A real-time fault intelligent detection method for an electric energy meter

Abstract

The present application relates to the technical field of data processing, and especially relates to a kind of real-time fault intelligent detection method of electric energy meter, the data fluctuation of the historical data sequence of any monitoring index of electric energy meter in each historical period is obtained according to the method, the historical neighborhood radius when each historical data sequence is clustered;According to the data rule similarity of any monitoring index under transformer in current period and each historical period, the neighborhood radius adaptation degree that each historical neighborhood radius is obtained as the neighborhood radius when the real-time data sequence of any monitoring index of electric energy meter in current period is clustered;According to each historical neighborhood radius and its neighborhood radius adaptation degree, the adaptive neighborhood radius corresponding to real-time data sequence is obtained, the adaptive neighborhood radius is used to cluster real-time data sequence, and the clustering result is obtained, according to the clustering result under each monitoring index of electric energy meter, the fault of electric energy meter is detected, and the accuracy of detecting the fault of electric energy meter is improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a real-time intelligent fault detection method for electricity meters. Background Technology

[0002] Electricity meter malfunctions refer to abnormal phenomena in monitoring data caused by factors such as aging, damage, or software abnormalities, resulting in jumps, excessively high or low values, or drastic fluctuations that deviate from normal data variation patterns. The health status of electricity meters is crucial for ensuring accurate metering, improving the operational efficiency of the power system, and enhancing user experience satisfaction. With technological advancements, electricity meter fault detection is becoming increasingly intelligent, enabling remote real-time monitoring. Compared to manual inspections, this not only improves detection efficiency but also effectively reduces electricity meter maintenance costs.

[0003] Because common anomalies in electricity meter faults manifest as sudden changes in monitoring data, showing significant differences from the original data, clustering algorithms such as DBSCAN are typically used for identification. However, electricity meter faults can also exhibit persistent anomalies such as data drift, severe fluctuations, or static patterns. There may be a large number of abnormal data points within a monitoring period. A key factor affecting the anomaly identification performance of the DBSCAN clustering algorithm lies in the selection of the neighborhood radius. When the proportion of abnormal data is high, the traditional DBSCAN clustering algorithm cannot effectively identify the anomalies, resulting in poor identification of these persistent anomalies and impacting the accuracy of electricity meter fault detection results.

[0004] Therefore, improving the accuracy of fault detection in electricity meters has become an urgent problem to be solved. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a real-time intelligent fault detection method for electricity meters to solve the problem of how to improve the accuracy of fault detection in electricity meters.

[0006] This invention provides a real-time intelligent fault detection method for electricity meters, which includes the following steps:

[0007] Obtain historical data sequences of any monitoring indicator of the electricity meter within a preset number of historical time periods. Based on the data fluctuations in each historical data sequence, obtain the historical neighborhood radius when clustering each historical data sequence.

[0008] Obtain the real-time sequence of the transformer under any of the monitoring indicators in the current time period and the historical sequence of the transformer in each historical time period. Based on the similarity of data patterns between each historical sequence and the real-time sequence of the transformer, obtain the similarity index of electricity consumption between each historical time period and the current time period.

[0009] Obtain the real-time data sequence of any of the monitoring indicators of the electricity meter in the current time period. Based on the similarity index of electricity consumption between each historical time period and the current time period, as well as the abnormal performance of each historical data sequence, obtain the neighborhood radius fit degree when using the neighborhood radius of each historical neighborhood as the neighborhood radius for clustering the real-time data sequence.

[0010] Based on the historical neighborhood radius and its corresponding neighborhood radius fit, an adaptive neighborhood radius is obtained when clustering the real-time data sequence. The real-time data sequence is then clustered based on the adaptive neighborhood radius to obtain the clustering results. Based on the clustering results under each monitoring indicator of the electricity meter, the faults of the electricity meter are detected in real time.

[0011] Preferably, the step of obtaining the historical neighborhood radius for clustering each historical data sequence based on data fluctuations in each historical data sequence includes:

[0012] For any historical data sequence, if the time interval between any data in the historical data sequence and the first data in the historical data sequence is greater than or equal to a preset time interval, and the time interval between any data and the last data in the historical data sequence is greater than or equal to a preset time interval, then the data is taken as the data to be analyzed. In the historical data sequence, a window of a preset length is constructed with the data to be analyzed as the center. The preset time interval is half of the difference between the preset length and the constant 1.

[0013] Calculate the absolute value of the difference between every two adjacent data points within the window, and linearly normalize the sum of all the absolute values ​​of the differences to obtain the degree of fluctuation of the data to be analyzed.

[0014] The fluctuation level of each data point to be analyzed in any historical data sequence is obtained. Based on the fluctuation level of each data point to be analyzed in any historical data sequence, the historical neighborhood radius is obtained when clustering the historical data sequence using the K-distance graph method.

[0015] Preferably, the step of obtaining a similarity index between electricity consumption in each historical period and the current period based on the data pattern similarity between each transformer's historical sequence and real-time sequence includes:

[0016] For any historical sequence of a transformer, the STL time series decomposition algorithm is used to obtain the trend terms of the historical sequence and the real-time sequence of the transformer respectively. Based on the similarity of the trend terms of the historical sequence and the real-time sequence of the transformer, the similarity of the changing trends between the historical sequence and the real-time sequence of the transformer is obtained.

[0017] Using the STL time series decomposition algorithm, the residual terms of any transformer historical sequence and transformer real-time sequence are obtained respectively. Based on the difference between the residual terms of any transformer historical sequence and transformer real-time sequence, the stability similarity between any transformer historical sequence and transformer real-time sequence is obtained.

[0018] Based on the similarity of the changing trends and the similarity of the stability, the similarity index of the electricity consumption between the historical period in which the historical sequence of any transformer is located and the current period is obtained.

[0019] Preferably, the step of obtaining the trend similarity between any historical transformer sequence and the real-time transformer sequence based on the similarity of trend terms includes:

[0020] The trend item of any transformer historical sequence is recorded as the historical trend item, and the trend item of the transformer real-time sequence is recorded as the real-time trend item. In the historical trend item, a preset number of historical trend values ​​are evenly selected and recorded as historical representative trend values.

