An electric energy meter abnormality detection method and system

Abstract

This invention relates to the field of electrical testing technology, specifically to a method and system for detecting anomalies in electricity meters. The method includes: activating a wideband programmable power source to apply a composite excitation signal to the electricity meter under test; acquiring electrical parameters such as discrete current and voltage sequences to generate a heterogeneous operating condition dataset; generating a physical field correction tensor through a thermomagnetic dynamic correction model, performing physical field mapping operations on the discrete current and voltage sequences to generate an environmental hysteresis loop trajectory; receiving the environmental hysteresis loop trajectory, performing piecewise linearization analysis to generate a core depth saturation distortion index; and projecting the core depth saturation distortion index onto a metering error tolerance evaluation interval to generate a magnetic saturation metering anomaly signal. This invention, by constructing a strong coupling mechanism between environment, physics, and geometry, ensures that the anomaly judgment logic no longer relies solely on electrical thresholds, resulting in more accurate judgments of minor metering deviations induced by environmental factors.

Description

Technical Field

[0001] This invention relates to the field of electrical testing technology, specifically to a method and system for detecting abnormalities in electricity meters. Background Technology

[0002] Monitoring metering errors and assessing the condition of smart meters in complex electromagnetic environments (such as industrial power grids containing high-frequency harmonics and DC component intrusion) is an important technological application area in the current power metering field. Existing technologies generally employ a standard source comparison method combined with time-domain waveform distortion analysis of electrical signals. This involves monitoring the amplitude and phase changes of voltage and current signals or calculating the total harmonic distortion rate. When relevant indicators deviate from the allowable error threshold specified by national standards, the system automatically identifies and determines that the metering is abnormal. This is currently the mainstream technology for online condition monitoring of smart meters.

[0003] However, existing technical solutions have significant shortcomings in detection sensitivity and environmental adaptability. Their evaluation systems for anomaly detection suffer from a disconnect between superficial observations and physical mechanisms. The systems only issue threshold-based fault alarms at the moment of significant waveform distortion (such as severe clipping), failing to continuously track and deeply analyze the core magnetization process. They cannot determine the nonlinear transition trend of the transformer, nor can they construct and confirm the true distortion trajectory of the environmental hysteresis loop. Furthermore, their anomaly outputs are too general, only providing a single conclusion of measurement deviation without adapting to the saturation area ratio or inflection point characteristics of the hysteresis loop. This results in fault diagnosis data lacking specificity and effectiveness in identifying physical causes (such as hard flux cutoff or soft overload). In addition, existing monitoring models are mostly based on pure electrical signal analysis under ideal operating conditions, ignoring the changes in physical characteristics of the transformer core caused by environmental temperature, DC bias, and grid frequency drift. This makes it difficult to accurately match the actual measurement performance under complex and variable operating conditions, leading to missed detections of minor measurement deviations induced by environmental factors.

[0004] Therefore, a method and system for detecting abnormalities in electricity meters are proposed. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for detecting abnormalities in electricity meters, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for detecting abnormalities in an electricity meter, comprising:

[0007] A wideband programmable power source is started to apply a composite excitation signal to the energy meter under test; the discrete current sequence on the secondary side of the current transformer of the energy meter under test, the discrete voltage sequence at the voltage sampling terminal, the feedback signal of the thermistor deployed on the surface of the transformer core, the DC bias component in the signal transmission link, and the grid frequency drift value are collected to generate a heterogeneous operating condition dataset.

[0008] The heterogeneous operating condition dataset is input into the thermomagnetic dynamic correction model. The permeability attenuation coefficient is generated using the thermistor feedback signal. The eddy current loss weighting factor is calculated using the grid frequency drift value. The hysteresis loop center point offset vector is determined using the DC bias component. The physical field correction tensor is constructed. The physical field mapping operation is performed on the current discrete sequence and the voltage discrete sequence to generate the environmental hysteresis loop trajectory.

[0009] Receive the environmental hysteresis loop trajectory, perform piecewise linearization analysis, and identify the inflection point coordinates; extract the area of ​​the saturation region above the inflection point coordinates and the area of ​​the linear region below the inflection point coordinates, calculate the area ratio, and generate the core depth saturation distortion index.

[0010] The magnetic core depth saturation distortion index is projected onto a preset measurement error tolerance evaluation range, and a magnetic saturation measurement anomaly signal is generated based on the anomaly judgment logic triggered by the landing point position.

[0011] Preferably, the specific process of loading the composite excitation signal includes: calling harmonic synthesis logic to generate a fundamental sine sequence and a high-frequency odd harmonic sequence; performing linear superposition operation to generate a digitally synthesized distorted waveform; driving a digital-to-analog converter to convert the digitally synthesized distorted waveform into an analog current excitation signal; injecting the analog current excitation signal into the current loop of the energy meter under test through a power amplification link; using a hardware phase-locked loop to lock the zero-crossing point of the analog current excitation signal; and triggering a synchronous sampling clock based on the zero-crossing point to perform timing alignment operations on the discrete current sequence and the discrete voltage sequence.

[0012] Preferably, the specific processing steps of the thermomagnetic dynamic correction model include: receiving a heterogeneous operating condition dataset and separating the thermistor feedback signal, the power grid frequency drift value, and the DC bias component; constructing a three-dimensional orthogonal correction matrix containing temperature, frequency, and magnetic bias dimensions as the thermomagnetic dynamic correction model; mapping the thermistor feedback signal to the temperature dimension and retrieving the core permeability temperature drift curve to obtain the permeability attenuation coefficient; mapping the power grid frequency drift value to the frequency dimension and retrieving the eddy current loss weighting factor using a variant of the Steinmetz formula; mapping the DC bias component to the magnetic bias dimension and retrieving the hysteresis loop center point offset vector; and constructing a physical field correction tensor by the permeability attenuation coefficient, the eddy current loss weighting factor, and the hysteresis loop center point offset vector.

[0013] Preferably, the specific generation process of the environmental hysteresis loop trajectory includes: receiving the physical field correction tensor; correcting the amplitude of the current discrete sequence using the permeability attenuation coefficient to generate an equivalent magnetic field strength sequence; performing phase compensation operation on the voltage discrete sequence using the eddy current loss weighting factor to generate an equivalent magnetic induction intensity sequence; mapping the equivalent magnetic field strength sequence to abscissa data; mapping the equivalent magnetic induction intensity sequence to ordinate data; performing coordinate system translation transformation on the abscissa data and ordinate data using the hysteresis loop center point offset vector; and corresponding the transformed abscissa data and ordinate data according to the timestamp to generate the environmental hysteresis loop trajectory.

[0014] Preferably, the specific identification process of the inflection point coordinates includes: receiving the environmental hysteresis loop trajectory and calculating the tangent slope sequence of adjacent discrete points; performing a second-order difference operation on the tangent slope sequence to generate a curvature mutation spectrum; setting a magnetic saturation critical curvature threshold; scanning the curvature mutation spectrum and locating the extreme point positions that exceed the magnetic saturation critical curvature threshold; locking the magnetic field strength coordinates and magnetic induction intensity coordinates corresponding to the extreme point positions as magnetic saturation inflection point coordinates; and using the magnetic saturation inflection point coordinates to physically divide the environmental hysteresis loop trajectory into a nonlinear saturation segment above the inflection point and a linear magnetization segment below the inflection point.

[0015] Preferably, the specific generation process of the core depth saturation distortion index includes: receiving a nonlinear saturation segment and a linear magnetization segment; performing a discrete Riemann integral operation on the nonlinear saturation segment to calculate the saturation area; performing a trapezoidal integral operation on the linear magnetization segment to calculate the linear area; constructing a dimensionless performance comparison function; inputting the saturation area as the numerator and the linear area as the denominator into the dimensionless performance comparison function; performing a division operation and logarithmically scaling the calculation result; and generating the core depth saturation distortion index.

