A method for detecting power consumption information of a power box

Abstract

The present application relates to the technical field of electric data processing, and specifically to a method for detecting electric power information of a power box, the method comprising: obtaining a cell where illegal electric power is used; the steps of detecting abnormal electric power users in any cell where illegal electric power is used are: obtaining the electric load data, voltage data and current data of each user in the cell at each acquisition time; determining the load abnormality coefficient at each acquisition time; determining the first electric power abnormality index and the second electric power abnormality index of each user at each acquisition time; comprehensively analyzing the distribution of the first electric power abnormality index and the second electric power abnormality index, determining the decomposition scale of the electric power load data of each user in the wavelet denoising algorithm, and denoising the electric power load data; determining abnormal electric power users based on the similarity of the fluctuation degree between the electric power load data of all users after denoising. The present application aims to improve the accuracy of detecting abnormal electric power users.

Description

Technical Field

[0001] The present application relates to the technical field of electrical data processing, and in particular to a method for detecting power consumption information of a power box. Background Art

[0002] As a key device in the power system, the power box plays an increasingly important role in the smart grid, especially in the data collection and analysis of electricity consumption information. Through electricity consumption information data, the operating status of the power system can be effectively monitored and managed, and various potential abnormal electricity consumption problems can be detected and handled in a timely manner, such as power grid failures, illegal electricity use, etc. In order to ensure the accuracy and reliability of the electricity consumption information data collected by the power box, the collected data needs to be denoised.

[0003] Currently, the commonly used data denoising method is the wavelet denoising algorithm. The denoising effect of the wavelet denoising algorithm is related to the decomposition scale in the algorithm, but the decomposition scale is usually selected manually. If the decomposition scale is set improperly, the noise data in the power consumption information data collected by the power box cannot be effectively filtered out, or the power consumption information data is over-smoothed, resulting in the loss of abnormal data caused by abnormal power consumption problems in the data, affecting the accuracy and reliability of the user's power consumption information data, and further affecting the accuracy of subsequent detection of abnormal power users. Summary of the invention

[0004] In view of the above, it is necessary to provide a method for detecting power consumption information of a power box, which improves the accuracy of detecting abnormal power users compared with traditional power consumption information detection methods.

[0005] The power consumption information detection method of a power box of the present application adopts the following technical solution:

[0006] An embodiment of the present application provides a method for detecting power consumption information of a power box, the method comprising the following steps:

[0007] The cells where the total electricity consumption collected is different from the actual total electricity consumption are regarded as cells with illegal electricity consumption;

[0008] The steps for detecting abnormal electricity users in any community with illegal electricity use are:

[0009] Obtain the power load data, voltage data and current data of each user in the cell at each collection time;

[0010] Based on all power load data of all users in the cell, a power load matrix is ​​constructed, each column vector in the power load matrix is ​​combined with multiple adjacent column vectors to form each local matrix, and based on the difference in the distribution of the eigenvalues ​​of the covariance matrix between each local matrix and all other local matrices, the outliers of each local matrix are determined;

[0011] Determine the load anomaly coefficient at the acquisition time corresponding to each column vector based on the distribution of the abnormal values ​​of all local matrices containing each column vector in the power load matrix;

[0012] The difference between the voltage data and the rated voltage is recorded as the voltage difference, and based on the load abnormality coefficient and the difference between the voltage difference of each user and all other users at each collection time, the first power consumption abnormality index of each user at each collection time is determined;

[0013] Determine the second power consumption abnormality index of each user at each collection time by combining the load abnormality coefficient, the voltage difference, the current, and the historical year-on-year current and the historical month-on-month current of each user at each collection time;

[0014] Comprehensively analyzing the distribution of the first power consumption abnormality index and the second power consumption abnormality index, determining the decomposition scale of the power load data of each user in the wavelet denoising algorithm, and denoising the power load data according to the decomposition scale;

[0015] Based on the similarity of fluctuations between the power load data of all users after denoising, abnormal power users are identified;

[0016] The process of determining the outlier is as follows:

[0017] The mean and range of all eigenvalue modulus lengths of the covariance matrix of each local matrix are recorded as the eigenvalue mean and eigenvalue range, respectively;

[0018] The expression of the outlier value of each local matrix is:

