Power distribution terminal with power failure anomaly monitoring and early warning function

Abstract

The present application relates to the technical field of data processing, and especially relates to a power distribution terminal with power failure anomaly monitoring and early warning function, which comprises: a data acquisition module for obtaining a K value range according to the difference between a real-time data sequence of the power distribution terminal in a current period and a historical data sequence in a historical period; a data analysis module for obtaining mutation data in the real-time data sequence; a data processing module for obtaining a mutation distribution density according to the distribution of the difference between the mutation data and data in a local range of the mutation data, and obtaining a time distribution density according to the data distribution in the local range of the mutation data; and an anomaly early warning module for obtaining an adaptive K value of the mutation data according to the mutation distribution density, the time distribution density and the K value range of the mutation data, obtaining a LOF value of the mutation data by using a LOF algorithm, and then monitoring and early warning the power failure anomaly of the power distribution terminal, thereby improving the accuracy of the anomaly early warning.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a power distribution terminal with power supply fault anomaly monitoring and early warning function. Background Technology

[0002] In the power system, the distribution terminal is an important component of the power grid. Its stability and reliability are directly related to the operational safety of the entire power system. Therefore, it is necessary to monitor the power supply data of the distribution terminal for anomalies in order to provide early warning of potential faults and prevent significant economic losses and safety hazards.

[0003] The traditional algorithm for anomaly detection in power supply data of distribution terminals is the Locator-of-Flight (LOF) algorithm. LOF is an anomaly detection algorithm based on local density of data, used to identify outliers in a dataset. In the process of early warning of power supply data anomalies in distribution terminals, the LOF algorithm can be effectively applied to monitor power supply data anomalies. The K value is a key parameter in the LOF algorithm, defining the number of nearest neighbor data points considered when calculating the local density of each data point. However, the traditional LOF algorithm uses a fixed K value. Power supply data is affected by various factors, such as residential electricity consumption times, distribution box temperature changes, and the operating status of surrounding electrical equipment. These factors may cause the power supply data to exhibit different characteristics at different times. Therefore, a fixed K value cannot be adjusted according to data changes, which may lead to errors in the identification of abnormal power supply data.

[0004] Therefore, how to adaptively obtain the K value in the LOF algorithm to improve the accuracy of fault and anomaly early warning for power supply data of distribution terminals has become an urgent problem to be solved. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a power distribution terminal with power supply fault anomaly monitoring and early warning function to solve the problem of how to adaptively obtain the K value in the LOF algorithm in order to improve the accuracy of fault anomaly early warning for power supply data of the power distribution terminal.

[0006] This invention provides a power distribution terminal with power supply fault anomaly monitoring and early warning function, the terminal comprising the following steps:

[0007] The data acquisition module is used to acquire real-time power supply data of the power distribution terminal at each sampling time in the current cycle to obtain the real-time data sequence, acquire the historical data sequence of the power distribution terminal in each historical cycle, and obtain the range of K values ​​in the LOF algorithm based on the difference between the real-time data sequence and each historical data sequence.

[0008] The data analysis module is used to obtain mutation data in the real-time data sequence based on the data fluctuations in the real-time data sequence, and to obtain the local change difference value of each mutation data based on the difference between each mutation data and each data in its local range.

[0009] The data processing module is used to obtain the mutation distribution density of each mutation data based on the distribution of all local variation difference values, obtain the window of each mutation data in the real-time data sequence according to the range of K values, and obtain the temporal distribution density of each mutation data based on the number of mutation data within the window of each mutation data.

[0010] The anomaly warning module is used to obtain the adaptive K value for each mutation data based on the mutation distribution density and time distribution density of each mutation data, as well as the range of K values. Based on the adaptive K value of each mutation data, the LOF algorithm is used to obtain the LOF value of each mutation data. Based on all LOF values, the power distribution terminal is monitored and warned of power supply fault anomalies.

[0011] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows:

[0012] This invention includes a data acquisition module for acquiring real-time power supply data of a power distribution terminal at each sampling moment within the current cycle to obtain a real-time data sequence, acquiring historical data sequences of the power distribution terminal within each historical cycle, and obtaining the K value range in the LOF algorithm based on the difference between the real-time data sequence and each historical data sequence; a data analysis module for acquiring abrupt changes in the real-time data sequence based on data fluctuations, and obtaining the local variation difference value of each abrupt change based on the difference between each abrupt change and each data within its local range; a data processing module for obtaining the abrupt change distribution density of each abrupt change based on the distribution of all local variation difference values, obtaining the window of each abrupt change in the real-time data sequence based on the K value range, and obtaining the temporal distribution density of each abrupt change based on the number of abrupt change within the window of each abrupt change; and an anomaly warning module for obtaining an adaptive K value for each abrupt change based on the abrupt change distribution density, temporal distribution density, and K value range, obtaining the LOF value of each abrupt change using the LOF algorithm based on the adaptive K value of each abrupt change, and performing power supply fault anomaly monitoring and warning for the power distribution terminal based on all LOF values. Specifically, based on the difference between the real-time power supply data sequence and the historical data sequence, the real-time power supply data is transformed into two-dimensional data, thereby obtaining the K value range in the LOF algorithm; based on the difference between each mutation data and each data within its local range, the local change difference value of each mutation data is obtained, which is used to transform the mutation data into two-dimensional data, thereby obtaining the mutation distribution density of each mutation data; based on the data distribution within the window of each mutation data, the temporal distribution density of each mutation data is obtained, which is used to compensate for the loss of temporal characteristics in the mutation distribution density of each mutation data. Combining the K value range, mutation distribution density, and temporal distribution density, an adaptive K value for each mutation data is obtained, making the anomaly detection results obtained by the LOF algorithm more accurate and reducing the probability of missed or false detections in the power distribution terminal. Attached Figure Description

[0013] 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.

[0014] Figure 1 This is a structural block diagram of a power distribution terminal with power supply fault abnormality monitoring and early warning function provided in Embodiment 1 of the present invention;

[0015] Figure 2This invention provides a clustering scatter plot after clustering the historical differences of all data in a real-time data sequence.

[0016] Figure 3 This is a two-dimensional scatter plot provided by an embodiment of the present invention after clustering the local variation differences of all mutation data. Detailed Implementation

[0017] 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.

[0018] 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.