[0016] Preferably, the specific generation process of the magnetic saturation metering anomaly signal includes: receiving the magnetic core depth saturation distortion index; retrieving a graded evaluation interval table constructed based on the electricity meter accuracy level standard; performing a numerical comparison operation between the magnetic core depth saturation distortion index and the safety threshold and fault threshold in the graded evaluation interval table; if it is less than the safety threshold, it falls into the normal operation interval; if it is between the safety threshold and the fault threshold, it falls into the accuracy degradation interval; if it is greater than the fault threshold, it falls into the functional failure interval; outputting the calibrated landing point interval position status code; and specifying the landing point interval position that falls into the functional failure interval. The status code is used to activate the interrupt request; a preset fault semantic mapping dictionary is invoked; the status code of the landing point interval is used as an index key for retrieval and matching; if the status code of the landing point interval position points to the accuracy degradation interval, the soft magnetic saturation overload tag is extracted as the transformer overload type; if the status code of the landing point interval position points to the functional failure interval, the hard magnetic flux truncation overload tag is extracted as the transformer overload type; the anomaly encoding generator is invoked to construct a fault message containing the current timestamp, the magnetic core depth saturation distortion index amplitude, and the transformer overload type; the fault message is encapsulated as a magnetic saturation metering anomaly signal.

[0017] An electricity meter anomaly detection system includes:

[0018] Data capture module: Starts a wideband programmable power source to load a composite excitation signal onto the energy meter under test; collects the discrete current sequence on the secondary side of the current transformer of the energy meter under test, the discrete voltage sequence at the voltage sampling terminal, the feedback signal of the thermistor deployed on the surface of the transformer core, the DC bias component in the signal transmission link, and the grid frequency drift value, and generates a heterogeneous operating condition dataset.

[0019] Magnetic field correction module: Input heterogeneous operating condition dataset into thermomagnetic dynamic correction model, generate permeability attenuation coefficient using thermistor feedback signal, calculate eddy current loss weighting factor using grid frequency drift value, determine hysteresis loop center point offset vector using DC bias component, construct physical field correction tensor, perform physical field mapping operation on current discrete sequence and voltage discrete sequence, and generate environmental hysteresis loop trajectory.

[0020] Deviation analysis module: Receives the environmental hysteresis loop trajectory, performs piecewise linearization analysis, identifies inflection point coordinates; extracts the area of ​​the saturation region above the inflection point coordinates and the area of ​​the linear region below the inflection point coordinates, calculates the area ratio, and generates the core depth saturation distortion index.

[0021] Anomaly detection module: Projects the magnetic core depth saturation distortion index onto the preset measurement error tolerance evaluation range, and generates a magnetic saturation measurement anomaly signal based on the anomaly judgment logic triggered by the landing point position.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0023] 1. A heterogeneous operating condition dataset is constructed by synchronously acquiring thermistor signals, DC bias, and frequency drift values ​​from the instrument transformers. Using a thermomagnetic dynamic correction model, these environmental parameters are analyzed as permeability attenuation coefficients, eddy current loss weighting factors, and hysteresis loop center point offset vectors, respectively, forming a physical field correction tensor. This tensor performs physical-level preprocessing on the original current and voltage sequences, reducing the interference of thermodynamic and electromagnetic factors on the metering signals at the data source, which is beneficial for subsequent analysis based on the actual physical operating conditions of the instrument transformers.

[0024] 2. By using a physical field correction tensor to perform amplitude correction and phase compensation on the current-voltage sequence, an equivalent magnetic field and magnetic induction intensity sequence is generated, which is then mapped to the environmental hysteresis loop trajectory through coordinate transformation. This process transforms abstract time-domain sampled values ​​into closed curves that intuitively reflect the magnetization process of the iron core, demonstrating the domain flipping behavior and energy loss of the iron core under non-ideal conditions. This physical space mapping method identifies nonlinear changes in magnetic flux that are difficult to detect through simple time-domain waveform analysis, providing a feature carrier with rich physical meaning for identifying subtle metrological deviations.

[0025] 3. The hysteresis loop trajectory is segmented and analyzed to identify the magnetic saturation inflection point and calculate the area ratio of the saturation region to the linear region. After dimensionless processing, a core depth saturation distortion index is generated. This index quantifies the degree to which the core deviates from the linear operating region from a geometric and topological perspective, sensitively reflecting soft saturation phenomena induced by the environment. Projecting this index onto the metering error evaluation interval, and based on the anomaly judgment logic triggered by the landing point, a numerical correlation between changes in physical characteristics and metering errors is established. This facilitates the accurate diagnosis of subtle performance degradation of electricity meters under complex operating conditions.

[0026] 4. The physical field correction tensor generated by the thermomagnetic dynamic correction model is deeply embedded in the hysteresis loop construction process, so that the final generated core depth saturation distortion index inherently contains the physical constraints of environmental conditions. This strong coupling mechanism of "environment-physics-geometry" means that the anomaly judgment logic no longer relies on electrical thresholds in isolation, but distinguishes between soft magnetic saturation overload caused by environmental drift and hard magnetic flux cutoff caused by physical damage based on the status code of the landing point interval. By calling the fault semantic mapping dictionary, the system transforms the abstract distortion index into a fault message containing overload type and physical cause, which can match the actual metering performance under complex and variable operating conditions and make more accurate judgments on slight metering deviations induced by environmental factors. Attached Figure Description

[0027] Figure 1 This is a flowchart of an energy meter anomaly detection method proposed in an embodiment of this invention application;

[0028] Figure 2 This is a flowchart of the thermomagnetic dynamic correction and hysteresis loop reconstruction proposed in an embodiment of this invention application;

[0029] Figure 3 This is a flowchart illustrating the calculation and grading of the magnetic saturation distortion index as proposed in an embodiment of this invention.

[0030] Figure 4 This is a structural diagram of an energy meter anomaly detection system proposed in an embodiment of this invention. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] Please see Figures 1-3 The present invention provides a method for detecting abnormalities in an electricity meter, the specific steps of which are as follows:

[0033] A wideband programmable power source is started to apply a composite excitation signal to the energy meter under test; the discrete current sequence on the secondary side of the current transformer of the energy meter under test, the discrete voltage sequence at the voltage sampling terminal, the feedback signal of the thermistor deployed on the surface of the transformer core, the DC bias component in the signal transmission link, and the grid frequency drift value are collected to generate a heterogeneous operating condition dataset.

[0034] The heterogeneous operating condition dataset is input into the thermomagnetic dynamic correction model. The permeability attenuation coefficient is generated using the thermistor feedback signal. The eddy current loss weighting factor is calculated using the grid frequency drift value. The hysteresis loop center point offset vector is determined using the DC bias component. The physical field correction tensor is constructed. The physical field mapping operation is performed on the current discrete sequence and the voltage discrete sequence to generate the environmental hysteresis loop trajectory.

[0035] Receive the environmental hysteresis loop trajectory, perform piecewise linearization analysis, and identify the inflection point coordinates; extract the area of ​​the saturation region above the inflection point coordinates and the area of ​​the linear region below the inflection point coordinates, calculate the area ratio, and generate the core depth saturation distortion index.

[0036] The magnetic core depth saturation distortion index is projected onto a preset measurement error tolerance evaluation range, and a magnetic saturation measurement anomaly signal is generated based on the anomaly judgment logic triggered by the landing point position.

[0037] The technical solution of the present invention will be further described in detail below with reference to specific embodiments.

