the error reduction is ≥80%, the correction is deemed effective. If the reduction is <80%, the compensation parameters are re-optimized until the requirements are met. Subsequently, the parasitic parameter measurement deviation is calculated and feedback correction is carried out: the real-time measured value of the parasitic parameter used to calculate the expected component error components is compared with the equivalent parasitic parameter derived from the error decomposition result. The back-calculation logic of the equivalent parasitic parameter is based on the overall error after the non-component error correction. It is substituted into the metering link error transfer function to calculate the theoretical value of the parasitic parameter that matches the current error, i.e., the equivalent parasitic parameter.

[0155] The difference between the two values ​​is calculated and integrated according to the parasitic parameter dimension to obtain the parasitic parameter measurement deviation vector. Each element of this vector corresponds to the measurement deviation of a type of parasitic parameter, such as resistance measurement deviation and loss tangent measurement deviation. Based on the deviation vector, feedback correction is performed on the parameter reconstruction model and signal control logic: the fitting coefficients of the parameter reconstruction model are adjusted, such as the cross-coupling coefficient of the temperature and humidity sensitive model, to reduce the parasitic parameter measurement deviation; the sampling trigger threshold and data filtering parameters in the signal control logic are optimized to improve the stability and accuracy of parasitic parameter measurement. After correction, it needs to be verified through multiple sets of test data. The parasitic parameter measurement deviation needs to be controlled within 1% to ensure the accuracy of parameter measurement.

[0156] Step 74: Apply the corrected parasitic parameter measurement, re-execute the steps of matching real-time measurements with the dependent model library and generating dynamic error judgment thresholds, and analyze the exceedance of real-time measurement errors in the new period. Simultaneously, calculate the residual sequence between the parasitic parameter measurements and the predicted values ​​of the dependent model library in this period and analyze its statistical characteristics. The specific operations are as follows:

[0157] First, the corrected parasitic parameter measurement results are applied, and the key steps from the previous stage are re-executed: the real-time measurement values ​​are re-matched with the dependent model library, and the calculation accuracy of the residual sequence is optimized based on the corrected parasitic parameters and the ideal response values ​​of the model library; the dynamic error judgment threshold is regenerated, and the upper and lower boundaries of the threshold are updated by combining the corrected parasitic parameter attenuation trajectory and the legal maximum permissible error to ensure that the threshold can accurately adapt to the corrected performance status of the electricity meter.

[0158] Subsequently, a new calibration and verification cycle is set. During the new cycle, the real-time metering error of the electricity meter is continuously monitored, and its exceeding of limits is analyzed. The real-time metering error is obtained by comparing it with the corrected dynamic error judgment threshold. The exceeding of limits analysis includes the frequency of exceeding limits, that is, the number of times the threshold is exceeded per unit time; the maximum exceeding limit value, that is, the maximum error amount exceeding the threshold; and the duration of exceeding limits, that is, the duration of a single exceeding limit. If the frequency of exceeding limits is ≤ 1 time / month, the maximum exceeding limit value is ≤ 10% of the dynamic threshold, and the duration of exceeding limits is ≤ 5 minutes, then the metering error control in the new cycle is deemed to be effective. If the above indicators are not met, the process returns to step 73 to re-correct parameters and optimize the model.

[0159] Simultaneously, the residual sequence between the measured values ​​of parasitic parameters and the predicted values ​​of the dependent model library within the new cycle is calculated, and its statistical characteristics are analyzed in depth. The analysis indicators include residual mean, representing systematic bias, which must be ≤0.5%; residual variance, representing random bias, which must be ≤0.1%²; normality test, using the KS test, with a significance level of 0.05 and a p-value >0.05; trend test, using the linear trend test, with no significant trend. If the statistical characteristics of the residual sequence meet the above requirements, it indicates that the accuracy of the corrected parasitic parameter measurement and model matching is up to standard, and the calibration effect is reliable. If the statistical characteristics do not meet the standards, the adaptability of the dependent model library needs to be rechecked, the model parameters adjusted, and the verification re-tested. Through the above process, step 74 realizes the closed-loop management of on-site calibration, error decomposition, parameter correction, and verification update, continuously ensuring the metering accuracy and reliability of the electricity meter in the high-altitude environment.

[0160] The above are merely preferred embodiments of the present invention; however, the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and its improved concept, should be covered within the scope of protection of the present invention.