[0019] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0020] See Figure 1 This is a structural block diagram of a power distribution terminal with power supply fault abnormality monitoring and early warning function provided in Embodiment 1 of the present invention, as shown below. Figure 1 As shown, the terminal may include:

[0021] The data acquisition module 11 is used to acquire the real-time power supply data of the power distribution terminal at each sampling time in the current cycle to obtain the real-time data sequence, acquire the historical data sequence of the power distribution terminal in each historical cycle, and obtain the range of K values ​​in the LOF algorithm based on the difference between the real-time data sequence and each historical data sequence.

[0022] The power supply data of the power distribution terminal includes voltage data and current data. In this embodiment of the invention, taking current data as an example, the real-time power supply data (i.e., current data) of the power distribution terminal at each sampling moment in the current cycle is obtained by the current sensor to obtain the real-time data sequence. One day is set as a cycle and the sampling frequency is 1Hz. There is no limitation here, and the implementer can set it according to the specific scenario.

[0023] The traditional algorithm for anomaly detection in power distribution terminal data is the Loop Flow (LOF) algorithm. LOF is an anomaly detection algorithm based on local density, identifying outliers by calculating the local density deviation of each data point relative to its neighborhood. In the process of power distribution terminal fault and anomaly early warning, the LOF algorithm can be effectively applied to power supply data anomaly monitoring. The K value in the LOF algorithm is a key parameter, defining the number of nearest neighbor data points considered when calculating the local density of each data point. However, the traditional LOF algorithm uses a fixed K value. Power supply data is affected by various factors, such as residential electricity consumption periods, distribution box temperature changes, and the operating status of surrounding electrical equipment, which may cause the power supply data to exhibit different data characteristics at different times. Therefore, in this embodiment of the invention, the K value in the LOF algorithm is adaptively obtained by combining the data characteristics of the real-time data sequence of the power distribution terminal in the current period to improve the accuracy of fault and anomaly early warning for the power distribution terminal's power supply data.

[0024] Since the LOF algorithm measures the anomaly of a data point based on the density of each data point and its neighbors, it is first necessary to determine the range of the K value based on the data distribution density of the real-time power supply data to prevent unreasonable K value settings. Considering that the real-time power supply data sequence is one-dimensional time-series data, while the data distribution density requires two-dimensional data analysis, it is necessary to convert the one-dimensional time-series data into two-dimensional data. Furthermore, considering that changes in the power supply data of the distribution terminal are mainly affected by users' electricity consumption habits, and that residential users' electricity consumption habits have certain regularities, in this embodiment of the invention, historical data sequences of the distribution terminal within each historical period are obtained. Based on the differences between the real-time data sequence and each historical data sequence, the real-time power supply data is converted into two-dimensional data. Specifically:

[0025] For any data in the real-time data sequence, based on the sampling time of the data, historical data with the same sampling time as the data are obtained in each historical data sequence and recorded as target historical data. The historical mean of all target historical data is calculated, and the absolute value of the difference between the data and the historical mean is recorded as the historical difference of the data.

[0026] A scatter plot is constructed based on all historical differences to obtain a two-dimensional scatter plot of real-time power supply data. The horizontal axis represents the real-time power supply data (current data in this embodiment of the invention), and the vertical axis represents the historical differences.

[0027] Furthermore, based on the distribution density of the historical differences of each real-time power supply data in the real-time data sequence, the range of K values ​​in the LOF algorithm is obtained. The specific steps are as follows:

[0028] (1) Cluster the data points in the two-dimensional scatter plot of real-time power supply data, and obtain the density distribution characteristic value based on the clustering result.

[0029] Specifically, density-based clustering algorithms, such as the DBSCAN algorithm, are used to cluster the data points (i.e., the historical differences of all data in the real-time data sequence) in the two-dimensional scatter plot of real-time power supply data to obtain the clustering results. No restrictions are placed here; implementers can choose the clustering algorithm according to the specific scenario. The DBSCAN algorithm is existing technology and will not be elaborated upon here; please refer to [reference needed]. Figure 2 This is a clustered scatter plot obtained by clustering the historical differences of all data in a real-time data sequence. Under normal circumstances, power supply data fluctuates within a certain range. If there are many data points within a certain range, it indicates that the power supply data within that range is normal. Figure 2 The data points in circle 1; due to uncontrollable user behavior, there may be situations where historical differences are large, such as... Figure 2 The data in circle number 2; some data also differ significantly from the other data, such as... Figure 2 The data in circle 3 are likely outliers, appearing as isolated points in the cluster scatter plot. Since isolated points are not considered when calculating the K-value range, they are removed based on the clustering results, resulting in at least one cluster. The number of elements in each cluster is recorded as the corresponding cluster density. The mean cluster density of all cluster densities is obtained, and the reciprocal of this mean density is subtracted from a constant 1 to obtain the first variable. The number of clusters with densities greater than or equal to the mean density is counted, and the reciprocal of this number is subtracted from a constant 1 to obtain the second variable. The average of the first and second variables is calculated to obtain the density distribution characteristic value of all clusters.

[0030] In one embodiment, the formula for calculating the density distribution characteristic value is:

[0031]

[0032] Where A represents the density distribution characteristic value, and M represents the number of all clusters. H represents the cluster density of the m-th cluster, and H represents the number of clusters that are greater than or equal to the mean cluster density.

[0033] It should be noted that the larger the mean cluster density, and the more clusters there are that are greater than or equal to the mean cluster density, the more concentrated the data distribution is and the larger the density distribution characteristic value is.

[0034] (2) Obtain the range of K values ​​in the LOF algorithm based on the density distribution characteristic values.

