A Method and System for Identifying Metering Anomalies in Electricity Meters Based on Multimodal Data

Abstract

This application relates to the field of electricity metering, and particularly to a method and system for identifying metering anomalies in electricity meters based on multimodal data. The method includes: acquiring voltage values ​​from multiple electricity meters and current values ​​from the meter under test; calculating the voltage deviation of each electricity meter and the current deviation of each meter under test; calculating the gain estimate of each meter under test using the voltage and current deviations; calculating the mean gain and variance of each meter under test, and calculating a common gain variance benchmark and mean gain benchmark for the electricity meters; calculating the anomaly index of each meter under test; determining the metering status of each meter under test using the anomaly index and anomaly threshold, and determining whether to trigger an alarm for each meter under test based on the metering status. The technical solution of this application solves the problem that existing technologies are easily affected by voltage drop in the incoming line and voltage fluctuations in the power grid, leading to false alarms and missed alarms. This application achieves real-time perception and early warning of the metering status of electricity meters.

Description

Technical Field

[0001] This application relates to the field of electricity metering, and in particular to a method and system for identifying metering anomalies in electricity meters based on multimodal data. Background Technology

[0002] Against the backdrop of the rapid development of smart grids and advanced measurement systems, electricity meters, as the core terminal for billing, directly affect the efficiency of power supply and the fairness of electricity use. Currently, IoT technology is driving the widespread deployment of smart meters, but long-term operation of electricity meters is susceptible to component aging, line voltage drop, and grid fluctuations. Traditional manual calibration or coarse-grained power analysis methods are insufficient for real-time and accurate online diagnosis of electricity meters.

[0003] Existing technologies propose multimodal data fusion methods for identifying meter anomalies. These methods collect actual power consumption data and device correlation information from smart devices, use machine learning to predict total power consumption, and dynamically adjust error thresholds based on load and environmental fluctuations. The results are then compared with the displayed power consumption to determine anomalies. However, existing methods suffer from the following problems: The total power consumption data collected by the meter on a daily or half-day basis cannot reflect short-term fluctuations or sudden metering anomalies. This makes existing meter detection methods insensitive to instantaneous or short-term faults, and they cannot perform real-time meter detection and diagnosis. Therefore, they lack real-time online continuous monitoring capabilities and cannot provide early warnings when anomalies occur. Furthermore, existing technologies do not consider the impact of external interference such as voltage drop in the incoming power line and grid voltage fluctuations on the accuracy of meter measurement. Summary of the Invention

[0004] This application provides a method and system for identifying metering anomalies in electricity meters based on multimodal data. It is used to solve the problem that existing technologies are easily affected by voltage drop of the incoming line and voltage fluctuations in the power grid, leading to false alarms and missed alarms, so as to achieve the purpose of real-time perception and early warning of the metering status of electricity meters.

[0005] Firstly, a method for identifying metering anomalies in electricity meters based on multimodal data includes the following steps: The sliding window method is used to obtain multiple voltage values ​​of multiple energy meters in multiple consecutive sampling windows, and multiple current values ​​of the meter under test in multiple consecutive sampling windows, wherein the multiple energy meters include the meter under test and a preset reference meter. The voltage values ​​are used to calculate multiple voltage deviations for each of the energy meters within each sampling window, and the current values ​​are used to calculate multiple current deviations for each of the meters under test within each sampling window. The gain estimate for each of the meters under test within each sampling window is calculated using the voltage deviation and current deviation, respectively. The mean and variance of the gain estimate for each of the test tables are calculated to obtain the mean gain and variance of the gain for each test table. The common gain mean benchmark of the energy meters is calculated using the gain mean, and the common gain variance benchmark of the energy meters is calculated using the gain variance. The anomaly index of each of the test tables is calculated using the aforementioned gain mean benchmark, gain variance benchmark, gain mean, and gain variance. The measurement status of each meter under test is determined by the anomaly index and a preset anomaly threshold, and an alarm for each meter under test is triggered based on the measurement status.

