Three-phase electric energy meter fault detection method and system based on zero sequence component analysis

Abstract

The present application relates to the field of electric energy meter fault detection, especially to a three-phase electric energy meter fault detection method and system based on zero sequence component analysis, comprising: synchronously collecting three-phase voltage and current data, extracting each phase fundamental electrical quantity to construct a complex vector containing amplitude and phase information; solving the zero sequence complex vector, obtaining the phase misalignment degree through conjugate operation and normalization processing; setting long and short windows, calculating the mean and standard deviation of the long window and constructing an asymmetric energy linkage operator, obtaining the energy cumulative value through the short-time cumulative window cumulative operator, calculating the adaptive gain based on the historical baseline window statistics, constructing the dynamic confidence boundary, and statistically determining the out-of-limit proportion of the characteristics in the period, thereby identifying the electric energy meter fault. The present application distinguishes power grid disturbance from internal electric meter fault through zero sequence phase characteristics, combines double windows and adaptive boundary, energy operator and multi-period proportion judgment, takes into account detection sensitivity and anti-interference ability, and accurately identifies electric meter hardware degradation.

Description

Technical Field

[0001] This invention relates to the field of electricity meter fault detection. In particular, it relates to a method and system for three-phase electricity meter fault detection based on zero-sequence component analysis. Background Technology

[0002] Three-phase electricity meters are the core equipment for electricity metering in power systems, and their operational reliability directly affects the fairness of electricity trade settlement and the safety of power grid operation. Components inside the equipment, such as voltage sampling circuits, current transformers, signal conditioning circuits, and analog-to-digital converters, are susceptible to performance degradation due to factors such as lightning overvoltage, temperature changes, component aging, and manufacturing defects, leading to metering deviations. If these faults are not diagnosed and resolved in a timely manner, they can easily cause metering inaccuracies and electricity disputes.

[0003] Currently, fault detection in three-phase electricity meters mainly employs methods such as hardware redundancy self-testing, periodic calibration of standard sources, and passive monitoring of electrical quantity thresholds. Some solutions also rely on harmonics and negative sequence components for fault diagnosis. Hardware redundancy self-testing solutions are costly and cannot cover all fault types; periodic calibration has long time intervals, making it difficult to detect sudden performance degradation of equipment; conventional passive monitoring mostly relies on fixed thresholds to determine abnormal electrical quantities.

[0004] Existing detection methods do not perform in-depth analysis of the phase relationship of electrical quantities, and cannot effectively distinguish between two types of problems: power grid load imbalance and single-phase phase drift inside the energy meter. Fluctuations in the normal operating conditions of the distribution network can easily cause false alarms from detection devices. At the same time, early weak phase shift faults of equipment are also easily missed, making it difficult to balance the reliability and sensitivity of fault detection. Summary of the Invention

[0005] To address the shortcomings of existing detection methods, such as the lack of in-depth analysis of the phase relationship of electrical quantities, difficulty in distinguishing between grid imbalance and internal phase drift of the meter, false alarms and missed alarms, and the inability to balance detection reliability and sensitivity, this invention provides solutions in the following aspects.

[0006] In the first aspect, the three-phase energy meter fault detection method based on zero-sequence component analysis includes: synchronously acquiring three-phase voltage and current sampling data, extracting the fundamental voltage and fundamental current of each phase, and converting them into complex vectors carrying amplitude and phase information respectively; solving the zero-sequence voltage complex vector and zero-sequence current complex vector according to the complex vectors of each phase voltage and current respectively, performing complex number operations by taking the conjugate of the zero-sequence voltage complex vector and the zero-sequence current complex vector, and then performing normalization processing by combining the amplitude of the zero-sequence voltage complex vector with a preset minimum constant to obtain the phase misalignment degree used to characterize the phase offset state of the zero-sequence vector; configuring two independent data windows with different durations: one of which is a sliding statistical window used to statistically analyze the background characteristics of the phase misalignment degree, and using the sliding statistical window to calculate the phase misalignment. The following methods are used to determine the degree of energy misalignment: First, the moving average and standard deviation of the phase misalignment are calculated. Second, the deviation of the phase misalignment degree from the moving average and the first-order difference of the phase misalignment degree are used to construct an asymmetric energy linkage operator. Third, a short-term accumulation window with a length shorter than the moving statistical window is used. All asymmetric energy linkage operators within the short-term accumulation window are extracted and summed to obtain the cumulative asymmetric energy value. Fourth, a historical baseline window composed of continuous historical data is extracted. The median and absolute median difference of the cumulative asymmetric energy values ​​within the historical baseline window are calculated. An adaptive gain coefficient is calculated based on the standard deviation and a preset constant. An adaptive dynamic confidence boundary is constructed based on the median, absolute median difference, and adaptive gain coefficient. Fifth, the proportion of cumulative asymmetric energy values ​​exceeding the adaptive dynamic confidence boundary within a preset period is statistically analyzed. The proportion is used to determine whether the three-phase energy meter has malfunctioned.