[0035] A larger density distribution characteristic value indicates a larger number of elements in each cluster. If the K value in the LOF algorithm is small in this case, the local density may not be accurately estimated due to insufficient points in the neighborhood, leading to inaccurate outlier identification using the LOF algorithm. Therefore, the K value needs to be increased appropriately. Conversely, the K value needs to be decreased appropriately to prevent data from clusters of different densities from being mixed together, which would lead to inaccurate local density estimation and, consequently, inaccurate outlier identification using the LOF algorithm. The specific method for obtaining the range of K values ​​in the LOF algorithm is as follows:

[0036] The minimum cluster density is used as the lower bound of the initial K value in the LOF algorithm, denoted as . The maximum cluster density is used as the upper limit of the initial K value in the LOF algorithm, denoted as . Calculate the difference between the upper limit and the lower limit of the initial K value to obtain the range difference of the initial K value, i.e. The product of the density distribution characteristic value and the difference in the initial K value range is rounded up to obtain the K value range adjustment coefficient, i.e. ;

[0037] In this embodiment of the invention, based on the concept of rounding, a preset density distribution characteristic value of 0.5 is set. This is not a limitation and can be set by the implementer according to the specific scenario. If the density distribution characteristic value is less than 0.5, the initial upper limit of K needs to be appropriately reduced, that is, the initial upper limit of K is subtracted from the K range adjustment coefficient to obtain the final upper limit of K, denoted as . Based on the initial lower limit of K and the final upper limit of K, the range of K values ​​in the LOF algorithm is obtained. The formula for calculating the upper limit of the final K value is as follows: ,in, This represents the upper limit of the final K value. This represents the upper limit of the initial K value. This represents the lower limit of the initial K value, and A represents the characteristic value of the density distribution. Indicates the rounding up symbol;

[0038] If the density distribution characteristic value is less than 0.5, the initial lower limit of K needs to be appropriately increased. This is achieved by calculating the sum of the initial lower limit of K and the K range adjustment coefficient to obtain the final lower limit of K. Based on the final lower limit of K and the initial upper limit of K, the K range in the LOF algorithm is obtained. The formula for calculating the lower limit of the final K value is as follows: ,in, This indicates the lower limit of the final K value. This represents the upper limit of the initial K value. This represents the lower limit of the initial K value, and A represents the characteristic value of the density distribution. The symbol indicates rounding up.

[0039] Thus, the range of K values ​​in the LOF algorithm is obtained, denoted as .

[0040] The data analysis module 12 is used to obtain mutation data in the real-time data sequence based on the data fluctuation in the real-time data sequence, and to obtain the local change difference value of each mutation data based on the difference between each mutation data and each data in its local range.

[0041] Since abnormal data generally deviates from normal or surrounding values, in this embodiment of the invention, data deviating from normal or surrounding values ​​is selected based on the data fluctuations in the real-time data sequence and denoted as mutation data. This mutation data is then further analyzed to obtain the adaptive K value for each mutation data point in the LOF algorithm. The steps for obtaining mutation data in the real-time data sequence are as follows:

[0042] (1) Based on the rate of change of each data in the real-time data sequence, the mutation degree of each data is obtained, and suspected mutation data is obtained by screening based on the mutation degree.

[0043] In this embodiment of the invention, taking the i-th data in a real-time data sequence as an example, the specific method for obtaining the mutation degree of the i-th data is as follows:

[0044] A scatter plot is constructed based on the real-time data sequence, where the horizontal axis of the scatter plot represents the sampling time and the vertical axis represents the real-time power supply data.

[0045] If the i-th data is neither the first nor the last data in the scatter plot, then calculate the left slope between the i-th data and its left neighbor, and the right slope between the i-th data and its right neighbor. Calculate the absolute value of the difference between the left and right slopes to obtain the rate of change of the i-th data. Calculate the constant 1 minus the reciprocal of the rate of change to obtain the degree of fluctuation of the i-th data.

[0046] In a real-time data sequence, if the number of data points preceding the i-th data point is greater than or equal to a first preset number, then the first preset number of data points preceding the i-th data point in the real-time data sequence are obtained and denoted as reference data. The position of the reference data points is continuous with the position of the i-th data point. In order to reduce the impact of fluctuations in other data points on the i-th data point, in this embodiment of the invention, the first preset number is set to 3. This is not limited here, and the implementer can set it according to the specific scenario. The mean of all reference data points is calculated to obtain the reference value of the i-th data point. The absolute value of the difference between the i-th data point and the reference value is calculated to obtain the reference difference. The constant 1 is subtracted from the reciprocal of the reference difference to obtain the mutation probability of the i-th data point. The average value between the mutation probability of the i-th data point and the degree of fluctuation is calculated to obtain the degree of mutation of the i-th data point.

[0047] In one embodiment, the formula for calculating the mutation degree of the i-th data is:

[0048]

[0049] in, This indicates the degree of mutation in the i-th data point. Indicates the right slope. Indicates the left slope. This represents the value of the i-th data element. This indicates the first preset quantity, which is also the quantity of reference data. This represents the value of the j-th reference data, where 1 represents a constant. Represents the absolute value symbol.

[0050] It should be noted that, This is the rate of change of the i-th data point. Specifically, if the i-th data point is the first data point in the scatter plot, then the right slope between the i-th data point and its right neighbor is calculated to obtain the rate of change of the i-th data point. If the i-th data point is the last data point in the scatter plot, then calculate the left slope between the i-th data point and its left neighbor to obtain the rate of change of the i-th data point. . This represents the mutation probability of the i-th data point. Specifically, if the number of data points preceding the i-th data point is less than a first preset number, then the mutation probability of the i-th data point is not calculated; instead, the fluctuation level of the i-th data point is used as its mutation degree. Therefore, the formula for calculating the mutation degree of the i-th data point is: The greater the rate of change, the more significant the turning point of the i-th data point, and the more pronounced the change in the i-th data point. The larger the value, the greater the degree of mutation in the i-th data. The larger the value, the greater the difference between the i-th data point and the preceding preset number of data points, and the more likely the i-th data point is to be a data point that has undergone a mutation. The larger the value, the greater the degree of mutation of the i-th data.

[0051] Considering that the power distribution terminal belongs to the power system and is inevitably subject to electromagnetic interference, there may be data superposition that causes abnormal power supply data performance. Therefore, in this embodiment of the invention, a preset mutation degree threshold of 0.6 is set according to experimental statistics. This is not limited here, and the implementer can set it according to the specific scenario. If the mutation degree of the i-th data is greater than 0.6, the i-th data is marked as suspected mutation data. Similarly, all suspected mutation data in the real-time data sequence are obtained to further analyze the suspected mutation data, filter out the interference data that may be affected by electromagnetic interference, and thus obtain the mutation data that actually shows abnormality, thereby improving the accuracy of the determination of mutation data.

[0052] (2) In the real-time data sequence, based on the change trend of each suspected mutation data in its local range, the mutation data in the real-time data sequence is obtained.