[0021] For any historical representative trend value, obtain the position of the historical representative trend value in the historical trend item, obtain the real-time trend value with the same position in the real-time trend item, calculate the absolute value of the difference between the historical representative trend value and the real-time trend value, and obtain the change trend difference value between the historical representative trend value and the real-time trend value.

[0022] Calculate the trend difference value corresponding to each historical representative trend value in any transformer historical sequence, accumulate all trend difference values ​​to obtain the accumulated value, calculate the reciprocal of the sum between the preset constant and the accumulated value, and obtain the trend similarity between the historical sequence of any transformer and the real-time sequence of the transformer.

[0023] Preferably, the step of obtaining the stability similarity between any historical transformer sequence and the real-time transformer sequence based on the difference in residual terms between the historical and real-time transformer sequences includes:

[0024] Calculate the average of all residual values ​​in the residual term of any transformer historical sequence, and denot it as the historical residual mean. Calculate the average of all residual values ​​in the residual term of the transformer real-time sequence, and denot it as the real-time residual mean.

[0025] Calculate the absolute value of the difference between the historical residual mean and the real-time residual mean to obtain the residual consistency index. Calculate the reciprocal of the sum of the preset constant and the residual consistency index to obtain the stability similarity between any transformer historical sequence and the transformer real-time sequence.

[0026] Preferably, the step of obtaining the similarity index of electricity consumption between the historical period and the current period of any transformer's historical sequence based on the similarity of the change trend and the similarity of the stability includes:

[0027] The product of the trend similarity and the stability similarity is calculated to obtain the similarity index of electricity consumption between the historical period and the current period of any transformer's historical sequence.

[0028] Preferably, the step of obtaining the neighborhood radius fit degree, which uses the neighborhood radius of each historical neighborhood as the neighborhood radius when clustering the real-time data sequence, based on the similarity index of electricity consumption between each historical time period and the current time period, and the abnormal performance of each historical data sequence, includes:

[0029] For any historical data sequence, calculate the standard deviation of all data in the historical data sequence, and linearly normalize the product of the similarity index of electricity consumption corresponding to the historical period in which the historical data sequence is located and the reciprocal of the standard deviation, to obtain the neighborhood radius fit degree, which uses the historical neighborhood radius corresponding to the historical data sequence as the neighborhood radius when clustering the real-time data sequence.

[0030] Preferably, the step of obtaining the adaptive neighborhood radius for clustering the real-time data sequence based on each historical neighborhood radius and its corresponding neighborhood radius fit includes:

[0031] For any historical neighborhood radius, calculate the proportion of the neighborhood radius fit degree corresponding to that historical neighborhood radius in the total neighborhood radius fit degree, and obtain the fit weight of that historical neighborhood radius.

[0032] Obtain the adaptation weights of all historical neighborhood radii, and calculate the weighted average of all historical neighborhood radii based on the adaptation weights of all historical neighborhood radii to obtain the adaptive neighborhood radius when clustering real-time data sequences.

[0033] Preferably, the step of clustering the real-time data sequence based on the adaptive neighborhood radius to obtain the clustering result includes:

[0034] For any real-time data in the real-time data sequence, if the time interval between the real-time data and the first data in the real-time data sequence is greater than or equal to a preset time interval, and the time interval between the real-time data and the last data in the real-time data sequence is greater than or equal to a preset time interval, then the real-time data is taken as the real-time data to be analyzed. In the real-time data sequence, a real-time window of a preset length is constructed with the real-time data to be analyzed as the center. The preset time interval is half of the difference between the preset length and the constant 1.

[0035] Calculate the absolute value of the difference between every two adjacent data points within the real-time window, and linearly normalize the sum of all absolute values ​​of the differences to obtain the degree of fluctuation of the real-time data to be analyzed.

[0036] Obtain the fluctuation degree of each real-time data to be analyzed in the real-time data sequence, and construct a scatter plot based on the fluctuation degree of each real-time data to be analyzed. The horizontal axis of the scatter plot represents the real-time data to be analyzed, and the vertical axis represents the fluctuation degree.

[0037] Based on the adaptive neighborhood radius corresponding to the real-time data sequence, the data points in the scatter plot are clustered to obtain the clustering results.

[0038] Preferably, the step of real-time detection of faults in the electricity meter based on the clustering results under each monitoring indicator of the electricity meter includes:

[0039] If an isolated point exists in the clustering results of any monitoring indicator, a fault warning will be issued for the electricity meter.

[0040] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows:

[0041] This invention acquires historical data sequences of any monitoring indicator of an electricity meter within a preset number of historical time periods. Based on the data fluctuations in each historical data sequence, it obtains the historical neighborhood radius for clustering each historical data sequence. It also acquires the real-time sequence of the transformer for any monitoring indicator under the transformer within the current time period and the historical sequence of the transformer within each historical time period. Based on the similarity of data patterns between each historical sequence and the real-time sequence, it obtains a similarity index of electricity consumption between each historical time period and the current time period. Furthermore, it acquires the real-time data sequence of any monitoring indicator of the electricity meter within the current time period. Based on the similarity index of electricity consumption between each historical time period and the current time period, as well as the abnormal behavior of each historical data sequence, it obtains a neighborhood radius fit degree, using each historical neighborhood radius as the neighborhood radius for clustering the real-time data sequence. Based on each historical neighborhood radius and its corresponding neighborhood radius fit degree, it obtains an adaptive neighborhood radius for clustering the real-time data sequence. It then clusters the real-time data sequence based on the adaptive neighborhood radius to obtain clustering results. Finally, based on the clustering results for each monitoring indicator of the electricity meter, it performs real-time fault detection of the electricity meter. Specifically, based on the neighborhood radius when clustering the monitoring data of electricity meters in historical periods, and the similarity of the data change patterns of the monitoring data at the transformer in historical periods and the current period, an adaptive neighborhood radius is obtained when clustering the monitoring data of electricity meters in the current period. Then, the monitoring data of electricity meters in the current period is clustered according to the adaptive neighborhood radius. This effectively reduces the problem that the directly obtained neighborhood radius cannot well identify anomalies because the data in the current period is greatly affected by electricity meter faults, thus improving the accuracy of electricity meter fault identification using the DBSCAN clustering algorithm. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a flowchart of a real-time intelligent fault detection method for electricity meters provided in Embodiment 1 of the present invention. Detailed Implementation

[0044] Embodiments of this disclosure are described in detail below, with examples of these embodiments illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting it.