[0007] Preferably, the step of constructing the complex vector is as follows: Using the effective value of the fundamental voltage at the current moment as the magnitude of the complex vector of each phase voltage and the phase angle of the fundamental voltage of each phase as the vector rotation angle, construct the complex vector of the fundamental voltage of the corresponding phase. Using the effective value of the fundamental current at the current moment as the modulus of the complex vector of each phase current and the phase angle of the fundamental current of each phase as the vector rotation angle, construct the complex vector of the fundamental current of the corresponding phase.

[0008] Preferably, the calculation method for the phase misalignment degree is as follows: The zero-sequence voltage complex vector is obtained by averaging the three-phase voltage complex vectors, and the zero-sequence current complex vector is obtained by averaging the three-phase current complex vectors. The conjugate vector of the zero-sequence voltage complex vector is calculated, and the conjugate vector is multiplied by the zero-sequence current complex vector. The imaginary part of the result is extracted and the absolute value is taken. The absolute value is divided by the sum of the zero-sequence voltage vector amplitude and the preset minimum constant to obtain the phase misalignment degree.

[0009] Preferably, the asymmetric energy linkage operator is constructed as follows: The average value of the phase misalignment is calculated in real time through a preset sliding statistical window. The difference between the phase misalignment at the current moment and the average value is calculated. The difference is then raised to the power of three, and the positive and negative directions of the difference are retained. Calculate the difference between the phase misalignment degree at the current moment and the previous moment and take the absolute value to obtain the instantaneous change amplitude of the phase misalignment degree; multiply the result of the cubic operation with the absolute value difference to obtain the asymmetric energy linkage operator.

[0010] Preferably, the method for obtaining the cumulative value of asymmetric energy is as follows: Set a short-term accumulation window with a preset duration, extract the asymmetric energy linkage operator values ​​corresponding to each sampling time in the window in sequence, and sum all the operator values ​​in the window to obtain the cumulative asymmetric energy value.

[0011] Preferably, the construction steps of the adaptive dynamic confidence boundary are as follows: Retrieve all cumulative asymmetric energy values ​​within the historical baseline window, and calculate the median and absolute median difference of all cumulative asymmetric energy values; combine the standard deviation of phase misalignment output by the sliding statistical window and the preset fixed constant to calculate the adaptive gain coefficient; Adding the median to the product of the gain coefficient and the absolute median difference yields the dynamic alarm upper limit, i.e., the adaptive dynamic confidence boundary.

[0012] Preferably, the step of determining whether a three-phase energy meter has malfunctioned based on the proportion includes: A fixed-duration statistical judgment period is set, which includes multiple sets of continuous sampling data. The cumulative asymmetric energy value corresponding to each set of sampling data is compared with the adaptive dynamic confidence boundary. When the cumulative asymmetric energy value exceeds the adaptive dynamic confidence boundary, it is judged as data exceeding the limit. The number of all out-of-limit data within the statistical judgment period is determined, and the proportion of out-of-limit data in all sampled data of the period is calculated. A judgment threshold is pre-configured. If the proportion of out-of-limit data is higher than the judgment threshold, it is determined that the internal components of the three-phase energy meter have experienced performance degradation and a fault alarm signal is output. If the proportion of out-of-limit data is lower than the judgment threshold, it is determined to be an instantaneous external disturbance to the power grid and no fault alarm signal is output.

[0013] Secondly, a three-phase energy meter fault detection system based on zero-sequence component analysis includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned three-phase energy meter fault detection method based on zero-sequence component analysis is implemented.

[0014] The present invention has the following effects: 1. This invention constructs a phase misalignment characteristic quantity by mining the amplitude and phase coupling relationship of the zero-sequence vectors of voltage and current. It can essentially distinguish between bidirectional random disturbances caused by the imbalance of three-phase loads in the power grid and unidirectional phase drift caused by the degradation of single-phase components inside the energy meter. This overcomes the shortcomings of traditional detection methods that rely solely on amplitude information and cannot distinguish between internal and external disturbances, and significantly reduces the probability of false alarms.

[0015] 2. This invention combines long and short window statistical modeling with adaptive dynamic confidence boundaries, which can dynamically adjust the discrimination threshold in real time according to the intensity of power grid fluctuations. When the power grid disturbance is severe, the boundary is automatically widened to resist interference, and when the power grid is in a steady state, the boundary is automatically tightened to identify weak anomalies. This effectively solves the problems that fixed thresholds cannot adapt to dynamic operating conditions and early small phase shift faults are easily missed, thus balancing detection reliability and sensitivity.