[0053] Because data affected by electromagnetic interference is instantaneously recoverable, the trends of change before and after the interference are similar, and the differences between the data are small. Based on these characteristics, taking the f-th suspected mutation data as an example, the interference probability of the f-th suspected mutation data is obtained, and then the truly abnormal mutation data is screened out. Specifically:

[0054] In a real-time data sequence, if the f-th suspected mutation data is not the first data and the number of data before the f-th suspected mutation data is less than a second preset number, then all data before the f-th suspected mutation data and the second preset number of data after the f-th suspected mutation data are recorded as target data. The position of the target data is continuous with the position of the f-th suspected mutation data. The reciprocal of the DTW distance between the target data before the f-th suspected mutation data and the target data after the f-th suspected mutation data is calculated to obtain the first trend similarity of the f-th suspected mutation data. The reciprocal of the absolute value of the difference between the left and right adjacent data of the f-th suspected mutation data is calculated to obtain the second trend similarity of the f-th suspected mutation data. The first trend similarity and the second trend similarity are weighted and summed to obtain the interference probability of the f-th suspected mutation data.

[0055] If the f-th suspected mutation data is not the last data, and the number of data after the f-th suspected mutation data is less than the second preset number, then all data after the f-th suspected mutation data and the second preset number of data before the f-th suspected mutation data are recorded as target data. The position of the target data is continuous with the position of the f-th suspected mutation data. The reciprocal of the DTW distance between the target data before the f-th suspected mutation data and the target data after the f-th suspected mutation data is calculated to obtain the first trend similarity of the f-th suspected mutation data. The reciprocal of the absolute value of the difference between the left and right adjacent data of the f-th suspected mutation data is calculated to obtain the second trend similarity of the f-th suspected mutation data. The first trend similarity and the second trend similarity are weighted and summed to obtain the interference probability of the f-th suspected mutation data.

[0056] If the number of data points before the f-th suspected mutation data point is greater than or equal to a second preset number, and the number of data points after the f-th suspected mutation data point is also greater than or equal to a second preset number, then the second preset number of data points after the f-th suspected mutation data point and the second preset number of data points before the f-th suspected mutation data point are recorded as target data. The position of the target data point is continuous with the position of the f-th suspected mutation data point. The reciprocal of the DTW distance between the target data points before and after the f-th suspected mutation data point is calculated to obtain the first trend similarity of the f-th suspected mutation data point. The reciprocal of the absolute value of the difference between the left and right adjacent data points of the f-th suspected mutation data point is calculated to obtain the second trend similarity of the f-th suspected mutation data point. The first trend similarity and the second trend similarity are weighted and summed to obtain the interference probability of the f-th suspected mutation data point. To reduce the impact of fluctuations in other data points on the f-th suspected mutation data point, in this embodiment of the invention, the second preset number is set to 3. This is not a limitation and can be set by the implementer according to the specific scenario.

[0057] In one implementation, the formula for calculating the interference probability of the f-th suspected mutation data is:

[0058]

[0059] in, This represents the interference probability of the f-th suspected mutation data. Indicates the first weight. Indicates the first-trend similarity. Indicates the second weight. This represents the left neighbor of the f-th suspected mutation data in the real-time data sequence. This represents the right neighbor of the f-th suspected mutation data in the real-time data sequence. Represents the absolute value symbol.

[0060] It should be noted that, The larger the value, the more similar the trend of change between the data before and after the f-th suspected mutation data, and thus... The larger the value, the more likely the f-th suspected mutation data is data affected by electromagnetic interference; The smaller the value, the smaller the difference between the data before and after the f-th suspected mutation data, and thus... The larger the value, the more likely the f-th suspected mutation data is to be data affected by electromagnetic interference. Since the trend of changes between the data before and after the f-th suspected mutation data better reflects the recoverability of electromagnetic interference, it is set to... , There are no restrictions here; implementers can set them according to the specific scenario.

[0061] Specifically, if the f-th suspected mutation data is the first data in the real-time data sequence, then a second preset number of data following the f-th suspected mutation data are obtained and denoted as target data. The position of the target data is continuous with the position of the f-th suspected mutation data. The absolute value of the difference between the f-th suspected mutation data and each of the target data is calculated to obtain the data difference. The mean of all data differences is calculated to obtain the average data difference. The constant 1 is subtracted from the average data difference to obtain the interference probability of the f-th suspected mutation data. That is, the formula for calculating the interference probability of the f-th suspected mutation data is: ,in, This represents the interference probability of the f-th suspected mutation data. Indicates the second preset quantity. This represents the f-th suspected mutation data point in the real-time data sequence. This represents the f-th suspected mutation data point in the real-time data sequence. One target data;

[0062] If the f-th suspected mutation data is the last data in the real-time data sequence, then the second preset number of data points preceding the f-th suspected mutation data are obtained and denoted as target data. The position of the target data is continuous with the position of the f-th suspected mutation data. The absolute value of the difference between the f-th suspected mutation data and each of the target data is calculated to obtain the data difference. The mean of all data differences is calculated to obtain the average data difference. The constant 1 is subtracted from the average data difference to obtain the interference probability of the f-th suspected mutation data. That is, the formula for calculating the interference probability of the f-th suspected mutation data is: ,in, This represents the interference probability of the f-th suspected mutation data. Indicates the second preset quantity. This represents the f-th suspected mutation data point in the real-time data sequence. This represents the h-th target data point preceding the f-th suspected mutation data point in the real-time data sequence.

[0063] According to experimental statistics, the preset interference probability threshold is set to 0.7. This threshold is not restricted and can be set by the implementer according to the specific scenario. If the interference probability of the f-th suspected mutation data is less than or equal to 0.7, then the f-th suspected mutation data is determined to be the actual abnormal mutation data, and the f-th suspected mutation data is recorded as mutation data. If the interference probability of the f-th suspected mutation data is greater than 0.7, then the f-th suspected mutation data is determined to be interference data affected by electromagnetic interference, and the f-th suspected mutation data is recorded as interference data.

[0064] Similarly, all mutation data in the real-time data sequence are obtained. If no mutation data is found, it is determined that the power distribution terminal is normal; if mutation data is found, further analysis of the mutation data is required to obtain the adaptive K value of each mutation data in the LOF algorithm.

[0065] Since the LOF algorithm measures the anomaly of a data point based on the density of each data point and its neighbors, it is first necessary to obtain its adaptive K value based on the data distribution density of the mutation data. Considering that the mutation data is one-dimensional time series data, while the data distribution density requires two-dimensional data for analysis, it is necessary to transform the one-dimensional time series data into two-dimensional data. Therefore, in this embodiment of the invention, based on the difference between each mutation data point and each data point within its local range, a local change difference value is obtained for each mutation data point. This local change difference value is then used to transform the mutation data into two-dimensional data, facilitating the subsequent analysis of the density distribution of the mutation data.

