An end-edge-cloud cooperation-based power meter fault early warning method and system

Abstract

The present application relates to the field of data processing, and more particularly to a kind of based on end-edge-cloud cooperation's electric energy meter fault early warning method and system, method includes: end side gathers voltage, current waveform, short time window is divided to calculate variance and is uploaded to edge node;Edge node constructs self-adapting suppression coefficient according to voltage variance variation and current variance variation, the variance of target window is transiently inhibited, and the variance after inhibition is obtained;Long time window is constructed, and environmental baseline is constructed according to the variance after inhibition, and the independent fluctuation energy of electric energy meter is calculated;Edge node constructs sliding window to real-time independent fluctuation energy, and health deviation index is calculated, and fault is judged according to health deviation index and early warning threshold value.The present application reduces data transmission and storage cost through end-edge-cloud cooperation, eliminates load switching interference, decouples environmental common mode drift, and realizes the precise early warning of electric energy meter hardware degradation.

Description

Technical Field

[0001] This invention relates to the field of data processing. More specifically, this invention relates to a method and system for early warning of electricity meter faults based on end-edge-cloud collaboration. Background Technology

[0002] As the core terminal for power grid metering and monitoring, the reliability of electricity meters is of great significance to the stable operation of the power grid and the electricity service for users.

[0003] Currently, early warning systems for electricity meter faults mainly rely on regular manual inspections and alarm mechanisms based on fixed thresholds. Manual inspections are inefficient and make it difficult to detect early potential faults in a timely manner; alarm mechanisms based on fixed thresholds typically only respond to events that exceed current or voltage limits.

[0004] In addition, in existing technologies, data is often uploaded to the cloud for centralized analysis, but the amount of raw waveform data is huge, communication bandwidth and storage costs are high, and cloud processing has transmission delays, which cannot meet the needs of on-site real-time early warning. Summary of the Invention

[0005] To address the technical problems in the prior art, where data is often uploaded to the cloud for centralized analysis, but the original waveform data is enormous, communication bandwidth and storage costs are high, and cloud processing suffers from transmission delays, thus failing to meet the needs of real-time early warning at the field level, this invention provides solutions in the following aspects.

[0006] In the first aspect, a method for early warning of electricity meter faults based on end-edge-cloud collaboration includes: The voltage and current waveforms of the energy meter are collected at the edge, short time windows are divided, the voltage variance and current variance within each short time window are calculated and uploaded to the edge computing node; For each short time window, the edge computing node takes that short time window as the target window, constructs an adaptive suppression coefficient based on the change in voltage variance and current variance between the target window and the previous short time window, and uses the adaptive suppression coefficient to perform transient suppression on the original voltage variance and original current variance of the target window to obtain the transiently suppressed voltage variance and current variance. The edge computing node constructs a long-term window, builds an environmental baseline based on the voltage variance and current variance after transient suppression within the long-term window, and calculates the independent fluctuation energy of the electricity meter. During the healthy period, the edge computing node reports the independent fluctuation energy to the cloud side; the edge computing node constructs a sliding window for the real-time input independent fluctuation energy, compares the independent fluctuation energy in the sliding window with the independent fluctuation energy during the healthy period, and calculates the real-time health deviation index of the electricity meter. The energy meter is judged to be faulty based on the health deviation index and the preset warning threshold.

[0007] Optionally, the voltage variance change is the difference between the voltage variance of the target window and the voltage variance of the previous short-time window, and the current variance change is the difference between the current variance of the target window and the current variance of the previous short-time window.

[0008] Optionally, the construction of the adaptive suppression coefficient includes: The voltage variance change and the current variance change are multiplied and the absolute value is taken to obtain the mutation intensity. The historical average value of the mutation intensity up to the previous short time window is calculated as the first parameter. The mutation intensity is divided by the sum of the mutation intensity and the first parameter to obtain the adaptive suppression coefficient.

[0009] Optionally, the transient suppression process includes: The adaptive suppression coefficient is multiplied by the voltage variance change and the current variance change, respectively, and the corresponding product is subtracted from the original voltage variance and the original current variance of the target window.

[0010] Optionally, the environmental baseline includes a voltage background baseline and a current background baseline; the voltage background baseline is the mean of the voltage variance after all transient suppression within the long time window; the current background baseline is the mean of the current variance after all transient suppression within the long time window.

[0011] Optionally, the calculation of the independent wave energy includes: The voltage relative volatility is obtained by comparing the voltage variance after transient suppression with the voltage background baseline, and the current relative volatility is obtained by comparing the current variance after transient suppression with the current background baseline. The absolute value of the difference between the current relative volatility and the voltage relative volatility is calculated, and then multiplied by a preset current scale coefficient to obtain the independent fluctuation energy.