[0006] By adopting the above technical solution, real-time sensing and early warning of the metering status of electricity meters have been achieved.

[0007] Preferably, the step of calculating multiple voltage deviations for each of the energy meters within each sampling window using the voltage value, and calculating multiple current deviations for each meter under test within each sampling window using the current value, includes: Based on the voltage value, a voltage sequence is constructed for each of the energy meters in each sampling window; based on the current value, a current sequence is constructed for each of the meters under test in each sampling window. The average value of each voltage sequence is calculated to obtain the average voltage value of each energy meter in each sampling window; the average value of each current sequence is calculated to obtain the average current value of each meter under test in each sampling window. The voltage deviation is calculated using the voltage value and the average voltage value, and the current deviation is calculated using the current value and the average current value.

[0008] By adopting the above technical solution, the static deviation of the electricity meter is eliminated, and the focus is on instantaneous dynamic fluctuations.

[0009] Preferably, the step of calculating the gain estimate of each of the meters under test in each sampling window using the voltage deviation and current deviation includes: The voltage deviation and current deviation are used to calculate the grid reference coefficient between the reference tables in each sampling window, the grid-load interaction coefficient between the reference table and the meter under test in each sampling window, the meter-grid coordination coefficient, and the fluctuation activity coefficient and voltage drop coupling coefficient of the meter under test itself in each sampling window. Using the power grid reference coefficient, grid-load interaction coefficient, meter-grid coordination coefficient, fluctuation activity coefficient, and voltage drop coupling coefficient, the gain estimate of each meter under test is calculated in each sampling window.

[0010] By adopting the above technical solution, the grid fluctuations, the voltage drop of the incoming line, and the internal gain of the electricity meter are decoupled, an unbiased gain estimate is obtained, and external interference is eliminated.

[0011] Preferably, the step of calculating the mean and variance of the gain estimate for each of the test tables to obtain the mean gain and variance of the gain for each test table includes: The gain sequence of each of the tables under test is constructed using the gain estimates. The mean of all gain estimates in each gain sequence is calculated to obtain the mean gain, and the gain variance is calculated using the gain sequence and the mean gain.

[0012] By adopting the above technical solution, we can suppress single-window random noise, improve the temporal stability of quantization gain estimates, and provide reliable statistical characteristics for anomaly detection.

[0013] Preferably, the step of calculating the common gain mean benchmark of the energy meters using the gain mean and calculating the common gain variance benchmark of the energy meters using the gain variance includes: In response to the completion of manual calibration of the instrument under test, its mean gain and variance of gain are recalculated. The mean gain benchmark is calculated using the mean gain of all manually calibrated meters under test, and the variance gain benchmark is calculated using the variance gain of all manually calibrated meters under test.

[0014] By adopting the above technical solutions, a common benchmark for electricity meters is established, common-mode interference such as grid voltage fluctuations and load fluctuations is eliminated, and the robustness of anomaly identification is improved.

[0015] Preferably, determining the measurement status of each meter under test using the anomaly index and a preset anomaly threshold includes: If the abnormality index is greater than or equal to the abnormality threshold, the metering status of the meter under test is determined to be abnormal; otherwise, the metering status of the meter under test is determined to be normal.

[0016] By adopting the above technical solution, the calculated anomaly index is the degree of deviation of the mean gain quantized by adaptive Mahalanobis distance, which realizes sensitive detection of small gain drops.

[0017] Preferably, the step of determining whether to trigger an alarm for each of the meters under test based on the metering status includes: If the metering status of the meter under test is determined to be abnormal, the meter under test will trigger an alarm; otherwise, the meter under test will not trigger an alarm.

[0018] By adopting the above technical solution, occasional interference such as communication interruptions and load switching is filtered out, false alarms are avoided through continuous multi-window judgment, and alarms are triggered in real time.