[0016] 3. This invention uses an asymmetric energy linkage operator to amplify the energy of persistent phase shift faults, a short-time window to filter out random noise, and a multi-cycle over-limit ratio statistical discrimination method to effectively eliminate interference from instantaneous and occasional disturbances such as lightning strikes and switch switching, thereby achieving accurate and stable identification of hardware degradation faults in electricity meters. Attached Figure Description

[0017] Figure 1 This is a flowchart of steps S1-S4 in the three-phase energy meter fault detection method based on zero-sequence component analysis according to an embodiment of the present invention.

[0018] Figure 2 This is a structural block diagram of a three-phase energy meter fault detection system based on zero-sequence component analysis according to an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0020] Reference Figure 1 The three-phase energy meter fault detection method based on zero-sequence component analysis includes steps S1-S4, as follows: S1: Synchronously collect three-phase voltage and current sampling data, extract the fundamental voltage and fundamental current of each phase, and convert them into complex vectors carrying amplitude and phase information respectively.

[0021] During the online operation of a three-phase energy meter, the built-in high-precision analog-to-digital converter module first completes the synchronous sampling of the three-phase voltage signals (A, B, and C) and the three-phase current signals. This synchronous sampling method ensures that the original waveforms of the three phases are acquired at the same time point, avoiding distortion of the relative phase relationship between the three phases caused by sampling timing misalignment.

[0022] The acquired analog electrical signals are processed by analog-to-digital conversion to generate digital sampling sequences. Then, the electrical parameters corresponding to the fundamental components of each phase are separated and extracted by Fourier analysis algorithm. The effective value of the fundamental voltage, the phase angle of the fundamental voltage, the effective value of the fundamental current, and the phase angle of the fundamental current are obtained respectively.

[0023] If only scalar data such as effective values ​​or phase angles are stored separately, the coupling relationship between the amplitude and phase of the electrical quantity in the same phase will be severed. This will result in the loss of crucial phase dimension information during subsequent three-phase vector synthesis and zero-sequence component calculations, ultimately leading to a deviation between the zero-sequence vector calculation results and the actual electrical vector state of the distribution network. Therefore, this invention uniformly maps the fundamental voltage and current of each phase to a complex plane to construct a complex vector, and uniformly defines the physical meaning of the vector: the magnitude of the vector in the complex plane corresponds to the effective value of the electrical quantity, and the vector rotation argument corresponds to the fundamental phase angle of the electrical quantity.

[0024] For each phase A, B, and C, two types of complex vectors are constructed: For the fundamental voltage vector, the effective value of the fundamental voltage of that phase at the current sampling time is selected as the magnitude of the complex plane vector, and the phase angle of the fundamental voltage of that phase at the same time is selected as the vector rotation argument, generating a complex vector specific to that phase's fundamental voltage; For the fundamental current vector, the effective value of the fundamental current of that phase at the current sampling time is selected as the magnitude of the complex plane vector, and the phase angle of the fundamental current of that phase at the same time is selected as the vector rotation argument, generating a complex vector specific to that phase's fundamental current. These vectors are expressed in complex numbers based on the Euler rotation relation, where the imaginary unit is used to characterize the phase orthogonal components. The rotation relation can decompose a single phase angle into a cosine real component and a sinusoidal imaginary component, fully restoring the vector characteristics of AC electrical quantities rotating over time.

[0025] This complex vector representation method can completely preserve the amplitude and phase information of each phase voltage and current, fully restoring the inherent amplitude and phase correlations of the three-phase electrical quantities A, B, and C. When performing related calculations using the symmetrical component method based on this set of vectors, the phase information is not lost throughout the entire process of three-phase vector superposition and zero-sequence component solution. This can accurately restore the true zero-sequence vector operating state of the distribution network, providing a reliable vector input foundation for high-precision calculation of subsequent fault characteristic quantities such as phase misalignment degree.

[0026] When the performance of various single-phase measuring components inside an electricity meter degrades, such as the performance degradation of the current transformer core, the drift of the sampling resistor value, or the offset of the operational amplifier output bias, zero-sequence voltage and zero-sequence current will be generated. At the same time, external conditions such as uneven distribution of three-phase loads on the distribution network side will also generate zero-sequence voltage and zero-sequence current. Existing conventional fault monitoring methods rely solely on the amplitude of the zero-sequence component and a fixed threshold for alarm judgment, which cannot distinguish between the two types of zero-sequence signals with completely different causes. Ultimately, this easily leads to frequent false alarms during normal load fluctuations in the power grid and missed detection of early, weak component degradation faults in the electricity meter.

