Electricity stealing user identification method and device and electronic equipment

Abstract

The invention provides an electricity stealing user identification method and device and electronic equipment, and the method comprises the steps: obtaining the power utilization data of a plurality of users; clustering the power consumption data of the plurality of users to obtain the power consumption type of each user; and performing electricity stealing behavior identification on the electricity consumption data of the users belonging to the same electricity consumption type to respectively obtain electricity stealing users in each electricity consumption type. The electricity consumption data of the multiple users are clustered to obtain the electricity consumption type of each user, then the electricity stealing users are determined in the electricity consumption data of the same electricity consumption type, misjudgment possibly occurring during comparison among different electricity consumption types is reduced, abnormal behavior recognition in similar electricity consumption modes is generally more accurate, and the accuracy of the electricity stealing users is improved. Due to the fact that data distribution is tighter, abnormal behaviors can be more easily highlighted from background noise, and recognition accuracy is improved. And meanwhile, because the power utilization types are distinguished, parallel calculation can be carried out when the electricity stealing users are determined by each power utilization type, and the efficiency of identifying the electricity stealing users is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of anti-electricity theft analysis, and in particular relates to an electricity theft user identification method, device and electronic equipment. Background Art

[0002] The power industry is a basic industry that promotes the sustainable development of the society's high-quality economy. Especially in the era of big data and artificial intelligence, as the advancement of science and technology changes people's way of production and life, the society's demand for electricity has increased significantly. However, many electricity users are driven by interests to steal electricity illegally. There are many ways to steal electricity, such as directly hooking cables or wires to steal electricity. In the era of intelligent high-tech, there are also theft techniques such as forging lead seals and even unsealing and restoring electricity, strong magnetic theft, wireless remote control theft, high-power wireless interference theft, and electricity meter programming theft. These theft incidents are endless, causing serious economic losses to the country and power companies.

[0003] In the related art, electricity stealing users are generally identified directly based on the trend or time series analysis of massive user electricity usage data. However, the massive user electricity usage data makes it difficult to identify electricity stealing users, resulting in poor identification effect of electricity stealing users. Summary of the invention

[0004] In view of this, an object of the present invention is to provide a method, device and electronic equipment for identifying electricity theft users to meet the needs.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] According to a first aspect, an embodiment of the present invention provides a method for identifying electricity theft users, comprising: acquiring electricity usage data of multiple users; clustering the electricity usage data of multiple users to obtain the electricity usage type of each user; identifying electricity theft behavior based on the electricity usage data of users belonging to the same electricity usage type, and obtaining electricity theft users in each electricity usage type respectively.

[0007] Optionally, a method for identifying electricity thieves also includes: determining the geographical coordinates of the electricity thieves in each electricity usage type based on smart meter identifiers in electricity usage data of the electricity thieves in each electricity usage type; dividing the electricity thieves in the same electricity usage type into regions according to density based on the geographical coordinates of the electricity thieves in the same electricity usage type to obtain high-incidence areas of electricity theft in each electricity usage type; calculating the average amount of electricity theft by the electricity thieves in the high-incidence areas of electricity theft in each electricity usage type, and sorting the average amount of electricity theft by the electricity thieves in the high-incidence areas of electricity theft in each electricity usage type; and querying the strategy mapping table based on the user's electricity usage type and the sorting result, and outputting anti-electricity theft strategies for the high-incidence areas of electricity theft in each electricity usage type.

[0008] Optionally, electricity theft behavior is identified on electricity usage data of users belonging to the same electricity usage type to obtain electricity theft users in each electricity usage type, including: extracting electricity usage data features of users belonging to the same electricity usage type, the data features including downward trend features, change features, and fluctuation features; reducing the dimension of the electricity usage data features of users belonging to the same electricity usage type according to the principal component analysis method; inputting the reduced electricity usage data features of users belonging to the same electricity usage type into a local outlier factor model to calculate the local reachable density of each user's electricity usage data; determining the local outlier factor score of each user based on the local reachable density of each user's electricity usage data; when the local outlier factor score of any user exceeds a preset score threshold, the user is identified as an electricity theft user of that electricity usage type.

[0009] Optionally, when the local outlier factor score of any user exceeds a preset score threshold, the user is identified as an electricity theft user of the electricity usage type, including: when the local outlier factor score of any user exceeds a preset score threshold, the user is identified as the first suspicious electricity theft user of the electricity usage type; the entropy weight method is used to weight the electricity theft behavior identification indicator set according to the electricity usage data of users of the same electricity usage type, and an observation weighted matrix of the electricity theft behavior identification indicator is obtained, wherein the electricity theft behavior identification indicator includes a current abnormality indicator, a voltage undervoltage indicator, and a power factor abnormality indicator; according to the observation weight matrix of the electricity theft behavior identification indicator The weighted matrix is ​​measured to determine the positive ideal solution and negative ideal solution of the user's electricity consumption data under the behavior of electricity theft; the Euclidean distance between the electricity consumption data of each user of the same electricity consumption type and the positive ideal solution and the negative ideal solution is calculated; according to the Euclidean distance between the electricity consumption data of each user and the positive ideal solution and the negative ideal solution, the closeness value of the electricity consumption data of each user and the positive ideal solution is determined; the closeness value of the electricity consumption data of each user and the positive ideal solution is sorted, and the user corresponding to the closeness value of the target proportion is identified as the second suspicious electricity theft user; the first suspicious electricity theft user and the second suspicious electricity theft user of the same electricity consumption type are intersected to obtain the electricity theft user.

[0010] Optionally, the electricity usage data of multiple users are clustered to obtain the electricity usage type of each user, including: extracting electricity usage feature data from the electricity usage data of multiple users; applying the elbow rule to the electricity usage feature data to determine the number of cluster centers; calculating the distance values ​​between the electricity usage feature data in the electricity usage data of multiple users and each cluster center; dividing the electricity usage data of multiple users into cluster centers with the smallest distance value, and taking the type represented by the divided cluster center as the user's electricity usage type.

[0011] Optionally, based on the geographic coordinates of electricity theft users in the same electricity usage type, the electricity theft users in the same electricity usage type are divided into regions according to density to obtain high-incidence areas of electricity theft in each electricity usage type, including: determining the distance value between any user and adjacent users based on the geographic coordinates of electricity theft users in the same electricity usage type; determining the average distance value based on the distance value between any user and adjacent users; when the average distance value of any user is less than a preset average distance value, the user is regarded as the central user of the current electricity usage type, and when there are overlapping adjacent users between multiple central users, each central user and the corresponding adjacent user are included in the same area; and determining the high-incidence areas of electricity theft in the current electricity usage type based on the geographic coordinates of users in the same area.