[0038] Example 1

[0039] This application discloses a method for detecting abnormalities in an electricity meter, see below. Figure 1The specific steps proposed in this invention include: S1, starting a wideband programmable power source to apply a composite excitation signal to the energy meter under test; collecting the discrete current sequence on the secondary side of the current transformer of the energy meter under test, the discrete voltage sequence at the voltage sampling terminal, the feedback signal of the thermistor deployed on the surface of the transformer core, the DC bias component in the signal transmission link, and the grid frequency drift value to generate a heterogeneous operating condition dataset; S2, inputting the heterogeneous operating condition dataset into a thermomagnetic dynamic correction model, using the thermistor feedback signal to generate a permeability attenuation coefficient, using the grid frequency drift value to calculate the eddy current loss weighting factor, and using the DC bias component to determine the... The offset vector of the center point of the hysteresis loop is used to form a physical field correction tensor. Physical field mapping operations are performed on the discrete current sequence and the discrete voltage sequence to generate the environmental hysteresis loop trajectory. S3: The environmental hysteresis loop trajectory is received, and a piecewise linearization analysis operation is performed to identify the inflection point coordinates. The area of ​​the saturation region above the inflection point coordinates and the area of ​​the linear region below the inflection point coordinates are extracted, and the area ratio is calculated to generate the magnetic core depth saturation distortion index. S4: The magnetic core depth saturation distortion index is projected onto the preset measurement error tolerance evaluation interval. According to the anomaly judgment logic triggered by the landing point position, a magnetic saturation measurement anomaly signal is generated.

[0040] Further, a wideband programmable power source is activated to apply a composite excitation signal to the energy meter under test; the discrete current sequence on the secondary side of the current transformer of the energy meter under test, the discrete voltage sequence at the voltage sampling terminal, the feedback signal of the thermistor deployed on the surface of the transformer core, the DC bias component in the signal transmission link, and the grid frequency drift value are collected to generate a heterogeneous operating condition dataset; corresponding to step S1 above; the specific implementation process includes:

[0041] The harmonic synthesis logic is invoked to generate a fundamental sine wave sequence and a high-frequency odd harmonic sequence, and a linear superposition operation is performed to generate a digitally synthesized distorted waveform. The digital-to-analog converter unit is driven to convert the digitally synthesized distorted waveform into an analog current excitation signal. The analog current excitation signal is injected into the current loop of the energy meter under test through a power amplification link. The zero-crossing point of the analog current excitation signal is locked using a hardware phase-locked loop. Based on the zero-crossing point, a synchronous sampling clock is triggered to perform timing alignment operations on the discrete current sequence and the discrete voltage sequence.

[0042] Specifically, the process of generating heterogeneous operating condition datasets is as follows:

[0043] The waveform synthesis engine of the wideband programmable power source is activated. Internally, this engine runs harmonic synthesis logic based on direct digital frequency synthesis technology. The preset reference sampling frequency is 12.8 kHz, corresponding to a time resolution of 78.125 microseconds. At the logic operation layer, a fundamental sinusoidal sequence is first defined, with its frequency parameter set to 50 Hz (power frequency) and its amplitude parameter set to the rated current, for example, 5 amps. Next, the high-frequency odd-order harmonic generation logic is invoked. Based on a preset test scheme, a third harmonic sequence at 150 Hz and a fifth harmonic sequence at 250 Hz are generated, with their amplitude coefficients set to 10% and 5% of the fundamental amplitude, respectively. The harmonic synthesis logic performs a linear superposition operation, algebraically summing the instantaneous values ​​of the fundamental sequence and each harmonic sequence at the same time domain point to generate a digitally synthesized distorted waveform sequence.

[0044] In this embodiment, a spectrum pre-scanning mechanism is initiated, calling historical power grid background noise data preset in the test system to simulate the current power grid background noise spectrum, calculating the signal-to-noise ratio margin, and adaptively adjusting the amplitude weighting coefficient of the high-frequency odd harmonic sequence accordingly to reconstruct the digitally synthesized distorted waveform. Before sending the digitally synthesized distorted waveform sequence into the digital-to-analog converter unit, a fast Fourier transform is performed on the power grid background voltage to obtain the background noise floor amplitude. The minimum effective signal-to-noise ratio threshold is set to 12 dB. If the noise amplitude of a certain harmonic (e.g., the fifth harmonic) in the background is detected to be too high, causing the ratio of the original injected signal amplitude to the background noise amplitude to be lower than the minimum effective signal-to-noise ratio threshold, an amplitude compensation algorithm will be automatically triggered: a gain coefficient is introduced for this specific harmonic. The calculation logic is as follows: multiply the background noise amplitude by a safety factor of 1.2, and then divide by the base injected amplitude to obtain the gain coefficient of the current harmonic. To prevent excessive injected energy from affecting power grid quality, the upper limit clamping value of this coefficient is set to 1.15 times. Conversely, if the background is extremely clean, the coefficients are reduced to 0.9 times to lower power consumption. The harmonics are then re-superimposed based on the updated weighting coefficients, refreshing the data buffer of the digitally synthesized distorted waveform. This process ensures the effectiveness of the excitation signal under noisy power grids while avoiding excessive injection under clean power grids.

[0045] The generated digitally synthesized distorted waveform is input to a 16-bit high-precision digital-to-analog converter (DAC) unit, which converts the discrete digital code into a continuous analog voltage signal. This analog signal is smoothed by a low-pass filter and then input to a transconductance power amplifier link. The power amplifier operates in constant current source mode, linearly mapping the voltage signal into an analog current excitation signal, which is then injected into the current loop of the energy meter under test. During this process, a hardware phase-locked loop (PLL) circuit monitors the zero-crossing point of the analog current excitation signal waveform in real time. When the signal is detected crossing zero potential from the negative half-cycle to the positive half-cycle, the PLL outputs a high-level pulse as a synchronization trigger signal. This signal is directly connected to the start-up terminal of the DAC, ensuring that the timing reference for data acquisition is strictly anchored to the physical phase of the excitation signal.

[0046] Driven by a synchronous clock, the multi-channel data acquisition system operates in parallel. The first and second channels acquire discrete current sequences from the secondary side of the transformer and discrete voltage sequences from the voltage sampling terminals, respectively, with a sampling length of 1024 points covering 10 complete power frequency cycles. Simultaneously, the third channel reads the feedback signal from the thermistor embedded in the core of the transformer in the energy meter under test, for example, acquiring a resistance of 2450 ohms; the fourth channel reads the DC bias component fed back by a Hall sensor in the signal transmission link, for example, 1.5 amps; and the fifth channel calculates the grid frequency drift value using the zero-crossing counting method, for example, -0.2 Hz. All the acquired physical quantities are encapsulated into a structured heterogeneous operating condition dataset. This structured heterogeneous operating condition dataset is encapsulated using a standardized data recording format, which strictly consists of a header information dictionary and a load data array. The header information dictionary stores the data alignment reference and metadata, including a unified clock timestamp, the base sampling frequency (12.8 kHz), the total length of sampling points (1024 points), and the unit identifiers for the physical quantities of each channel. The payload data array contains a five-dimensional synchronous data queue: the first and second dimensions are one-dimensional arrays of discrete current and voltage sequences, respectively; the third to fifth dimensions are single feature values ​​extracted within the same sampling time window, namely the thermistor resistance, DC bias current, and frequency drift. During the encapsulation process, the same zero-crossing point triggered by the hardware phase-locked loop is used as the absolute time zero point, forcing each discrete sampling point of the current and voltage sequences to be strictly bound to the unified clock timestamp in the header information dictionary; simultaneously, the single feature values ​​of the third to fifth dimensions are appended as global attributes of the time window to the end of the payload data array.

[0047] By generating digitally synthesized distorted waveforms containing high-frequency odd harmonics through linear superposition, the real power grid environment under nonlinear load connection can be simulated, exposing potential frequency response defects in instrument transformers. Simultaneously, using a zero-crossing trigger synchronous sampling clock to perform timing alignment of the current and voltage sequences eliminates phase errors caused by asynchronous sampling, which is beneficial to the accuracy of subsequent calculations of hysteresis loop closure and area, providing a high-precision time-domain data foundation for physical field correction.