[0027] To address the aforementioned identification confusion issue, this invention utilizes the inherent electrical laws of power distribution networks to construct characteristic features: the zero-sequence impedance angle of the distribution network system remains essentially constant. When only external load imbalance exists and the energy meter is fault-free, the relative angle between the zero-sequence voltage and zero-sequence current is uniquely determined by the system's zero-sequence impedance, and their phase difference remains stable over a long period. However, if a phase drift fault occurs in any phase measurement channel within the energy meter, the voltage and current phase information collected and output by the meter will deviate, directly altering the inherent phase matching relationship between the zero-sequence voltage and zero-sequence current. The ratio of the area of ​​the parallelogram corresponding to the vector cross product of the two to the amplitude of the zero-sequence voltage will change significantly. This change can serve as the core basis for distinguishing between internal and external disturbances. The specific implementation steps are as follows: S2: Solve the zero-sequence voltage complex vector and zero-sequence current complex vector according to the complex vectors of each phase voltage and current respectively. After taking the conjugate of the zero-sequence voltage complex vector and performing complex number operations with the zero-sequence current complex vector, and then combining the amplitude of the zero-sequence voltage complex vector with the preset minimum constant to complete the normalization process, the phase misalignment degree used to characterize the phase shift state of the zero-sequence vector is obtained.

[0028] First, the zero-sequence vector is solved based on the three-phase complex vectors: the arithmetic mean of the complex vectors of the fundamental voltages of all three phases A, B, and C is obtained to get the zero-sequence voltage complex vector; the arithmetic mean of the complex vectors of the fundamental currents of all three phases A, B, and C is obtained to get the zero-sequence current complex vector.

[0029] Based on two sets of zero-sequence complex vectors, characteristic quantity calculations are performed to generate the phase misalignment degree: First, the conjugate vector corresponding to the zero-sequence voltage complex vector is obtained. The conjugate vector and the zero-sequence current complex vector are multiplied by a complex number. The imaginary part of the multiplication result is extracted and its absolute value is obtained. The absolute value is used as the numerator. Then, the amplitude corresponding to the zero-sequence voltage complex vector is obtained. The amplitude is added to a preset minimum constant to obtain the denominator. The numerator value is divided by the denominator value to obtain the phase misalignment degree used to distinguish between internal and external zero-sequence disturbances.

[0030] Among them, the operation of multiplying complex numbers and taking the imaginary part and the absolute value is geometrically equivalent to the magnitude of the cross product of the zero-sequence voltage vector and the zero-sequence current vector, which corresponds to the area of ​​the parallelogram enclosed by the two sets of vectors. The purpose of the preset minimum constant is to deal with the extreme working conditions of the three-phase power grid being almost completely symmetrical and the zero-sequence voltage amplitude approaching zero, so as to avoid the denominator value approaching zero infinitely, causing the calculated value to diverge and the program to overflow. The preset minimum constant is taken as one percent of the rated voltage of the equipment.

[0031] Specifically, the degree of phase misalignment satisfies the following relationship: ; In the formula, Indicates the first The degree of phase misalignment at any given moment Indicates the first The conjugate vector of the zero-sequence voltage complex vector at time t. Indicates the first The zero-sequence voltage complex vector is obtained by averaging the three-phase fundamental voltage complex vectors at time 1. Indicates the first The zero-sequence voltage vector obtained by averaging the three-phase current vectors at time 1. A very small positive constant representing the magnitude of the rated voltage. This function represents the operation of taking the imaginary part of a complex number, which is equivalent to the modulus of the original cross product. ,in, This represents the real part of the zero-sequence voltage complex vector. This represents the imaginary part of the zero-sequence voltage complex vector; The real part of the complex vector corresponding to the zero-sequence current. This represents the imaginary part of the zero-sequence current complex vector.

[0032] From a physical perspective, the degree of phase misalignment represents the proportion of the projected component of the zero-sequence current in the direction perpendicular to the zero-sequence voltage under a unit zero-sequence voltage. This characteristic exhibits differentiated response properties: amplitude disturbances caused by load imbalances in the distribution network are naturally suppressed, while minute phase shifts caused by aging of internal components of the meter are significantly amplified.

[0033] The operating conditions can be divided into two scenarios: First, when there is only an imbalance in the three-phase load of the power grid and the meter is fault-free, the zero-sequence impedance angle of the system is fixed, the phase relationship between the zero-sequence voltage and the zero-sequence current is stable, and the phase misalignment degree remains within a stable range for a long time; Second, when the aging of devices such as the current transformer, sampling circuit, and operational amplifier inside the meter causes phase drift, the phase of the collected voltage and current deviates, the zero-sequence vector angle changes, and the phase misalignment degree increases significantly.