[0012] Optionally, a method for identifying electricity theft users also includes: receiving electricity usage data of high-incidence areas of electricity theft for each type of electricity usage within a target time period, and evaluating the effectiveness of strategy implementation based on the electricity usage data; when the strategy implementation effect does not achieve the expected implementation effect, sending information on high-incidence areas of electricity theft for the corresponding type of electricity usage to the target client.

[0013] According to a second aspect, an embodiment of the present invention provides an electricity theft user identification device, comprising: an electricity usage data acquisition module, used to acquire electricity usage data of multiple users; an electricity usage type determination module, used to cluster the electricity usage data of multiple users to obtain the electricity usage type of each user; and an electricity theft user determination module, used to identify electricity theft behavior based on the electricity usage data of users belonging to the same electricity usage type, and obtain electricity theft users in each electricity usage type.

[0014] According to a third aspect, an embodiment of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the steps of the method for identifying electricity theft users described in the first aspect or any embodiment of the first aspect.

[0015] According to a fourth aspect, an embodiment of the present invention provides a computer storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the method for identifying electricity theft users described in the first aspect or any embodiment of the first aspect.

[0016] The embodiment of the present invention provides a method for identifying electricity theft users, which clusters the electricity usage data of multiple users to obtain the electricity usage type of each user, and then determines the electricity theft users in the electricity usage data of the same electricity usage type, thereby reducing the possible misjudgment when comparing different electricity usage types, and identifying abnormal behavior in similar electricity usage patterns is usually more accurate, because the data distribution is more compact, and abnormal behavior is more likely to stand out from background noise. At the same time, since the present invention can train a specific electricity theft detection model for the electricity usage pattern of the specific type for users of the same electricity usage type, which is simpler than training a general model, the model training can be simplified in this way, and since the electricity usage types are distinguished, each electricity usage type can be operated in parallel when determining the electricity theft users, thereby improving the efficiency of identifying the electricity theft users.

[0017] An embodiment of the present invention provides a method for identifying electricity theft users. By determining the average amount of electricity theft in high-incidence areas of electricity theft for each type of electricity usage, serious electricity theft areas can be determined, which is helpful for analyzing the spatial distribution characteristics of electricity theft behavior and facilitating unified management. In addition, according to the user's electricity usage type and the sorting result, a strategy mapping table is queried to output the corresponding anti-electricity theft strategy, thereby realizing the automation of strategy provision.

[0018] An electricity theft user identification method provided by an embodiment of the present invention adopts a local outlier factor model to determine electricity theft users in various electricity usage types, and can identify users who show abnormal electricity usage patterns, which is helpful to timely discover and prevent electricity theft behaviors. The local outlier factor model has good adaptability to different electricity usage types and data distributions, which improves the versatility of the model, and performs dimensionality reduction processing in data processing, which reduces misjudgments caused by data noise or irrelevant features, and the data processing speed after dimensionality reduction is faster, which helps to improve the efficiency of electricity theft detection.

[0019] An embodiment of the present invention provides a method for identifying electricity theft users, which assigns weights through the entropy weight method. As an objective weight allocation method, the entropy weight method can determine the weights of various indicators according to the characteristics of the data itself, reducing the influence of subjective judgment. In addition, this embodiment takes the intersection of the two methods to integrate the recognition results of the two methods. The two methods detect electricity theft behavior from different angles, respectively, thereby improving the accuracy of recognition.

[0020] Other advantages, objectives and features of the present invention will be described in the following description and will be apparent to those skilled in the art to some extent, or those skilled in the art may be taught from the practice of the present invention. The objectives and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to make the purpose, technical solution and beneficial effects of the present invention clearer, the present invention provides the following drawings for illustration:

[0022] Figure 1 This is a specific example flow chart of identifying electricity theft users in the present invention;

[0023] Figure 2 A schematic diagram of calculating the fifth distance of a data point P in the local outlier factor algorithm of the present invention;

[0024] Figure 3 A schematic diagram of reachable distance in the local outlier factor algorithm of the present invention;

[0025] Figure 4 It is a schematic diagram of the local outlier factor score of the present invention;

[0026] Figure 5 It is a scatter plot of the local outlier factor score of the present invention;

[0027] Figure 6 A schematic diagram of determining a high-incidence area of ​​electricity theft according to the present invention;

[0028] Figure 7 It is a principle block diagram of a specific example of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, it can also be the internal connection of two components, it can be a wireless connection, or it can be a wired connection. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0031] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0032] The embodiment of the present invention provides a method for identifying electricity stealing users. Figure 1 As shown, including:

[0033] S101, obtaining electricity consumption data of multiple users; the electricity consumption data of the users may include the electricity consumption, electricity consumption time, peak electricity consumption, etc. of each user.

[0034] S103, clustering the electricity consumption data of multiple users to obtain the electricity consumption type of each user;

[0035] For example, cluster analysis is an unsupervised learning method for grouping data points into different clusters so that the data points within the same cluster have high similarity, while the data points between different clusters have low similarity. The clustering algorithm can be K-means clustering, hierarchical clustering, DBSCAN, etc. In power data analysis, clustering can help identify different power consumption patterns, such as residential users, commercial users, industrial users, and aquaculture power consumption.

[0036] The process of clustering the electricity consumption data of multiple users to obtain the electricity consumption type of each user includes: extracting electricity consumption feature data from the electricity consumption data of multiple users; applying the elbow rule to the electricity consumption feature data to determine the number of cluster centers; calculating the distance values ​​between the electricity consumption feature data in the electricity consumption data of multiple users and each cluster center; dividing the electricity consumption data of multiple users into the cluster center with the smallest distance value, and taking the type represented by the divided cluster center as the user's electricity consumption type.

[0037] Specifically, first, according to the acquired user's electricity consumption data, the data is cleaned and normalized to eliminate the influence of different dimensions, and then the features in the electricity consumption data are extracted, such as maximum electricity consumption, minimum electricity consumption, average electricity consumption, peak time of electricity consumption, frequency of electricity consumption, etc. The extracted electricity consumption data features are a multi-dimensional vector. For example, if 5 features are extracted, each data point for clustering can be represented as a 5-dimensional vector.