[0019] Secondly, a metering anomaly identification system based on multimodal data is provided, comprising: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the metering anomaly identification method based on multimodal data described above is implemented.

[0020] This application has the following effects: The electricity meter anomaly identification method based on multimodal data proposed in this application uses a reference meter and the meter under test to simultaneously collect voltage and current data, and calculates voltage deviation and current deviation. Because voltage deviation and current deviation can reflect the actual fluctuations of the power grid, the internal gain of the electricity meter, and the voltage drop of the incoming line, the external interference of power grid voltage fluctuations and the voltage drop of the incoming line to the electricity meter is eliminated. A common gain mean benchmark and gain variance benchmark are calculated using the gain mean and gain variance. Because the common gain mean benchmark and gain variance benchmark of the electricity meter eliminate common-mode interference such as power grid voltage fluctuations and load fluctuations, and the voltage drop coupling coefficient is specifically calculated for the voltage drop of the incoming line, it can effectively resist the interference of the voltage drop of the incoming line and power grid voltage fluctuations on anomaly identification. Because the sliding window continuously outputs the gain and filters out occasional random noise in conjunction with the anomaly judgment rules, real-time perception and early warning of the electricity meter's metering status are realized.

[0021] The above and other objects, advantages and features of this application will become more apparent to those skilled in the art from the following detailed description of specific embodiments of this application in conjunction with the accompanying drawings. Attached Figure Description

[0022] The following sections will describe some specific embodiments of this application in detail by way of example and not limitation, with reference to the accompanying drawings. The same reference numerals in the drawings denote the same or similar parts or components. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings: Figure 1 This is a schematic flowchart of a method for identifying metering anomalies based on multimodal data according to an embodiment of this application; Figure 2 This is a schematic flowchart illustrating the calculation of the voltage deviation of each energy meter in each sampling window and the current deviation of each meter under test in each sampling window according to an embodiment of this application. Figure 3 This is a schematic diagram of an energy meter metering anomaly identification system based on multimodal data according to an embodiment of this application. Detailed Implementation

[0023] The following reference Figures 1 to 3 This application describes a method and system for identifying metering anomalies in electricity meters based on multimodal data. In this description, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature, that is, include one or more of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. When a feature "includes or contains" one or more of the features it encompasses, unless otherwise specifically described, this indicates that other features are not excluded and may be further included.

[0024] In the description of this embodiment, the terms "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0025] This embodiment provides a method for identifying metering anomalies in electricity meters based on multimodal data. This method can achieve real-time perception and early warning of the metering status of electricity meters while resisting interference from power grid fluctuations and voltage drop of the incoming line.

[0026] Please see Figure 1 , Figure 1 This is a schematic flowchart of a method for identifying metering anomalies in electricity meters based on multimodal data, according to an embodiment of this application. The method generally includes: Step S101: Using the sliding window method, multiple voltage values ​​of multiple energy meters in multiple consecutive sampling windows and multiple current values ​​of the meter under test in multiple consecutive sampling windows are obtained, wherein the multiple energy meters include the meter under test and a preset reference meter. Step S102: Calculate multiple voltage deviations of each energy meter in each sampling window using the voltage value of each energy meter in each sampling window; calculate multiple current deviations of each meter under test in each sampling window using the current value of each meter under test in each sampling window. Step S103: Using the voltage deviation of each energy meter in each sampling window and the current deviation of each meter under test in each sampling window, calculate the gain estimate of each meter under test in each sampling window. Step S104: Calculate the mean and variance of the gain estimate for each test table to obtain the mean gain and variance of the gain for each test table. Step S105: Calculate the common average gain benchmark of the energy meters using the average gain of each meter under test, and calculate the common gain variance benchmark of the energy meters using the gain variance of each meter under test. Step S106: Calculate the anomaly index of each meter under test using the common gain mean and gain variance benchmarks of the electricity meters, as well as the gain mean and gain variance of each meter under test. Step S107: Use the abnormal index of each test meter and the preset abnormal threshold to determine the metering status of each test meter, and determine whether to trigger an alarm for each test meter based on the metering status of each test meter.