[0034] The phase misalignment characteristic constructed in this invention normalizes the vector coupling relationship between zero-sequence voltage and zero-sequence current into an orthogonal projection component corresponding to a unit zero-sequence voltage. This characteristic inherently possesses two types of differentiated response characteristics: it has a suppressive effect on the amplitude fluctuation of zero-sequence current caused by load imbalance in the distribution network and is unaffected by conventional asymmetric disturbances in the power grid; it is highly sensitive to minute phase shifts caused by the aging of single-phase components inside the energy meter, where even a very slight phase deviation can cause a significant change in the vector cross product result. This characteristic fundamentally overcomes the shortcomings of traditional monitoring schemes in distinguishing between external power grid imbalance disturbances and single-phase measurement faults inside the energy meter, significantly improving the fault detection discrimination capability; at the same time, this characteristic has extremely low background noise under normal operating conditions, and can serve as a highly sensitive basic characteristic quantity to support the subsequent extraction and accurate identification of fault energy characteristics.

[0035] During the long-term operation of electricity meters, the load of the distribution network and the on-site operating environment will change slowly. Affected by these factors, the steady-state benchmark value of the phase misalignment degree will drift synchronously. If a fixed judgment threshold is used for fault identification, it cannot adapt to the dynamically changing normal operating condition benchmark, and is prone to misjudgment and missed judgment. To address this, this invention introduces a sliding statistical window to statistically analyze the mean and standard deviation of the phase misalignment degree in real time, dynamically constructing an adaptive background benchmark for the normal operation of the power grid in the current period, and providing a reference standard that is updated synchronously with the operating conditions for subsequent fault feature extraction.

[0036] The aging and degradation of internal hardware in electricity meters exhibits clear time-domain characteristics: it is a unidirectional, continuous phase shift process. For example, the continuous decay of the permeability of a single-phase current transformer will cause the phase lag of the corresponding phase current to gradually increase. This type of hardware failure will bring two identifiable features: first, the phase misalignment will deviate continuously in a fixed direction relative to the background mean, forming a directional drift; second, there will be stable small-amplitude numerical fluctuations between adjacent sampling periods, forming periodic transient changes.

[0037] Normal load fluctuations in the power grid are bidirectional random symmetrical disturbances and do not possess sustained unidirectional offset characteristics. This invention combines directional deviation and periodic transient changes to construct a dedicated asymmetric energy linkage operator, which can directionally amplify the fault energy corresponding to hardware degradation, while filtering out the symmetrical random jitter caused by normal load fluctuations in the power grid, thus achieving effective separation of the two types of disturbances. The specific steps are as follows: S3: Configure two independent data windows with different durations: one is a sliding statistical window used to statistically analyze the background features of phase misalignment, which calculates the moving mean and standard deviation of phase misalignment; and the asymmetric energy linkage operator is constructed by combining the deviation of phase misalignment from the moving mean and the first-order difference of phase misalignment. The other is a short-time accumulation window with a shorter length than the sliding statistical window, which extracts all asymmetric energy linkage operators within the short-time accumulation window and sums them to obtain the cumulative asymmetric energy value.

[0038] A fixed-duration sliding statistical window is configured, with 200 sampling points per window and a total duration of 4 seconds. This duration covers the complete fluctuation cycle of the distribution network's conventional load, enabling the complete capture of slow load change patterns. Based on this sliding statistical window, the moving mean and moving standard deviation corresponding to the degree of phase misalignment are calculated in real time, thus representing the baseline of normal grid operation at the current moment.

[0039] In the initial stage of power-on startup of the electricity meter, when the number of data collected in the buffer has not reached the number of complete sampling points of the sliding statistical window, the mean and standard deviation are updated in real time using a local statistical method that increases point by point. When the buffer sampling data fills the window length, it switches to a first-in-first-out continuous sliding update mode, removing the earliest data frame by frame and incorporating the latest sampling data, continuously and dynamically refreshing the background benchmark.

[0040] Based on a pre-configured sliding statistical window, the system continuously collects data on the degree of phase misalignment and calculates the average value corresponding to the degree of phase misalignment in real time. This average value serves as a benchmark reference for the current normal operating conditions of the power grid. The difference between the degree of phase misalignment at the current sampling time and this average value is calculated. This difference represents the overall deviation of the current sampling point relative to the normal operating background. The obtained deviation is then raised to the third power, and the mathematical properties of odd powers are used to fully preserve the positive and negative offset directions corresponding to the deviation, thereby distinguishing between two types of operating conditions: positive continuous offset and negative offset.

[0041] The numerical difference between the phase misalignment degree of the current sampling moment and the previous sampling moment is calculated synchronously, and the absolute value of the difference is taken. The result represents the instantaneous fluctuation amplitude of the phase misalignment degree within adjacent sampling periods, which is used to capture the perturbation change characteristics on a short time scale. The directional deviation obtained by the cubic operation is multiplied by the absolute value difference representing the instantaneous fluctuation amplitude to obtain the asymmetric energy linkage operator corresponding to a single sampling moment.