[0038] For K-means clustering, the number of clusters K can be determined in advance by the elbow rule. Then, K data points are randomly selected as the initial cluster centers, and each data point is assigned to the nearest cluster center to form K clusters. The distance calculation method can be Euclidean distance, Manhattan distance, etc. Once all data points are assigned to clusters, the cluster center of each cluster will be updated according to the feature vectors of all data points in the cluster, usually by calculating the mean vector. Repeat the above steps of assigning and updating cluster centers until the stopping condition is met, that is, clustering is completed. The stopping condition can be that the change of the cluster center is less than a certain threshold or reaches a preset number of iterations.

[0039] S105, identifying electricity theft behaviors using electricity consumption data of users belonging to the same electricity consumption type, and obtaining electricity theft users in each electricity consumption type.

[0040] Exemplarily, after the electricity usage data of multiple users are clustered, the multiple users are divided into different electricity usage types according to the clusters. Since the electricity usage data of users of the same electricity usage type are similar, the misjudgment that may occur when comparing different electricity usage types can be reduced. Therefore, the embodiment of the present invention regards the electricity usage data of users belonging to the same electricity usage type as a set, and determines the electricity theft users corresponding to the electricity usage type in the set.

[0041] A method for determining the electricity theft user corresponding to the electricity usage type in the set can be to input the electricity usage data of different electricity usage types into a deep learning model corresponding to the electricity usage type. The deep learning model can be a convolutional neural network (CNN) or a recurrent neural network (RNN). The deep learning model of each electricity usage type is pre-trained based on the electricity usage data of different electricity usage types. This embodiment does not limit the type and network structure of the deep learning model, and those skilled in the art can determine it as needed.

[0042] During the training process, we first need to establish a data set of electricity theft behaviors for each type of electricity user. This can be done by obtaining electricity consumption data for multiple days for each type of electricity user and processing missing values. Then, based on the audit results, we add a label of whether electricity theft occurs to each user of each type of electricity user. This forms a labeled electricity theft behavior data set for users who have committed electricity theft. Finally, we divide the electricity theft behavior data set into three parts: training set, test set, and validation set. We train with the training set, test with the test set, and validate with the validation set.

[0043] In addition to inputting the user data of different electricity types into the corresponding deep learning to obtain the electricity stealing users in each electricity type, the electricity consumption data of different electricity types can also be input into the local outlier factor model, and the outlier factor can be used as the electricity stealing user; the electricity stealing user can also be identified according to the multi-standard decision analysis method, by constructing an evaluation index system for electricity consumption behavior, converting the user's electricity consumption data into a feature vector, and then applying the TOPSIS method to evaluate and sort the users to find possible electricity stealing behaviors. This embodiment does not limit the method of determining the electricity stealing user corresponding to each electricity type in the electricity consumption data of the electricity type, and those skilled in the art can determine it as needed.

[0044] The embodiment of the present invention provides a method for identifying electricity theft users, which clusters the electricity usage data of multiple users to obtain the electricity usage type of each user, and then determines the electricity theft users in the electricity usage data of the same electricity usage type, thereby reducing the possible misjudgment when comparing different electricity usage types, and identifying abnormal behavior in similar electricity usage patterns is usually more accurate, because the data distribution is more compact, and abnormal behavior is more likely to stand out from background noise. At the same time, since the present invention can train a specific electricity theft detection model for the electricity usage pattern of the specific type for users of the same electricity usage type, which is simpler than training a general model, the model training can be simplified in this way, and since the electricity usage types are distinguished, each electricity usage type can be operated in parallel when determining the electricity theft users, thereby improving the efficiency of identifying the electricity theft users.

[0045] As an optional implementation, a method for identifying a power theft user further includes:

[0046] Determine the geographic coordinates of the electricity stealing users in each electricity usage type according to the smart meter identifiers carried in the electricity usage data of the electricity stealing users in each electricity usage type;

[0047] According to the geographic coordinates of electricity theft users in the same electricity consumption type, electricity theft users in the same electricity consumption type are divided into regions according to density, and the high incidence areas of electricity theft in each electricity consumption type are obtained;

[0048] Calculate the average amount of electricity stolen by electricity thieves in high-incidence areas of each type of electricity consumption, and rank the average amount of electricity stolen by electricity thieves in high-incidence areas of each type of electricity consumption;

[0049] According to the user's electricity usage type and the sorting result, the strategy mapping table is queried to output the anti-electricity theft strategy for the high-incidence areas of electricity theft for each electricity usage type.

[0050] Exemplarily, the smart meter identifier is a unique identification number of the smart meter, which is used to distinguish different smart meters. These identifiers can be in the form of serial numbers, meter numbers, bar codes, QR codes, or RFID tags. According to the smart meter identifier carried by the electricity consumption data of the electricity theft users in each electricity consumption type, the method of determining the geographical coordinates of the electricity theft users in each electricity consumption type can be to use the smart meter identifier, through database query, to associate the electricity theft users with the corresponding geographical locations, and the database records the installation address corresponding to the smart meter identifier, and the installation address is stored in the form of geographical coordinates.

[0051] According to the geographic coordinates of electricity theft users in the same electricity consumption type, the electricity theft users in the same electricity consumption type are divided into regions according to density, and the high incidence areas of electricity theft behavior in each electricity consumption type can be determined by the density-based spatial clustering of applications with noise (DBSCAN) method.

[0052] The specific steps include: first, collect the geographic coordinate data of electricity stealing users of the same type of electricity consumption. Secondly, determine the two key parameters of the DBSCAN algorithm: ε (Epsilon, neighborhood radius) and MinPts (the minimum number of neighbors of the core object). ε and MinPts can be set according to user needs or determined according to the guidance of domain experts. ε is the neighborhood radius, that is, the distance threshold used to measure whether two points are close enough. It can be set to be slightly larger than the average distance of the data points in the cluster. MinPts is the minimum number of data points required to form a cluster. It should be set large enough to avoid misidentifying noise as a cluster, but not too large to ignore small clusters. Then, according to the geographic coordinates, calculate the Euclidean distance between electricity stealing users or use other distance measurement methods according to actual needs. For each electricity stealing user, determine whether it is a core object. If a electricity stealing user has no less than MinPts neighbors within the ε distance, the user is considered to be a core object. Starting from each core object, find all its density-reachable neighbors to form a cluster. Density-reachable means that the number of core objects within the ε distance reaches MinPts. The clusters formed are the high-incidence areas of electricity theft for this type of electricity consumption.

[0053] Calculate the average amount of electricity stolen by electricity thieves in high-incidence areas of each type of electricity consumption, sort the average amount of electricity stolen by electricity thieves in high-incidence areas of each type of electricity consumption, query the strategy mapping table according to the user's electricity consumption type and the sorting result, and output the anti-electricity theft strategy for high-incidence areas of each type of electricity consumption.