[0027] In step S101 above, this embodiment acquires all voltage values ​​of each energy meter within a preset window and all current values ​​of each meter under test within a preset window using a preset sampling period, including: Because the voltage values ​​from multiple reference meters are needed in subsequent calculations, this application sets at least two reference meters. In this embodiment, two energy meters are selected from all energy meters within a transformer substation area as reference meters, and the other energy meters are used as meters to be tested. The reference meters are denoted as follows: , The time indicates the first reference table. The time indicates the second reference table.

[0028] The selection criteria for the reference table are as follows: the reference table is a manually calibrated electricity meter. Since the reference table has been manually calibrated, the error of the voltage value it collects is small, so it is used as the reference table.

[0029] In this embodiment, the total number of sampling windows included in the sliding window is set to M, and the sliding window is composed of M sampling windows.

[0030] In this embodiment, each electricity meter synchronously collects the voltage value of its corresponding electrical device within each sampling window. Furthermore, each meter under test synchronously collects the current value of its corresponding electrical device within each sampling window at exactly the same timeframe as the voltage value collection. This embodiment preferably uses a sampling period of 1 second. If the sampling period is too long, rapid fluctuations in current and voltage will be lost; if the sampling period is too short, the amount of sampled data will surge, significantly increasing the computational demands. This embodiment preferably uses a sampling time of 15 minutes for each sampling window. If the total sampling time is too long, an anomaly will require a long waiting period for identification, failing to meet the real-world need for timely alarms; if the total sampling time is too short, the number of sampling points will be insufficient to support the accuracy of the required physical quantities.

[0031] Since the sampling period is 1 second and the sampling time of each sampling window is 15 minutes, each sampling window includes 900 sampling points, that is, each sampling window is 900. Each reference meter corresponds to a voltage value for each sampling point in each sampling window, and each meter under test corresponds to a voltage value and a current value for each sampling point in each sampling window.

[0032] In step S102 above, calculating the voltage deviation of each energy meter in each sampling window using the voltage value of each energy meter in each sampling window means calculating the voltage deviation of each reference meter and each meter under test in each sampling window using the voltage value of each sampling window; calculating the multiple current deviations of each meter under test in each sampling window using the current value of each meter under test in each sampling window means calculating only the multiple current deviations of each meter under test in each sampling window.

[0033] Please see Figure 2 , Figure 2 This is a schematic flowchart illustrating the calculation of the voltage deviation of each energy meter and the current deviation of each meter under test within each sampling window in one embodiment of this application. The specific process for calculating the voltage deviation of each energy meter and the current deviation of each meter under test within each sampling window in this embodiment is as follows: Step S112 involves constructing a voltage sequence for each energy meter using all voltage values ​​within each sampling window, and constructing a current sequence for each meter under test using all current values ​​within each sampling window, including: In this embodiment, all voltage and current values ​​collected within each sampling window are constructed into voltage and current sequences, respectively. Missing values ​​are then processed for each voltage and current sequence. The specific method is as follows: In this embodiment, if the number of consecutively missing sampling points in the current or voltage sequence is less than 1% of the sampling window length, linear interpolation is used to fill the gap; otherwise, the sampling window is discarded and re-acquired. The reason for selecting the number of consecutively missing sampling points to be less than 1% of the sampling window length is as follows: if the proportion of consecutively missing sampling points to the sampling window length is too high, the proportion of missing segments in the sampling window increases, and the error introduced by linear interpolation accumulates and amplifies accordingly; if the proportion of consecutively missing sampling points to the sampling window length is too low, external environmental factors may also affect the number of missing sampling points, leading to a significant increase in the discard rate of the sampling window.