[0048] Furthermore, the heterogeneous operating condition dataset is input into the thermomagnetic dynamic correction model. The permeability attenuation coefficient is generated using the thermistor feedback signal. The eddy current loss weighting factor is calculated using the grid frequency drift value. The hysteresis loop center point offset vector is determined using the DC bias component, forming a physical field correction tensor. Physical field mapping operations are then performed on the discrete current and voltage sequences to generate the environmental hysteresis loop trajectory; this corresponds to step S2 above. (See also...) Figure 2 The specific implementation process includes:

[0049] The system receives heterogeneous operating condition datasets and separates the thermistor feedback signal, grid frequency drift value, and DC bias component. A three-dimensional orthogonal correction matrix containing temperature, frequency, and magnetic bias dimensions is constructed as a thermomagnetic dynamic correction model. The thermistor feedback signal is mapped to the temperature dimension, and the permeability attenuation coefficient is analyzed by retrieving the core permeability temperature drift curve. The grid frequency drift value is mapped to the frequency dimension, and the eddy current loss weighting factor is analyzed using a variant of the Steinmetz formula. The DC bias component is mapped to the magnetic bias dimension, and the hysteresis loop center point offset vector is analyzed. The physical field correction tensor is constructed from the permeability attenuation coefficient, the eddy current loss weighting factor, and the hysteresis loop center point offset vector.

[0050] The system receives heterogeneous operating condition datasets containing discrete current sequences, discrete voltage sequences, thermistor resistance values, DC bias current, and frequency drift. The core of the thermomagnetic dynamic correction model is a three-dimensional orthogonal correction matrix based on a hybrid physical mechanism and data-driven approach. Its training and calibration processes must be completed during the calibration phase before the energy meter leaves the factory. During training, a gold standard dataset covering the entire temperature range (minus 40°C to 85°C), the entire frequency band (45 Hz to 55 Hz), and the entire DC bias (0 to 10 amps) needs to be constructed.

[0051] Specifically, the generation process of the physics field correction tensor is as follows:

[0052] The system receives feedback signals from a thermistor deployed on the surface of the transformer core, for example, a resistance of 2450 ohms. Using a simplified form of the Steinhardt-Hart equation, the material constant of the thermistor is set to 3950 Kelvin, and the nominal resistance is 10000 ohms at 25 degrees Celsius. The calculation formula is described as: the reciprocal of the current thermodynamic temperature equals the reciprocal of the reference thermodynamic temperature (298.15 Kelvin), plus the quotient obtained by dividing the natural logarithm of the ratio of the current resistance to the nominal resistance by the material constant. The calculated core surface temperature is 333.5 Kelvin, or 60.35 degrees Celsius. Subsequently, a preset PC40 manganese-zinc ferrite permeability temperature drift curve is retrieved, which shows that at 60 degrees Celsius, the relative permeability of the material decreases from 510 at 25 degrees Celsius to 459. Based on this, the permeability attenuation coefficient is calculated, which is equal to the relative permeability at the current temperature divided by the relative permeability at the reference temperature. The calculated result is approximately 0.90.

[0053] In the frequency dimension, the grid frequency drift value is read, for example, -0.2 Hz. Combined with the rated power frequency of 50 Hz, the actual frequency is calculated to be 49.8 Hz. A modified variant of the Steinmetz formula is used to calculate the eddy current loss weighting factor. Since the eddy current loss power of the transformer core is approximately proportional to the square of the frequency in the low-frequency range, the formula is described as: the eddy current loss weighting factor equals the square of the ratio of the actual frequency to the rated frequency. Substituting the values, the square of the quotient of 49.8 divided by 50.0 is approximately equal to 0.9920. This factor quantifies the reduction in the resistive component of eddy current loss caused by a slight frequency drop.

[0054] In the magnetic bias dimension, the DC bias component in the signal link is obtained, for example, 1.5 amperes. Based on Ampere's circuital law, the DC offset of the magnetic field strength is calculated, described by the formula: the number of turns on the primary side of the transformer multiplied by the DC bias current, and the product divided by the effective magnetic path length of the core. Assuming the primary side has 1 turn, the secondary winding has 2000 turns, the effective cross-sectional area of ​​the core is 2.5 square centimeters, and the effective magnetic path length is 0.15 meters, the calculated offset is 10 amperes per meter. Based on this, a hysteresis loop center point offset vector is constructed, with an abscissa component of 10 and a ordinate component of 0. This vector quantitatively describes the biasing effect of the static magnetic field on the dynamic AC magnetic field. The structural parameters, such as the number of turns on the primary side and the effective magnetic path length, can be obtained by scanning the QR code on the energy meter nameplate or pre-entered by the operator through the system's human-machine interface. The permeability attenuation coefficient (0.90), eddy current loss weighting factor (0.9920), and hysteresis loop center point offset vector (10, 0) generated in the above steps are combined into a physical field correction tensor. A three-dimensional orthogonal correction matrix containing temperature, frequency, and magnetic bias dimensions is constructed. This matrix is ​​specifically a third-order diagonal matrix, where the three elements on the main diagonal are the normalized permeability attenuation coefficient, eddy current loss weighting factor, and magnitude of the hysteresis loop center point offset vector, respectively. The orthogonalization process adopts the Gram-Schmidt orthogonalization method based on physical dimension normalization.

[0055] Complex environmental disturbances are orthogonally decomposed into three independent dimensions: temperature, frequency, and magnetic bias. Thermal effects are corrected using the permeability-temperature drift curve, eddy current losses caused by frequency fluctuations are quantified using a variant of the Steinmetz formula, and the magnetic bias dimension is used to locate the magnetic flux offset caused by the DC component. This multi-dimensional analytical mechanism can calculate each component of the physical field correction tensor, which is beneficial for improving the analytical accuracy of the correction model under concurrent multi-physics disturbances.

[0056] The system receives the physical field correction tensor; corrects the amplitude of the current discrete sequence using the permeability attenuation coefficient to generate an equivalent magnetic field strength sequence; performs phase compensation operation on the voltage discrete sequence using the eddy current loss weighting factor to generate an equivalent magnetic induction intensity sequence; maps the equivalent magnetic field strength sequence to abscissa data; maps the equivalent magnetic induction intensity sequence to ordinate data; performs coordinate system translation transformation on the abscissa and ordinate data using the hysteresis loop center point offset vector; and maps the transformed abscissa and ordinate data according to timestamps to generate the environmental hysteresis loop trajectory.

[0057] Specifically, the process of generating the environmental hysteresis loop trajectory is as follows:

[0058] First, using the physical field correction tensor combined with excitation component separation logic, a multidimensional physical quantity algebraic mapping operation is performed on the original discrete sequence. Addressing the physical constraint that the sampled secondary current of the current transformer cannot be directly equated to the main magnetic flux source inside the core due to load influence, a pre-set excitation component separation model is invoked. The collected discrete secondary current sequence is mapped and converted into an equivalent primary excitation current sequence by combining the known transformer ratio parameters. Subsequently, amplitude correction is performed on this equivalent primary excitation current sequence to generate the magnetic field strength. The specific calculation logic is as follows: each discrete instantaneous value in the equivalent primary excitation current sequence is multiplied by the number of turns of the primary coil of the current transformer, and the resulting product is divided by the effective magnetic path length of the core to calculate the basic magnetic field strength sequence under ideal conditions. Then, each value in this basic magnetic field strength sequence is divided by the permeability attenuation coefficient in the physical field correction tensor to compensate for the decrease in permeability caused by high-temperature environments, ultimately generating an equivalent magnetic field strength sequence that accurately reflects the actual physical operating conditions.

[0059] Secondly, phase compensation and integral transformation based on calibration mapping are performed on the discrete voltage sequence. To accurately correct the eddy current phase shift caused by frequency drift, a phase mapping data dictionary pre-built during the equipment's factory calibration phase is invoked.

[0060] The specific construction process of this data dictionary is as follows: During the calibration phase, a test excitation with a frequency continuously stepping between 45 Hz and 55 Hz is injected into the current transformer using a high-precision programmable power supply. Simultaneously, a standard flux measuring instrument without phase shift is used to capture the actual flux waveform of the iron core. The corresponding eddy current loss weighting factor sequence is calculated based on the injected frequency, and the measured hysteresis angles of the fundamental wave and each harmonic are extracted from the standard measuring instrument. Subsequently, key-value pairs are constructed and persistently stored using the eddy current loss weighting factor as the index key and the measured hysteresis angle as the mapping value, thus establishing the empirical correspondence between the two under various operating conditions. For example, when the calibration frequency results in an eddy current loss weighting factor calculation of 0.992, if the standard instrument measures the actual fundamental wave hysteresis angle to be 0.15 radians, the system generates a record in the dictionary with 0.992 as the index key and 0.15 radians as the precise mapping value.