[0066] Considering that interfering data can have varying degrees of impact, it is necessary to determine whether interfering data exists in the real-time data sequence before obtaining the local variation difference value for each mutation data. If interfering data exists, its values ​​need to be modified. Specifically:

[0067] Obtain the target number of all target data corresponding to any interfering data. In the real-time data sequence, obtain the mean of the target number of the interfering data and its nearest non-interfering data. Replace the interfering data with the mean of all its nearest non-interfering data to obtain a new real-time data sequence.

[0068] If there is no interfering data, the real-time data sequence will be used as the new real-time data sequence.

[0069] Furthermore, based on the new real-time data sequence, the local variation difference value of each mutation data is obtained, specifically:

[0070] In the new real-time data sequence, for any mutated data, a third preset number of nearest neighbor data are obtained. That is, based on the position of each data in the new real-time data sequence, the distance between the mutated data and other data in the new real-time data sequence is calculated, and all distances are sorted in ascending order to obtain a distance sequence. The data in the new real-time data sequence corresponding to the first third preset number of distances in the distance sequence are selected as the nearest neighbor data of the mutated data. In order to reduce the impact of fluctuations in other data on the mutated data, in this embodiment of the invention, the third preset number is set to 4. This is not limited here, and the implementer can set it according to the specific scenario. The absolute value of the difference between the mutated data and each of the nearest neighbor data is calculated, and the average value of all the absolute values ​​of the difference is calculated to obtain the local change difference value of the mutated data.

[0071] Similarly, obtain the local variation value for each mutation data.

[0072] The data processing module 13 is used to obtain the mutation distribution density of each mutation data according to the distribution of all local change difference values, obtain the window of each mutation data in the real-time data sequence according to the range of K values, and obtain the time distribution density of each mutation data according to the number of mutation data in the window of each mutation data.

[0073] Since the LOF algorithm measures the anomaly of a data point based on the density of each data point and its neighbors, this embodiment of the invention uses a density clustering algorithm, such as the DBSCAN algorithm, to cluster all local variation differences, obtaining at least one cluster. No limitation is imposed here; implementers can choose the clustering algorithm according to the specific scenario. The DBSCAN algorithm is existing technology and will not be elaborated upon here; see [reference]. Figure 3 It is a two-dimensional scatter plot that clusters the local variation differences of all mutation data. Figure 3 In the diagram, the horizontal axis represents mutation data, and the vertical axis represents local variation values. For the local variation value of the t-th mutation data, the number of members in the cluster containing the local variation value of the t-th mutation data is taken as the mutation distribution density of the t-th mutation data, denoted as . Similarly, obtain the mutation distribution density of all mutation data, and obtain its adaptive K value based on the mutation distribution density of each mutation data: the larger the mutation distribution density, the higher the possibility of abnormal power supply data of the distribution terminal. In this case, a smaller K value should be selected to capture subtle changes; conversely, a larger K value should be selected to avoid false detections.

[0074] Considering that the scatter plot after clustering the local variation differences of all mutation data loses the time characteristics, in this embodiment of the invention, a window for each mutation data is constructed according to the range of K values ​​in the LOF algorithm obtained by the data acquisition module 11, so as to obtain the time distribution density of each mutation data according to the data distribution within the window.

[0075] Taking the t-th mutation data as an example, the specific method for obtaining the window for the t-th mutation data based on the range of K values ​​is as follows:

[0076] Calculate the difference between the maximum and minimum values ​​in the range of K values ​​to obtain the K value range difference. If the K value range difference is even, the sum of the constant 1 and the K value range difference is recorded as the window size. If the K value range difference is odd, the K value range difference is recorded as the window size.

[0077] In the new real-time data sequence, if the number of data before the t-th mutation data is greater than or equal to the fourth preset number, and the number of data after the t-th mutation data is greater than or equal to the fourth preset number, then in the new real-time data sequence, a window centered on the t-th mutation data is constructed according to the window size, and the fourth preset number is half of the difference between the K value range difference and the constant 1.

[0078] If the number of data before the t-th mutation data is less than the fourth preset number, then in the new real-time data sequence, a window centered on the t-th mutation data is constructed based on the window size, all data before the t-th mutation data, and some data after the t-th mutation data.

[0079] If the number of data after the t-th mutation data is less than the fourth preset number, then in the new real-time data sequence, a window centered on the t-th mutation data is constructed based on the window size, all data after the t-th mutation data, and some data before the t-th mutation data.

[0080] Furthermore, based on the data distribution within the window of the t-th mutation data, the temporal distribution density of the t-th mutation data is obtained, specifically:

[0081] The number of other mutation data besides the t-th mutation data is obtained within the window of the t-th mutation data. If the number is 0, the time distribution density of the t-th mutation data is set to 0.

[0082] If the quantity is not zero, calculate the constant 1 minus the reciprocal of the quantity to obtain the first density; calculate the time interval between the sampling times of every two adjacent mutation data within the window of the t-th mutation data, and use the reciprocal of the mean of all time intervals as the second density; calculate the reciprocal of the variance of all time intervals to obtain the third density; calculate the average of the first density, the second density, and the third density to obtain the time distribution density of the t-th mutation data.

[0083] In one embodiment, the formula for calculating the temporal distribution density of the t-th mutation data is:

[0084]

[0085] in, Let N represent the temporal distribution density of the t-th mutation data, and let N represent the number of other mutation data within the window containing the t-th mutation data. The window representing the t-th mutation data The mutation data and the first The time interval between sampling times corresponding to each mutation data point, where S represents the variance of all time intervals.

[0086] It should be noted that the larger N is, the more mutations are present within the window containing the t-th mutation data; that is, the more mutations are present within the time range corresponding to the window containing the t-th mutation data. The larger the value, the greater the temporal distribution density of the t-th mutation data. The smaller the value, the smaller the time interval between two adjacent mutation data points, and thus... The larger S is, the greater the temporal distribution density of the t-th mutation data; the smaller S is, the more similar the time intervals between each pair of adjacent mutation data, and the more balanced the distribution of mutation data within the window of the t-th mutation data. The larger the value, the greater the temporal distribution density of the t-th mutation data.

[0087] Thus, the mutation distribution density and time distribution density of the t-th mutation data have been obtained.

