Nothing Special   »   [go: up one dir, main page]

WO2023109527A1 - 窃电行为的检测方法、装置、计算机设备和存储介质 - Google Patents

窃电行为的检测方法、装置、计算机设备和存储介质 Download PDF

Info

Publication number
WO2023109527A1
WO2023109527A1 PCT/CN2022/136079 CN2022136079W WO2023109527A1 WO 2023109527 A1 WO2023109527 A1 WO 2023109527A1 CN 2022136079 W CN2022136079 W CN 2022136079W WO 2023109527 A1 WO2023109527 A1 WO 2023109527A1
Authority
WO
WIPO (PCT)
Prior art keywords
electricity
sample
target
current
user cluster
Prior art date
Application number
PCT/CN2022/136079
Other languages
English (en)
French (fr)
Inventor
钟健强
余沛谦
李伟
温硕伟
黄儒
罗子勉
Original Assignee
广东电网有限责任公司东莞供电局
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 广东电网有限责任公司东莞供电局 filed Critical 广东电网有限责任公司东莞供电局
Publication of WO2023109527A1 publication Critical patent/WO2023109527A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R11/00—Electromechanical arrangements for measuring time integral of electric power or current, e.g. of consumption
    • G01R11/02—Constructional details
    • G01R11/24—Arrangements for avoiding or indicating fraudulent use
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere

Definitions

  • the present application relates to the technical field of electric power, for example, to a detection method, device, computer equipment, and storage medium for stealing electricity.
  • the method of monitoring electricity theft is mainly to consider the electricity consumption of electricity users, and analyze the electricity consumption of electricity users through self-defined rules or time series models, so as to detect whether electricity theft occurs.
  • Electricity consumption is the result of electricity consumption by electricity users, and there is a lot of information lost.
  • the same electricity consumption may be consumed by an air conditioner, or it may be consumed by a rice cooker and a TV, or it may be an electric heater. Consumed by appliances and hair dryers, this will result in lower accuracy in detecting electricity theft.
  • the present application proposes a method, device, computer equipment and storage medium for detecting electricity theft, in order to solve the problem of low accuracy in detecting electricity theft by using the electricity consumption of electricity consumers.
  • the embodiment of the present application provides a method for detecting electricity stealing behavior, including:
  • the sample electricity users with the same electricity stealing type in the electricity stealing behavior are divided into the same sample user cluster;
  • K-means clustering is performed on the target electricity users and the sample electricity users according to the target current characteristics and the sample current characteristics to obtain K target user clusters;
  • the target electricity user who has committed electricity stealing behavior is detected in the target user cluster according to the sample electricity user.
  • the embodiment of the present application also provides a detection device for stealing electricity, including:
  • the inrush current receiving module is configured to receive the inrush current detected by a plurality of electric meters installed in the places of the target electric users;
  • a current feature extraction module configured to extract the original current feature from the surge current
  • the current feature dimension reduction module is configured to perform principal component analysis on the original current feature, and take dimensionality reduction as the target current feature;
  • the current feature acquisition module is configured to acquire a sample current feature in the same format as the target current feature, and the sample current feature is associated with a sample electricity user who has stolen electricity;
  • the sample user cluster clustering module is configured to divide the sample electricity users with the same stealing type in the electricity stealing behavior into the same sample user cluster according to the sample current characteristics;
  • the clustering parameter setting module is configured to set multiple K values
  • the target user clustering module is set to perform K-means clustering on the target electricity users and the sample electricity users according to the target current characteristics and the sample current characteristics for each K value, to obtain K target user clusters;
  • a similarity calculation module configured to calculate the similarity between the sample user cluster and the target user cluster according to the distribution information of the sample electricity users
  • the validity detection module is configured to detect the validity of the target user cluster according to the similarity and the distribution information of the target power users;
  • the electricity stealing behavior detection module is configured to detect, in the target user cluster, the target electricity user who has occurred the electricity stealing behavior according to the sample electricity user based on the effective detection result of the target user cluster.
  • the embodiment of the present application also provides a computer device, the computer device comprising:
  • the processor When the program is executed by the processor, the processor implements the method for detecting electricity stealing behavior as described in the first aspect.
  • the embodiment of the present application also provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the electricity stealing described in the first aspect is implemented. Behavioral detection method.