[0042] Specifically, the asymmetric energy linkage operator satisfies the following subordinate relation: ; In the formula, Indicates the first Moment-asymmetric energy linkage operator, Indicates the first The degree of phase misalignment at any given moment Indicates the first The moving average of the phase misalignment at any given time. Difference Indicates the degree of deviation; odd powers retain the direction of deviation. When the deviation is positive ( When the signal is above the background level and persists, the cube amplifies the signal rapidly; when the deviation is negative, the result is negative. The bidirectional random oscillations caused by external loads will cancel each other out over a long period of averaging.

[0043] In other words, normal load fluctuations in the power grid are bidirectional random oscillations, and the operator calculation results cancel each other out under long-term statistics; the degradation of the electricity meter hardware brings about a unidirectional continuous phase shift, and the operator will continuously amplify the fault characteristic signal.

[0044] To eliminate the fluctuation interference of single-point operator values ​​caused by random sampling noise and instantaneous calculation errors, and to completely retain the cumulative change trend of continuous unidirectional fault energy, a short-term accumulation window with a fixed duration is pre-set. The short-term accumulation window corresponds to 5 consecutive sampling periods with a total duration of 100ms, and the window length is much smaller than the 4-second sliding statistical window. The asymmetric energy linkage operator values ​​corresponding to each sampling moment within the coverage of the short-term accumulation window are extracted sequentially. All operator values ​​within the window are superimposed and summed to finally obtain the asymmetric energy accumulation value used to characterize the total energy of short-term directional disturbance.

[0045] The purpose of setting a short-term accumulation window is twofold: first, to eliminate numerical fluctuations in the operator caused by random sampling noise and algorithm calculation errors at a single sampling moment, thus smoothly suppressing transient clutter; second, to fully preserve the trend of continuous accumulation of hardware degradation fault energy. A 100ms window duration is shorter than the duration of common transient disturbances such as lightning strikes in the distribution network and the starting of high-power motors, ensuring rapid capture of energy changes when early hardware phase shifts occur, thereby improving fault detection response speed.

[0046] The physical meaning of asymmetric energy accumulation is the sum of disturbance energy with unidirectional offset characteristics within a short-term window. The numerical performance varies significantly under different operating conditions: when the distribution network is operating smoothly and normally, the bidirectional load fluctuations cancel each other out, and the accumulation value remains in a low baseline range for a long time; when the internal components of the meter continue to age and degrade, the unidirectional phase offset exists stably, the fault energy continues to accumulate within the short-term window, and the accumulation value steadily increases; when encountering short-term severe external disturbances such as lightning strikes or motor switching, the accumulation value will briefly surge, but such disturbances are mostly bidirectional and short-term, and the energy cannot continue to accumulate, so it will not form a long-term upward trend.

[0047] After obtaining the cumulative value of asymmetric energy, an adaptive dynamic confidence boundary is constructed based on long-term historical operating data. This boundary is then dynamically adjusted in conjunction with real-time fluctuations in the power grid's operation. Simultaneously, a multi-cycle over-limit percentage statistical method is used to accurately identify faults, effectively distinguishing between instantaneous external disturbances to the power grid and persistent hardware degradation faults within the energy meter. The specific boundary construction process is as follows: S4: Extract a historical baseline window composed of continuous historical data, calculate the median and absolute median difference of the cumulative asymmetric energy value within the historical baseline window, calculate the adaptive gain coefficient based on the standard deviation and preset constant, construct the adaptive dynamic confidence boundary based on the median, absolute median difference and adaptive gain coefficient, and count the proportion of the cumulative asymmetric energy value exceeding the adaptive dynamic confidence boundary within the preset period, and determine whether the three-phase energy meter has malfunctioned based on the proportion.

[0048] The cumulative asymmetric energy values ​​for all moments within a preset historical baseline window are retrieved. Based on this set of steady-state characteristic data, the corresponding median and absolute median difference are statistically calculated. Both the median and absolute median difference are robust statistical features with strong anti-interference capabilities, effectively mitigating interference from occasional extreme data such as lightning strikes, switching operations, and sudden load changes. They accurately characterize the baseline level and normal fluctuation range of the cumulative asymmetric energy value under normal steady-state operation of the power grid.

[0049] Furthermore, by combining the standard deviation of phase misalignment output in real time from the sliding statistical window with a preset fixed adjustment constant and a preset minimum constant, a real-time adaptive gain coefficient is calculated. The magnitude of this gain coefficient is positively correlated with the intensity of power grid operation fluctuations and can dynamically follow changes in power grid operating conditions in real time. Finally, the statistically obtained median is superimposed with the product of the adaptive gain coefficient and the absolute median difference to calculate the dynamic alarm upper limit. This dynamic alarm upper limit constitutes the adaptive dynamic confidence boundary for fault diagnosis in this invention.