[0054] Each type of electricity user has a different strategy for anti-electricity theft, and different severity levels of electricity theft have different warning and punishment strengths for anti-electricity theft. Therefore, this embodiment proposes to calculate the average amount of electricity theft in the high-incidence areas of electricity theft for each type of electricity user, sort the average amount of electricity theft in the high-incidence areas of electricity theft for each type of electricity user, query the strategy mapping table according to the user's electricity type and the sorting results, and output the anti-electricity theft strategy for the high-incidence areas of electricity theft for each type of electricity user. According to the user's electricity type and the sorting results, the strategy mapping table is queried, and the specific method of outputting the anti-electricity theft strategy for the high-incidence areas of electricity theft for each type of electricity user can be to pre-store multiple strategy tables, each strategy table corresponds to each type of electricity user, and each strategy table records the corresponding response strategy for the severity level of electricity theft for this type of electricity user. The severity level of electricity theft is determined by the average amount of electricity theft sorting structure. For example, the sorting position is the top 25% of the total sorting number, which corresponds to a high level of electricity theft. According to the user's electricity type, the strategy table corresponding to the electricity type is queried to obtain the anti-electricity theft strategy corresponding to the severity level of electricity theft.

[0055] The following are examples of anti-electricity theft strategies at different levels of severity for four types of electricity theft: commercial electricity theft, aquaculture electricity theft, industrial electricity theft, and residential electricity theft:

[0056] Under different levels of electricity theft severity, corresponding anti-electricity theft strategies can be formulated for commercial electricity, aquaculture electricity, industrial electricity and residential electricity. The following is an example of a strategy customized for electricity theft behaviors of different levels of electricity theft severity under the electricity usage type of aquaculture electricity:

[0057] If it is a low-level theft, the output anti-theft strategy is: educate farmers on electricity regulations, emphasize the importance of legal electricity use, regularly inspect farms, and check whether the electricity equipment is compliant. If it is a medium-level theft, the output anti-theft strategy is: use electricity consumption statistics to analyze, identify abnormal electricity usage patterns, and warn farmers. If it is a high-level theft, the output anti-theft strategy is: punish farmers who repeatedly or seriously steal electricity, strengthen supervision, use remote disconnect devices, and immediately cut off power to theft.

[0058] An embodiment of the present invention provides a method for identifying electricity theft users. By determining the average amount of electricity theft in high-incidence areas of electricity theft for each type of electricity usage, serious electricity theft areas can be determined, which is helpful for analyzing the spatial distribution characteristics of electricity theft behavior and facilitating unified management. In addition, according to the user's electricity usage type and the sorting result, a strategy mapping table is queried to output the corresponding anti-electricity theft strategy, thereby realizing the automation of strategy provision.

[0059] As an optional implementation, electricity theft behavior is identified by using electricity consumption data of users belonging to the same electricity consumption type, and electricity theft users in each electricity consumption type are obtained respectively, including:

[0060] The electricity consumption data features of users belonging to the same electricity consumption type are extracted, and the data features include downward trend features, change features, and fluctuation features; the electricity consumption data features of users belonging to the same electricity consumption type are reduced in dimension according to the principal component analysis method; the electricity consumption data features of users belonging to the same electricity consumption type after dimension reduction are input into the local outlier factor model to calculate the local reachable density of each user's electricity consumption data; the local outlier factor score of each user is determined according to the local reachable density of each user's electricity consumption data; when the local outlier factor score of any user exceeds the preset score threshold, the user is identified as an electricity theft user of this electricity consumption type.

[0061] Exemplarily, the local outlier factor algorithm is a density-based local outlier algorithm, which does not need to consider the global distribution of data, but only focuses on the distribution of data points around the data object to achieve outlier detection of data. It is suitable for data with different density distributions.

[0062] The method of extracting the downward trend characteristics of the electricity consumption data of users of the same electricity consumption type is as follows: if the electricity consumption data of user n is q n = {q n 1 ,qn 2 ,...,q n h}, the trend factor of user n at time t is the average moving average of k items before time t Compare sequence q n and the relative size of F at each time point, if q n There are v segments below F, each containing points b1, b2, ..., b v , then the calculation formula of the downward trend feature V1 is:

[0063]

[0064] The variation characteristics of power data can be measured by the difference between the beginning and the end of the user's power data. The absolute deviation V2 of the average load of the previous cycle (7 days) and the next cycle (7 days) on the same day is calculated by the user, and the ratio V3 of the average load of the next cycle (7 days) to the average load of the entire cycle (15 days) is expressed as:

[0065]

[0066] Wherein, r=7, and t represents the statistical day.

[0067] The fluctuation characteristics can be measured by the standard deviation of the power series in the statistical period. This article uses the standard deviation V4 of the power load series in the entire period (15 days), the standard deviation V5 of the power load series in the previous period (7 days), and the standard deviation V6 of the power load series in the next period (7 days) to express:

[0068] V4=sd(q n (t-r) ,q n (t-(r-1)) ,...,q n (t+(r-1)) ,q n (t+r) )

[0069] V5=sd(q n (t-r) ,q n (t-(r-1)) ,...,q n (t-1) )

[0070] V6=sd(q n (t+1) ,q n (t+2) ,...,q n (t+r) )

[0071] Wherein, r=7, and t represents the statistical day.

[0072] The electricity consumption data of 263 breeding users from April 2022 to June 2022 were collected in a high-loss area of ​​a certain power grid station. n ,n=1,2,...,263},q n = {q n (h) ,h=1,2,...,91} as an example, six features, including trend feature V1, change feature V2, change feature V3, fluctuation feature V4, fluctuation feature V5 and fluctuation feature V6, are extracted from these electricity consumption data sets Q. The results are shown in Table 1.

[0073] Table 1 Extracting user power characteristics

[0074]

[0075] The six features extracted from the electricity data are not independent of each other. Some are correlated and have high dimensions, making them difficult to identify. In order to reduce the correlation and identification difficulty of the features, the principal component analysis method can be used to reduce the dimension of the electricity features. The factor loading matrix obtained is shown in Table 2.

[0076] Table 2 Factor loading matrix

[0077]

[0078] According to the cumulative variance contribution rate of 81%, the principal components are extracted to obtain two principal components, PC1 and PC2. These two principal components are used as new power characteristics. The calculation relationship between the principal component and the power characteristics is:

[0079] PC1=-0.058×V1+0.512×V2+0.363×V3+0.494×V4+0.421×V5+0.424×V6

[0080] PC2=-0.868×V1-0.111×V2+0.439×V3-0.151×V4-0.105×V5-0.082×V6

[0081] According to the calculation formula of the principal component, the principal component values ​​of the user's electricity consumption are obtained as shown in Table 3.