[0061] The specific compensation process is as follows: The eddy current loss weighting factor calculated in real time is used as a search keyword and input into the phase mapping data dictionary. If the search keyword accurately matches an existing key value, the corresponding mapping angle is directly extracted; if the search keyword lies between two adjacent discrete key values ​​in the dictionary, a first-order linear interpolation operation is performed to calculate the highly matched true eddy current hysteresis angle, which is then strictly defined as the phase compensation angle required for subsequent processing. Next, a fast Fourier transform is performed on the discrete voltage sequence to convert it to the frequency domain, extracting the fundamental wave and each harmonic component; the phase compensation angle extracted by the above table lookup or interpolation is accurately superimposed onto the phase spectrum of each frequency component, and an inverse fast Fourier transform is performed to restore it to the time domain waveform, thereby completing the phase correction of the voltage sequence in an objective data-driven manner.

[0062] Subsequently, trapezoidal numerical integration is performed on the corrected voltage sequence, which involves calculating the cumulative sum of the trapezoidal surface formed by adjacent voltage sampling points and time intervals. The accumulated result is then divided by the product of the number of turns in the secondary winding of the transformer and the effective cross-sectional area of ​​the core, generating an equivalent magnetic flux density sequence. Finally, the equivalent magnetic flux density sequence is mapped to the x-axis, and the equivalent magnetic flux density sequence is mapped to the y-axis. A coordinate system translation transformation is then performed using the hysteresis loop center point offset vector (the x-axis is incremented by 10). The transformed point sets are connected in time stamp order, ultimately generating a closed curve offset towards the first quadrant with a "flattened" top, i.e., the environmental hysteresis loop trajectory.

[0063] By utilizing the physical field correction tensor, the voltage and current data, which originally only represented port characteristics, are transformed into an equivalent magnetic field strength and magnetic induction intensity sequence that reflects the movement law of magnetic domains inside the iron core. Furthermore, through coordinate system translation transformation, the asymmetry of the hysteresis loop caused by DC bias is effectively restored, enabling the generated trajectory to realistically reproduce the dynamic behavior of the transformer in the nonlinear operating region. This provides a physically accurate geometric object for subsequent graphical topology analysis.

[0064] Further, the environmental hysteresis loop trajectory is received, and a piecewise linearization analysis operation is performed to identify the inflection point coordinates; the area of ​​the saturation region above the inflection point coordinates and the area of ​​the linear region below the inflection point coordinates are extracted, and the area ratio is calculated to generate the core depth saturation distortion index; this corresponds to step S3 above; see [link to relevant documentation]. Figure 3 The specific implementation process includes:

[0065] The system receives the environmental hysteresis loop trajectory and calculates the tangent slope sequence of adjacent discrete points. It then performs a second-order difference operation on the tangent slope sequence to generate a curvature mutation spectrum. A magnetic saturation critical curvature threshold is set. The curvature mutation spectrum is scanned, and the extreme points exceeding the magnetic saturation critical curvature threshold are located. The magnetic field strength coordinates and magnetic induction intensity coordinates corresponding to the extreme point positions are locked as magnetic saturation inflection point coordinates. Using the magnetic saturation inflection point coordinates, the environmental hysteresis loop trajectory is physically divided into a nonlinear saturation segment above the inflection point and a linear magnetization segment below the inflection point.

[0066] Specifically, the process of identifying the inflection point coordinates and generating the curvature abrupt change spectrum of the environmental hysteresis loop trajectory is as follows:

[0067] The system receives environmental hysteresis loop trajectory data containing 1024 discrete coordinate points. Since the hysteresis loop is a closed curve, to avoid slope calculation overflow at extreme endpoints due to the zero rate of change of the abscissa, the closed hysteresis loop trajectory is first physically divided into an upper branch with decreasing magnetic field strength and a lower branch with increasing magnetic field strength, based on the changing trend of the magnetic field strength at adjacent discrete points. Subsequent analytical operations will be performed independently and in parallel for the upper and lower branches. First, a first-order difference operation is performed to obtain the tangent slope sequence of adjacent discrete points. The calculation logic is described as follows: the tangent slope of the current point equals the difference between the ordinate (magnetic flux density) of the next point and the ordinate of the current point, divided by the difference between the abscissa (magnetic flux density) of the next point and the abscissa of the current point. Next, a second-order difference operation is performed on this tangent slope sequence, and the absolute value of the difference between adjacent slopes is taken to generate a curvature abrupt change spectrum. Since the first derivative of the magnetization curve changes little near the magnetic saturation inflection point, this embodiment uses a simplified second-order differential modulus as an approximate representation of curvature. The spectrum reflects the rate of change of permeability. In the linear region of the magnetization curve, the permeability is basically constant and the curvature spectrum value is close to 0; while in the transition region entering saturation, the permeability drops sharply and the curvature spectrum shows a significant peak.

[0068] In this embodiment, Gaussian smoothing filtering is performed on the curvature abrupt spectrum to remove high-frequency quantization noise. Then, the energy density integral within a local window is calculated to generate an energy density-weighted smoothed curvature spectrum. Since second-order difference operations amplify analog-to-digital conversion quantization noise during sampling, the directly generated curvature abrupt spectrum often contains numerous spurious spikes. Therefore, a Gaussian convolution kernel with a standard deviation of 1.5 is introduced to smooth the original curvature spectrum, utilizing the exponential decay characteristic of the Gaussian function to eliminate glitches. Next, a sliding window with a width of 5 sampling points is constructed, and the root mean square of all curvature values ​​within the window is calculated as the energy feature value of the window's center point. Subsequently, the amplitude of the current point in the original curvature spectrum is multiplied by this energy feature value to obtain the energy-weighted curvature abrupt spectrum. Only when the smoothed curvature amplitude and its corresponding energy density value simultaneously exhibit extreme value characteristics is the point marked as a valid candidate point. This step, by introducing energy dimension verification, eliminates pseudo-mutation points that have large amplitudes but extremely short durations, ensuring that the inflection points located subsequently are real physical magnetic domain reversal regions, rather than circuit noise interference.

[0069] A preset critical curvature threshold for magnetic saturation is invoked, for example, set to 50. This threshold is obtained by calling a calibration dataset pre-built in a standard laboratory environment, calculating the average curvature value of the transformer material in the transition region between the linear and saturation regions, and taking 1.5 times this average curvature value as the current critical curvature threshold for magnetic saturation. By scanning the curvature abrupt change spectrum, a point is identified as a local maximum only if the curvature abrupt change value at a certain point is greater than the critical curvature threshold for magnetic saturation, and the curvature abrupt change value at that point is simultaneously strictly greater than the curvature abrupt change values ​​at both its preceding and following adjacent points. For example, if a curvature abrupt change peak is detected at a magnetic field strength of 450 amperes per meter, the coordinates corresponding to that location, i.e., magnetic field strength of 450 amperes per meter and magnetic induction intensity of 400 millitalas, are locked as the coordinates of the magnetic saturation inflection point. The magnetic field strength coordinates and magnetic induction intensity coordinates corresponding to the first local maximum point locations in the upper and lower branches are locked as the coordinates of the magnetic saturation inflection points of their respective branches.

[0070] Using the identified coordinates of the magnetic saturation inflection point, the closed hysteresis loop trajectory is physically divided into two independent geometric segments. The segmentation logic is based on the absolute value of the abscissa: the set of discrete points whose absolute abscissa values ​​are greater than the absolute value of the inflection point (e.g., 450 amperes per meter) is defined as the nonlinear saturation segment, which characterizes the flat response characteristics of the magnetic core after entering deep saturation, where the magnetic induction intensity changes very little with the increase of the magnetic field; the set of discrete points whose absolute abscissa values ​​are less than or equal to the absolute value of the inflection point is defined as the linear magnetization segment, which characterizes the linear magnetic permeability of the magnetic core in the unsaturated state.