[0088] The anomaly warning module 14 is used to obtain the adaptive K value of each mutation data according to the mutation distribution density and time distribution density of each mutation data, as well as the range of K values. Based on the adaptive K value of each mutation data, the LOF value of each mutation data is obtained using the LOF algorithm. Based on all LOF values, the power distribution terminal is monitored and warned of power supply fault anomalies.

[0089] After obtaining the mutation distribution density and time distribution density of the t-th mutation data through the data processing module 13, the adaptive K value of the t-th mutation data in the LOF algorithm is obtained based on the mutation distribution density and time distribution density of the t-th mutation data and the K value range in the LOF algorithm obtained by the data acquisition module 11: The larger the mutation distribution density and the larger the time distribution density, the higher the probability of abnormal power supply data in the distribution terminal. In this case, a smaller K value should be selected to capture subtle changes; conversely, a larger K value should be selected to avoid false detections. Specifically:

[0090] Subtract the mutation distribution density of the t-th mutation data from the constant 1 to obtain the first feature value of the t-th mutation data. Subtract the time distribution density of the t-th mutation data from the constant 1 to obtain the second feature value of the t-th mutation data. Perform a weighted summation on the first feature value and the second feature value to obtain the weighted summation result.

[0091] Calculate the difference between the maximum and minimum values ​​in the range of K values ​​to obtain the K value range difference. Round down the product of the K value range difference and the weighted summation result to obtain the K value adjustment coefficient. Calculate the sum between the minimum value in the range of K values ​​and the K value adjustment coefficient to obtain the adaptive K value of the t-th mutation data.

[0092] In one implementation, the adaptive K value for the t-th mutation data is calculated using the following formula:

[0093]

[0094] in, This represents the adaptive K value for the t-th mutation data. This represents the minimum value in the range of K values. This represents the difference between the maximum and minimum values ​​within the range of K values, i.e., the range difference of K values. Indicates the third weight. This represents the mutation distribution density of the t-th mutation data. Indicates the fourth weight. This represents the time distribution density of the t-th mutation data. This indicates the floor function.

[0095] It should be noted that the higher the mutation distribution density and time distribution density of the t-th mutation data, the higher the probability of anomalies in the power supply data of the distribution terminal. In this case, a smaller K value should be selected to capture subtle changes. The smaller the value, the better. Since the mutation distribution density is obtained based on the difference between mutation data and each data point within its local range, it is more valuable for reference. Therefore, in this embodiment of the invention, it is set... , There are no restrictions here; implementers can set them according to the specific scenario.

[0096] Similarly, the adaptive K value for each mutation data point is obtained. Further, based on the adaptive K value for each mutation data point, the LOF value for each mutation data point is obtained using the LOF algorithm.

[0097] Since LOF values ​​are usually non-negative real numbers, if the LOF value is greater than 1, it indicates a higher probability of outliers. Therefore, in this embodiment of the invention, a preset anomaly threshold of 1 is set. This is not limited here, and the implementer can set it according to the specific scenario. If there is at least one LOF value greater than 1 among all LOF values, an abnormal power supply fault warning will be issued to the power distribution terminal to notify relevant personnel in a timely manner.

[0098] If all LOF values ​​are less than or equal to 1, the power supply data of the distribution terminal at future times is predicted using the ARIMA algorithm based on the real-time data sequence. This predicted data is then added to the real-time data sequence as the last data point. According to data processing module 13, the mutation distribution density and time distribution density of the predicted data are obtained. Then, based on the K value range, the mutation distribution density, and the time distribution density of the predicted data, an adaptive K value is obtained. Further, based on the adaptive K value of the predicted data, the LOF algorithm is used to obtain the LOF value of the predicted data. If the LOF value of the predicted data is greater than 1, an abnormal power supply fault warning is issued to the distribution terminal to promptly notify relevant personnel. The LOF algorithm and ARIMA algorithm are existing technologies and will not be elaborated upon here.

[0099] In summary, this invention includes a data acquisition module for acquiring real-time power supply data of the power distribution terminal at each sampling moment within the current cycle to obtain a real-time data sequence, acquiring historical data sequences of the power distribution terminal within each historical cycle, and obtaining the K value range in the LOF algorithm based on the difference between the real-time data sequence and each historical data sequence; a data analysis module for acquiring abrupt changes in the real-time data sequence based on data fluctuations, and obtaining the local variation difference value of each abrupt change based on the difference between each abrupt change and each data within its local range; a data processing module for obtaining the abrupt change distribution density of each abrupt change based on the distribution of all local variation difference values, obtaining the window of each abrupt change in the real-time data sequence based on the K value range, and obtaining the temporal distribution density of each abrupt change based on the number of abrupt change within the window of each abrupt change; and an anomaly warning module for obtaining the adaptive K value of each abrupt change based on the abrupt change distribution density, temporal distribution density, and K value range, obtaining the LOF value of each abrupt change using the LOF algorithm based on the adaptive K value of each abrupt change, and performing power supply fault anomaly monitoring and warning on the power distribution terminal based on all LOF values. Specifically, based on the difference between the real-time power supply data sequence and the historical data sequence, the real-time power supply data is transformed into two-dimensional data, thereby obtaining the K value range in the LOF algorithm; based on the difference between each mutation data and each data within its local range, the local change difference value of each mutation data is obtained, which is used to transform the mutation data into two-dimensional data, thereby obtaining the mutation distribution density of each mutation data; based on the data distribution within the window of each mutation data, the temporal distribution density of each mutation data is obtained, which is used to compensate for the loss of temporal characteristics in the mutation distribution density of each mutation data. Combining the K value range, mutation distribution density, and temporal distribution density, an adaptive K value for each mutation data is obtained, making the anomaly detection results obtained by the LOF algorithm more accurate and reducing the probability of missed or false detections in the power distribution terminal.