[0082] Table 3 Main components of user power consumption

[0083]

[0084] The electricity consumption curve of electricity thieves is generally quite different from that of normal electricity users, and there are generally fewer electricity thieves. Therefore, electricity thieves should be regarded as outliers, and the local outlier factor algorithm is used to calculate the local outlier factor score. If the ratio of the average density of points around data point P to the density of point P is bounded by 1, the greater the ratio is, the greater the density of points around point P is than the density of this point, and point P is more likely to be an outlier; the smaller the ratio is, the smaller the density of points around point P is than the density of this point, and point P is more likely to be a normal point.

[0085] Assume that P and O are points in n-dimensional space, and use d(P,O) to represent the distance between point P and point O. k (P) represents the kth distance of P. This distance refers to the distance from point P to other points, which are sorted from small to large. The distance from the kth point to P is called the kth distance of P. With P as the center, d k (P) is the radius of the neighborhood where at least k points are less than or equal to d k (P); at the same time, there are at most k-1 points, not including P, whose distance to P is less than d k (P); Taking two-dimensional space as an example, the fifth distance d5(P) of point P is to radiate outward from P until it covers the fifth neighboring point, such as Figure 2 shown.

[0086] With P as the center, d k (P) is the radius of the neighborhood including the distance equal to d k The point set consisting of the points on (P) is called the kth neighborhood of point P, denoted by N k (p), the number of points contained in the neighborhood is denoted by |N k (P)|, obviously there is|N k (p)|≥k holds;

[0087] The kth reachable distance from point O to point P is reachdist k (P,O) is defined as the kth distance d from point O k (O) and the distance d(P,O) from point P to point O, whichever is greater, that is: reachdist k (P,O)=max{d k (O),d(P,O)}

[0088] Figure 3 As shown, the 5th reachable distance from point o1 to point p is d(p,o1), and the 5th reachable distance from o2 to p is d5(o2).

[0089] The local reachability density lrd of point P k (P) is defined as the reciprocal of the average k-th reachable distance from all points in the k-th distance neighborhood of point P to point P, that is:

[0090]

[0091] From the above formula, we can see that the greater the density of point P, the more likely it is that point P belongs to the same cluster as the surrounding neighboring points. The lower the density, the more likely P is an outlier. In fact, if P and the surrounding neighboring points are in the same cluster, then the reachable distance reachdist k (P,O) is more likely to be smaller d k (P), the sum of the reachable distances is small and the density is high; if P is far away from the surrounding neighborhood points, the reachable distance may take a larger value d(p,o), the sum of the reachable distances is large, the density is low, and it is more likely to be an outlier.

[0092] Local Outlier Factor Score LOF k (P) is defined as the local outlier factor LOF of point P k (P) is essentially the k-neighborhood N of the data point P. k The ratio of the average local reachability density of all points in (P) to the local reachability density of point P, that is:

[0093]

[0094] From the above formula, we can see that: LOF k (P) is greater than 1, indicating that the density around data point P is greater than the density of point P, and the point is more likely to be an outlier; LOF k The smaller (P) is than 1, the smaller the density around data point P is, the more likely the point is a normal point. See the schematic diagram of outlier factor score for details. Figure 4 .

[0095] The values ​​of the two principal components in Table 3 are used to calculate the local outlier factor score of each aquaculture electricity customer by defining the number of points k included in the user neighborhood as 30, and Table 4 is obtained.

[0096] Table 4 Identification results of the local outlier factor algorithm model for electricity theft

[0097]

[0098] Make a scatter plot of the local outlier factor scores of 263 users, such as Figure 5 shown.

[0099] Figure 5The 12 extremely large LOF scores are shown, which correspond to the 12 users with a high suspicion of electricity theft. From the distribution of the scatter plot, it can be seen that these 12 users are mainly distributed in two intervals. Therefore, these 12 suspected electricity theft users can be divided into two levels of electricity theft suspicion. The highest suspicion level is indicated by "**", and the lower suspicion level is indicated by "*". They are marked in the last column of Table 8. The rest are normal electricity users.

[0100] An electricity theft user identification method provided by an embodiment of the present invention adopts a local outlier factor model to determine electricity theft users in various electricity usage types, and can identify users who show abnormal electricity usage patterns, which is helpful to timely discover and prevent electricity theft behaviors. The local outlier factor model has good adaptability to different electricity usage types and data distributions, which improves the versatility of the model, and performs dimensionality reduction processing in data processing, which reduces misjudgments caused by data noise or irrelevant features, and the data processing speed after dimensionality reduction is faster, which helps to improve the efficiency of electricity theft detection.

[0101] As an optional implementation, when the local outlier factor score of any user exceeds a preset score threshold, the user is identified as an electricity theft user of the electricity usage type, including:

[0102] When the local outlier factor score of any user exceeds the preset score threshold, the user is identified as the first suspected electricity theft user of the electricity usage type; the entropy weight method is used to weight the electricity theft behavior identification index set according to the electricity usage data of users of the same electricity usage type, and the observation weighted matrix of the electricity theft behavior identification index is obtained, wherein the electricity theft behavior identification index includes a current abnormality index, a voltage undervoltage index, and a power factor abnormality index; the positive ideal solution and the negative ideal solution of the user's electricity usage data under the electricity theft behavior are determined according to the observation weighted matrix of the electricity theft behavior identification index; the Euclidean distance between the electricity usage data of each user of the same electricity usage type and the positive ideal solution and the negative ideal solution is calculated; according to the Euclidean distance between the electricity usage data of each user and the positive ideal solution and the negative ideal solution, the closeness value of the electricity usage data of each user to the positive ideal solution is determined; the closeness value of the electricity usage data of each user to the positive ideal solution is sorted, and the user corresponding to the closeness value of the target proportion is identified as the second suspected electricity theft user; the first suspected electricity theft user and the second suspected electricity theft user of the same electricity usage type are intersected to obtain the electricity theft user.