[0045] It should be noted that the terms "first," "second," etc., used in this disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.

[0046] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0047] See Figure 1 This is a flowchart of a real-time intelligent fault detection method for electricity meters provided in Embodiment 1 of the present invention. Figure 1 As shown, the method may include:

[0048] Step S101: Obtain the historical data sequence of any monitoring indicator of the electricity meter within a preset number of historical time periods, and obtain the historical neighborhood radius when clustering each historical data sequence based on the data fluctuations in each historical data sequence.

[0049] A common anomaly in electricity meter malfunctions is sudden changes in monitoring data, which is traditionally identified using clustering algorithms such as DBSCAN. However, electricity meter malfunctions can also exhibit persistent anomalies such as data drift, severe fluctuations, or static patterns. There may be a large number of abnormal data points within a monitoring period. A key factor affecting the anomaly identification effectiveness of the DBSCAN clustering algorithm is the choice of neighborhood radius. The neighborhood radius should aim to cluster similar normal data into the same cluster while ensuring that abnormal data is separated. When clustering data with a high proportion of abnormal data, the neighborhood radius determined solely by the current data segment is significantly affected by the abnormal data, leading to lower accuracy in identifying electricity meter anomalies.

[0050] Considering that electricity meters are used to monitor users' electricity consumption, if a user's historical electricity consumption is similar to that of the current period, and the electricity meter malfunctions during the current period, using the neighborhood radius from DBSCAN clustering of the historical data segment as the neighborhood radius for DBSCAN clustering of the current data segment can effectively reduce the impact of the overall anomaly of the current data segment on the clustering results and improve the accuracy of electricity meter fault detection. Therefore, in this embodiment of the invention, the neighborhood radius for DBSCAN clustering of the historical data segment of the electricity meter is adjusted based on the similarity of electricity consumption between the user's historical and current periods to obtain the applicable neighborhood radius for DBSCAN clustering of the current data segment of the electricity meter.

[0051] First, set the monitoring indicators for electricity meter fault detection as current, voltage, and power. Record the current, voltage, and power detected by the electricity meter every 1 second. At the same time, set the time period to 5 minutes, that is, evaluate the fault status of the electricity meter every 5 minutes. There is no limit here. Implementers can set the monitoring indicators, recording frequency, and time period according to the specific scenario.

[0052] Next, data on various monitoring indicators of the electricity meter within historical time periods are obtained. Since the numerical range and fluctuations of the electricity meter's monitoring data should be basically consistent under normal circumstances (no circuit adjustments or changes in electricity demand), only data from the two hours preceding the current time period are analyzed. This is not a limitation; implementers can set it according to specific scenarios. Each time period (5 minutes long) within the two hours preceding the current time period is recorded as a historical time period, totaling 24 historical time periods. The neighborhood radius for DBSCAN clustering of the data within the current time period is obtained by adjusting the neighborhood radius when performing DBSCAN clustering on the data from each historical time period. Taking current as an example, the current value monitored by the electricity meter at each moment within the two hours preceding the current time period is linearly normalized to obtain historical current data. Linear normalization is an existing technology and will not be elaborated upon here.

[0053] Finally, the historical current data within each historical period are combined into a historical data sequence, and the neighborhood radius for DBSCAN clustering of the current of the electricity meter in each historical period is obtained. Considering that abnormal data caused by electricity meter faults may include not only excessively high or low values, but also drastic data fluctuations, this embodiment of the invention performs DBSCAN clustering on the monitoring data of the electricity meter based on the fluctuations in the monitoring data to more accurately detect electricity meter faults. Therefore, the neighborhood radius for DBSCAN clustering of the current data of the electricity meter in each historical period also needs to be obtained based on the fluctuations in the historical data sequence of the current of the electricity meter in each historical period. This allows for adjustment of the neighborhood radius for DBSCAN clustering of the current data of the electricity meter in the current period based on the neighborhood radius for DBSCAN clustering of the historical current data of the electricity meter in each historical period. Taking the m-th historical time period as an example, that is, taking the m-th historical data sequence (the historical current data of the electricity meter in the m-th historical time period) as an example, the steps to obtain the neighborhood radius when performing DBSCAN clustering on the m-th historical data sequence are as follows:

[0054] (1) Obtain the data to be analyzed in the m-th historical data sequence, and obtain the fluctuation degree of each data to be analyzed based on the fluctuation in the neighborhood range of each data to be analyzed.

[0055] The time interval is set to 5 seconds, but this is not a limitation and can be set by the implementer according to the specific scenario. Taking each data point in the m-th historical data sequence as the center, the data fluctuation within 5 seconds of each data point is analyzed. Since the monitoring data of the electricity meter is recorded every 1 second in this embodiment, a window of length 5 is constructed centered on each data point in the m-th historical data sequence. The data within each data point's window represents the data within 5 seconds. Because the number of data points before the first two data points and the number of data points after the last two data points in the m-th historical data sequence are less than 2, i.e., a window cannot be constructed, if the time interval between any data point in the m-th historical data sequence and the first data point in the m-th historical data sequence is greater than or equal to 2, and the time interval between any data point and the last data point in the m-th historical data sequence is greater than or equal to 2, then this data point is taken as the data to be analyzed. A window of length 5 is constructed centered on this data point to obtain the degree of fluctuation of the data to be analyzed based on the data fluctuation within the window.

[0056] Taking the i-th data point to be analyzed in the m-th historical data sequence as an example, the specific method for obtaining the fluctuation level of the i-th data point to be analyzed is as follows:

[0057] Calculate the absolute value of the difference between every two adjacent data points within the window of the i-th data point to be analyzed, and then linearly normalize the sum of all the absolute values ​​of the differences to obtain the degree of fluctuation of the i-th data point to be analyzed.

[0058] In one embodiment, the formula for calculating the fluctuation level of the i-th data to be analyzed is:

[0059]

[0060] in, This indicates the degree of fluctuation of the i-th data point to be analyzed. This represents the number of all data points within the window containing the i-th data to be analyzed (in this embodiment of the invention, ...). ), This represents the (j+1)th data point within the window containing the i-th data to be analyzed. This represents the j-th data point within the window containing the i-th data to be analyzed. Represents a linear normalization function. Represents the absolute value symbol.