[0034] This embodiment takes a voltage sequence from one of the energy meters as an example, and the specific method for linear interpolation filling is as follows: In this embodiment, the last sampling point before the missing segment of the voltage sequence is identified as the previous valid point, and the first sampling point after the missing segment is identified as the next valid point. The voltage values ​​of the previous and next valid points are obtained. Based on these voltage values ​​and the corresponding times of the previous and next valid points, the voltage value to be filled at the missing sampling time is calculated. The calculation formula used is as follows: In equation (1), This represents the missing voltage value to be filled at the t-th sampling time, where t is the index of the sampling time. This is the voltage value of the last valid point before the missing segment; This is the voltage value of the first valid point after the missing segment; The sampling time corresponding to the previous valid point; The sampling time corresponding to the subsequent valid point; For the missing segment to satisfy The sampling time.

[0035] Step S122, in this embodiment, the voltage and current sequences after linear interpolation are filtered, including: In this embodiment, the mean and standard deviation of the unfiltered voltage of the voltage sequence corresponding to each sampling window are calculated, as well as the mean and standard deviation of the unfiltered current of the current sequence corresponding to each sampling window.

[0036] In this embodiment, the voltage value corresponding to each sampling point in each voltage sequence is subtracted from the average value of the unfiltered voltage in that voltage sequence to obtain the difference. The standard deviation of the unfiltered voltage is multiplied by 3 to obtain the product. If the absolute value of the difference is greater than the product, the sampling point corresponding to the difference is determined to be a gross error, and the sampling point is removed from the voltage sequence. After removing all gross errors from the voltage sequence, the filtering process of the voltage sequence is completed.

[0037] In this embodiment, the filtering method for each current sequence is the same as that for the voltage sequence. After filtering each voltage sequence and each current sequence, the missing values ​​of each voltage sequence and each current sequence are processed again by using a strategy that if the number of consecutive missing sampling points is less than 1% of the sampling window length, linear interpolation is used to fill them in; otherwise, the sampling window is discarded and re-acquired. This fills in the missing sampling points caused by the removal of gross errors, thus completing the preprocessing of all voltage sequences and all current sequences.

[0038] Step S132: In this embodiment, the voltage deviation of each energy meter at each sampling moment within each sampling window is calculated using the voltage value of each sampling point in each voltage sequence and the average voltage of each voltage sequence. Similarly, the current deviation of each meter under test at each sampling moment within each sampling window is calculated using the current value of each sampling point in each current sequence and the average current of each current sequence. This includes: In this embodiment, the voltage deviation of each energy meter at each sampling time within each sampling window is calculated using the following formula: In equation (2), Let m be the voltage deviation of the x-th energy meter at the t-th sampling time in the m-th sampling window, where x is the index of the energy meter. Since each sampling window corresponds to a voltage sequence or current sequence, m is the index of the sampling window or voltage sequence or current sequence. Let be the voltage value of the x-th energy meter at the t-th sampling time within the m-th sampling window; Let be the average voltage of the m-th voltage sequence to which the sampling point corresponding to the t-th sampling time of the x-th energy meter belongs.

[0039] In this embodiment, the current deviation of each meter under test is calculated at each sampling time within each sampling window using the following formula: In equation (3), Let be the current deviation of the k-th meter under test at the t-th sampling time in the m-th sampling window, where k is the index of the meter under test; Let be the current value of the k-th meter under test at the t-th sampling time in the m-th sampling window; It is the mean current of the m-th current sequence to which the sampling point at the t-th sampling time of the k-th test table belongs.

[0040] In step S103 above, this embodiment uses each voltage deviation and each current deviation to calculate the grid reference coefficient between the reference tables in each sampling window, the grid-load interaction coefficient and the meter-grid coordination coefficient between the reference table and the meter under test in each sampling window, and the fluctuation activity coefficient and voltage drop coupling coefficient of the meter under test itself in each sampling window, including: In this embodiment, the calculation of the power grid reference coefficient between two reference tables within the m-th sampling window is taken as an example. The calculation formula used is as follows: In equation (4), This represents the power grid reference coefficient between the two reference tables within the m-th sampling window. This is the length of the sampling window; This represents the voltage deviation of the sampling point corresponding to the t-th sampling time of the m-th voltage sequence in the first reference table. This represents the voltage deviation at the sampling point corresponding to the t-th sampling time of the m-th voltage sequence in the second reference table.