[0103] For example, users who use electricity normally usually have relatively stable voltage, current changes that can basically maintain three-phase balance, power factor that does not change suddenly within a certain range, power fluctuation trends that remain within a certain range, and line loss changes that do not change too much in the substation area. When a user is suspected of stealing electricity, the power data will show abnormal changes: current abnormalities are manifested as undercurrent and three-phase current imbalance. Undercurrent means "insufficient current". When the current of any phase of the energy meter is lower than 0.5% of the rated current, and at least one of the other two phases has a current of not less than 10% of the rated current, it is considered to be undercurrent, and the current undercurrent indicator is marked as current abnormality indicator X1; three-phase current imbalance means that the active power and reactive power of each phase load are not completely equivalent, and the phase sum of the three-phase current is not 0. For the low-voltage distribution network, when the three-phase voltage and the three-phase load are symmetrical at the same time, the three-phase current should be symmetrical. If the three-phase load is unbalanced, it will cause the line power loss to increase. The current three-phase imbalance indicator is marked as current abnormality indicator X2; "undervoltage" means "insufficient voltage". Although electrical equipment has a rated voltage, it is difficult to ensure that it works constantly at the rated voltage during actual operation. As long as the equipment is at the rated If the voltage is within the fluctuation range of a certain voltage, it can be considered normal, and when the voltage is lower than this range, it is called "undervoltage". If the voltage is lower than 60% of the rated voltage of the energy meter and the current is greater than 15% of the rated current of the energy meter, it is considered as undervoltage, and the voltage undervoltage index is marked as X3; abnormal power factor means that its value is low, because the power factor is an important coefficient of the efficiency of electrical equipment. The lower the power factor, the higher the reactive power used by the circuit in the conversion of the alternating magnetic field, and the greater the line power supply loss. Relative to the AC circuit, the cosine value of the phase difference between voltage and current is called the power factor. The "Power Factor Adjustment Electricity Fee Method" stipulates that the power factor of agricultural users must be higher than 0.8. Combined with the experience of power grid experts, a power factor less than 0.6 is considered to be abnormal, and the power factor index abnormality is recorded as X4. Therefore, the indicator set for electricity theft identification is:

[0104] I={current abnormality index X1; current abnormality index X2; voltage undervoltage index X3; power factor abnormality index X4}

[0105] The entropy weight method is an objective weighting method. The idea of ​​this method is to use information entropy to measure the degree of dispersion of the indicator and determine the weight of the indicator in the comprehensive evaluation system by the degree of dispersion. The smaller the information entropy value of the indicator, the greater its degree of dispersion and the greater the amount of information it contains. It is more significant for the comprehensive evaluation system, and thus the greater the weight of this indicator, and vice versa.

[0106] Let X ij (1≤i≤n j ,1≤j≤4) represents the observed value of the jth indicator in the ith sample, n jIndicates the sample size for obtaining the j index, represents the minimum value of the j-th index, represents the maximum value of the j-th index sample observation, z ij Represents the standardized value of the i-th sample of the j-th index.

[0107]

[0108] Use p ij represents the probability value of the i-th sample of the j-th index, which needs to meet the two conditions of non-negative and summing to 1, p ij The following calculation formula can be used:

[0109]

[0110] The information entropy e of the jth index j Defined as:

[0111]

[0112] The weight w of indicator j j for:

[0113]

[0114] The weights assigned to the indicators in the electricity theft characteristic indicator set according to the above entropy weighting method are shown in Table 5.

[0115] Table 5 Entropy weights of indicators in the electricity theft characteristic indicator set

[0116]

[0117] The TOPSIS method is a multi-objective decision-making method based on the distance between good and bad solutions. It introduces the concepts of "positive ideal solution" and "negative ideal solution". The sample observation value of each indicator has a maximum and a minimum. The maximum is called the positive ideal solution, and the minimum is called the negative ideal solution. A user is actually a sample point. Each user has 4 characteristic indicators. Each indicator has a sample observation sequence. The maximum value of each sample sequence constitutes a positive ideal solution vector. The minimum value of the sample sequence constitutes a negative ideal solution vector Sample x i =(x i1 ,x i2 ,x i3 ,x i4 )The closer it is to the positive ideal solution, the greater the suspicion. Since each indicator is given a certain weight, it is necessary to consider the weight given by the entropy weight method. The specific calculation steps are as follows

[19]

[20]

[22] :

[0118] Step 1: Remember yij is the standardized data of user i at the jth index, and the observation sample matrix of the user's four characteristic indicators is obtained:

[0119]

[0120] Step 2: The entropy weight method is used to assign weights to the four characteristic indicators;

[0121] W=(w1,w2,w3,w4)=(0.4410.0160.3320.211);

[0122] Step 3: Construct the sample observation weight matrix Z = (r ij ) n×4 :

[0123]

[0124] According to the data of four characteristic indicators of electricity users in the breeding industry under the jurisdiction of a certain power grid company and the weights assigned by the entropy weight method, the sample observation weighted matrix in Table 6 is obtained.

[0125] Table 6 Weighting matrix of characteristic indicators

[0126]

[0127] Step 4: Determine the positive ideal solution based on the weighted matrix and negative ideal solution

[0128]

[0129] According to the sample data in Table 2, we can get:

[0130] S + =(0.239,0.298,0.255,0.207)

[0131] S - =(0.024,0.030,0.026,0.021)

[0132] Step 5: Calculate the sample value and and Euclidean distance and

[0133]

[0134] Step 6: Calculate the closeness P between the sample value of the i-th user and the positive ideal solution i :

[0135]

[0136] The posting progress of each farming user calculated according to the weighted matrix in Table 6 is shown in Table 7.

[0137] Table 7 The closeness P of user characteristic index samples to electricity theft t

[0138]

[0139] According to the closeness P of the abnormal power consumption index sample in Table 7 t The size of each user is sorted, and the closeness P t The larger the value is, the more suspicious the user is of electricity theft. The top 3% of users in terms of proximity are selected as suspected electricity theft users as the second suspicious electricity theft users, and Table 8 is obtained:

[0140] Table 8 TOPSIS model electricity theft identification results

[0141]

[0142] “**” indicates that there is a strong suspicion of electricity theft, and “*” indicates that there is a suspicion of electricity theft.

[0143] In the above embodiment, the user whose local outlier factor score exceeds the preset score threshold is identified as the first suspicious electricity theft user of this electricity usage type, corresponding to the user with major suspicion of electricity theft in Table 4, and the two identification results are displayed in a table, as shown in Table 9.

[0144] Table 9 Recognition results of two models

[0145]

[0146] In Table 9, the intersection of the first suspicious electricity theft user and the second suspicious electricity theft user is taken as the final electricity theft user, as shown in Table 10.

[0147] Table 10 Final recognition results of the two models

[0148]

[0149] An embodiment of the present invention provides a method for identifying electricity theft users, which assigns weights through the entropy weight method. As an objective weight allocation method, the entropy weight method can determine the weights of various indicators according to the characteristics of the data itself, reducing the influence of subjective judgment. In addition, this embodiment takes the intersection of the two methods to integrate the recognition results of the two methods. The two methods detect electricity theft behavior from different angles, respectively, thereby improving the accuracy of recognition.