Claims

1. A method for detecting abnormal metering performance of a smart energy meter, characterized in that, The steps include: Step 1, establishing a reference response surface for the parasitic parameters of the voltage sampling resistor and current transformer inside the electricity meter with respect to temperature and humidity. Step 2: During the operation of the electricity meter, the real-time measurement value of the parasitic parameter is acquired in situ. Step 3: Match the real-time measured values ​​with the reference response surface to determine the theoretical performance state under the current environmental stress; based on the theoretical performance state and the preset component stress life correlation model, calculate and generate the expected performance degradation trajectory for a predetermined future period. Step 4: Based on the expected performance degradation trajectory, determine the theoretical measurement error component caused by environmental stress; combine the theoretical measurement error component with the legal maximum permissible error to obtain a dynamic error judgment threshold. Step 5: If the real-time metering error of the electricity meter exceeds the dynamic error judgment threshold, an environmental stress alarm is triggered; if the deviation of the real-time measurement value from the expected performance degradation trajectory exceeds a predetermined number of times and exceeds the preset process deviation threshold, a component aging alarm is triggered. Step 6: When the component aging alarm is triggered, obtain the actual performance trajectory of the parasitic parameters under multiple historical environmental cycles, and analyze the irreversible offset characteristics of the actual performance trajectory relative to the reference response surface. Step 7: Periodically calibrate the energy meter on-site using an external standard, and use the calibration results to verify and calibrate the benchmark of the real-time measurement value obtained in situ.

2. The method for detecting abnormal metering performance of a smart energy meter according to claim 1, characterized in that, Establish a baseline response surface for its parasitic parameters with respect to temperature and humidity, including: Step 11: Design and execute a time-series change path for temperature and humidity that simulates the sudden changes in outdoor conditions at high altitudes. Simultaneously capture the transient fluctuation sequence of the resistance value of the voltage sampling resistor inside the power meter and the relaxation curve of the loss tangent of the current transformer to obtain the dynamic response sequence of the path. Step 12: Based on the path dynamic response sequence, identify and extract the cooperative response mode determined by the temperature change rate, humidity change rate and their phase difference, and establish a mapping rule set from the temperature and humidity change path feature vector to the parasitic parameter dynamic response feature vector. Step 13: Combine the mapping rule set with the underlying physical model of the components to construct an active matching model library. The active matching model library is used to generate the corresponding dynamic evolution trajectory of parasitic parameters based on temperature and humidity time series data. Step 14: Run the active matching model library, input virtual environmental time series data covering the full temperature and humidity range and having randomly changing path characteristics, batch calculate the dynamic trajectory of parasitic parameters, extract the measured values ​​and characteristic peak points after reaching steady state under a specific temperature and humidity combination, and form a dependency model library.

3. The method for detecting abnormal metering performance of a smart energy meter according to claim 2, characterized in that, Obtain real-time measurements of parasitic parameters, including: Step 21: Analyze the metering cycle, determine the idle time window for analog-to-digital conversion and digital processing, plan the microsecond-level deterministic detection time slots, and allocate exclusive and periodically repeating specific time slots for the voltage sampling resistor and current transformer in the energy meter, and generate a time slot mapping table. Step 22: For the exclusive time slot, design and inject an orthogonal characteristic detection signal that has a predetermined correspondence with the predetermined parasitic parameter changes of the component. The signal is orthogonal to the main metering signal in the time domain and frequency domain. Step 23: In the same time slot of injecting the orthogonal feature detection signal, synchronously acquire the response waveform of the detected component, and perform coherent demodulation processing with the injected signal as a reference to extract the complex number of the baseband response. Step 24: Based on the complex baseband response, compensation is performed using a pre-stored parasitic effect model of the detection loop, and the real-time measured value of the parasitic parameter is calculated according to the excitation response transfer function of the component.

4. The method for detecting abnormal metering performance of a smart energy meter according to claim 3, characterized in that, Determine the theoretical performance state under current environmental stress, including: Step 31: Based on the dependent model library, the feature vector of temperature and humidity change path is discretized into a set of actions, and the feature vector of parasitic parameter dynamic response is discretized into a set of states. The state transition probability matrix and observation likelihood model are statistically generated. Step 32: Obtain the real-time measurement sequence of historical parasitic parameters and the synchronized temperature and humidity data sequence. Convert the temperature and humidity data sequence into the action sequence. Based on the state transition probability matrix and the observation likelihood model, decode the action sequence and the measurement sequence to obtain the hidden state sequence and its terminal state as the current theoretical performance state.