[0061] It should be noted that, The larger the value, the more drastic the current fluctuation, and thus... The larger the value, the greater the fluctuation of the i-th data to be analyzed.

[0062] Similarly, obtain the fluctuation level of each data point to be analyzed in the m-th historical data sequence.

[0063] (2) Based on the fluctuation level of each data to be analyzed in the m-th historical data sequence, the neighborhood radius when performing DBSCAN clustering on the data in the m-th historical data sequence is obtained and denoted as the historical neighborhood radius.

[0064] K-distance graph method is a commonly used method in the DBSCAN clustering algorithm to determine the neighborhood radius. Based on the fluctuation level of each data point to be analyzed in the m-th historical data sequence, the K-distance graph method is used to obtain the neighborhood radius when clustering the m-th historical data sequence. The general process is as follows: (a) Construct a coordinate graph based on each data point to be analyzed in the m-th historical data sequence and its fluctuation level. The horizontal axis of the coordinate graph represents the fluctuation level, and the vertical axis represents the data to be analyzed; (b) Calculate the distance from each data point in the coordinate graph to its k-th nearest neighbor. Set k to 4; this is not a restriction, and the implementer can set it according to the specific scenario; (c) After sorting all distances in descending order, construct a line graph. The distance corresponding to the first valley point in the line graph is taken as the neighborhood radius when clustering the m-th historical data sequence, denoted as the historical neighborhood radius. The K-distance graph method is existing technology and will not be elaborated upon here.

[0065] Similarly, obtain the neighborhood radius when performing DBSCAN clustering on each historical data sequence, which is to obtain the historical neighborhood radius when performing DBSCAN clustering on the historical current data of the electricity meter in each historical period.

[0066] Thus, the historical neighborhood radius for DBSCAN clustering of the current data of the electricity meters in each historical period is obtained. By adjusting the historical neighborhood radius, the neighborhood radius for DBSCAN clustering of the current data of the electricity meters in the previous period can be obtained.

[0067] Step S102: Obtain the real-time sequence of the transformer under any of the monitoring indicators in the current time period and the historical sequence of the transformer in each historical time period. Based on the data pattern similarity between each historical sequence and the real-time sequence of the transformer, obtain the similarity index of electricity consumption between each historical time period and the current time period.

[0068] In this embodiment of the invention, the appropriate neighborhood radius for the current time period is calculated based on the historical neighborhood radius corresponding to the historical time period. The neighborhood radius corresponding to the historical time period is obtained based on the fluctuation level of historical monitoring data. Since the fluctuation performance of different time periods may be different, such as the current, voltage and power values ​​during peak electricity consumption periods compared to other time periods, the more similar the user's electricity consumption in the historical time period and the current time period, the better the fit of using the historical neighborhood radius corresponding to the historical time period as the neighborhood radius corresponding to the current time period.

[0069] Since the fault status of the electricity meter in the current time period is unknown, the similarity of monitoring data between historical time periods and the current time period cannot be used to determine the similarity of a user's electricity consumption between historical time periods and the current time period. Considering that an electricity meter fault only affects the accuracy of the data monitored by the user and does not affect the data of other nodes, and that the transformer node connected to the user node is the upstream node of the electricity meter, the monitoring data at the transformer will not be affected by the electricity meter fault, in this embodiment of the invention, a resistive voltage divider, a current transformer, and a metering chip are installed at the transformer to monitor the voltage, current, and power at the transformer location, recording the voltage, current, and power values ​​at the transformer every 1 second, according to the recording frequency of the electricity meter monitoring data.

[0070] Since step S101 takes current as an example to obtain the historical neighborhood radius corresponding to the historical data sequence of the current of the energy meter in each historical period, we still take current as an example, and linearly normalize the current value monitored at the transformer in the current period and the current value monitored at the transformer in the two hours before the current period to obtain the transformer historical sequence of the current monitored at the transformer in each historical period, and the transformer real-time sequence of the current monitored at the transformer in the current period.

[0071] Furthermore, based on the similarity between the historical sequence and the real-time sequence of each transformer, a similarity index of the user's electricity consumption regarding current is obtained between each historical time period and the current time period. This index is used to determine the similarity of the user's electricity consumption between the historical time period and the current time period. Taking the m-th historical time period as an example, that is, taking the m-th transformer historical sequence as an example, the steps to obtain the similarity index of the current at the transformer in the m-th historical time period and the current time period are as follows:

[0072] (1) By comparing the data change trends between the historical sequence of the m-th transformer (current data at the transformer in the m-th historical period) and the real-time sequence of the transformer (current data at the transformer in the current period), the similarity of the change trends between the historical sequence of the m-th transformer and the real-time sequence of the transformer is obtained.

[0073] Specifically, since current data fluctuates under normal circumstances, the similarity of current data in different time periods cannot be analyzed by comparing them one by one. However, the overall trend of two current data segments with the same change pattern is the same. Therefore, the STL time series decomposition algorithm is used to obtain the trend items of the m-th transformer historical sequence and the transformer real-time sequence respectively. If the data change trends of the historical period and the current period are similar, their corresponding trend items should also be consistent. The STL time series decomposition algorithm is an existing technology and will not be described in detail here. The trend item of the m-th transformer historical sequence is recorded as the historical trend item, and the trend item of the transformer real-time sequence is recorded as the real-time trend item. In the historical trend item, a preset number of historical trend values ​​are evenly selected and recorded as historical representative trend values. The preset number of representatives is set to 30, which is not limited here. The implementer can set it according to the specific scenario.

[0074] For any historical representative trend value, obtain the position of the historical representative trend value in the historical trend item, obtain the real-time trend value with the same position in the real-time trend item, calculate the absolute value of the difference between the historical representative trend value and the real-time trend value, and obtain the change trend difference value between the historical representative trend value and the real-time trend value.

[0075] Calculate the trend difference value corresponding to each historical representative trend value in the historical sequence of the m-th transformer, accumulate all trend difference values ​​to obtain the accumulated value, calculate the reciprocal of the sum between the preset constant and the accumulated value, and obtain the trend similarity between the historical sequence of the m-th transformer and the real-time sequence of the transformer.