[0100] 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 power distribution terminal having a power failure anomaly monitoring and warning function, characterized by, The power distribution terminal with the power supply fault anomaly monitoring and early warning function comprises: a data acquisition module, configured to acquire real-time power supply data of the power distribution terminal at each sampling time in a current period to obtain a real-time data sequence, acquire historical data sequences of the power distribution terminal in each historical period, and obtain a K value range in the LOF algorithm according to differences between the real-time data sequence and each historical data sequence; a data analysis module, configured to acquire mutant data in the real-time data sequence according to data fluctuation in the real-time data sequence, and obtain a local change difference value of each mutant data according to a difference between each mutant data and each data in a local range of the mutant data; a data processing module, configured to obtain a mutant distribution density of each mutant data according to a distribution of all local change difference values, obtain a window of each mutant data in the real-time data sequence according to the K value range, and obtain a time distribution density of each mutant data according to a number of mutant data in the window of each mutant data; an anomaly early warning module, configured to obtain an adaptive K value of each mutant data according to the mutant distribution density and the time distribution density of each mutant data and the K value range, obtain an LOF value of each mutant data by using the LOF algorithm according to the adaptive K value of each mutant data, and perform power supply fault anomaly monitoring and early warning on the power distribution terminal according to all LOF values; In the data acquisition module, the K value range in the LOF algorithm is obtained according to differences between the real-time data sequence and each historical data sequence, and the method comprises the following steps: for any data in the real-time data sequence, a historical data same as a sampling time of the any data is acquired in each historical data sequence as a target historical data according to the sampling time of the any data, a historical mean value of all target historical data is calculated, and an absolute value of a difference between the any data and the historical mean value is recorded as a historical difference value of the any data; the historical difference values of all data in the real-time data sequence are clustered to obtain a clustering result, isolated points are removed according to the clustering result to obtain at least one clustering cluster, and a number of elements in each clustering cluster is recorded as a corresponding cluster density; a cluster density mean value of all cluster densities is acquired, a first variable is obtained by subtracting a reciprocal of the cluster density mean value from a constant 1; a number of clustering clusters greater than or equal to the cluster density mean value is counted, a second variable is obtained by subtracting a reciprocal of the number of clustering clusters from the constant 1; an average value between the first variable and the second variable is calculated to obtain a density distribution characteristic value of all clustering clusters; a minimum cluster density is taken as an initial K value lower limit in the LOF algorithm, a maximum cluster density is taken as an initial K value upper limit in the LOF algorithm, an initial K value range difference is obtained by calculating a difference between the initial K value upper limit and the initial K value lower limit, and a K value range adjustment coefficient is obtained by rounding up a product between the density distribution characteristic value and the initial K value range difference. If the density distribution characteristic value is less than a preset density distribution characteristic value, a lower limit of the initial K value is subtracted by the K value range adjustment coefficient to obtain a final upper limit of the K value, and a K value range in the LOF algorithm is obtained according to the initial lower limit of the K value and the final upper limit of the K value; If the density distribution characteristic value is greater than or equal to a preset density distribution characteristic value, a sum of the initial lower limit of the K value and the K value range adjustment coefficient is obtained to obtain a final lower limit of the K value, and a K value range in the LOF algorithm is obtained according to the final lower limit of the K value and the initial upper limit of the K value; In the data processing module, a time distribution density of each mutation data is obtained according to a number of mutation data in a window of each mutation data, including: For any mutation data, a number of other mutation data in a window of the any mutation data except the any mutation data is obtained, if the number is 0, the time distribution density of the any mutation data is set to 0; If the number is not 0, a first density is obtained by calculating a reciprocal of the number from a constant 1, a second density is obtained by calculating a reciprocal of a mean value of all time intervals between sampling time points corresponding to each adjacent two mutation data in the window of the any mutation data, and a third density is obtained by calculating a reciprocal of a variance of all time intervals, and an average value between the first density, the second density and the third density is calculated to obtain the time distribution density of the any mutation data.

2. The power distribution terminal with power failure anomaly monitoring and warning function according to claim 1, characterized in that, In the data analysis module, mutation data in a real-time data sequence is obtained according to a data fluctuation in the real-time data sequence, including: A scatter plot is constructed according to the real-time data sequence, a horizontal coordinate of the scatter plot represents a sampling time point, and a vertical coordinate represents real-time power supply data; For any data in the scatter plot, if the any data is a first data in the scatter plot, a right slope between the any data and a right adjacent data is calculated to obtain a change rate of the any data; If the any data is a last data in the scatter plot, a left slope between the any data and a left adjacent data is calculated to obtain the change rate of the any data; If the any data is not the first data in the scatter plot and is not the last data in the scatter plot, a left slope between the any data and the left adjacent data and a right slope between the any data and the right adjacent data are calculated, and an absolute value of a difference between the left slope and the right slope is calculated to obtain the change rate of the any data; A fluctuation degree of the any data is obtained by calculating a reciprocal of the change rate from a constant 1. In the real-time data sequence, if the data before the any data is greater than or equal to a first preset number, the first preset number of data before the any data in the real-time data sequence is obtained as reference data, the position of the reference data is continuous with the position of the any data, the mean value of all the reference data is calculated to obtain the reference value of the any data, the absolute value of the difference between the any data and the reference value is calculated to obtain the reference difference value, the reciprocal of the reference difference value is subtracted by the constant 1 to obtain the mutation probability of the any data, and the average value between the mutation probability of the any data and the fluctuation degree is calculated to obtain the mutation degree of the any data; If the data before the any data is less than the first preset number, the fluctuation degree of the any data is taken as the mutation degree of the any data; If the mutation degree of the any data is greater than a preset mutation degree threshold, the any data is taken as suspected mutation data; In the real-time data sequence, according to the change trend of each suspected mutation data in its local range, the mutation data in the real-time data sequence is obtained.