[0019] ; Where dn represents the outlier value of the local matrix An; represents the mean of the differences between the means of the eigenvalues ​​of the local matrix An and all other local matrices; Indicates the extreme eigenvalue of the local matrix An; It represents the mean of the eigenvalue extreme values ​​of all local matrices of the power load matrix; exp() represents an exponential function with a natural constant as the base;

[0020] The expression of the first power consumption abnormality index is:

[0021] ; In the formula, represents the first power consumption anomaly index of user p at the i-th collection time; represents the load anomaly coefficient at the i-th collection moment; represents the voltage difference of user p at the i-th acquisition time; represents the mean value of the voltage difference between user p and all other users at the i-th acquisition time; β represents a value preset to be greater than 0; norm() represents a normalization function;

[0022] The process of determining the second power consumption abnormality index is as follows:

[0023] ; In the formula, represents the second power consumption abnormality index of user p at the i-th collection time; represents the load anomaly coefficient at the i-th collection moment; They respectively represent the voltage data and current data of user p at the i-th collection time; U represents the rated voltage of the transmission line where the cell is located; c1 represents the current data of user p at the same time as the i-th collection time within the previous day of the i-th collection time; c2 represents the current data of user p at the previous collection time of the i-th collection time; ε represents a preset value greater than 0; norm() represents a normalization function; exp() represents an exponential function with a natural constant as the base;

[0024] The process of determining the decomposition scale is as follows:

[0025] The average of the first power consumption abnormality index and the second power consumption abnormality index of each user at each collection time is used as the load abnormality index of each user at each collection time;

[0026] Determining the load abnormality degree of each user based on the distribution of the load abnormality index;

[0027] The expression of the decomposition scale of each user's power load data in the wavelet denoising algorithm is:

[0028] ; In the formula, It represents the decomposition scale of the power load data of user p in the wavelet denoising algorithm; All represent integers greater than 0; Indicates the abnormal load degree of user p; round() indicates the rounding function.

[0029] In one embodiment, the row vectors in the power load matrix are composed of power load data of the same user at all collection times arranged in time series; and the column vectors are composed of power load data of all users at the same collection time.

[0030] In one embodiment, the load anomaly coefficient is the average of the anomaly values ​​of all local matrices containing the column vectors.

[0031] In one embodiment, the process of determining the load abnormality degree is as follows:

[0032] Perform threshold segmentation on the load anomaly index of all users at all collection times;

[0033] The ratio of the number of all load anomaly indexes corresponding to any user that are greater than the segmentation threshold to the total number at the collection time is taken as the load anomaly degree of the any user.

[0034] In one embodiment, the process of determining the abnormal electricity user is as follows:

[0035] Arrange all the power load data of each user after denoising in time sequence to form a power load time series;

[0036] The normalized values ​​of the discrete coefficient, load rate and load peak-to-valley difference rate of the power load data in each power load time series are used to form each load characteristic vector;

[0037] Cluster all load feature vectors;

[0038] Based on the distance between different load feature vectors in each cluster, abnormal electricity users are identified.

[0039] In one embodiment, the distance between any two load characteristic vectors in each cluster is calculated and recorded as the abnormal distance, and the users corresponding to all load characteristic vectors in the cluster where the two load characteristic vectors corresponding to the maximum value of the abnormal distance are located are regarded as abnormal electricity users.

[0040] This application has at least the following beneficial effects:

[0041] The present application determines the abnormal value of each local matrix by analyzing the distribution range of the eigenvalues ​​of the covariance matrix of the local matrix when there is abnormal data in the local matrix, relative to the difference in the distribution range of the eigenvalues ​​when there is no abnormal data, reflecting the possibility of abnormal data in each local matrix, and then determining the possibility of abnormal data in the power load data collected at each collection time. It fully considers the influence of the randomness of power grid failures, illegal power use by users, and noise in the data collection and transmission process on the distribution characteristics of power load data under normal power consumption mode, and improves the accuracy of evaluating the possibility of abnormal data in the power load data collected at each collection time.