[0012] Optionally, the calculation of the health deviation index includes: The difference between the mean of independent fluctuation energy within the sliding window and the mean of independent fluctuation energy during the healthy period is calculated, and the difference is divided by the standard deviation of the independent fluctuation energy during the healthy period to obtain the health deviation index.

[0013] Secondly, a power meter fault early warning system based on end-edge-cloud collaboration 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 power meter fault early warning method based on end-edge-cloud collaboration described in any one of the claims is implemented.

[0014] The present invention has the following beneficial effects: 1. This invention significantly reduces communication bandwidth and cloud storage costs by compressing the original waveform through end-side variance calculation; by constructing an adaptive suppression coefficient, it automatically smooths out variance spikes caused by high-power load switching by utilizing the synchronous abrupt changes in voltage and current, while fully preserving the small degradation signals inside the meter, thus avoiding false alarms and missed alarms of the fixed threshold method; by constructing a long-term window environmental baseline and calculating independent fluctuation energy, it decouples common-mode drift such as ambient temperature and gradual changes in transformer load; and by comparing the health deviation index with the sliding window, it achieves quantitative judgment of the degree of hardware degradation of the energy meter, improving the accuracy and generalization ability of fault early warning.

[0015] 2. This invention reports independent fluctuation energy during the healthy period to the cloud, where the healthy period data is aggregated and the health deviation index is calculated using edge nodes. This achieves a collaborative architecture that combines centralized establishment of health benchmarks with distributed execution of real-time deviation detection. This ensures the statistical reliability of benchmark parameters and reduces the dependence of real-time early warning on cloud computing and transmission bandwidth, giving the system better real-time performance and scalability. Attached Figure Description

[0016] Figure 1 This is a flowchart of steps S1-S4 in an embodiment of the present invention, which describes a method for early warning of electricity meter faults based on end-edge-cloud collaboration.

[0017] Figure 2 This is a structural block diagram of an energy meter fault early warning system based on end-edge-cloud collaboration according to an embodiment of the present invention. Detailed Implementation

[0018] 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.

[0019] The electricity meter fault early warning method of this invention is deployed in a smart grid distribution area environment. The terminal side consists of the computing module built into the electricity meter, responsible for raw waveform acquisition and preliminary compression; the edge side consists of edge computing nodes in the distribution area, such as smart fusion terminals or edge gateways, responsible for real-time data processing and fault identification; the cloud side consists of a remote server cluster, responsible for aggregating historical data from multiple distribution areas, establishing health benchmarks, and continuously optimizing algorithm parameters. The three layers interact via power communication networks such as HPLC and 4G / 5G, forming a collaborative working mode of terminal-side compressed uploading, edge-side real-time analysis, and cloud-side global modeling.

[0020] Reference Figure 1 A method for early warning of electricity meter faults based on end-edge-cloud collaboration includes steps S1-S4, as detailed below: S1: The end-side module collects the voltage and current waveforms of the energy meter and divides them into short time windows, calculating the variance within each short time window to compress the amount of data.

[0021] The electricity meter acquires voltage and current waveforms in real time at a fixed sampling frequency, such as 6.4kHz. Uploading the raw waveforms directly would consume significant communication bandwidth and cloud storage resources, and the raw waveforms contain a large amount of redundant information, such as the power frequency fundamental wave, making it difficult to directly extract fault characteristics. Therefore, the data is first compressed within the electricity meter's built-in end-side module.

[0022] Because the original sampling points are too dense, it is necessary to group the continuous sampling points according to a fixed duration to form continuous short time windows. In this embodiment of the invention, the length of each short time window is 1 second, and adjacent short time windows do not overlap, that is, the step size is also 1 second. The 1-second short time window can capture the fluctuations caused by load switching and hardware degradation, without smoothing out transient features due to the window being too long.

[0023] In addition, the collected voltage and current values ​​are divided by their corresponding rated values ​​to convert them into dimensionless per-unit values.

[0024] For all sampling points within each short time window, calculate the variance of voltage and the variance of current. Variance describes the degree of fluctuation of voltage or current around its average value within that one second; the greater the fluctuation, the greater the variance. Anomalies in the power grid, such as load switching and meter internal hardware degradation, can lead to increased variance.

[0025] After the above operations, the original high-frequency waveform is compressed into one voltage variance value and one current variance value per second, reducing the data volume by thousands of times. The variance values ​​of all short time windows are arranged in chronological order to form voltage variance sequences and current variance sequences.