  • FIG. 1 is a flow chart of a method for detecting electricity stealing behavior provided in Embodiment 1 of the present application;
  • FIG. 2 is a schematic structural diagram of a detection device for stealing electricity provided in Embodiment 2 of the present application;
  • FIG. 3 is a schematic structural diagram of a computer device provided in Embodiment 3 of the present application.
  • FIG. 1 is a flow chart of a method for detecting electricity theft provided by Embodiment 1 of the present application.
  • This embodiment is applicable to situations where surge currents are used to detect electricity theft, and the method can be implemented by a detection device for electricity theft.
  • the detection device of the electricity stealing behavior can be implemented by software and/or hardware, and can be configured in a computer device as a server, for example, a server, a workstation, a personal computer, etc., the method includes the following steps:
  • Step 101 Receive surge currents detected by multiple ammeters.
  • the electric meter is installed in the place of the target electricity user, such as a residence, an apartment, etc., and the electric meter is usually equipped with a surge current protector, and elements can be added to the protector for detecting Inrush current.
  • the surge current is also called the switch-on surge or input surge current, which refers to the instantaneous high input current consumed by the power supply or electrical equipment when it is switched on, that is, the current flowing into the power supply equipment at the moment the power supply or electrical equipment is switched on. Peak current due to the high initial current required to charge capacitors and inductors or transformers.
  • discharging capacitors in the power supply provide low impedance, allowing high currents to flow into the circuit as they charge from zero to maximum. These currents can be as high as 20 times the steady state current. Even though it only lasts about 10 milliseconds, it still takes 30 to 40 cycles for the current to stabilize to its normal operating value. If not limited, high currents can damage equipment and cause other equipment powered by the same power source to fail, in addition to creating voltage dips on the power line.
  • the meter continuously detects the inrush current generated by each target user when using electrical equipment, and transmits the information of the inrush circuit to the server through the Internet of Things and other means.
  • Step 102 extracting original current features from the surge current.
  • a feature used to distinguish normal power consumption behavior from power stealing behavior can be extracted from the surge current, which is recorded as the original current feature.
  • the first time period (such as 1 day) may be divided into multiple second time periods (such as 5 minutes).
  • a flag is generated in each second time period, which indicates whether the inrush current is received, for example, if the inrush current is received in the second time period, the flag is 1, and the inrush current is not received in the second time period , it is marked as 0.
  • the frequency of receiving the surge current within the second time period is counted.
  • the maximum value, minimum value, and average value of the surge current within the second time period are respectively counted.
  • the frequency of receiving surge current within the second time period is set to a specified value, such as 0.
  • All the original character strings in the first time period are spliced according to the order of the second time period to form the original current feature.
  • the above original current characteristics are only examples.
  • other original current characteristics can be set according to the actual situation.
  • the total current value is obtained by accumulating the surge current in the second time period, and The inrush current calculation variance in the period, etc.
  • those skilled in the art may also adopt other original current characteristics according to actual needs.
  • Step 103 performing principal component analysis on the original current features, and taking dimensionality reduction as the target current features.
  • the original current features are expressed in the form of vectors. Since the original current features have many dimensions, the dimension of the original current features can be reduced by PCA (Principal Component Analysis) algorithm while maintaining the main feature components of the original current features. , which is recorded as the target current feature, so that the dimension of the target current feature is reduced to the order of one or ten, thereby reducing the amount of computation.
  • PCA Principal Component Analysis
  • PCA transforms the original data (original current features) into a set of linearly independent representations of each dimension (target current features) through linear transformation.
  • the original current features may be combined into a matrix to obtain a first current matrix X with n rows and m columns.
  • the covariance matrix C is calculated for the first current matrix X, Among them, X T is the transpose matrix of X.
  • Step 104 Obtain a sample current signature whose format is the same as that of the target current signature.
  • other means can be used to pre-screen the sample electricity users who have stolen electricity, and set the sample current characteristics for the sample users, that is, the sample current characteristics are associated with the sample electricity consumers who have electricity theft behaviors. household.
  • the method of constructing sample current features is the same as that of constructing target current features, so that the format of sample current features (such as the number and meaning of dimensions, etc.) and the format of constructing target current features (such as the number and meaning of dimensions, etc.), namely , to obtain the inrush current detected by multiple meters installed in the places of sample electricity users, extract the original current characteristics from the inrush current, conduct principal component analysis on the original current characteristics, and use dimensionality reduction as the sample current characteristics.