[0050] Before the historical baseline window is filled with a preset amount of data, an incremental statistical method is used to update the median and absolute median difference of all collected asymmetric energy cumulative value sequences in real time. Each time a new set of valid sampled data is added, the valid sampled data is incorporated into the statistical sequence, and the median and absolute median difference are recalculated iteratively to continuously update the steady-state baseline statistics. Only when the total amount of continuous valid data in the cache reaches the full length of the historical baseline window is the full-window sliding statistical mode fixed, and the adaptive dynamic confidence boundary and over-limit proportion fault discrimination logic are officially enabled.

[0051] Specifically, the adaptive gain coefficients satisfy the following relationship: ; In the formula, Indicates the first Time-adaptive gain coefficient This represents a preset constant; in this embodiment, the value is [value to be filled in]. You can choose according to the actual situation. This represents the standard deviation of the phase misalignment calculated using a sliding statistical window. Represents the positive minimum constant. This is used to avoid the divergence of calculated values ​​under extreme conditions where the standard deviation value approaches zero, thus ensuring the stability of the algorithm operation.

[0052] The adaptive dynamic confidence boundary possesses adaptive adjustment characteristics, adapting to the complex and ever-changing operating conditions of the power grid: when the power grid load fluctuates drastically and disturbances occur frequently, the standard deviation of the phase misalignment degree increases significantly, the adaptive gain coefficient is amplified synchronously, and the dynamic confidence boundary range is automatically widened. This can effectively accommodate small overshoots of characteristic quantities caused by short-term drastic fluctuations in the power grid, and prevent false alarms caused by instantaneous disturbances. When the power grid operates stably and disturbances are weak, the standard deviation of the phase misalignment degree is small, the adaptive gain coefficient decreases accordingly, and the dynamic confidence boundary is automatically tightened. This significantly improves the algorithm's sensitivity to identifying early and minor hardware degradation faults in the energy meter, and accurately captures abnormal changes in characteristic quantities caused by minute phase shifts.

[0053] After constructing the adaptive dynamic confidence boundary, steady-state fault identification is carried out: a fixed-duration statistical judgment period is preset, which includes feature data from multiple consecutive sampling times to avoid random errors in single-point instantaneous data. The cumulative asymmetric energy value corresponding to each set of sampled data within the statistical judgment period is compared with the real-time updated adaptive dynamic confidence boundary. When the cumulative asymmetric energy value exceeds the dynamic confidence boundary at a certain moment, the sampled data is determined to be out-of-bounds abnormal data.

[0054] The total number of all out-of-limit data within the statistical judgment period is combined with the total amount of all sampled data within the period to calculate the effective proportion of out-of-limit data. A fixed fault judgment threshold is pre-configured, and fault differentiation is completed by the relationship between the proportion of out-of-limit data and the judgment threshold: if the statistically obtained proportion of out-of-limit data is higher than the preset judgment threshold, it indicates that the characteristic quantity is in a continuous abnormal state, not caused by instantaneous disturbances, and it can be determined that the internal measuring element of the three-phase energy meter has undergone continuous performance degradation, and a fault alarm signal is output immediately; if the proportion of out-of-limit data is lower than the preset judgment threshold, it indicates that there are only a few occasional out-of-limit events, which are caused by instantaneous external disturbances such as power grid impacts, and there is no continuous hardware fault, so no fault alarm output is triggered.

[0055] This invention also provides a three-phase energy meter fault detection system based on zero-sequence component analysis. For example... Figure 2 As shown, the system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the three-phase energy meter fault detection method based on zero-sequence component analysis according to the first aspect of the present invention. The system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, the settings and functions of which are known in the art and will not be described further here.

[0056] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A fault detection method for three-phase energy meters based on zero-sequence component analysis, characterized in that, include: Simultaneously collect three-phase voltage and current sampling data, extract the fundamental voltage and fundamental current of each phase, and convert them into complex vectors carrying amplitude and phase information respectively; The zero-sequence voltage complex vector and the zero-sequence current complex vector are solved by taking the conjugate of the zero-sequence voltage complex vector and performing complex number operations with the zero-sequence current complex vector. Then, the normalization process is completed by combining the magnitude of the zero-sequence voltage complex vector with a preset minimum constant, so as to obtain the phase misalignment degree used to characterize the phase shift state of the zero-sequence vector. Two independent data windows with different durations are configured: one is a sliding statistical window used to statistically analyze the background features of phase misalignment, which calculates the moving mean and standard deviation of phase misalignment; and an asymmetric energy linkage operator is constructed by combining the deviation of phase misalignment from the moving mean and the first-order difference of phase misalignment. The other is a short-time accumulation window with a shorter length than the sliding statistical window, which extracts all asymmetric energy linkage operators within the short-time accumulation window and sums them to obtain the cumulative asymmetric energy value. A historical baseline window composed of continuous historical data is extracted. The median and absolute median difference of the cumulative asymmetric energy value within the historical baseline window are calculated. The adaptive gain coefficient is calculated based on the standard deviation and a preset constant. An adaptive dynamic confidence boundary is constructed based on the median, absolute median difference, and adaptive gain coefficient. The proportion of the cumulative asymmetric energy value exceeding the adaptive dynamic confidence boundary within a preset period is statistically analyzed. The three-phase energy meter is judged to have malfunctioned based on the proportion.