[0042] Furthermore, by performing feature analysis on abnormal power load data caused by two abnormal power usage problems, namely, power grid failure and illegal power usage by users, the first and second power usage abnormality indexes are respectively constructed in combination with the load abnormality coefficient, and then the load abnormality degree of each user is determined, thereby improving the accuracy of identifying the abnormal data of the user's power load data caused by power grid failure and illegal power usage by users, and improving the accuracy of evaluating the proportion of abnormal data caused by power grid failure and illegal power usage by users in all power load data of users;

[0043] Furthermore, based on the degree of load abnormality, the decomposition scale in the wavelet denoising algorithm is obtained, the power load data is denoised, and abnormal electricity users are detected. The decomposition scale in the wavelet denoising algorithm can be adaptively obtained according to the characteristics of the power load data. While effectively denoising the power load data, the possibility of abnormal data in the power load data caused by power grid failures and illegal electricity use being over-smoothed is reduced, the accuracy and reliability of the power load data are improved, and the accuracy of subsequent detection of abnormal electricity users is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0045] Figure 1 A flowchart of the steps of a method for detecting power consumption information of a power box provided in this application;

[0046] Figure 2 It is a schematic diagram of the process of obtaining the decomposition scale;

[0047] Figure 3 Schematic diagram of power load curve before denoising;

[0048] Figure 4 Schematic diagram of power load curve for wavelet denoising algorithm with fixed decomposition scale;

[0049] Figure 5 Schematic diagram of power load curve of wavelet denoising algorithm with adaptive decomposition scale;

[0050] Figure 6 The figure is a flowchart for obtaining abnormal electricity users. DETAILED DESCRIPTION

[0051] In the description of the embodiments of the present application, words such as "exemplary", "or", "for example" and the like are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "or", "for example" and the like is intended to present related concepts in a concrete manner.

[0052] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art in the present application. The terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. It should be understood that, unless otherwise specified, " / " means or.

[0053] It should also be noted that the terms "first" and "second" in the present application are used to distinguish similar objects rather than to describe a specific order or sequence.

[0054] The specific scheme of the method for detecting power consumption information of a power box provided by the present application is described in detail below with reference to the accompanying drawings.

[0055] An embodiment of the present application provides a method for detecting power usage information of a power box. Specifically, the following method for detecting power usage information of a power box is provided. Figure 1 , the method comprises the following steps:

[0056] Step 1: The cells where the collected total power consumption is different from the actual total power consumption are regarded as cells with illegal power consumption.

[0057] The smart meter in the power box is used to collect the user's electricity consumption information data, including electricity consumption data. The sampling interval of the electricity consumption information data is T minutes.

[0058] In this embodiment, the value of T is 15 minutes. The value of T is preset manually and can be set by the implementer. This application does not impose any special restrictions.

[0059] Obtain the sum of the power consumption of all users in any cell within a preset time period, and the total power consumption provided to the cell by the regional power grid. If the absolute value of the difference between the sum of the power consumption and the total power consumption exceeds g% of the total power consumption, there are users using electricity illegally in the cell.

[0060] In this embodiment, the value of the preset time period is one month. The value of the preset time period is preset manually and can be set by the implementer. This application does not impose any special restrictions.

[0061] In this embodiment, the value of g is 5. The value of g is preset manually and can be set by the implementer. This application does not impose any special restrictions.

[0062] Step 2: Detect any abnormal electricity users in a cell where illegal electricity use exists.

[0063] The steps for detecting abnormal electricity users in any community with illegal electricity use are:

[0064] Step 2.1, obtaining the power load data, voltage data and current data of each user in the cell at each collection time.

[0065] This application takes any cell with illegal electricity use as an example, and obtains the electricity usage information data of each user in the cell, including power load data, voltage data and current data. The sampling time interval of power load data, voltage data and current data is T minutes.

[0066] Step 2.2, construct a power load matrix based on all power load data of all users in the cell, combine each column vector in the power load matrix with multiple adjacent column vectors to form local matrices, and determine the outliers of each local matrix based on the difference in the distribution of the eigenvalues ​​of the covariance matrix between each local matrix and all other local matrices.

[0067] A power load matrix is ​​constructed based on the power load data of all users in the cell at all collection times. The row vectors in the power load matrix are composed of the power load data of the same user at all collection times arranged in time series; the column vectors are composed of the power load data of all users at the same collection time.

[0068] Taking the nth column in the power load matrix as an example, the matrix composed of α column vectors from the nth column to the (n+α-1)th column is recorded as the nth local matrix An of the power load matrix.

[0069] In this embodiment, the value of α is 100. The value of α is preset manually and can be set by the implementer. This application does not impose any special restrictions.