  • the format of sample current features such as the number and meaning of dimensions, etc.
  • the format of constructing target current features such as the number and meaning of dimensions, etc.
  • Step 105 according to the characteristics of the sample current, classify the sample electricity users with the same electricity stealing behavior into the same sample user cluster.
  • stealing electricity for example, stealing electricity by under-voltage method, stealing electricity by under-current method, stealing electricity by disconnecting the neutral wire in the method of under-voltage method, stealing electricity by disconnecting the neutral wire in the method of under-current method
  • stealing electricity is usually not carried out continuously, but selectively selected for some time periods, so that different types of electricity stealing behaviors will have different effects on the surge current. Influence.
  • disconnecting the neutral wire to steal electricity is to disconnect the neutral wire of the circuit, that is, the neutral wires at the front and back of the meter, so that the voltage coil of the meter cannot be looped according to the voltage line when it is energized, and thus does not work normally. As a result, the meter cannot measure the electricity consumption of the electricity consumer.
  • shunting power stealing in the meter is to connect a shunt line in parallel between the phase wire current inlet and outlet holes flowing through the meter, so that when the user uses electricity, the current coil of the meter has no current or partial current through it.
  • phase-to-zero swap stealing is to swap the entrance and exit of the phase line and neutral line in the ammeter. After the power is cut off, the neutral line connected to the ammeter is replaced with other zero line or grounding line, and the ammeter has no current flow.
  • clustering algorithms such as K-Means (K-Means), Spectral Clustering (Spectral Clustering), and Hierarchical Clustering (Hierarchical Clustering) can be called to cluster sample electricity users using sample current characteristics, and The sample electricity users are divided into different sample user clusters. These sample user clusters can be manually checked by the operation and maintenance personnel to log in to the client for accurate clustering. If not accurately clustered, the operation and maintenance personnel will send adjustments to the server on the client Instructions to notify the server to add and delete sample electricity users in a sample user cluster. If the clustering is accurate, it can be determined that the electricity stealing behavior of the sample electricity users in the same sample user cluster is the same type of electricity theft .
  • K-Means K-Means
  • Spectral Clustering Spectral Clustering
  • Hierarchical Clustering Hierarchical Clustering
  • Step 106 setting multiple K values.
  • Step 107 for each K value, perform K-means clustering on target electricity users and sample electricity users according to target current characteristics and sample current characteristics to obtain K target user clusters.
  • K-Means (K-Means) algorithm is used to cluster the target electricity consumers
  • K (K is a positive integer) value represents the number of clusters clustered
  • the selection of the K value has a significant impact on the K-means clustering.
  • the results (the target electricity users included in the cluster) are obviously affected, and the results obtained by different K values are different. If you manually set the K value and manually observe the pros and cons of the K-means clustering results, the operation is cumbersome and the efficiency is low.
  • this embodiment can set multiple reasonable K values, such as 3-100, and perform K-means clustering once for a K value to obtain K target user clusters. Since the calculation of K-means clustering is simple, it can ensure multiple For the normal execution of secondary K-means clustering, the sample electricity users, sample current characteristics, and sample user clusters are used as prior knowledge, and the prior knowledge is used as supervision. Electric users perform K-means clustering to obtain K target user clusters, and select appropriate target user clusters from the target user clusters corresponding to different K values.
  • step 107 may include the following steps:
  • Step 1071 for each K value, initialize K target user clusters.
  • K target user clusters can be initialized, and each target user cluster has a center.
  • the center can be initially set randomly, or K can be selected as far away from each other as possible.
  • Point as the center it is also possible to first cluster the data (target current characteristics, sample current characteristics) with hierarchical clustering algorithm or Canopy algorithm, after obtaining K clusters, select a point from each cluster, the point can be The center point of the cluster, or the point closest to the center point of the cluster, etc.
  • Step 1072 calculate the distance between the combined current feature and the center.
  • the combined electricity user is regarded as a point, and the distance between the point and the center is calculated using its combined current characteristics, such as Euclidean distance, cosine distance, etc.