[0041] In this embodiment, the calculation of the network-load interaction coefficient between the first reference table and the k-th test table within the m-th sampling window is taken as an example. The calculation formula used is as follows: In equation (5), is the network load interaction coefficient between the first reference table and the k-th test table within the m-th sampling window.

[0042] In this embodiment, the calculation of the table-network coordination coefficient between the first reference table and the k-th test table within the m-th sampling window is taken as an example. The calculation formula used is as follows: In equation (6), The table network coordination coefficient between the first reference table and the k-th test table within the m-th sampling window; Let be the voltage deviation of the k-th meter under test at the t-th sampling time within the m-th sampling window.

[0043] In this embodiment, the calculation of the fluctuation activity coefficient of the k-th test table within the m-th sampling window is taken as an example. The calculation formula used is as follows: In equation (7), It represents the fluctuation activity coefficient of the k-th test table within the m-th sampling window.

[0044] In this embodiment, the calculation of the voltage drop coupling coefficient of the k-th test meter within the m-th sampling window is taken as an example. The calculation formula used is as follows: In equation (8), Let be the voltage drop coupling coefficient of the k-th test meter within the m-th sampling window.

[0045] In this embodiment, the calculation of the gain estimate of the k-th test table within the m-th sampling window is taken as an example. The calculation formula used is as follows: In equation (9), The gain estimate of the k-th meter under test in the m-th sampling window is used to quantify the measurement scaling factor of the voltage or current of the meter under test's internal metering channel, thereby reflecting whether there is drift or failure in the internal hardware of the meter under test. It is a very small constant. This is used to prevent the denominator of the formula from being 0.

[0046] In this embodiment, under fault-free conditions, the voltage measurement channel and the current measurement channel of the meter under test should maintain a fixed proportional relationship, that is, the gain estimate should be constant. After the meter under test has been running for a long time, the gain estimate will deviate due to internal hardware factors such as aging of the shunt, saturation of the current transformer core, drift of the sampling resistor, and temperature drift of the metering chip. Therefore, the gain estimate can reflect, to a certain extent, whether there are problems with the internal hardware of the meter under test.

[0047] In step S104 above, this embodiment constructs a gain sequence for each table under test using all gain estimates of each table under test, calculates the mean of all gain estimates in each gain sequence to obtain the mean gain of each table under test, and calculates the gain variance of each table under test using the gain sequence and mean gain of each table under test, including: In this embodiment, all gain estimates for each test table are constructed into a gain sequence. The mean of all gain estimates in the gain sequence is calculated to obtain the mean gain of each test table. Taking the calculation of the mean gain of the kth test table as an example, the calculation formula used is as follows: In equation (10), Let be the average gain of the k-th test table.

[0048] This embodiment calculates the gain variance of each test table using its gain sequence and mean gain. The specific formula is as follows: In equation (11), Let be the gain variance of the k-th test table.

[0049] In step S105 above, in this embodiment, it is preferable to manually calibrate the number of meters to be tested obtained by rounding down to 15% of the total number of all electricity meters in the distribution area, and record the total number as B, where B > 2, to form a set. The mean gain and variance of the test instrument after calibration are recalculated according to steps S101-S104.

[0050] In this embodiment, if the number of manually calibrated meters under test is less than 2, the calculation requirements for the mean gain benchmark and the variance gain benchmark are not met. If B is less than 15% of the total number of all energy meters, the calculation of the mean gain and variance gain of the manually calibrated meters under test will be greatly affected by noise, causing the mean gain of the manually calibrated meters under test to deviate from the value under fault-free conditions. If B is greater than 15% of the total number of all energy meters, if there is a regional system deviation within the distribution area, such as different voltage levels at the beginning and end of the distribution area, too many manually calibrated meters under test will average out this spatial difference, leading to misjudgments by meters sensitive to installation location.