[0070] Since users in the same community are under the same climatic conditions and most users have similar work and rest schedules, most users in the same community usually have similar electricity consumption patterns, that is, in the same time period, most users in the community usually have similar power load changes, and abnormal data in the power load data is usually caused by various uncertain factors such as power grid failures, illegal power consumption, and noise during data collection and transmission. The randomness of these factors will destroy the data distribution characteristics under the normal power consumption pattern. When there is no abnormal data in the collected power load data, the power load matrix composed of the power load data can be regarded as a random matrix. The covariance matrix constructed by the local matrix of the power load matrix satisfies the single-ring law in the random matrix, that is, all eigenvalues ​​of the covariance matrix of all local matrices without abnormal power load data should be distributed in a circular ring in the complex domain, so that all the eigenvalues ​​have a similar distribution range in the complex domain. When there is abnormal data in the local matrix, the eigenvalues ​​of the covariance matrix of the local matrix will be distributed outside the circular ring range.

[0071] Based on the above analysis, taking the local matrix An as an example, the covariance matrix En of the local matrix An is obtained, all eigenvalues ​​of the covariance matrix En are extracted, and the mean of all eigenvalue moduli is recorded as the eigenvalue mean of the local matrix An, which is used to characterize the average distance between all eigenvalues ​​of the covariance matrix En and the origin in the complex domain; and the range of all eigenvalue moduli is recorded as the eigenvalue range of the local matrix An, which is used to characterize the distribution range of all eigenvalues ​​of the covariance matrix En in the complex domain, wherein the acquisition of the covariance matrix and the extraction of eigenvalues ​​are well-known technologies and will not be repeated in this application.

[0072] Furthermore, based on the difference in the mean eigenvalues ​​between each local matrix and all other local matrices, as well as the distribution of the eigenvalue ranges of all local matrices, the outliers of each local matrix are determined to characterize the possibility of the existence of abnormal data in each local matrix. The expression is:

[0073] ; Where dn represents the outlier value of the local matrix An; represents the mean of the differences between the means of the eigenvalues ​​of the local matrix An and all other local matrices; Indicates the extreme eigenvalue of the local matrix An; It represents the mean of the eigenvalue ranges of all local matrices of the power load matrix; exp() represents an exponential function with a natural constant as the base.

[0074] In this embodiment, the difference between the means of the eigenvalues ​​is the absolute value of the difference. As other implementation methods, based on the measurable difference between the means of the eigenvalues, the implementer may use other calculation methods to measure the difference between the means of the eigenvalues, such as ratio, square of the difference, etc. This application does not impose any special restrictions.

[0075] It should be noted that: in the complex domain, the greater the difference between the average distance between the eigenvalues ​​of the covariance matrix En and the origin relative to the average distance between all eigenvalues ​​of the covariance matrices of the remaining local matrices and the origin, that is, the greater the distribution range of the eigenvalues ​​of the covariance matrix En in the complex domain is compared with the mean difference of the distribution range of all eigenvalues ​​of the covariance matrices of all local matrices in the complex domain, that is, The larger it is, the less similar the eigenvalue distribution range of the covariance matrix En and the covariance matrices of all other local matrices in the complex domain are, and the more likely it is that the distribution range of all eigenvalues ​​of the covariance matrix En in the complex domain will exceed the circular range determined by all eigenvalues ​​of the covariance matrices of all other local matrices in the complex domain, then the greater the possibility of abnormal data existing in the local matrix An.

[0076] Step 2.3, based on the distribution of abnormal values ​​of all local matrices containing each column vector in the power load matrix, determine the load abnormality coefficient at the collection time corresponding to each column vector.

[0077] Taking the i-th collection moment as an example, all local matrices containing the i-th column vector in the power load matrix are obtained, and the mean of the abnormal values ​​of all local matrices containing the i-th column vector is recorded as the load abnormality coefficient at the i-th collection moment, which is used to characterize the possibility of the existence of abnormal data in the power load data of all users at the i-th collection moment. The greater the possibility of the existence of abnormal data in the local matrix of the i-th column vector, the greater the load abnormality coefficient.

[0078] Step 2.4, record the difference between the voltage data and the rated voltage as the voltage difference, and determine the first power consumption abnormality index of each user at each collection time based on the load abnormality coefficient and the difference in the voltage difference between each user and all other users at each collection time.