[0026] Finally, the calculated voltage variance sequence and current variance sequence are uploaded to the edge computing node of the substation via local communication, such as RS485 or HPLC.

[0027] S2: After receiving the voltage variance sequence and the current variance sequence, the edge computing node uses the voltage synchronous change characteristics and the current synchronous change characteristics to construct an adaptive suppression coefficient to eliminate load switching interference.

[0028] After receiving the voltage and current variance sequences, edge nodes must also consider the following: the switching of high-power loads in the power grid, such as air conditioners and elevators, can cause simultaneous and drastic jumps in voltage and current, forming high-amplitude spikes in the variance sequences, which can easily be misjudged as faults. Traditional fixed threshold methods are prone to false alarms for these spikes, or the threshold may be set too high to avoid spikes, causing the missed true, minute degradation signals.

[0029] Therefore, it is necessary to utilize the physical law that any high-power load switching will inevitably cause simultaneous and violent fluctuations in voltage and current, and construct an adaptive suppression coefficient to automatically identify and smooth out the variance jump spikes caused by high-power load switching, while fully preserving the small degradation signals inside the meter.

[0030] First, using the k-th short-time window as the target window and the (k-1)-th short-time window as the reference window, calculate the difference in voltage variance between the target window and the reference window to obtain the change in voltage variance. The value of k is 2, 3, 4…, meaning the calculation starts from the second short-time window. Similarly, calculate the change in current variance.

[0031] Next, the voltage variance change and the current variance change are multiplied and their absolute values ​​are taken to obtain the intensity of the abrupt change in the target window. This calculation shows that the product only increases significantly when both voltage and current change drastically simultaneously and in the same direction (i.e., both positive or both negative). If only one changes while the other remains stable, or if their directions of change are opposite (e.g., voltage decreases while current increases), this may occur under certain special load conditions, resulting in a small or even negative product. Therefore, by performing this product operation, it is possible to effectively distinguish between synchronous voltage and current abrupt changes and unilateral abnormal fluctuations, thereby improving the accuracy of load switching identification.

[0032] Then, the historical average value of the abrupt change intensity at the cutoff reference window, i.e., the corresponding k-1 time, is calculated and recorded as the first parameter. This first parameter represents the normal disturbance level of the environment where the energy meter is located under normal operating conditions.

[0033] Furthermore, the adaptive suppression coefficient of the target window is obtained by dividing the abrupt change intensity of the target window by the sum of the abrupt change intensity of the target window and the first parameter. When a large load switching occurs, the abrupt change intensity of the target window is much greater than the first parameter, and the adaptive suppression coefficient of the target window approaches 1; under stable operating conditions, the abrupt change intensity of the target window is much less than the first parameter, and the adaptive suppression coefficient of the target window approaches 0.

[0034] Finally, the edge nodes multiply the adaptive suppression coefficients calculated above by the voltage variance change and the current variance change, respectively, and subtract the product from the variance of the target window to obtain the voltage variance and current variance of the target window after transient suppression, which can be expressed by the following formula: In the formula, Let V be the voltage variance after transient suppression in the k-th short time window. The original voltage variance for the k-th short time window. Let be the adaptive suppression coefficient for the k-th short time window. Let V be the change in voltage variance during the k-th short-time window. Let Variance be the current variance after transient suppression in the k-th short time window. Let be the original current variance for the k-th short-time window. Let be the change in current variance during the k-th short time window.

[0035] Through the above calculations, while preserving the original variance trend, the instantaneous spikes caused by high-power load switching are selectively smoothed out. When the values ​​approach 1, the k-th short-time window is backed up to the voltage variance and current variance values ​​of the (k-1)-th short-time window, thus smoothing out the jump spikes; when When the values ​​approach 0, the original voltage variance and current variance values ​​of the k-th short time window are completely preserved.

[0036] When slow degradation occurs inside the meter, the changes in the variances of voltage and current are slow and asynchronous. For example, only the current variance gradually increases while the voltage variance remains unchanged. In this case, the intensity of the abrupt change is relatively small. When the value is close to 0, the degradation signal will not be suppressed. This effectively suppresses false anomalies caused by load switching without affecting the minute degradation signals inside the meter.

[0037] S3: Construct a long-term window, and calculate the mean values ​​of voltage variance and current variance after transient suppression within the long-term window at the side nodes to decouple environmental common-mode drift, and calculate the independent fluctuation energy of the energy meter.