[0051] This embodiment calculates the gain mean benchmark using the gain mean of all manually calibrated meters under test (TMTs), and calculates the gain variance benchmark using the gain variance of all manually calibrated TMTs, including: In this embodiment, the calculation formula used to calculate the common average gain benchmark of the electricity meter is as follows: In equation (12), This is the common average gain benchmark for electricity meters; 0 represents the benchmark.

[0052] In this embodiment, the calculation formula used to calculate the common gain variance benchmark of the electricity meter is as follows: In equation (13), This serves as a common gain variance benchmark for electricity meters.

[0053] In step S106 above, this embodiment takes the calculation of the anomaly index of the k-th test table as an example, and the calculation formula used is as follows: In equation (14), is the anomaly index of the k-th meter under test, used to represent the adaptive Mahalanobis distance by which the mean gain of the meter under test deviates from the benchmark of the mean gain of the energy meter under normal conditions.

[0054] In step S107 above, if the abnormality index of the meter under test is greater than or equal to the abnormality threshold, the metering status of the meter under test is determined to be abnormal; otherwise, the metering status of the meter under test is determined to be normal. The abnormality threshold is obtained as follows: Multiple abnormal indices are obtained for each fault-free energy meter within the time period corresponding to the total number of multiple sampling windows. That is, the time period includes the total number of multiple sampling windows. Within each sampling window, one abnormal index is obtained. The mean and standard deviation of all these abnormal indices are calculated. The standard deviation is multiplied by 3 to obtain the product. The mean is added to the product to obtain the abnormal threshold.

[0055] In this embodiment, if the metering status of the meter under test is determined to be abnormal, the meter under test will trigger an alarm; otherwise, the meter under test will not trigger an alarm.

[0056] In this embodiment, before the metering status of the meter under test that triggered the alarm is determined to be normal again, even if the abnormal index of the meter under test is still greater than or equal to the abnormal threshold, the meter will not trigger the alarm again. Therefore, it can prevent the meter under test from continuously triggering invalid alarms and generating redundant alarm data.

[0057] The flowchart provided in this embodiment is not intended to indicate that the operations of the method will be performed in any particular order, or that all operations of the method are included in every case. Furthermore, the method may include additional operations. Within the scope of the technical concept provided by the method in this embodiment, additional variations can be made to the above method.

[0058] It should be understood that in some embodiments, the components may be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods may be implemented using software or firmware stored in memory and executed by a suitable instruction execution system.

[0059] This embodiment also provides a metering anomaly identification system for electricity meters based on multimodal data, such as... Figure 3 As shown, it includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the computer program is executed by the processor, it implements the steps of the electricity meter metering anomaly identification method based on multimodal data in any of the above embodiments.

[0060] The computer program used to perform the operations of this application may be assembly instructions, Instruction Set Architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, integrated circuit configuration data, or source code or object code written in any combination of one or more programming languages ​​and procedural programming languages. The computer program may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer may be connected to the user's computer via any type of network (including a Local Area Network (LAN) or Wide Area Network (WAN)) or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, to perform aspects of this application, electronic circuits, including, for example, programmable logic circuits, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), may execute computer-readable program instructions using status information from computer-readable program instructions to personalize the electronic circuits.

[0061] Therefore, those skilled in the art should recognize that although many exemplary embodiments of this application have been shown and described in detail herein, many other variations or modifications conforming to the principles of this application can be directly determined or derived from the disclosure of this application without departing from the spirit and scope of this application. Thus, the scope of this application should be understood and construed as covering all such other variations or modifications.