[0079] Grid failures and illegal electricity use by users will not only cause abnormalities in users' power load data, but will also usually cause abnormalities in users' voltage and current data. For example, in the event of a grid failure, there may be voltage lead failures, abnormal sampling voltage, or abnormal reference voltage. These phenomena may cause abnormal voltage readings of smart meters used by most users in the community. For illegal electricity use by users, such as electricity theft, they may deliberately bypass the meter or tamper with the meter parameters, which will usually cause the voltage data of users who use electricity illegally to be far lower than the rated voltage of the transmission line in the community. At the same time, the current data collected by the smart meter will be much smaller than the user's historical year-on-year and month-on-month current.

[0080] Based on the above analysis, the rated voltage of the transmission line in the cell is obtained. Taking user p in the cell as an example, the difference between the voltage data of user p at the i-th collection time and the rated voltage is recorded as the voltage difference of user p at the i-th collection time.

[0081] In this embodiment, the difference between the voltage data and the rated voltage is the absolute value of the difference. As other implementation methods, based on the measurable difference between the voltage data and the rated voltage, the implementer may use other calculation methods to measure the difference between the voltage data and the rated voltage, such as a ratio, the square of the difference, etc. This application does not impose any special restrictions.

[0082] Furthermore, based on the voltage difference of each user at each collection moment, and the difference in voltage difference between each user and all other users, combined with the load anomaly coefficient at each collection moment, the first power consumption anomaly index of each user at each collection moment is determined to characterize the possibility that the power load data of each user at each collection moment is abnormal data caused by a power grid fault. The expression is:

[0083] ; In the formula, represents the first power consumption anomaly index of user p at the i-th collection time; represents the load anomaly coefficient at the i-th collection moment; represents the voltage difference of user p at the i-th acquisition time; represents the mean value of the voltage difference between user p and all other users at the i-th collection time; β represents a preset value greater than 0, the purpose of which is to prevent the denominator from being 0. The value of β is preset manually and can be set by the implementer. In this embodiment, the value of β is 1; norm() represents a normalization function.

[0084] In this embodiment, the difference between the voltage differences is the absolute value of the difference. As other implementation methods, based on the difference between the measurable voltage differences, the implementer may use other calculation methods to measure the difference between the voltage differences, such as ratio, square of the difference, etc. This application does not impose any special restrictions.

[0085] It should be noted that the greater the possibility of abnormal power load data appearing in the local matrix containing the i-th column vector, that is, The larger the value is, the greater the possibility that the power load data of user p at the i-th collection time is abnormal; at the same time, the greater the difference between the voltage data of user p at the i-th collection time and the rated voltage, that is, The larger the voltage difference between user p and all other users at the i-th acquisition moment, the smaller the voltage difference between user p and all other users at the i-th acquisition moment, that is, The larger the value, the greater the possibility that the power load data of user p at the i-th collection time is abnormal data caused by power grid failure, that is, The bigger.

[0086] Step 2.5, combining the load anomaly coefficient, and the voltage difference, current, and historical year-on-year current and historical quarter-on-quarter current of each user at each collection time, determine the second power consumption anomaly index of each user at each collection time.

[0087] Combined with the load anomaly coefficient at each collection moment, the difference between the voltage data of the user at each collection moment and the rated voltage, and the difference between the current data at each collection moment and its historical year-on-year current and its historical month-on-month current, the second power consumption anomaly index of each user at each collection moment is determined to characterize the possibility that the power load data of each user at each collection moment is abnormal data caused by illegal power consumption by the user. The expression is:

[0088] ; In the formula, represents the second power consumption abnormality index of user p at the i-th collection time; represents the load anomaly coefficient at the i-th collection moment; They respectively represent the voltage data and current data of user p at the ith collection moment; U represents the rated voltage of the transmission line where the cell is located; c1 represents the current data of user p at the same time as the ith collection moment within the previous day of the ith collection moment; c2 represents the current data of user p at the previous collection moment of the ith collection moment; ε represents a value preset to be greater than 0, in order to prevent the denominator from being 0. The value of ε is preset manually and can be set by the implementer. In this embodiment, the value of ε is 1; norm() represents a normalization function.

[0089] It should be noted that: The larger the value is, the greater the possibility that the power load data of user p at the i-th collection time is abnormal; at the same time, the voltage The greater the degree of being less than the rated voltage b, that is, The larger the current is, the greater the degree to which the current is less than the historical year-on-year current c1 and the historical year-on-year current c2, that is, The larger the value, the greater the possibility that the power load data of user p at the i-th collection moment is abnormal data caused by the user's illegal power consumption, that is, The bigger.