[0038] Even after the transient suppression in S2, common-mode interference such as changes in ambient temperature and slow fluctuations in the total load of the distribution area still exists in the data. For example, rising summer temperatures can increase line resistance, causing voltage variance and current variance to rise slowly and synchronously. This gradual drift overlaps with the abnormal characteristics caused by internal hardware degradation of the meter, such as solder joint oxidation and loose connectors, in the frequency domain, and conventional filtering methods cannot separate the two.

[0039] Therefore, it is necessary to first construct a long time window to calculate the mean of the variance after transient suppression within the long time window as the environmental baseline, thereby decoupling common-mode drift, and further calculate the independent fluctuation energy that only reflects the internal hardware degradation of the electricity meter.

[0040] First, a long time window is defined. In this embodiment of the invention, the long time window is 10 minutes. Since each short time window is 1 second and does not overlap, 10 minutes corresponds to 600 short time windows. For each time k, the edge node takes the 600 windows preceding the current short time window, i.e., k-599 to k, as the long time window for time k. The mean of the voltage variance and the mean of the current variance after transient suppression are calculated within this long time window. These are used as the voltage background baseline and the current background baseline, respectively. These two means represent the background baseline of the voltage variance and current variance under the current environment, mainly reflecting slow external changes.

[0041] Then, the voltage variance after transient suppression at time k is compared with the voltage background baseline. That is, the difference between the voltage variance after transient suppression and the voltage background baseline is calculated, and then divided by the voltage background baseline to obtain the voltage relative volatility at time k. Similarly, the current relative volatility at time k is obtained.

[0042] Next, calculate the absolute value of the difference between the relative fluctuation rate of current and the relative fluctuation rate of voltage. If external factors cause voltage and current to change synchronously, their relative fluctuation rates are close, and the difference is close to 0. If the internal current loop of the electricity meter deteriorates, such as increased contact resistance while the voltage loop is normal, the relative fluctuation rate of current will be greater than the relative fluctuation rate of voltage, and the difference will be significantly positive.

[0043] Finally, it is also considered that under heavy load, i.e., when the current variance is large, even a small relative deviation corresponds to a large absolute abnormal energy, and should be given a higher weight; while under light load or no load, the same percentage deviation has a very small actual absolute energy, and should be given a lower weight. In addition, if the current fluctuation is particularly drastic in the current short time window, it indicates the existence of normal load fluctuations, and the weight should be appropriately suppressed to avoid false amplification. For this reason, a current scale coefficient needs to be constructed to weight and adjust the absolute value of the difference between the relative current fluctuation rate and the relative voltage fluctuation rate.

[0044] The above current scale factor can be expressed by the following formula: In the formula, For current scale factor, Let k be the current background baseline. Let be the original current variance for the k-th short-time window. This is to prevent division by zero for extremely small positive numbers.

[0045] Furthermore, by multiplying the aforementioned current scale coefficient by the absolute value of the difference between the relative fluctuation rate of current and the relative fluctuation rate of voltage, the independent fluctuation energy of the energy meter at time k is obtained.

[0046] During the healthy period of an electricity meter, the independent fluctuation energy is approximately zero-mean white noise; as the internal hardware of the electricity meter gradually deteriorates, the independent fluctuation energy exhibits a slow positive drift.

[0047] S4: The cloud side gathers health period data based on the calculated independent fluctuation energy of the electricity meter and establishes benchmark parameters. The edge nodes calculate the health deviation index based on the benchmark parameters to determine the electricity meter fault.

[0048] The independent fluctuation energy of the electricity meter at each time point calculated from S3 above has a near-zero mean when the electricity meter is healthy, and the mean slowly shifts positively when it deteriorates. The traditional fixed threshold method has poor generalization ability because the fluctuation amplitude of independent fluctuation energy varies greatly during the healthy periods of different electricity meters; some electricity meters have large fluctuations during the healthy period, while others have small fluctuations.

[0049] First, during the initial health period of the electricity meter's operation, such as the first 30 days, the side-side nodes will report the daily independent fluctuations in energy to the cloud side.

[0050] The cloud-based system aggregates the independent fluctuation energy throughout the entire healthy period and calculates the mean and standard deviation of the independent fluctuation energy. The cloud-based system then distributes the mean and standard deviation of the independent fluctuation energy throughout the entire healthy period to the edge nodes of the corresponding transformer areas.

[0051] The edge nodes maintain a sliding window for the real-time input independent fluctuation energy. In this embodiment of the invention, a 100-second window is set, meaning that a sliding window is constructed by extracting 99 seconds of data from the current moment forward. The difference between the mean of the independent fluctuation energy within the current sliding window and the mean of the independent fluctuation energy throughout the entire healthy period is calculated. Then, this difference is divided by the standard deviation of the independent fluctuation energy throughout the entire healthy period to obtain the health deviation index at the current moment.