5. The method for detecting abnormal metering performance of a smart energy meter according to claim 4, characterized in that, Calculate and generate the expected performance degradation trajectory for future predetermined periods, including: Step 33: Starting from the terminal state, combining the action distribution of future temperature and humidity prediction data transformation and the state transition probability matrix, perform Monte Carlo simulation, randomly generate multiple state transition paths, and call the active matching model library to calculate the performance degradation of each path to obtain a set of random trajectories. Step 34: Perform statistical analysis on the random trajectory set, calculate the mean value of performance degradation at each future time point and the preset confidence interval, and form the expected performance degradation trajectory.

6. The method for detecting abnormal metering performance of a smart energy meter according to claim 5, characterized in that, The dynamic error determination threshold is obtained, including: Step 41: For the expected performance degradation trajectory, establish an inverse mapping function to inversely analyze the trajectory into a time-varying fault mode parameter sequence; Step 42: Based on the fault parameter sequence, run the pre-stored metering link error transfer function to obtain the expected error trend line and the error risk envelope. Step 43: Analyze the time-domain characteristics of the error risk envelope, and based on its volatility and trend correlation, apply an asymmetric compression function to process it and generate a dynamic error component baseline. Step 44: Perform asymmetric synthesis of the dynamic error component baseline and the statutory maximum permissible error under hard limiting to determine the dynamic error judgment threshold.

7. The method for detecting abnormal metering performance of a smart energy meter according to claim 6, characterized in that, The analysis includes the irreversible offset characteristics of the actual performance trajectory relative to the reference response surface, including: Step 61: When the component aging alarm is triggered, obtain the time series of actual measured values ​​of parasitic parameters within the historical environmental cycle, query the dependency model library based on the temperature and humidity values ​​at each time point in the series to obtain the ideal response value series, and vector subtract the actual measured value series from the ideal response value series to obtain the aging and noise residual series. Step 62: Extract the aging trend component from the aging and noise residual sequence, identify the pattern features of the trend component, and obtain the aging pattern feature vector.

8. The method for detecting abnormal metering performance of a smart energy meter according to claim 7, characterized in that, Also includes: Step 63: Select common temperature and humidity state points in multiple historical environmental cycles as anchor points. For each anchor point, calculate its aging residual in different historical cycles and the current cycle to form a historical evolution dataset of aging residual for each anchor point. Step 64: Analyze the aging residual historical evolution dataset, examine its monotonicity or cumulativeity, calculate the offset vector of the current aging residual relative to the historical value, and combine it with the aging mode feature vector to confirm the aging type and quantization offset.

9. A method for detecting abnormal metering performance of a smart energy meter according to claim 8, characterized in that, The energy meter is periodically calibrated on-site using an external standard, and the calibration results are used to verify and calibrate the benchmark of the real-time measurement values ​​acquired in situ, including: Step 71: During the normal power supply and metering of the electricity meter, apply a differential test signal with an amplitude lower than a predetermined proportion of the rated electrical quantity corresponding to the current operating condition of the electricity meter, and record the difference between the standard value and the sampling response to obtain a micro-differential standard value-sampling response difference dataset. Step 72: Based on the dataset, combined with the temperature and humidity data during calibration and the real-time measurement values ​​of the parasitic parameters, a hierarchical error recursive decomposition model is constructed. The overall error is decomposed and the calibration chain error and the expected component error component calculated by the real-time measurement values ​​of the parasitic parameters through the metering link error transfer function are deducted in sequence to obtain the non-component error component. Step 73: Correct the internal compensation parameters of the energy meter using the non-component error components; The parasitic parameter measurement values ​​used to calculate the expected component error are compared with the equivalent parasitic parameters derived by error decomposition to calculate the deviation vector of the parasitic parameter measurement. The deviation vector is used to perform feedback correction on the parameter reconstruction model and the signal control logic.

10. A method for detecting abnormal metering performance of a smart energy meter according to claim 9, characterized in that, Also includes: Step 74: Apply the corrected parasitic parameter measurement, re-execute the steps of matching the real-time measurement value with the dependent model library and generating the dynamic error judgment threshold, and analyze the over-limit situation of real-time measurement error in the new period. At the same time, calculate the residual sequence between the parasitic parameter measurement value and the predicted value of the dependent model library in this period and analyze its statistical characteristics.

Images