  • the combined electricity users include target electricity users and sample electricity users
  • the combined current characteristics include target current characteristics and sample current characteristics
  • Step 1073 classify combined electricity users into the target user cluster with the smallest distance.
  • the distance between the combined power user and the center of each target user cluster can be compared, and the target user cluster with the smallest distance is selected as the target user cluster to which the combined power user belongs, so that Classify combined electricity users into the target user cluster with the smallest distance.
  • Step 1074 in each target user cluster, calculate the average value of the combined current characteristics of the combined power users, so as to update the center of the target user cluster.
  • each target user cluster contains multiple combined power users, and the average value of the combined current characteristics of multiple combined power users is calculated as the new value of the target user cluster. center of.
  • Step 1078 judge whether the range of change of the center during update is less than or equal to the first threshold; based on the judgment result that the range of change of the center during update is less than or equal to the first threshold, execute step 1079, based on the fact that the range of change of the center during update is greater than or equal to the first threshold The judgment result of the first threshold returns to step 1072 .
  • Step 1079 determine that the target user cluster converges.
  • the difference between the center before the update and the center after the update can be calculated as the range of change during the update, and the range of change during the update is compared with the preset first threshold.
  • the change range during the update is less than or equal to the first threshold, it means that the change of the center update is small, it can be confirmed that the target user cluster converges, and the K-means clustering is completed.
  • the range of change during updating is greater than the first threshold, it means that the change of the center update is large, and it can be confirmed that the target user cluster has not converged, and enters the next round of training, and re-executes steps 1072-1077 until the target user cluster converges.
  • Step 108 calculating the similarity between the sample user cluster and the target user cluster according to the distribution information of the sample electricity users.
  • the sample electricity user has a predetermined type of stealing electricity. Therefore, the distribution information of the sample electricity user in the target user cluster can be compared with the distribution information of the sample electricity user in the sample user cluster, so as to calculate the sample user
  • the similarity between the cluster and the target user cluster is used to characterize the similarity of the clustering of the sample electricity users.
  • step 108 may include the following steps:
  • Step 1081 Count the total number of sample electricity users belonging to each type of electricity theft in the target user cluster.
  • sample power users are divided into different target user clusters, and the division of target user clusters may not be consistent with the division of sample user clusters.
  • the total number of sample power users of different types of stealing electricity in the target user cluster is counted.
  • Step 1082 setting the target user cluster with the largest total number as the characteristic user cluster belonging to the electric stealing type.
  • the sample electricity users under this electricity stealing type may be divided into different target user clusters, and the total number of sample electricity users under this electricity stealing type in different target user clusters is compared, Therefore, select the target user cluster with the largest total number, mark the electricity stealing type, and record it as a characteristic user cluster representing the electricity stealing type.
  • Step 1083 for the specified electricity theft type, calculate the overlapping degree between the characteristic user cluster and the sample user cluster on the sample electricity user, as the similarity between the characteristic user cluster and the sample user cluster.
  • the sample electricity users are given.
  • the clustering is the sample user cluster first, and the later clustering is the characteristic user cluster. At this time, the difference between the characteristic user cluster and the sample user cluster can be calculated.
  • the degree of overlap on the electronic account is used as the similarity between the feature user cluster and the sample user cluster.
  • the first sub-number X1 of all sample power users in the statistical characteristic user cluster the sample power user does not distinguish which type of electricity theft belongs to, and the statistical characteristic user cluster belongs to the specified
  • the second sub-number X2 of the sample electricity-stealing users of the specified electricity-stealing type and on the other hand, count the third sub-number X3 of the sample electricity-stealing users belonging to the specified electricity-stealing type in the sample user cluster.
  • Step 109 Detect the validity of the target user cluster according to the similarity and the distribution information of the target electricity users.
  • the sample user cluster is used as prior knowledge, and the performance of clustering is tested.
  • the sample user cluster can be used as a reference to evaluate the similarity between the sample user cluster and the target user cluster in the distribution of sample electricity users, so as to test the target user cluster Clustering performance.
  • this embodiment can combine the similarity between the sample user cluster and the target user cluster in the distribution of the sample power users and the distribution information of the target power users to detect the effectiveness of the target user cluster, that is, to detect the target user cluster is valid or invalid.
  • the proportion of the target power users in the characteristic user cluster to all target power users can be counted.
  • the similarity is compared with a preset second threshold, and the proportion is compared with a preset third threshold.