[0076] In one embodiment, the formula for calculating the similarity of the changing trends between the historical sequence of the m-th transformer and the real-time sequence of the transformer is:

[0077]

[0078] in, This represents the similarity in changing trends between the historical sequence of the m-th transformer and the real-time sequence of the transformer. This indicates the preset representative quantity, which is the total number of historical representative trend values. This represents the k-th historical trend value in the trend term (historical trend term) of the m-th transformer historical sequence. This refers to the real-time trend value in the real-time trend term of the transformer's real-time sequence that is in the same position as the k-th historical representative trend value. This represents a preset constant used to ensure that the fraction is meaningful. There are no restrictions here; implementers can set them according to the specific scenario. Represents the absolute value symbol.

[0079] It should be noted that, This represents the difference between two trend values ​​at the same position in the trend term corresponding to the m-th historical time period and the current time period. The smaller the value, the more consistent the trend of current change at the transformer during the m-th historical period with the current trend during the current period, and thus... The larger the value, the more similar the changing trends between the historical sequence of the m-th transformer and the real-time sequence of the transformer.

[0080] (2) By comparing the fluctuations between the historical sequence of the m-th transformer and the real-time sequence of the transformer, the stability similarity between the historical sequence of the m-th transformer and the real-time sequence of the transformer is obtained.

[0081] Because current data has a certain degree of fluctuation, and the degree of fluctuation reflects the stability of the current power consumption, if there is a large difference in the stability of power consumption between the m-th historical period and the current period, it is possible that the user frequently adjusts the electrical equipment in one period, while the electrical equipment is working continuously in another period. In this case, the consistency of the user's power consumption between the m-th historical period and the current period is low.

[0082] Therefore, in this embodiment of the invention, the STL time series decomposition algorithm is used to obtain the residual terms of the m-th transformer historical sequence and the transformer real-time sequence, respectively. The residual terms reflect the fluctuation between the m-th transformer historical sequence and the transformer real-time sequence, and then the stability similarity between the m-th transformer historical sequence and the transformer real-time sequence is obtained, which is used to reflect the consistency of the user's electricity consumption in the m-th historical period and the current period.

[0083] Specifically, calculate the average of all residual values ​​in the residual term of the m-th transformer historical sequence, denoted as the historical residual mean; calculate the average of all residual values ​​in the residual term of the transformer real-time sequence, denoted as the real-time residual mean.

[0084] Calculate the absolute value of the difference between the historical residual mean and the real-time residual mean to obtain the residual consistency index. Calculate the reciprocal of the sum of the preset constant and the residual consistency index to obtain the stability similarity between the m-th transformer historical sequence and the transformer real-time sequence.

[0085] In one embodiment, the formula for calculating the stability similarity between the m-th transformer historical sequence and the transformer real-time sequence is:

[0086]

[0087] in, This represents the stability similarity between the historical sequence of the m-th transformer and the real-time sequence of the transformer. This represents the average of all residual values ​​in the residual term of the m-th transformer historical sequence, which is also the historical residual mean. This represents the average of all residual values ​​in the residual term of the transformer real-time sequence, which is also known as the real-time residual mean. This represents a preset constant used to ensure that the fraction is meaningful. There are no restrictions here; implementers can set them according to the specific scenario. Represents the absolute value symbol.

[0088] It should be noted that, The smaller the value, the more consistent the data stability between the m-th historical period and the current period, and thus... The larger the value, the more similar the stability is between the historical sequence of the m-th transformer and the real-time sequence of the transformer.

[0089] (3) Based on the similarity of the changing trends and the stability of the historical sequence of the m-th transformer and the real-time sequence of the transformer, the similarity index of the user's electricity consumption in the m-th historical period and the current period is obtained.

[0090] Specifically, the product of the trend similarity and the stability similarity is calculated to obtain the similarity index of electricity consumption with respect to current between the historical period in which the m-th transformer historical sequence is located (i.e., the m-th historical period) and the current period.

[0091] In one embodiment, the formula for calculating the similarity index of electricity consumption between the m-th historical time period and the current time period is:

[0092]

[0093] in, This represents the similarity index between the current and electricity consumption patterns in the m-th historical time period. This represents the similarity in trend between the historical sequence of the m-th transformer (current data at the transformer during the m-th historical period) and the real-time sequence of the transformer. This represents the stability similarity between the historical sequence of the m-th transformer and the real-time sequence of the transformer.

[0094] It should be noted that, The larger the value, the more consistent the trend of current change at the transformer in the m-th historical period with the current trend in the current period, and thus... The larger the value, the more similar the user's electricity consumption in the m-th historical period is to the current period. The better the fit between the historical neighborhood radius corresponding to the current data of the electricity meter in the m-th historical period (the m-th historical data sequence) and the neighborhood radius corresponding to the current data of the electricity meter in the current period. The larger the value, the more consistent the data stability between the current at the transformer in the m-th historical period and the current period, and thus... The larger the value, the more similar the user's electricity consumption in the m-th historical period is to the current period. The better the fit between the historical neighborhood radius corresponding to the current data of the electricity meter in the m-th historical period (the m-th historical data sequence) and the neighborhood radius corresponding to the current data of the electricity meter in the current period.

[0095] Thus, the similarity index of the user's electricity consumption in the m-th historical time period and the current time period has been obtained. Similarly, the similarity index of the user's electricity consumption in each historical time period and the current time period has been obtained. Based on the similarity index of the electricity consumption in each historical time period and the current time period, the historical neighborhood radius corresponding to each historical data sequence (the historical current data of the electricity meter in each historical time period) is adjusted to obtain the neighborhood radius when performing DBSCAN clustering on the current data of the electricity meter in the current time period.

[0096] Step S103: Obtain the real-time data sequence of any of the monitoring indicators of the electricity meter in the current time period. Based on the similarity index of electricity consumption between each historical time period and the current time period, as well as the abnormal performance of each historical data sequence, obtain the neighborhood radius fit degree when using the neighborhood radius of each historical neighborhood as the neighborhood radius for clustering the real-time data sequence.