[0052] Finally, a warning threshold is set. In this embodiment of the invention, the warning threshold is set to 3. If the health deviation index is greater than 3 for multiple consecutive times, the side node determines that the electricity meter has a hardware degradation risk and immediately issues a warning message.

[0053] In addition, the cloud-based system can periodically collect historical data and early warning records from all distribution areas, analyze the degradation patterns of different models and batches of electricity meters, and remotely adjust edge-side algorithm parameters (such as long-term window length, sliding window length, and early warning thresholds) to achieve continuous optimization of the system model. For newly installed electricity meters, the cloud-based system can provide initial baseline parameters based on historical data from the same model and distribution area, avoiding waiting during the health period.

[0054] Thus, this invention achieves accurate early warning of electricity meter faults through an edge-cloud collaborative architecture.

[0055] This invention also provides a power meter fault early warning system based on end-edge-cloud collaboration. For example... Figure 2 As shown, the system includes a processor and a memory. The memory stores computer program instructions. When the computer program instructions are executed by the processor, the method for early warning of electricity meter faults based on end-edge-cloud collaboration according to the first aspect of the present invention is implemented.

[0056] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0057] 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 method for early warning of electricity meter faults based on end-edge-cloud collaboration, characterized in that, include: The voltage and current waveforms of the energy meter are collected at the edge, short time windows are divided, the voltage variance and current variance within each short time window are calculated and uploaded to the edge computing node; For each short time window, the edge computing node uses that short time window as the target window and constructs an adaptive suppression coefficient based on the voltage variance change and current variance change between the target window and the previous short time window. This includes: multiplying the voltage variance change and the current variance change and taking the absolute value to obtain the catastrophe intensity; calculating the historical average of the catastrophe intensity up to the previous short time window as a first parameter; and dividing the catastrophe intensity by the sum of the catastrophe intensity and the first parameter to obtain the adaptive suppression coefficient. The original voltage variance and original current variance of the target window are transiently suppressed using the adaptive suppression coefficient to obtain the transiently suppressed voltage variance and current variance; The edge computing node constructs a long-term window, builds an environmental baseline based on the transient-suppressed voltage variance and current variance within the long-term window, and calculates the independent fluctuation energy of the electricity meter. The environmental baseline includes a voltage background baseline and a current background baseline. The calculation of the independent fluctuation energy of the electricity meter includes: obtaining the relative voltage fluctuation rate based on the transient-suppressed voltage variance and the voltage background baseline, obtaining the relative current fluctuation rate based on the transient-suppressed current variance and the current background baseline, calculating the absolute value of the difference between the relative current fluctuation rate and the relative voltage fluctuation rate, and then multiplying it by a preset current scale coefficient to obtain the independent fluctuation energy. During the healthy period, the edge computing node reports independent fluctuation energy to the cloud side; the edge computing node constructs a sliding window for the real-time input independent fluctuation energy, compares the independent fluctuation energy within the sliding window with the independent fluctuation energy during the healthy period, and calculates the real-time health deviation index of the electricity meter, including: calculating the difference between the mean of the independent fluctuation energy within the sliding window and the mean of the independent fluctuation energy during the healthy period, and dividing the difference by the standard deviation of the independent fluctuation energy during the healthy period to obtain the health deviation index; The energy meter is judged to be faulty based on the health deviation index and the preset warning threshold.

2. The method for early warning of electricity meter faults based on end-edge-cloud collaboration according to claim 1, characterized in that, The voltage variance change is the difference between the voltage variance of the target window and the voltage variance of the previous short-time window, and the current variance change is the difference between the current variance of the target window and the current variance of the previous short-time window.

3. The method for early warning of electricity meter faults based on end-edge-cloud collaboration according to claim 1, characterized in that, The transient suppression process includes: The adaptive suppression coefficient is multiplied by the voltage variance change and the current variance change, respectively, and the corresponding product is subtracted from the original voltage variance and the original current variance of the target window.

4. The method for early warning of electricity meter faults based on end-edge-cloud collaboration according to claim 1, characterized in that, The voltage background baseline is the mean of the voltage variance after all transient suppression within the long time window; the current background baseline is the mean of the current variance after all transient suppression within the long time window.

5. A power meter fault early warning system based on end-edge-cloud collaboration, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the power meter fault early warning method based on end-edge-cloud collaboration according to any one of claims 1-4.

Images