3. The power distribution terminal having a power failure abnormality monitoring and warning function according to claim 2, characterized by, In the data analysis module, according to the change trend of each suspected mutation data in its local range, the mutation data in the real-time data sequence is obtained, including: For any suspected mutation data, in the real-time data sequence, if the any suspected mutation data is the first data, the second preset number of data after the any suspected data is obtained as target data, the position of the target data is continuous with the position of the any suspected mutation data, the absolute value of the difference between the any suspected mutation data and each target data is calculated to obtain a data difference, the mean value of all the data differences is calculated to obtain an average data difference, the average data difference is subtracted by the constant 1 to obtain the interference probability of the any suspected mutation data; If the any suspected mutation data is the last data, the second preset number of data before the any suspected data is obtained as target data, the position of the target data is continuous with the position of the any suspected mutation data, the absolute value of the difference between the any suspected mutation data and each target data is calculated to obtain a data difference, the mean value of all the data differences is calculated to obtain an average data difference, the average data difference is subtracted by the constant 1 to obtain the interference probability of the any suspected mutation data; If the any suspected mutation data is not the first data, and the number of data before the any suspected mutation data is less than the second preset number, all data before the any suspected mutation data and the second preset number of data after the any suspected mutation data are recorded as target data, the position of the target data is continuous with the position of the any suspected mutation data, the reciprocal of the DTW distance between the target data before the any suspected mutation data and the target data after the any suspected mutation data is calculated to obtain the first trend similarity of the any suspected mutation data, the reciprocal of the absolute value of the difference between the left adjacent data and the right adjacent data of the any suspected mutation data is calculated to obtain the second trend similarity of the any suspected mutation data, and the first trend similarity and the first trend similarity are weighted and summed to obtain the interference probability of the any suspected mutation data; If the any suspected mutation data is not the last data, and the number of data after the any suspected mutation data is less than the second preset number, all data after the any suspected mutation data and the second preset number of data before the any suspected mutation data are recorded as target data, the position of the target data is continuous with the position of the any suspected mutation data, the reciprocal of the DTW distance between the target data before the any suspected mutation data and the target data after the any suspected mutation data is calculated to obtain the first trend similarity of the any suspected mutation data, the reciprocal of the absolute value of the difference between the left adjacent data and the right adjacent data of the any suspected mutation data is calculated to obtain the second trend similarity of the any suspected mutation data, and the first trend similarity and the first trend similarity are weighted and summed to obtain the interference probability of the any suspected mutation data; If the number of data before the any suspected mutation data is greater than or equal to the second preset number, and the number of data after the any suspected mutation data is greater than or equal to the second preset number, the second preset number of data after the any suspected mutation data and the second preset number of data before the any suspected mutation data are recorded as target data, the position of the target data is continuous with the position of the any suspected mutation data, the reciprocal of the DTW distance between the target data before the any suspected mutation data and the target data after the any suspected mutation data is calculated to obtain the first trend similarity of the any suspected mutation data, the reciprocal of the absolute value of the difference between the left adjacent data and the right adjacent data of the any suspected mutation data is calculated to obtain the second trend similarity of the any suspected mutation data, and the first trend similarity and the first trend similarity are weighted and summed to obtain the interference probability of the any suspected mutation data; If the interference probability of the any suspected mutation data is less than or equal to the preset interference probability threshold, the any suspected mutation data is recorded as mutation data.

4. The power distribution terminal having a power failure abnormality monitoring and warning function according to claim 3, characterized by, The data analysis module, according to the difference between each mutation data and each data in its local range, obtains a local change difference value of each mutation data, comprising: In the real-time data sequence, if the interference probability of any suspected mutation data is greater than a preset interference probability threshold, the any suspected mutation data is recorded as interference data, if there is interference data in the real-time data sequence, the target quantity of all target data corresponding to any interference data is obtained, the mean value of the target quantity of near neighbor non-interference data of the any interference data in the real-time data sequence is obtained, the any interference data is replaced by the mean value of all near neighbor non-interference data, and a new real-time data sequence is obtained, if there is no interference data in the real-time data sequence, the real-time data sequence is taken as the new real-time data sequence; In the new real-time data sequence, for any mutation data, the third preset quantity of near neighbor data of the any mutation data is obtained, the absolute value of the difference between the any mutation data and each near neighbor data is calculated, the average value of all difference absolute values is calculated, and the local change difference value of the any mutation data is obtained.

5. The power distribution terminal with power failure anomaly monitoring and warning function according to claim 1, characterized in that, The data processing module, according to the distribution of all local change difference values, obtains a mutation distribution density of each mutation data, comprising: All local change difference values are clustered to obtain at least one cluster, for any local change difference value, the member quantity of the cluster where the any local change difference value is located is obtained, and the mutation distribution density of the mutation data corresponding to the any local change difference value is obtained.

6. The power distribution terminal having a power failure abnormality monitoring and warning function according to claim 4, characterized by, The data processing module, according to the K value range, obtains a window of each mutation data in the real-time data sequence, comprising: The difference between the maximum value and the minimum value in the K value range is calculated to obtain a K value range difference, if the K value range difference is even, the sum between the constant 1 and the K value range difference is taken as the window size, if the K value range difference is odd, the K value range difference is taken as the window size; In the new real-time data sequence, for any mutation data, if the quantity of data before the any mutation data is greater than or equal to a fourth preset quantity, and the quantity of data after the any mutation data is greater than or equal to the fourth preset quantity, according to the window size, a window centered on the any mutation data is constructed in the new real-time data sequence, the fourth preset quantity is half of the difference between the K value range difference and the constant 1; If the quantity of data before the any mutation data is less than the fourth preset quantity, according to the window size, all data before the any mutation data and part of data after the any mutation data, a window centered on the any mutation data is constructed in the new real-time data sequence; If the quantity of data after the any mutation data is less than the fourth preset quantity, according to the window size, all data after the any mutation data and part of data before the any mutation data, a window centered on the any mutation data is constructed in the new real-time data sequence.

7. The power distribution terminal having a power failure abnormality monitoring and warning function according to claim 1, characterized by, The anomaly early warning module, according to the mutation distribution density and the time distribution density of each mutation data, and the K value range, obtains the adaptive K value of each mutation data, comprising: For any mutation data, subtract the mutation distribution density of the any mutation data from a constant 1 to obtain a first eigenvalue of the any mutation data, subtract the time distribution density of the any mutation data from the constant 1 to obtain a second eigenvalue of the any mutation data, and perform weighted summation on the first eigenvalue and the second eigenvalue to obtain a weighted summation result; Calculate the difference between the maximum value and the minimum value in the K value range to obtain a K value range difference, and take the product between the K value range difference and the weighted summation result to the nearest integer to obtain a K value adjustment coefficient, and calculate the sum between the minimum value in the K value range and the K value adjustment coefficient to obtain the adaptive K value of the any mutation data.

8. The power distribution terminal with power failure anomaly monitoring and warning function according to claim 1, characterized in that, The power supply fault anomaly monitoring and early warning of the power distribution terminal according to all LOF values, comprising: In all LOF values, if there is at least one LOF value greater than the preset anomaly threshold, the power supply fault anomaly of the power distribution terminal is early warned; If all LOF values are less than or equal to the preset anomaly threshold, the power supply data of the power distribution terminal at the future time is predicted according to the real-time data sequence by using the preset prediction algorithm to obtain the predicted data, the adaptive K value of the predicted data is obtained, the LOF value of the predicted data is obtained according to the adaptive K value of the predicted data, and if the LOF value of the predicted data is greater than the preset anomaly threshold, the power supply fault anomaly of the power distribution terminal is early warned.

Images