[0097] According to the method of acquiring historical current data of the electricity meter in each historical period, the real-time current data of the electricity meter in the current period is acquired to obtain the real-time data sequence. After obtaining the similarity index of electricity consumption between each historical period and the current period through step S102, the neighborhood radius fit degree is obtained based on the similarity index of electricity consumption between each historical period and the current period, which is used as the neighborhood radius when performing DBSCAN clustering on the real-time data sequence (real-time current data of the electricity meter in the current period).

[0098] Considering that if the data anomaly level within a historical period is significant, the accuracy of the historical neighborhood radius corresponding to that historical period obtained through step S101 will also decrease. To reduce the interference of abnormal historical periods, the impact of historical periods with significant anomalies on the calculation results should be reduced. Therefore, in this embodiment of the invention, the standard deviation of all data in each historical data sequence is calculated to characterize the anomaly level of historical current data within each historical period. Then, based on the similarity index of current consumption between each historical period and the current period, and the standard deviation of each historical data sequence, the neighborhood radius fit is obtained, which uses the neighborhood radius corresponding to each historical data sequence as the neighborhood radius when performing DBSCAN clustering on the real-time data sequence.

[0099] Taking the m-th historical time period as an example, that is, taking the m-th historical data sequence as an example, the specific method for obtaining the neighborhood radius fit degree of using the historical neighborhood radius corresponding to the m-th historical data sequence as the neighborhood radius when performing DBSCAN clustering on the real-time data sequence is as follows:

[0100] The product of the similarity index of electricity consumption corresponding to the historical time period (i.e., the similarity index of electricity consumption between the current period and the historical time period) of the m-th historical data sequence and the reciprocal of the standard deviation of the m-th historical data sequence is linearly normalized to obtain the neighborhood radius fit degree, which uses the historical neighborhood radius corresponding to the m-th historical data sequence as the neighborhood radius when performing DBSCAN clustering on the real-time data sequence.

[0101] In one embodiment, the formula for calculating the neighborhood radius fit when using the historical neighborhood radius corresponding to the m-th historical data sequence as the neighborhood radius for DBSCAN clustering of the real-time data sequence is as follows:

[0102]

[0103] in, This represents the neighborhood radius fit when using the historical neighborhood radius corresponding to the m-th historical data sequence as the neighborhood radius for clustering the real-time data sequence. This represents the similarity index between the current and electricity consumption patterns in the m-th historical time period. This represents the standard deviation of all data in the m-th historical data sequence. This represents the linear normalization function.

[0104] It should be noted that, The larger the value, the more similar the user's electricity consumption patterns in the m-th historical period are to the current period, and thus... The larger the value, the better the fit of using the historical neighborhood radius corresponding to the m-th historical data sequence as the neighborhood radius when performing DBSCAN clustering on the real-time data sequence; The smaller the value, the less abnormal the m-th historical data sequence is, and thus... The larger the value, the better the fit of using the historical neighborhood radius corresponding to the m-th historical data sequence as the neighborhood radius when performing DBSCAN clustering on the real-time data sequence.

[0105] Similarly, the neighborhood radius fit is obtained by using the historical neighborhood radius corresponding to each historical data sequence as the neighborhood radius when performing DBSCAN clustering on the real-time data sequence.

[0106] Step S104: Based on each historical neighborhood radius and its corresponding neighborhood radius fit, obtain the adaptive neighborhood radius when clustering the real-time data sequence. Cluster the real-time data sequence according to the adaptive neighborhood radius to obtain the clustering result. Based on the clustering result under each monitoring indicator of the electricity meter, detect the fault of the electricity meter in real time.

[0107] After obtaining the neighborhood radius fit degree (which uses the historical neighborhood radius corresponding to each historical data sequence as the neighborhood radius when performing DBSCAN clustering on the real-time data sequence) through step S103, the adaptive neighborhood radius for performing DBSCAN clustering on the real-time data sequence is obtained based on the historical neighborhood radius corresponding to each historical data sequence and its neighborhood radius fit degree. Specifically:

[0108] For any historical neighborhood radius, calculate the proportion of the neighborhood radius fit degree corresponding to that historical neighborhood radius in the total neighborhood radius fit degree, and obtain the fit weight of that historical neighborhood radius.

[0109] Obtain the adaptation weights of all historical neighborhood radii, and calculate the weighted average of all historical neighborhood radii based on the adaptation weights of all historical neighborhood radii to obtain the adaptive neighborhood radius when performing DBSCAN clustering on the real-time data sequence.

[0110] In one embodiment, the formula for calculating the adaptive neighborhood radius when performing DBSCAN clustering on a real-time data sequence is as follows:

[0111]

[0112] in, This represents the adaptive neighborhood radius when performing DBSCAN clustering on a real-time data sequence. This represents the number of all historical time periods, that is, the number of all historical data sequences. This represents the historical neighborhood radius corresponding to the m-th historical data sequence. This represents the neighborhood radius fit, where the historical neighborhood radius corresponding to the m-th historical data sequence is used as the neighborhood radius when performing DBSCAN clustering on the real-time data sequence. This represents the neighborhood radius fit when using the historical neighborhood radius corresponding to the q-th historical data sequence as the neighborhood radius for DBSCAN clustering of the real-time data sequence.

[0113] It should be noted that, The larger the value, the better the fit between the historical neighborhood radius corresponding to the m-th historical data sequence and the neighborhood radius when performing DBSCAN clustering on the real-time data sequence, thus enabling the adaptive neighborhood radius when performing DBSCAN clustering on the real-time data sequence. Neighborhood radius when performing DBSCAN clustering on historical data sequences The closer.

[0114] After obtaining the adaptive neighborhood radius for DBSCAN clustering of the real-time data sequence, the fluctuation degree of each real-time data point in the real-time data sequence is obtained in the same way as in step S101, to obtain the fluctuation degree of each real-time data point in the real-time data sequence. Based on the fluctuation degree of each real-time data point in the real-time data sequence and the adaptive neighborhood radius of the real-time data sequence, the DBSCAN clustering algorithm is used to cluster the real-time data points in the real-time data sequence, obtaining the clustering result of the current of the electricity meter in the current time period. The specific method for obtaining the fluctuation degree of each real-time data point in the real-time data sequence is as follows:

[0115] For any real-time data in the real-time data sequence, if the time interval between the real-time data and the first data in the real-time data sequence is greater than or equal to 2, and the time interval between the real-time data and the last data in the real-time data sequence is greater than or equal to 2, then the real-time data is taken as the real-time data to be analyzed, and a real-time window of length 5 is constructed with the real-time data to be analyzed as the center in the real-time data sequence.