[0090] Step 2.6, comprehensively analyzing the distribution of the first power consumption abnormality index and the second power consumption abnormality index, determining the decomposition scale of each user's power load data in the wavelet denoising algorithm, and denoising the power load data according to the decomposition scale.

[0091] Furthermore, the average of the first power consumption anomaly index and the second power consumption anomaly index of each user at each collection moment is used as the load anomaly index of each user at each collection moment, which is used to characterize the possibility that the power load data of each user at each collection moment is abnormal data caused by power grid failure or illegal power consumption by the user. The greater the possibility, the greater the load anomaly index.

[0092] For the load anomaly index of each user in the cell at each collection time, the load anomaly index of each user in the cell at each collection time is calculated according to the same calculation method as the load anomaly index of user p at the i-th collection time.

[0093] The load anomaly index of all users at all collection times is used as the input of the threshold segmentation algorithm, and the segmentation threshold is output. Taking user p as an example, the number of all load anomaly indexes corresponding to user p that are greater than the segmentation threshold is counted, and the ratio of the number to the total number at the collection time is used as the load anomaly degree of user p, which is used to characterize the proportion of abnormal data caused by the two abnormal power consumption problems of power grid failure and illegal power consumption of users in the power load data of user p during the data collection period.

[0094] In this embodiment, the threshold segmentation algorithm is the Otsu threshold segmentation method. As other implementation methods, on the basis of being able to obtain the segmentation threshold, the implementer can adopt other existing technologies to obtain the segmentation threshold, such as the global threshold segmentation method, the iterative threshold segmentation method, etc., and this application does not make any special restrictions.

[0095] Based on the abnormal load degree of each user, the decomposition scale of each user's power load data in the wavelet denoising algorithm is determined, and the expression is:

[0096] ; In the formula, It represents the decomposition scale of the power load data of user p in the wavelet denoising algorithm; Both represent integers greater than 0, which are used to adjust the decomposition scale of the power load data of user p in the wavelet denoising algorithm; represents the abnormal load degree of user p; round() represents the rounding function. The flowchart of obtaining the decomposition scale is as follows: Figure 2 shown.

[0097] In this embodiment, The values ​​of are 3 and 5 respectively. The value of is preset manually and can be set by the implementer. This application does not impose any special restrictions.

[0098] The power load data of user p at all acquisition moments are arranged in time sequence to form a power load time series. The power load time series is used as the input of the wavelet denoising algorithm, and the denoised power load time series of user p is output, wherein the threshold of each wavelet coefficient is determined using the Stein unbiased risk threshold method, and the soft threshold function is selected to process the wavelet coefficient. The wavelet denoising algorithm is a well-known technology and will not be described in detail in this application.

[0099] The schematic diagram of the power load curve before denoising is as follows: Figure 3 The schematic diagram of the power load curve of the wavelet denoising algorithm with a fixed decomposition scale is shown in Figure 4 The schematic diagram of the power load curve of the adaptive decomposition scale wavelet denoising algorithm is shown in Figure 5 shown. Figure 3 , Figure 4 and Figure 5 The horizontal axis in is time, and the vertical axis is load, that is, power load data. The power load curve is a curve composed of the lines between any two adjacent collection time points of power load data. Figure 3 The power load data in is the power load data before denoising. Figure 4 The power load data in is the power load data denoised by the wavelet denoising algorithm with a fixed decomposition scale. Figure 5 The power load data in is the power load data denoised by the wavelet denoising algorithm using adaptive decomposition scale. Figure 4 and Figure 5 It can be seen that compared with the wavelet denoising algorithm with a fixed decomposition scale, the wavelet denoising algorithm with an adaptive decomposition scale has a better denoising effect.

[0100] For each user, the denoised power load time series of each user is obtained using the same method as the denoised power load time series of user p.

[0101] Step 2.7, based on the similarity of fluctuations between the power load data of all users after denoising, determine the abnormal power users.

[0102] For the denoised power load time series of each user, the discrete coefficient, load rate and load peak-to-valley difference rate of the power load data in the power load time series are extracted, and the discrete coefficient, load rate and load peak-to-valley difference rate are normalized respectively, and the normalized value of the discrete coefficient, the normalized value of the load rate and the normalized value of the load peak-to-valley difference rate corresponding to each user are used to form the load characteristic vector of each user, wherein the extraction of the discrete coefficient, load rate and load peak-to-valley difference rate of the power load data are all well-known technologies and will not be repeated in this application.