[0150] As an optional implementation, according to the geographic coordinates of electricity theft users in the same electricity usage type, electricity theft users in the same electricity usage type are divided into regions according to density, and high-incidence areas of electricity theft behavior in each electricity usage type are obtained, including:

[0151] According to the geographical coordinates of electricity stealing users of the same electricity type, determine the distance between any user and adjacent users;

[0152] Determine an average distance value based on the distance value between any user and the adjacent user;

[0153] When the average distance value of any user is less than the preset average distance value, the user is regarded as the central user of the current electricity consumption type. When there are overlapping adjacent users among multiple central users, each central user and the corresponding adjacent user are included in the same area;

[0154] Based on the geographic coordinates of users in the same area, determine the areas with high incidence of electricity theft for the current electricity usage type.

[0155] For example, the determination method of adjacent users may be: when the line connecting user A and user B is used as the axis and user A is used as the origin, and there is no intermediate user within the preset angle range, so that user A and user B can be connected by transit, then user B is defined as an adjacent user of user A. Figure 6 Point A is used as an example for user A. Assuming that the preset angle range is 30 degrees, take the user at point D as an example. Point A is connected to point D, and the AD line is used as the axis and point A is used as the origin. A 30-degree angle is divided on both sides of the axis. See the dotted parts on both sides of the AD line. There is an angle formed by two dotted parts. Within the angle range, there is no user that is closer to point A. Therefore, the user at point D is the neighboring user of the user at point A. Similarly, the line between the user at point G and the user at point A has a closer point F within the preset angle range, and point G can be reached via the line at point F. Therefore, point G is not a neighboring user of the user at point A.

[0156] According to the above definition, users at points B, C, D, E, and F are all neighbors of user A. Based on the coordinate positions of users at points B, C, D, E, and F and the coordinate positions of user A, the distance values ​​between users at points B, C, D, E, and F and user A are calculated. The distance value is calculated as follows:

[0157]

[0158] Among them, x i represents the horizontal coordinate position of user i, x a Indicates the horizontal coordinate position of user A, y i Indicates the vertical coordinate position of user i, y a Indicates the vertical coordinate position of user A.

[0159] According to the user's geographic coordinates and the average distance value to the adjacent users, the method of determining the high-incidence area of ​​electricity theft for the current type of electricity consumption can be that when the average distance value to the adjacent users is less than the preset threshold, it means that the electricity theft users near the user's geographic coordinates are relatively dense. The size of the preset threshold is determined according to the actual distribution of the location of the electricity theft users. After finding all users whose average distance values ​​to the adjacent users are less than the preset threshold, all users whose average distance values ​​to the adjacent users are less than the preset threshold are taken as the central users, and it is determined whether the adjacent users of each central user overlap. If there is overlap, the two central users and the corresponding adjacent users are included in the same high-incidence area of ​​electricity theft, such as Figure 6 As shown in the figure, it is assumed that point A and point E, point L and point H are all central users, point E is a neighboring user of point A, and the neighboring users of point E include point M and point N in addition to the central user of point A. The neighboring users of the central user of point L are point F and point user, point F is also a neighboring user of the central user of point A, point M is also a neighboring user of the central user of point E, points B, G, K, I, J are neighboring users of the central user of point H, and point B is also a neighboring user of the central user of point A. It can be seen that there are overlapping neighboring users between point A and point E, point L and point H central users. Therefore, point A and point E, point L and point H central users and their neighboring users are all included in the same high-incidence area of ​​electricity theft. If the neighboring users of the central user do not overlap, each central user and its corresponding neighboring users are regarded as independent high-incidence areas of electricity theft.

[0160] The geographic coordinates of the area with high incidence of electricity theft are determined by the area surrounded by the above multiple points. Specifically, it is necessary to combine the user geographic coordinates corresponding to each node to determine which users are edge users, connect the edge users to form a closed area, and the closed area is the area with high incidence of electricity theft, such as Figure 6 As shown, the high-incidence area of ​​electricity theft is a closed area surrounded by straight lines between the NMLFGKJIBCDN nodes.

[0161] The embodiment of the present invention provides a method for identifying electricity theft users. The method determines the high-incidence area of ​​electricity theft, does not require complex clustering algorithm clustering, reduces the amount of calculation, and saves computing resources.

[0162] As an optional implementation, a method for identifying electricity theft users also includes: receiving electricity usage data of high-incidence areas of electricity theft for each type of electricity usage within a target time period, and evaluating the effectiveness of strategy implementation based on the electricity usage data; when the strategy implementation effect does not achieve the expected implementation effect, sending the corresponding information of high-incidence areas of electricity theft for each type of electricity usage to the target client.

[0163] Exemplarily, the method of evaluating the implementation effect of the strategy based on the power consumption data may be to construct a prediction model to predict the power consumption after the implementation of the strategy, compare the prediction results with the actual data, and evaluate the accuracy and effectiveness of the strategy, or to identify the power theft users again based on the power consumption data of the high-incidence areas of power theft of each type of power consumption in the target time period, and determine the overlap rate between the power theft users after the implementation of the strategy and the power theft users before the implementation of the strategy. This embodiment does not limit the method of evaluating the implementation effect of the strategy based on the power consumption data, and those skilled in the art can determine it as needed.

[0164] The target client can be a mobile terminal or PC terminal of relevant personnel of the power bureau, or a mobile terminal or PC terminal of relevant personnel in charge of areas with high incidence of electricity theft of this type of electricity consumption, so as to prompt relevant personnel and enable them to respond in time.

[0165] The present invention provides a device for identifying electricity stealing users, comprising: an electricity usage data acquisition module, used for acquiring electricity usage data of multiple users; an electricity usage type determination module, used for clustering the electricity usage data of multiple users to obtain the electricity usage type of each user; and an electricity stealing user determination module, used for identifying electricity stealing behaviors based on the electricity usage data of users belonging to the same electricity usage type, and obtaining electricity stealing users in each electricity usage type respectively.

[0166] The present application also provides an electronic device, such as Figure 7 As shown, a processor 501 and a memory 502, wherein the processor 501 and the memory 502 may be connected via a bus or other means.

[0167] The processor 501 may be a central processing unit (CPU). The processor 501 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or a combination of the above chips.

[0168] The memory 502 is a non-transitory computer-readable storage medium that can be used to store non-transitory software programs, non-transitory computer executable programs and modules, such as program instructions / modules corresponding to the method for identifying and accessing electricity theft users in the embodiment of the present invention. The processor executes various functional applications and data processing of the processor by running the non-transitory software programs, instructions and modules stored in the memory.