[0116] Calculate the absolute value of the difference between every two adjacent data points within the real-time window, and linearly normalize the sum of all the absolute values ​​of the differences to obtain the degree of fluctuation of the real-time data to be analyzed.

[0117] Similarly, the fluctuation level of each real-time data to be analyzed in the real-time data sequence is obtained. Further, a scatter plot is constructed based on the fluctuation level of each real-time data to be analyzed, where the horizontal axis of the scatter plot represents the real-time data to be analyzed and the vertical axis represents the fluctuation level. Based on the adaptive neighborhood radius corresponding to the real-time data sequence, DBSCAN clustering is performed on the data points in the scatter plot to obtain the clustering result of the current of the electricity meter in the current time period.

[0118] Similarly, the adaptive neighborhood radius for DBSCAN clustering of the electricity meter voltage within the current time period is obtained, denoted as... And the adaptive neighborhood radius when performing DBSCAN clustering on the power of the electricity meters in the current time period, denoted as Furthermore, using the DBSCAN clustering algorithm, based on the adaptive neighborhood radius corresponding to each monitoring indicator within the current time period, DBSCAN clustering is performed on the monitoring data of each monitoring indicator within the current time period to obtain the clustering results of each monitoring indicator of the electricity meter within the current time period. DBSCAN clustering is an existing technology and will not be elaborated here. If any monitoring indicator has an outlier in the clustering results within the current time period, a fault warning is issued for the electricity meter to notify relevant technical personnel to troubleshoot the fault. In particular, since power is obtained based on current and voltage, if the power clustering results within the current time period are normal (i.e., no outliers), but the current or voltage clustering results within the current time period are abnormal (i.e., outliers exist), it may be due to interference or other anomalies in the data transmission process of the electricity meter. In this case, a fault warning should also be issued for the electricity meter to notify relevant technical personnel to troubleshoot the interference or other anomalies in the data transmission process of the electricity meter.

[0119] It is worth noting that the key point of this invention is to adaptively generate the neighborhood radius for DBSCAN clustering of the monitoring data for the current period using historical monitoring data of the electricity meter, thereby improving the accuracy of identifying electricity meter faults using the DBSCAN algorithm. Identifying the fault type of the electricity meter based on the clustering results is existing technology and will not be elaborated here.

[0120] In summary, this invention acquires historical data sequences of any monitoring indicator of an electricity meter within a preset number of historical time periods. Based on the data fluctuations in each historical data sequence, it obtains the historical neighborhood radius for clustering each historical data sequence. It also acquires the real-time transformer sequence and the historical transformer sequence for each historical time period for any monitoring indicator of the transformer. Based on the similarity of data patterns between each historical and real-time transformer sequence, it obtains a similarity index of electricity consumption between each historical and current time period. Furthermore, it acquires the real-time data sequence of any monitoring indicator of the electricity meter within the current time period. Based on the similarity index of electricity consumption between each historical and current time period, and the abnormal behavior of each historical data sequence, it obtains a neighborhood radius fit degree, using each historical neighborhood radius as the neighborhood radius for clustering the real-time data sequence. Based on each historical neighborhood radius and its corresponding neighborhood radius fit degree, it obtains an adaptive neighborhood radius for clustering the real-time data sequence. Clustering of the real-time data sequence is performed based on the adaptive neighborhood radius to obtain the clustering results. Finally, based on the clustering results for each monitoring indicator of the electricity meter, real-time fault detection of the electricity meter is performed. Specifically, based on the neighborhood radius when clustering the monitoring data of electricity meters in historical periods, and the similarity of the data change patterns of the monitoring data at the transformer in historical periods and the current period, an adaptive neighborhood radius is obtained when clustering the monitoring data of electricity meters in the current period. Then, the monitoring data of electricity meters in the current period is clustered according to the adaptive neighborhood radius. This effectively reduces the problem that the directly obtained neighborhood radius cannot well identify anomalies because the data in the current period is greatly affected by electricity meter faults, thus improving the accuracy of electricity meter fault identification using the DBSCAN clustering algorithm.

[0121] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A real-time intelligent fault detection method for electricity meters, characterized in that, The method for real-time intelligent fault detection of electricity meters includes: Obtain historical data sequences of any monitoring indicator of the electricity meter within a preset number of historical time periods. Based on the data fluctuations in each historical data sequence, obtain the historical neighborhood radius when clustering each historical data sequence. Obtain the real-time sequence of the transformer under any of the monitoring indicators in the current time period and the historical sequence of the transformer in each historical time period. Based on the similarity of data patterns between each historical sequence and the real-time sequence of the transformer, obtain the similarity index of electricity consumption between each historical time period and the current time period. Obtain the real-time data sequence of any of the monitoring indicators of the electricity meter in the current time period. Based on the similarity index of electricity consumption between each historical time period and the current time period, as well as the abnormal performance of each historical data sequence, obtain the neighborhood radius fit degree when using the neighborhood radius of each historical neighborhood as the neighborhood radius for clustering the real-time data sequence. Based on each historical neighborhood radius and its corresponding neighborhood radius fit, an adaptive neighborhood radius is obtained when clustering the real-time data sequence. The real-time data sequence is then clustered based on the adaptive neighborhood radius to obtain the clustering results. Based on the clustering results under each monitoring indicator of the electricity meter, the faults of the electricity meter are detected in real time. Based on the similarity index of electricity consumption between each historical time period and the current time period, and the abnormal performance of each historical data sequence, the neighborhood radius fit is obtained when using the neighborhood radius of each historical neighborhood as the neighborhood radius for clustering the real-time data sequence, including: For any historical data sequence, calculate the standard deviation of all data in the historical data sequence, and linearly normalize the product of the similarity index of electricity consumption corresponding to the historical period in which the historical data sequence is located and the reciprocal of the standard deviation to obtain the neighborhood radius fit degree, which uses the historical neighborhood radius corresponding to the historical data sequence as the neighborhood radius when clustering the real-time data sequence. Based on each historical neighborhood radius and its corresponding neighborhood radius fit, the adaptive neighborhood radius for clustering real-time data sequences is obtained, including: For any historical neighborhood radius, calculate the proportion of the neighborhood radius fit degree corresponding to that historical neighborhood radius in the total neighborhood radius fit degree, and obtain the fit weight of that historical neighborhood radius. Obtain the adaptation weights of all historical neighborhood radii, and calculate the weighted average of all historical neighborhood radii based on the adaptation weights of all historical neighborhood radii to obtain the adaptive neighborhood radius when clustering real-time data sequences.