[0103] A clustering algorithm is used to cluster all load feature vectors, and the distance between any two load feature vectors in each cluster is calculated, recorded as the abnormal distance, and the cluster where the two load feature vectors corresponding to the maximum value of the abnormal distance are located is obtained, and the users corresponding to the load feature vectors in the cluster are regarded as abnormal power users. The schematic diagram of the process of obtaining abnormal power users is shown in the figure. Figure 6 shown.

[0104] In this embodiment, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm is used to cluster all load feature vectors, wherein the distance between the load feature vectors is used as the metric distance in the DBSCAN algorithm. As other implementation methods, on the basis of being able to cluster all load feature vectors, the implementer may use other existing technologies to cluster all load feature vectors, such as the Density peak Clustering Algorithm (DpC), the K-Means algorithm, etc., and this application does not impose any special restrictions.

[0105] In this embodiment, the distance between the load feature vectors is the Euclidean distance. As other implementation methods, on the basis of being able to measure the distance between the load feature vectors, the implementer may adopt other existing technologies to measure the distance between the load feature vectors, such as DTW (dynamic time warping) distance, Manhattan distance, etc., and this application does not impose any special restrictions.

[0106] In summary, the present application determines the abnormal value of each local matrix by analyzing the distribution range of the eigenvalues ​​of the covariance matrix of the local matrix when there is abnormal data in the local matrix, relative to the difference in the distribution range of the eigenvalues ​​when there is no abnormal data, reflecting the possibility of abnormal data in each local matrix, and then determining the possibility of abnormal data in the power load data collected at each collection moment, fully considering the influence of the randomness of power grid failures, illegal power consumption by users, and noise in the data collection and transmission process on the distribution characteristics of power load data under normal power consumption mode, and improving the accuracy of evaluating the possibility of abnormal data in the power load data collected at each collection moment;

[0107] Furthermore, by performing feature analysis on abnormal power load data caused by two abnormal power usage problems, namely, power grid failure and illegal power usage by users, the first and second power usage abnormality indexes are respectively constructed in combination with the load abnormality coefficient, and then the load abnormality degree of each user is determined, thereby improving the accuracy of identifying the abnormal data of the user's power load data caused by power grid failure and illegal power usage by users, and improving the accuracy of evaluating the proportion of abnormal data caused by power grid failure and illegal power usage by users in all power load data of users;

[0108] Furthermore, based on the degree of load abnormality, the decomposition scale in the wavelet denoising algorithm is obtained, the power load data is denoised, and abnormal electricity users are detected. The decomposition scale in the wavelet denoising algorithm can be adaptively obtained according to the characteristics of the power load data. While effectively denoising the power load data, the possibility of abnormal data in the power load data caused by power grid failures and illegal electricity use being over-smoothed is reduced, the accuracy and reliability of the power load data are improved, and the accuracy of subsequent detection of abnormal electricity users is improved.

[0109] The flowchart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to the embodiment of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. In the description corresponding to the flowchart and the block diagram in the accompanying drawings, the operations or steps corresponding to different boxes can also occur in a different order from the order disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. Each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented by a dedicated hardware-based system that performs the specified functions or actions, or may be implemented by a combination of dedicated hardware and computer instructions.

[0110] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the basic features of the present application. Therefore, no matter from which point of view, the above embodiments of the present application should be regarded as exemplary and non-restrictive.