[0071] By setting a critical curvature threshold for magnetic saturation, the mathematical inflection point of the transition from the linear region to the nonlinear region on the hysteresis loop can be locked, dividing the continuous magnetization trajectory into saturation regions and linear regions with different properties, laying the topological boundary foundation for subsequent refined quantitative evaluation based on area ratio.

[0072] The algorithm receives nonlinear saturation and linear magnetization regions; performs discrete Riemann integration on the nonlinear saturation region to calculate the saturation area; performs trapezoidal integration on the linear magnetization region to calculate the linear area; constructs a dimensionless performance comparison function; inputs the saturation area as the numerator and the linear area as the denominator into the dimensionless performance comparison function; performs division and logarithmically scales the calculation results; and generates the core depth saturation distortion index.

[0073] Specifically, the process of generating the core depth saturation distortion index is as follows:

[0074] For the two segments after segmentation, a differentiated numerical integration strategy is adopted. For the nonlinear saturation segment, due to its gentle curve, the discrete Riemann integral method is used for calculation. The calculation formula is described as follows: the area of ​​the saturation region is equal to the sum of the absolute values ​​of the magnetic induction intensity at all discrete points in the segment multiplied by the corresponding infinitesimal components of the magnetic field intensity (i.e., the magnetic field increment within the sampling interval). For example, the normalized area of ​​the saturation region is calculated to be 0.5.

[0075] For the linear magnetization region, due to its large slope and inclusion of zero-crossing points, a trapezoidal integral method is used to reduce truncation error. The calculation formula is as follows: the area of ​​the linear region is equal to the sum of the magnetic induction intensities of any two adjacent discrete points within the region multiplied by half the difference in magnetic field intensity between the two points, and then the areas of all infinitesimal trapezoidal elements are summed. For example, the normalized area of ​​the linear region is calculated to be 3.2.

[0076] First, a dimensionless efficiency comparison function is constructed, defined as the ratio of the saturated region area to the linear region area. Using the calculated saturated region area (0.5) as the numerator and the linear region area (3.2) as the denominator, a division operation is performed to obtain the area ratio, which is approximately 0.156. This ratio physically reflects the proportion of "distortion energy" in the total magnetization energy. To expand the dynamic range of the data and improve the resolution at low saturation, this ratio is logarithmically scaled. The calculation formula is described as: the core depth saturation distortion index equals 10 multiplied by the logarithm of the area ratio to the base 10. Substituting 0.156 into the calculation, its commonly used logarithm is approximately -0.806. Multiplying by 10 yields the final core depth saturation distortion index of -8.06 dB. This index intuitively quantifies the severity of magnetic saturation; the closer the index value is to zero, the higher the proportion of saturation distortion and the greater the risk of measurement error.

[0077] By constructing a dimensionless performance comparison function and performing logarithmic scaling, the problems of single area index being greatly affected by load size and having insufficient dynamic range are solved. The interference of current amplitude fluctuations on the evaluation index is eliminated, allowing the index to focus on the degree of waveform distortion itself. Simultaneously, logarithmic processing greatly expands the resolution of the index in the weak saturation stage, enabling more sensitive detection of subtle nonlinear trends when the core is just entering soft saturation.

[0078] Furthermore, the core depth saturation distortion index is projected onto a preset measurement error tolerance evaluation range, and a magnetic saturation measurement anomaly signal is generated based on the anomaly judgment logic triggered by the landing point position; corresponding to step S4 above; the specific implementation process includes:

[0079] Receive the core depth saturation distortion index; retrieve the graded evaluation interval table constructed according to the accuracy level standard of the electricity meter; perform a numerical comparison operation between the core depth saturation distortion index and the safety threshold and fault threshold in the graded evaluation interval table; if the index is less than the safety threshold, it falls into the normal operation interval; if the index is between the safety threshold and the fault threshold, it falls into the accuracy degradation interval; if the index is greater than the fault threshold, it falls into the functional failure interval; output the calibrated landing point interval position status code; activate the interrupt request for the landing point interval position status code that falls into the functional failure interval; call the preset fault semantic mapping dictionary; use the landing point interval position status code as the index key for retrieval and matching; if the landing point interval position status code points to the accuracy degradation interval, extract the soft magnetic saturation overload tag as the transformer overload type; if the landing point interval position status code points to the functional failure interval, extract the hard magnetic flux truncation overload tag as the transformer overload type; call the anomaly code generator to construct a fault message containing the current timestamp, the amplitude of the core depth saturation distortion index, and the transformer overload type; encapsulate the fault message into a magnetic saturation metering anomaly signal.

[0080] Specifically, the construction and calibration process of the graded evaluation interval table is as follows:

[0081] The grading evaluation interval table was constructed based on a calibration experiment on the "physical-error" correlation of a 0.5S-class energy meter transformer. During the calibration phase, a test set was constructed containing different DC biases (0 amps to 5 amps) and temperatures (25°C to 85°C). With minimizing the "false alarm rate" as the optimization objective, the loss function was defined as the number of samples where the actual measurement error exceeded the 0.5% limit but did not trigger the threshold. Through regression analysis of a large number of samples, the nonlinear mapping relationship between the core depth saturation distortion index and the measurement error was determined. The final parameters of the grading evaluation interval table are as follows: the safety threshold is set at -25 dB, corresponding to the linear operating region where the measurement error is less than 0.1%; the fault threshold is set at -10 dB, corresponding to the critical saturation point where the measurement error reaches 0.5%. The region between these two values ​​is defined as the accuracy degradation interval. For example, when the calibration experiment showed that the distortion index reached -15 dB when the DC bias was 1.2 amperes, the error of the electricity meter was about 0.3%. Although it was not out of tolerance, it had deviated from the ideal linear range, so it was classified as a warning range.

[0082] In this embodiment, the cumulative operating time and historical overload count of the energy meter under test are read to calculate the device aging drift factor. This factor is then used to dynamically correct the thresholds at each level in the graded evaluation interval table. Considering that the core material of the transformer (such as permalloy or nanocrystalline) will naturally age over time, leading to a decrease in the magnetic permeability baseline, the cumulative power-on time (in hours) in the energy meter's memory is read. The aging drift calculation logic is defined as follows: take the material aging constant (set to 0.02), multiply it by the natural logarithm of the ratio of the cumulative power-on time to the baseline constant (set to 8760 hours). The calculated aging drift value is then superimposed on the safety threshold and warning threshold of the graded evaluation interval table, and a correction operation is performed; that is, the new threshold equals the old threshold multiplied by 1 plus the aging drift value. This means that for old meters that have been in operation for more than five years, the judgment criteria will be automatically relaxed. The above process can distinguish between reasonable performance degradation caused by natural material aging and magnetic saturation anomalies caused by sudden faults, preventing false alarms for old equipment.

[0083] Specifically, the generation process of the magnetic saturation metering anomaly signal is as follows:

[0084] The system receives the core depth saturation distortion index for the current period, for example, -8.06 dB. First, it compares this value with the graded evaluation interval table level by level. The logical judgment order is as follows: First, it checks if the index is less than the safety threshold (-25 dB). If so, it is considered normal; if not, it continues to check if the index is greater than the fault threshold (-10 dB). In this example, -8.06 is greater than -10, so the judgment logic directly triggers the "functional failure" branch. Then, the corresponding landing point interval position status code is generated, defining the normal interval status code as 0x01, the accuracy degradation interval status code as 0x02, and the functional failure interval status code as 0x03. For this example, the output status code is 0x03. For this critical status code, the highest priority hardware interrupt request is immediately activated, the current regular metering task is suspended, and the system forcibly enters the fault handling subroutine to ensure that the abnormal event is responded to immediately, with the delay time controlled within milliseconds.