2. The method for real-time intelligent fault detection of an electricity meter according to claim 1, characterized in that, The step of obtaining the historical neighborhood radius for clustering each historical data sequence based on data fluctuations in each historical data sequence includes: For any historical data sequence, if the time interval between any data in the historical data sequence and the first data in the historical data sequence is greater than or equal to a preset time interval, and the time interval between any data and the last data in the historical data sequence is greater than or equal to a preset time interval, then the data is taken as the data to be analyzed. In the historical data sequence, a window of a preset length is constructed with the data to be analyzed as the center. The preset time interval is half of the difference between the preset length and the constant 1. Calculate the absolute value of the difference between every two adjacent data points within the window, and linearly normalize the sum of all the absolute values ​​of the differences to obtain the degree of fluctuation of the data to be analyzed. The fluctuation level of each data point to be analyzed in any historical data sequence is obtained. Based on the fluctuation level of each data point to be analyzed in any historical data sequence, the historical neighborhood radius is obtained when clustering the historical data sequence using the K-distance graph method.

3. The method for real-time intelligent fault detection of an electricity meter according to claim 1, characterized in that, The similarity index of electricity consumption between each historical period and the current period is obtained based on the data pattern similarity between the historical sequence and the real-time sequence of each transformer, including: For any historical sequence of a transformer, the STL time series decomposition algorithm is used to obtain the trend terms of the historical sequence and the real-time sequence of the transformer respectively. Based on the similarity of the trend terms of the historical sequence and the real-time sequence of the transformer, the similarity of the changing trends between the historical sequence and the real-time sequence of the transformer is obtained. Using the STL time series decomposition algorithm, the residual terms of any transformer historical sequence and transformer real-time sequence are obtained respectively. Based on the difference between the residual terms of any transformer historical sequence and transformer real-time sequence, the stability similarity between any transformer historical sequence and transformer real-time sequence is obtained. Based on the similarity of the changing trends and the similarity of the stability, the similarity index of the electricity consumption between the historical period in which the historical sequence of any transformer is located and the current period is obtained.

4. The method for real-time intelligent fault detection of an electricity meter according to claim 3, characterized in that, The step of obtaining the trend similarity between any historical transformer sequence and the real-time transformer sequence based on the similarity of the trend terms of any historical transformer sequence and the real-time transformer sequence includes: The trend item of any transformer historical sequence is recorded as the historical trend item, and the trend item of the transformer real-time sequence is recorded as the real-time trend item. In the historical trend item, a preset number of historical trend values ​​are evenly selected and recorded as historical representative trend values. For any historical representative trend value, obtain the position of the historical representative trend value in the historical trend item, obtain the real-time trend value with the same position in the real-time trend item, calculate the absolute value of the difference between the historical representative trend value and the real-time trend value, and obtain the change trend difference value between the historical representative trend value and the real-time trend value. Calculate the trend difference value corresponding to each historical representative trend value in any transformer historical sequence, accumulate all trend difference values ​​to obtain the accumulated value, calculate the reciprocal of the sum between the preset constant and the accumulated value, and obtain the trend similarity between the historical sequence of any transformer and the real-time sequence of the transformer.

5. The method for real-time intelligent fault detection of an electricity meter according to claim 3, characterized in that, The step of obtaining the stability similarity between any historical transformer sequence and the real-time transformer sequence based on the difference in residual terms includes: Calculate the average of all residual values ​​in the residual term of any transformer historical sequence, and denot it as the historical residual mean. Calculate the average of all residual values ​​in the residual term of the transformer real-time sequence, and denot it as the real-time residual mean. Calculate the absolute value of the difference between the historical residual mean and the real-time residual mean to obtain the residual consistency index. Calculate the reciprocal of the sum of the preset constant and the residual consistency index to obtain the stability similarity between any transformer historical sequence and the transformer real-time sequence.

6. The method for real-time intelligent fault detection of an electricity meter according to claim 3, characterized in that, The method of obtaining the similarity index of electricity consumption between the historical period and the current period of any transformer's historical sequence based on the similarity of the change trend and the similarity of the stability includes: The product of the trend similarity and the stability similarity is calculated to obtain the similarity index of electricity consumption between the historical period and the current period of any transformer's historical sequence.

7. The method for real-time intelligent fault detection of an electricity meter according to claim 1, characterized in that, The clustering of real-time data sequences based on adaptive neighborhood radius to obtain clustering results includes: For any real-time data in the real-time data sequence, if the time interval between the real-time data and the first data in the real-time data sequence is greater than or equal to a preset time interval, and the time interval between the real-time data and the last data in the real-time data sequence is greater than or equal to a preset time interval, then the real-time data is taken as the real-time data to be analyzed. In the real-time data sequence, a real-time window of a preset length is constructed with the real-time data to be analyzed as the center. The preset time interval is half of the difference between the preset length and the constant 1. Calculate the absolute value of the difference between every two adjacent data points within the real-time window, and linearly normalize the sum of all absolute values ​​of the differences to obtain the degree of fluctuation of the real-time data to be analyzed. Obtain the fluctuation degree of each real-time data to be analyzed in the real-time data sequence, and construct a scatter plot based on the fluctuation degree of each real-time data to be analyzed. The horizontal axis of the scatter plot represents the real-time data to be analyzed, and the vertical axis represents the fluctuation degree. Based on the adaptive neighborhood radius corresponding to the real-time data sequence, the data points in the scatter plot are clustered to obtain the clustering results.

8. The method for real-time intelligent fault detection of an electricity meter according to claim 1, characterized in that, The method of real-time fault detection of electricity meters based on clustering results under each monitoring indicator of the electricity meter includes: If an isolated point exists in the clustering results of any monitoring indicator, a fault warning will be issued for the electricity meter.

Images