Claims

1. A method for detecting power consumption information of a power box, characterized in that: The method comprises the following steps: The cells where the total electricity consumption collected is different from the actual total electricity consumption are regarded as cells with illegal electricity consumption; The steps for detecting abnormal electricity users in any community with illegal electricity use are: Obtain the power load data, voltage data and current data of each user in the cell at each collection time; Based on all power load data of all users in the cell, a power load matrix is ​​constructed, each column vector in the power load matrix is ​​combined with multiple adjacent column vectors to form each local matrix, and based on the difference in the distribution of the eigenvalues ​​of the covariance matrix between each local matrix and all other local matrices, the outliers of each local matrix are determined; Determine the load anomaly coefficient at the acquisition time corresponding to each column vector based on the distribution of the abnormal values ​​of all local matrices containing each column vector in the power load matrix; The difference between the voltage data and the rated voltage is recorded as the voltage difference, and based on the load abnormality coefficient and the difference between the voltage difference of each user and all other users at each collection time, the first power consumption abnormality index of each user at each collection time is determined; Determine the second power consumption abnormality index of each user at each collection time by combining the load abnormality coefficient, the voltage difference, the current, and the historical year-on-year current and the historical month-on-month current of each user at each collection time; Comprehensively analyzing the distribution of the first power consumption abnormality index and the second power consumption abnormality index, determining the decomposition scale of the power load data of each user in the wavelet denoising algorithm, and denoising the power load data according to the decomposition scale; Based on the similarity of fluctuations between the power load data of all users after denoising, abnormal power users are identified; The process of determining the outlier is as follows: The mean and range of all eigenvalue modulus lengths of the covariance matrix of each local matrix are recorded as the eigenvalue mean and eigenvalue range, respectively; The expression of the outlier value of each local matrix is: ; Where dn represents the outlier value of the local matrix An; represents the mean of the differences between the means of the eigenvalues ​​of the local matrix An and all other local matrices; Indicates the extreme eigenvalue of the local matrix An; It represents the mean of the eigenvalue extreme values ​​of all local matrices of the power load matrix; exp() represents an exponential function with a natural constant as the base; The expression of the first power consumption abnormality index is: ; In the formula, represents the first power consumption anomaly index of user p at the i-th collection time; represents the load anomaly coefficient at the i-th collection moment; represents the voltage difference of user p at the i-th acquisition time; represents the mean value of the voltage difference between user p and all other users at the i-th acquisition time; β represents a value preset to be greater than 0; norm() represents a normalization function; The process of determining the second power consumption abnormality index is as follows: ; In the formula, represents the second power consumption abnormality index of user p at the i-th collection time; represents the load anomaly coefficient at the i-th collection moment; They respectively represent the voltage data and current data of user p at the i-th collection time; U represents the rated voltage of the transmission line where the cell is located; c1 represents the current data of user p at the same time as the i-th collection time within the previous day of the i-th collection time; c2 represents the current data of user p at the previous collection time of the i-th collection time; ε represents a preset value greater than 0; norm() represents a normalization function; exp() represents an exponential function with a natural constant as the base; The process of determining the decomposition scale is as follows: The average of the first power consumption abnormality index and the second power consumption abnormality index of each user at each collection time is used as the load abnormality index of each user at each collection time; Determining the load abnormality degree of each user based on the distribution of the load abnormality index; The expression of the decomposition scale of each user's power load data in the wavelet denoising algorithm is: ; In the formula, It represents the decomposition scale of the power load data of user p in the wavelet denoising algorithm; All represent integers greater than 0; Indicates the abnormal load degree of user p; round() indicates the rounding function.

2. A method for detecting power consumption information of a power box according to claim 1, characterized in that: The row vectors in the power load matrix are composed of the power load data of the same user at all collection times arranged in time series; the column vectors are composed of the power load data of all users at the same collection time.

3. The method for detecting power consumption information of a power box according to claim 1, characterized in that: The load anomaly coefficient is the average of the anomaly values ​​of all local matrices containing the column vectors.

4. The method for detecting power consumption information of a power box according to claim 1, characterized in that: The process of determining the load abnormality degree is as follows: Perform threshold segmentation on the load anomaly index of all users at all collection times; The ratio of the number of all load anomaly indexes corresponding to any user that are greater than the segmentation threshold to the total number at the collection time is taken as the load anomaly degree of the any user.

5. The method for detecting power consumption information of a power box according to claim 1, characterized in that: The process of determining the abnormal electricity user is as follows: Arrange all the power load data of each user after denoising in time sequence to form a power load time series; The normalized values ​​of the discrete coefficient, load rate and load peak-to-valley difference rate of the power load data in each power load time series are used to form each load characteristic vector; Cluster all load feature vectors; Based on the distance between different load feature vectors in each cluster, abnormal electricity users are identified.

6. A method for detecting power consumption information of a power box according to claim 5, characterized in that: The distance between any two load characteristic vectors in each cluster is calculated and recorded as abnormal distance. In the cluster where the two load characteristic vectors corresponding to the maximum value of the abnormal distance are located, all users corresponding to the load characteristic vectors are regarded as abnormal electricity users.

Images