Claims

1. A method for identifying metering anomalies in electricity meters based on multimodal data, characterized in that, Includes the following steps: The sliding window method is used to obtain multiple voltage values ​​of multiple energy meters in multiple consecutive sampling windows, and multiple current values ​​of the meter under test in multiple consecutive sampling windows, wherein the multiple energy meters include the meter under test and a preset reference meter. The voltage values ​​are used to calculate multiple voltage deviations for each of the energy meters within each sampling window, and the current values ​​are used to calculate multiple current deviations for each of the meters under test within each sampling window. The gain estimate for each of the meters under test (DUT) within each sampling window is calculated using the voltage and current deviations, including: calculating the grid reference coefficient between the reference meters, the grid-load interaction coefficient between the reference meters and the DUT within each sampling window, and the fluctuation activity coefficient and voltage drop coupling coefficient of the DUT itself within each sampling window; and calculating the gain estimate for each DUT within each sampling window using the grid reference coefficient, grid-load interaction coefficient, DUT-grid coordination coefficient, fluctuation activity coefficient, and voltage drop coupling coefficient. The mean and variance of the gain estimate for each of the test tables are calculated to obtain the mean gain and variance of the gain for each test table. The common gain mean benchmark of the energy meters is calculated using the gain mean, and the common gain variance benchmark of the energy meters is calculated using the gain variance. The anomaly index of each of the test tables is calculated using the aforementioned gain mean benchmark, gain variance benchmark, gain mean, and gain variance. The measurement status of each meter under test is determined by the anomaly index and a preset anomaly threshold, and an alarm for each meter under test is triggered based on the measurement status.

2. The method for identifying metering anomalies in electricity meters based on multimodal data according to claim 1, characterized in that, The step of calculating multiple voltage deviations for each of the energy meters within each sampling window using the voltage value, and calculating multiple current deviations for each meter under test within each sampling window using the current value, includes: Based on the voltage value, a voltage sequence is constructed for each of the energy meters in each sampling window; based on the current value, a current sequence is constructed for each of the meters under test in each sampling window. The average value of each voltage sequence is calculated to obtain the average voltage value of each energy meter in each sampling window; the average value of each current sequence is calculated to obtain the average current value of each meter under test in each sampling window. The voltage deviation is calculated using the voltage value and the average voltage value, and the current deviation is calculated using the current value and the average current value.

3. The method for identifying metering anomalies in electricity meters based on multimodal data according to claim 1, characterized in that, The step of calculating the mean and variance of the gain estimate for each of the test tables to obtain the mean gain and variance of the gain for each test table includes: The gain sequence of each of the tables under test is constructed using the gain estimates. The mean of all gain estimates in each gain sequence is calculated to obtain the mean gain, and the gain variance is calculated using the gain sequence and the mean gain.

4. The method for identifying metering anomalies in electricity meters based on multimodal data according to claim 1, characterized in that, The step of calculating the common gain mean benchmark of the energy meters using the gain mean and calculating the common gain variance benchmark of the energy meters using the gain variance includes: In response to the completion of manual calibration of the instrument under test, its mean gain and variance of gain are recalculated. The mean gain benchmark is calculated using the mean gain of all manually calibrated meters under test, and the variance gain benchmark is calculated using the variance gain of all manually calibrated meters under test.

5. The method for identifying metering anomalies in electricity meters based on multimodal data according to claim 1, characterized in that, The step of determining the metering status of each meter under test using the anomaly index and a preset anomaly threshold includes: If the abnormality index is greater than or equal to the abnormality threshold, the metering status of the meter under test is determined to be abnormal; otherwise, the metering status of the meter under test is determined to be normal.

6. The method for identifying metering anomalies in electricity meters based on multimodal data according to claim 5, characterized in that, The step of determining whether to trigger an alarm for each of the meters under test based on the metering status includes: If the metering status of the meter under test is determined to be abnormal, the meter under test will trigger an alarm; otherwise, the meter under test will not trigger an alarm.

7. A metering anomaly identification system based on multimodal data, characterized in that, It includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the steps of the method for identifying anomalies in electricity metering based on multimodal data according to any one of claims 1-6.

Images