[0085] In the interrupt service routine, a fault semantic mapping dictionary pre-loaded in read-only memory is invoked. This dictionary is a key-value pair structure, using the status code as the index key. Using the current status code 0x03 as the key input, the retrieved overload type for the current transformer is "hard flux cutoff overload." This clearly indicates a physical phenomenon where the waveform top is severely clipped and the flux change rate approaches zero within half a cycle, significantly different from "soft magnetic saturation overload" (corresponding to status code 0x02), which only causes slight waveform asymmetry. Subsequently, an anomaly encoding generator is invoked to construct a fault message. This message contains three core data fields: first, the current timestamp generated using a real-time clock; second, the core depth saturation distortion index amplitude, retained to two decimal places, i.e., "-8.06 dB"; and third, the retrieved fault type text. Finally, these three data fields are encapsulated into a 128-byte magnetic saturation metering anomaly signal frame using a CRC32 checksum algorithm, ready to be sent to the master system via the RS-485 bus.

[0086] The fault semantic mapping dictionary is constructed based on destructive test data from energy meter type tests. In a laboratory environment, long-term weak DC injection (corresponding to soft saturation) and instantaneous large current impact (corresponding to hard cutoff) are simulated respectively, and the distortion index characteristics under different fault modes are recorded to establish a unique mapping relationship between status codes and physical causes.

[0087] The numerical distortion index is mapped to specific fault semantics, distinguishing between accuracy degradation caused by environmental drift and functional failure caused by physical damage or severe overcurrent. This hierarchical classification mechanism not only provides abnormal status codes but also generates fault messages containing specific overload types, providing maintenance personnel with a clearer decision-making basis for taking targeted maintenance measures.

[0088] This invention provides a method for detecting anomalies in electricity meters. It introduces multi-dimensional variables such as thermistor, frequency drift, and DC bias to construct a corrected model incorporating thermodynamic and electromagnetic constraints. Through physical field mapping calculations, the distorted current-voltage sequence affected by environmental factors is restored to an environmental hysteresis loop trajectory that accurately reflects the magnetization state of the iron core. This method transforms abstract metering errors into visualized hysteresis loop distortion, enabling not only the identification of hard truncation caused by large currents but also the sensitive detection of permeability decay and soft saturation caused by temperature increases or DC intrusion. The resulting core depth saturation distortion index quantifies the degree to which the iron core deviates from the linear region, thus achieving accurate determination of metering anomalies under complex and variable field conditions.

[0089] Example 2

[0090] This embodiment describes the specific implementation process of an energy meter anomaly detection system applied to an energy meter at a photovoltaic power generation grid connection point. This energy meter is subjected to a combined operating condition of high-frequency switching noise from the inverter and DC component injection over a long period. (See also...) Figure 4 The specific implementation process is as follows:

[0091] First, a wideband programmable power source is activated to simulate the output characteristics of a photovoltaic inverter under non-ideal grid-connected conditions. The internal harmonic synthesis logic is invoked, setting the fundamental frequency to 50 Hz and the amplitude to 5 amps, while superimposing a high-frequency ripple component with a frequency of 2 kHz and an amplitude of 0.2 amps. Simultaneously, a 1.2 amp DC bias current is artificially introduced into the waveform. After performing linear superposition, the generated digitally synthesized distorted waveform is sent to a 16-bit high-precision digital-to-analog converter unit, converted into an analog current signal, and injected into the 0.5-level precision energy meter under test via a power amplification link. At this point, the effective value of the discrete current sequence on the secondary side of the transformer is 5.1 amps, and the effective value of the discrete voltage sequence is 57.7 volts. The thermistor deployed on the surface of the transformer core is read, with a feedback resistance of 2200 ohms, indicating that the core temperature has risen to 65 degrees Celsius due to environmental and self-heating effects. The Hall sensor in the signal link detects the 1.2 amp DC bias component. The grid frequency monitoring module captures a 0.1 Hz negative frequency drift. All the above physical quantities are tagged with a unified timestamp to generate a heterogeneous operating condition dataset.

[0092] First, the thermistor signal was analyzed, and the absolute temperature of the iron core was calculated to be 338.15 Kelvin using the thermistor material constant. Referring to a pre-set data table of manganese-zinc ferrite magnetic properties, the saturation magnetic induction intensity of this material at 65 degrees Celsius decreased by approximately 10% compared to its nominal value at 25 degrees Celsius. Based on this, the system generated a permeability attenuation coefficient of 0.90. For the actual frequency of 49.9 Hz, the eddy current loss weighting factor was calculated to be 0.998 using the modified Steinmetz formula, and a slight lead compensation was applied to the voltage phase. Most importantly, for a DC bias of 1.2 amperes, combined with the single-turn coil on the primary side of the transformer and an effective magnetic circuit length of 0.15 meters, a significant shift of 8 amperes per meter in the magnetic field strength along the horizontal axis was calculated. After the physical field correction tensor is constructed, the system uses the permeability attenuation coefficient to perform amplitude weighted amplification of the current sequence to simulate the decrease in the energy storage capacity of the magnetic core at high temperature. The system also uses the offset vector to perform a translation transformation of the coordinate system, and finally generates an environmental hysteresis loop trajectory that is severely offset to the first quadrant and has a flat-top clipping feature at the top of the positive half-cycle.

[0093] Next, by calculating the tangent slope sequence of discrete points on the trajectory and performing second-order difference operations, the system detected a curvature abrupt change exceeding the set critical threshold of 50 at a magnetic field strength of 420 amperes per meter, thus accurately locking the coordinates of the magnetic saturation inflection point. This inflection point is used to physically divide the trajectory into a nonlinear saturation segment and a linear magnetization segment. For the nonlinear saturation segment, the discrete Riemann integral algorithm is used to calculate its normalized area as 0.72; for the linear magnetization segment, the trapezoidal integral algorithm is used to calculate its area as 3.05. The system constructs a dimensionless efficiency comparison function, dividing the saturation area by the linear area to obtain an area ratio of 0.236. Subsequently, this ratio is logarithmically scaled, i.e., the base-10 logarithm of the ratio is calculated and multiplied by 10, ultimately generating a core depth saturation distortion index of -6.27 dB.

[0094] Finally, the system retrieves the preset metering error tolerance evaluation interval table, compares the calculated -6.27 dB with the threshold, and the result shows that the current index is significantly higher than the fault threshold, indicating that the transformer is in a deep saturation state. At this time, the theoretically estimated metering error exceeds 0.8%, far exceeding the allowable error limit of a 0.5 class energy meter. The judgment logic then triggers the status code of the functional failure interval. After the interrupt handler is activated, it searches the fault semantic mapping dictionary and extracts the fault type label "hard flux cutoff overload" based on the status code. The system calls the anomaly code generator, encapsulates the current time, the distortion index amplitude of -6.27 dB, and the fault type label into a standard fault message, and outputs a magnetic saturation metering anomaly signal, prompting the control center that there is a serious DC bias and thermal saturation combined fault at the photovoltaic metering point, requiring immediate equipment maintenance.

[0095] Through the collaborative operation of four modules—"data acquisition," "magnetic field correction," "deviation analysis," and "anomaly detection"—the anomaly detection process has been engineered. This system architecture solidifies complex physical field correction algorithms into independent hardware or logic modules, ensuring the efficiency and stability of data stream processing. In particular, the pipelined design of the magnetic field correction and deviation analysis modules ensures that the system can extract environmental hysteresis loop characteristics during the operation of the electricity meter, achieving fully automated monitoring from signal acquisition to fault diagnosis.