[0169] The memory 502 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created by the processor, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory 502 may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0170] The one or more modules are stored in the memory 502, and when executed by the processor 501, the following is performed: Figure 1 The method for identifying electricity stealers in the illustrated embodiment.

[0171] For details of the above drone equipment, please refer to Figure 1 The corresponding related descriptions and effects in the illustrated embodiments can be understood and will not be repeated here.

[0172] This embodiment also provides a computer storage medium, which stores computer executable instructions, and the computer executable instructions can execute the electricity theft user identification method in any of the above method embodiments. The storage medium can be a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (Flash Memory), a hard disk (HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above types of memory.

[0173] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present invention.

Claims

1. A method for identifying electricity theft users, characterized in that: include: Obtain electricity consumption data of multiple users; Clustering the electricity consumption data of multiple users to obtain the electricity consumption type of each user; The electricity consumption data of users belonging to the same electricity consumption type are used to identify electricity theft behaviors, and the electricity theft users in each electricity consumption type are obtained respectively.

2. A method for identifying electricity theft users according to claim 1, characterized in that: Also includes: Determine the geographic coordinates of the electricity stealing users in each electricity usage type according to the smart meter identifiers in the electricity usage data of the electricity stealing users in each electricity usage type; According to the geographic coordinates of electricity theft users in the same electricity consumption type, electricity theft users in the same electricity consumption type are divided into regions according to density, and the high incidence areas of electricity theft in each electricity consumption type are obtained; Calculate the average amount of electricity stolen by electricity thieves in high-incidence areas of each type of electricity consumption, and rank the average amount of electricity stolen by electricity thieves in high-incidence areas of each type of electricity consumption; According to the user's electricity usage type and the sorting result, the strategy mapping table is queried to output the anti-electricity theft strategy for the high-incidence areas of electricity theft for each electricity usage type.

3. A method for identifying electricity theft users according to claim 1, characterized in that: The electricity consumption data of users belonging to the same electricity consumption type are used to identify electricity theft behaviors, and the electricity theft users in each electricity consumption type are obtained, including: Extract the electricity consumption data features of users belonging to the same electricity consumption type, including downward trend features, change features, and fluctuation features; The characteristics of electricity consumption data of users belonging to the same electricity consumption type are reduced in dimension according to the principal component analysis method; The reduced-dimensional power consumption data features of users of the same power consumption type are input into the local outlier factor model to calculate the local reachable density of each user's power consumption data; Determine the local outlier factor score of each user according to the local reachability density of each user's electricity consumption data; When the local outlier factor score of any user exceeds the preset score threshold, the user is identified as an electricity theft user of this electricity usage type.

4. A method for identifying electricity theft users according to claim 3, characterized in that: When the local outlier factor score of any user exceeds the preset score threshold, the user is identified as an electricity theft user of the electricity usage type, including: When the local outlier factor score of any user exceeds the preset score threshold, the user is identified as the first suspected electricity theft user of the electricity usage type; The entropy weight method is used to weight the electricity theft behavior identification index set according to the electricity consumption data of users of the same electricity consumption type, and the observation weighted matrix of the electricity theft behavior identification index is obtained, where the electricity theft behavior identification index includes the current abnormality index, the voltage undervoltage index, and the power factor abnormality index; According to the observation weighted matrix of the electricity theft behavior identification index, the positive ideal solution and the negative ideal solution of the user's electricity consumption data under the electricity theft behavior are determined; Calculate the Euclidean distance between the electricity consumption data of each user of the same electricity consumption type and the positive ideal solution and the negative ideal solution; According to the Euclidean distance between each user's power consumption data and the positive ideal solution and the negative ideal solution, the closeness value of each user's power consumption data to the positive ideal solution is determined; The closeness values ​​of the electricity consumption data of each user to the positive ideal solution are sorted, and the user corresponding to the closeness value of the target proportion is identified as the second suspected electricity theft user; The first suspicious electricity theft user and the second suspicious electricity theft user of the same electricity usage type are intersected to obtain the electricity theft user.

5. A method for identifying electricity theft users according to claim 1, characterized in that: Cluster the electricity consumption data of multiple users to obtain the electricity consumption type of each user, including: Extracting power consumption characteristic data from power consumption data of multiple users; The elbow rule is used for the electricity consumption characteristic data to determine the number of cluster centers; Calculate the distance between the power consumption characteristic data in the power consumption data of multiple users and each cluster center; The electricity consumption data of multiple users are divided into cluster centers with the smallest distance values, and the types represented by the divided cluster centers are used as the electricity consumption types of the users.

6. A method for identifying electricity theft users according to claim 2, characterized in that: According to the geographic coordinates of electricity theft users in the same electricity consumption type, electricity theft users in the same electricity consumption type are divided into regions according to density, and the high incidence areas of electricity theft in each electricity consumption type are obtained, including: According to the geographical coordinates of electricity stealing users of the same electricity type, determine the distance between any user and adjacent users; Determine an average distance value based on the distance value between any user and the adjacent user; When the average distance value of any user is less than the preset average distance value, the user is regarded as the central user of the current electricity consumption type. When there are overlapping adjacent users among multiple central users, each central user and the corresponding adjacent user are included in the same area; Based on the geographic coordinates of users in the same area, determine the areas with high incidence of electricity theft for the current electricity usage type.

7. A method for identifying electricity theft users according to any one of claims 2 to 4, characterized in that: Also includes: Receive electricity consumption data from high-incidence areas of electricity theft for each type of electricity use during the target time period, and evaluate the effectiveness of strategy implementation based on the electricity consumption data; When the policy implementation effect does not achieve the expected implementation effect, the information of high-incidence areas of electricity theft of the corresponding type of electricity consumption will be sent to the target client.

8. A device for identifying electricity theft users, characterized in that: include: An electricity consumption data acquisition module, used to acquire electricity consumption data of multiple users; An electricity usage type determination module is used to cluster the electricity usage data of multiple users to obtain the electricity usage type of each user; The electricity stealing user identification module is used to identify electricity stealing behaviors based on the electricity usage data of users belonging to the same electricity usage type, and obtain electricity stealing users in each electricity usage type.

9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the steps of the method for identifying electricity theft users as described in any one of claims 1 to 7.

10. A computer storage medium having computer instructions stored thereon, characterized in that: When the instruction is executed by the processor, the steps of a method for identifying an electricity stealing user described in any one of claims 1 to 7 are implemented.

Images