2. The three-phase energy meter fault detection method based on zero-sequence component analysis according to claim 1, characterized in that, The steps for constructing the complexized vector are as follows: Using the effective value of the fundamental voltage at the current moment as the magnitude of the complex vector of each phase voltage and the phase angle of the fundamental voltage of each phase as the vector rotation angle, construct the complex vector of the fundamental voltage of the corresponding phase. Using the effective value of the fundamental current at the current moment as the modulus of the complex vector of each phase current and the phase angle of the fundamental current of each phase as the vector rotation angle, construct the complex vector of the fundamental current of the corresponding phase.

3. The three-phase energy meter fault detection method based on zero-sequence component analysis according to claim 1, characterized in that, The calculation method for the degree of phase misalignment is as follows: The zero-sequence voltage complex vector is obtained by averaging the three-phase voltage complex vectors, and the zero-sequence current complex vector is obtained by averaging the three-phase current complex vectors. The conjugate vector of the zero-sequence voltage complex vector is calculated, and the conjugate vector is multiplied by the zero-sequence current complex vector. The imaginary part of the result is extracted and the absolute value is taken. The absolute value is divided by the sum of the zero-sequence voltage vector amplitude and the preset minimum constant to obtain the phase misalignment degree.

4. The three-phase energy meter fault detection method based on zero-sequence component analysis according to claim 1, characterized in that, The asymmetric energy linkage operator is constructed as follows: The average value of the phase misalignment is calculated in real time through a preset sliding statistical window. The difference between the phase misalignment at the current moment and the average value is calculated. The difference is then raised to the power of three, and the positive and negative directions of the difference are retained. Calculate the difference between the phase misalignment degree at the current moment and the previous moment and take the absolute value to obtain the instantaneous change amplitude of the phase misalignment degree; multiply the result of the cubic operation with the absolute value difference to obtain the asymmetric energy linkage operator.

5. The three-phase energy meter fault detection method based on zero-sequence component analysis according to claim 1, characterized in that, The method for obtaining the cumulative value of asymmetric energy is as follows: Set a short-term accumulation window with a preset duration, extract the asymmetric energy linkage operator values ​​corresponding to each sampling time in the window in sequence, and sum all the operator values ​​in the window to obtain the cumulative asymmetric energy value.

6. The three-phase energy meter fault detection method based on zero-sequence component analysis according to claim 1, characterized in that, The steps for constructing the adaptive dynamic confidence boundary are as follows: Retrieve all cumulative asymmetric energy values ​​within the historical baseline window, and calculate the median and absolute median difference of all cumulative asymmetric energy values. The adaptive gain coefficient is calculated by combining the standard deviation of phase misalignment output by the sliding statistical window and a preset fixed constant. Adding the median to the product of the gain coefficient and the absolute median difference yields the dynamic alarm upper limit, i.e., the adaptive dynamic confidence boundary.

7. The three-phase energy meter fault detection method based on zero-sequence component analysis according to claim 1, characterized in that, The method of determining whether a three-phase energy meter is malfunctioning based on the proportion includes: A fixed-duration statistical judgment period is set, which includes multiple sets of continuous sampling data. The cumulative asymmetric energy value corresponding to each set of sampling data is compared with the adaptive dynamic confidence boundary. When the cumulative asymmetric energy value exceeds the adaptive dynamic confidence boundary, it is judged as data exceeding the limit. The number of all out-of-limit data within the statistical judgment period is determined, and the proportion of out-of-limit data in all sampled data of the period is calculated. A judgment threshold is pre-configured. If the proportion of out-of-limit data is higher than the judgment threshold, it is determined that the internal components of the three-phase energy meter have experienced performance degradation and a fault alarm signal is output. If the proportion of out-of-limit data is lower than the judgment threshold, it is determined to be an instantaneous external disturbance to the power grid and no fault alarm signal is output.

8. A three-phase energy meter fault detection system based on zero-sequence component analysis, characterized in that, include: The processor and memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the three-phase energy meter fault detection method based on zero-sequence component analysis according to any one of claims 1-7.

Images