[0096] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for detecting anomalies in an electricity meter, characterized in that, include: A wide-band programmable power source is activated to apply a composite excitation signal to the energy meter under test. Collect the discrete current sequence on the secondary side of the current transformer of the energy meter under test, the discrete voltage sequence at the voltage sampling terminal, the feedback signal of the thermistor deployed on the surface of the transformer core, the DC bias component in the signal transmission link, and the grid frequency drift value to generate a heterogeneous operating condition dataset. The heterogeneous operating condition dataset is input into the thermomagnetic dynamic correction model. The permeability attenuation coefficient is generated using the thermistor feedback signal. The eddy current loss weighting factor is calculated using the grid frequency drift value. The hysteresis loop center point offset vector is determined using the DC bias component. The physical field correction tensor is constructed. The physical field mapping operation is performed on the current discrete sequence and the voltage discrete sequence to generate the environmental hysteresis loop trajectory. The specific generation process includes: receiving the physical field correction tensor; The amplitude of the current discrete sequence is corrected using the permeability attenuation coefficient to generate an equivalent magnetic field strength sequence; the phase compensation operation of the voltage discrete sequence is performed using the eddy current loss weighting factor to generate an equivalent magnetic induction intensity sequence; the equivalent magnetic field strength sequence is mapped to abscissa data; the equivalent magnetic induction intensity sequence is mapped to ordinate data; the coordinate system translation transformation of the abscissa and ordinate data is performed using the hysteresis loop center point offset vector; the transformed abscissa and ordinate data are matched according to timestamps to generate the environmental hysteresis loop trajectory. Receive the environmental hysteresis loop trajectory, perform piecewise linearization analysis, and identify the inflection point coordinates; extract the area of ​​the saturation region above the inflection point coordinates and the area of ​​the linear region below the inflection point coordinates, calculate the area ratio, and generate the core depth saturation distortion index. The magnetic core depth saturation distortion index is projected onto a preset measurement error tolerance evaluation range, and a magnetic saturation measurement anomaly signal is generated based on the anomaly judgment logic triggered by the landing point position.

2. The method for detecting anomalies in an electricity meter according to claim 1, characterized in that, The specific process of loading the composite excitation signal includes: calling harmonic synthesis logic to generate a fundamental sine sequence and a high-frequency odd harmonic sequence; performing linear superposition operation to generate a digitally synthesized distorted waveform; driving a digital-to-analog converter to convert the digitally synthesized distorted waveform into an analog current excitation signal; injecting the analog current excitation signal into the current loop of the energy meter under test through a power amplification link; using a hardware phase-locked loop to lock the zero-crossing point of the analog current excitation signal; and triggering a synchronous sampling clock based on the zero-crossing point to perform timing alignment operations on the discrete current sequence and the discrete voltage sequence.

3. The method for detecting anomalies in an electricity meter according to claim 1, characterized in that, The specific processing steps of the thermomagnetic dynamic correction model include: receiving heterogeneous operating condition datasets and separating the thermistor feedback signal, grid frequency drift value, and DC bias component; constructing a three-dimensional orthogonal correction matrix containing temperature, frequency, and magnetic bias dimensions as the thermomagnetic dynamic correction model; mapping the thermistor feedback signal to the temperature dimension and retrieving the core permeability temperature drift curve to obtain the permeability attenuation coefficient; mapping the grid frequency drift value to the frequency dimension and retrieving the eddy current loss weighting factor using a variant of the Steinmetz formula; mapping the DC bias component to the magnetic bias dimension and retrieving the hysteresis loop center point offset vector; and constructing a physical field correction tensor by the permeability attenuation coefficient, eddy current loss weighting factor, and hysteresis loop center point offset vector.

4. The method for detecting anomalies in an electricity meter according to claim 1, characterized in that, The specific identification process of the inflection point coordinates includes: receiving the environmental hysteresis loop trajectory and calculating the tangent slope sequence of adjacent discrete points; performing a second-order difference operation on the tangent slope sequence to generate a curvature mutation spectrum; setting a magnetic saturation critical curvature threshold; scanning the curvature mutation spectrum and locating the extreme point positions that exceed the magnetic saturation critical curvature threshold; locking the magnetic field strength coordinates and magnetic induction intensity coordinates corresponding to the extreme point positions as magnetic saturation inflection point coordinates; and using the magnetic saturation inflection point coordinates to physically divide the environmental hysteresis loop trajectory into a nonlinear saturation segment above the inflection point and a linear magnetization segment below the inflection point.

5. The method for detecting abnormalities in an electricity meter according to claim 1, characterized in that, The specific generation process of the core depth saturation distortion index includes: receiving a nonlinear saturation segment and a linear magnetization segment; performing a discrete Riemann integral operation on the nonlinear saturation segment to calculate the saturation area; performing a trapezoidal integral operation on the linear magnetization segment to calculate the linear area; constructing a dimensionless performance comparison function; inputting the saturation area as the numerator and the linear area as the denominator into the dimensionless performance comparison function; performing a division operation and logarithmically scaling the calculation result; and generating the core depth saturation distortion index.

6. The method for detecting anomalies in an electricity meter according to claim 1, characterized in that, The specific generation process of the magnetic saturation metering anomaly signal includes: receiving the magnetic core depth saturation distortion index; retrieving the graded evaluation interval table constructed according to the accuracy level standard of the electricity meter; performing a numerical comparison operation between the magnetic core depth saturation distortion index and the safety threshold and fault threshold in the graded evaluation interval table; if it is less than the safety threshold, it falls into the normal operation interval; if it is between the safety threshold and the fault threshold, it falls into the accuracy degradation interval; if it is greater than the fault threshold, it falls into the functional failure interval; outputting the calibrated landing point interval position status code; and checking the landing point interval position status code for those falling into the functional failure interval. The code activates the interrupt request; it calls the preset fault semantic mapping dictionary; it uses the landing point interval position status code as the index key for retrieval and matching; if the landing point interval position status code points to the accuracy deterioration interval, it extracts the soft magnetic saturation overload tag as the current transformer overload type; if the landing point interval position status code points to the functional failure interval, it extracts the hard magnetic flux truncation overload tag as the current transformer overload type; it calls the anomaly encoding generator to construct a fault message containing the current timestamp, the magnetic core depth saturation distortion index amplitude, and the current transformer overload type; and it encapsulates the fault message into a magnetic saturation metering anomaly signal.

7. An electricity meter anomaly detection system, characterized in that, include: Data capture module: Starts a wideband programmable power source to load a composite excitation signal onto the energy meter under test; collects the discrete current sequence on the secondary side of the current transformer of the energy meter under test, the discrete voltage sequence at the voltage sampling terminal, the feedback signal of the thermistor deployed on the surface of the transformer core, the DC bias component in the signal transmission link, and the grid frequency drift value, and generates a heterogeneous operating condition dataset. Magnetic field correction module: Input heterogeneous operating condition dataset into thermomagnetic dynamic correction model, generate permeability attenuation coefficient using thermistor feedback signal, calculate eddy current loss weighting factor using grid frequency drift value, determine hysteresis loop center point offset vector using DC bias component, construct physical field correction tensor, perform physical field mapping operation on current discrete sequence and voltage discrete sequence, and generate environmental hysteresis loop trajectory. The specific generation process includes: receiving the physical field correction tensor; The amplitude of the current discrete sequence is corrected using the permeability attenuation coefficient to generate an equivalent magnetic field strength sequence; the phase compensation operation of the voltage discrete sequence is performed using the eddy current loss weighting factor to generate an equivalent magnetic induction intensity sequence; the equivalent magnetic field strength sequence is mapped to abscissa data; the equivalent magnetic induction intensity sequence is mapped to ordinate data; the coordinate system translation transformation of the abscissa and ordinate data is performed using the hysteresis loop center point offset vector; the transformed abscissa and ordinate data are matched according to timestamps to generate the environmental hysteresis loop trajectory. Deviation analysis module: Receives the environmental hysteresis loop trajectory, performs piecewise linearization analysis, identifies inflection point coordinates; extracts the area of ​​the saturation region above the inflection point coordinates and the area of ​​the linear region below the inflection point coordinates, calculates the area ratio, and generates the core depth saturation distortion index. Anomaly detection module: Projects the magnetic core depth saturation distortion index onto the preset measurement error tolerance evaluation range, and generates a magnetic saturation measurement anomaly signal based on the anomaly judgment logic triggered by the landing point position.

Images