  • Step 110 based on the effective detection results of the target user cluster, detect target electricity users who have committed electricity theft in the target user cluster according to the sample electricity users.
  • the sample user clusters can be used as a reference, and the target power users who have electricity stealing behaviors can be detected in the target user clusters according to the sample power users.
  • the characteristic user cluster after the characteristic user cluster is determined, it may be determined that the target power user in the characteristic user cluster has a power stealing behavior corresponding to the type of power theft belonging to the characteristic user cluster.
  • the inrush current detected by a plurality of electric meters is received, and the electric meter is installed in the place of the target electric user, the original current characteristics are extracted from the inrush current, and the principal component analysis is performed on the original current characteristics, and dimensionality reduction is used as
  • the target current feature the sample current feature with the same format as the target current feature is obtained, the sample current feature is associated with the sample electricity users who have stolen electricity, and the sample electricity users with the same type of electricity stealing behavior are divided into
  • For the same sample user cluster set multiple K values, and for each K value, perform K-means clustering on the target electricity users and sample electricity users according to the target current characteristics and sample current characteristics, and obtain K target user clusters, according to
  • the distribution information of the sample electricity users calculates the similarity between the sample user cluster and the target user cluster, and detects the validity of the target user cluster according to the similarity and the distribution information of the target user cluster.
  • the target electricity users who have electricity theft behavior are detected according to the sample electricity users.
  • the inrush current is the process of electricity consumption by the electricity users, which can reflect the electricity consumption behavior of the electricity users to a certain extent. Rich information can improve the quality of features, thereby improving the accuracy of clustering.
  • using sample electricity users and their sample user clusters as prior knowledge to test the results of K-means clustering can not only Guarantee the accuracy of K-means clustering results, thereby ensuring the accuracy of detecting electricity stealing behavior for target electricity users, and avoid manually observing the performance of K-means clustering when manually setting the K value, greatly reducing workload and improving The efficiency of detection reduces the cost of detection.
  • the K-means clustering operation is simple, the amount of calculation is small, and the resources occupied are small, which can meet the detection of electricity theft behavior of a large number of electricity users.
  • FIG. 2 is a block diagram of a detection device for stealing electricity provided in Embodiment 2 of the present application, the device may include the following modules:
  • the inrush current receiving module 201 is configured to receive the inrush current detected by a plurality of electric meters installed in places of target electric users;
  • the current feature extraction module 202 is configured to extract the original current feature from the surge current
  • the current feature dimension reduction module 203 is configured to perform principal component analysis on the original current feature, and take dimensionality reduction as the target current feature;
  • the current feature acquisition module 204 is configured to acquire a sample current feature in the same format as the target current feature, and the sample current feature is associated with a sample electricity user who has stolen electricity;
  • the sample user cluster clustering module 205 is configured to divide the sample electricity users with the same electricity stealing type in the electricity stealing behavior into the same sample user cluster according to the sample current characteristics;
  • the clustering parameter setting module 206 is configured to set a plurality of K values
  • the target user clustering module 207 is configured to perform K-means clustering on the target electricity users and the sample electricity users according to the target current characteristics and the sample current characteristics for each K value, Get K target user clusters;
  • the similarity calculation module 208 is configured to calculate the similarity between the sample user cluster and the target user cluster according to the distribution information of the sample electricity users;
  • the validity detection module 209 is configured to detect the validity of the target user cluster according to the similarity and the distribution information of the target power users;
  • the electricity stealing behavior detection module 210 is configured to detect, in the target user cluster according to the sample electricity users, the target electricity user who has committed the electricity stealing behavior based on the effective detection result of the target user cluster.
  • the current feature extraction module 202 is further configured to:
  • the current feature dimensionality reduction module 203 is also set to:
  • the eigenvectors are arranged in rows from top to bottom, and the first q rows are taken to form the second current matrix;
  • the target user cluster clustering module 207 is also set to:
  • K target user clusters For each K value, initialize K target user clusters, each of which has a center in the target user cluster;
  • Classifying combined electricity users into the target user cluster with the smallest distance the combined electricity users include the target electricity users and the sample electricity users;
  • the similarity calculation module 208 is also set to:
  • For the specified electricity stealing type calculate the degree of overlap between the characteristic user cluster and the sample user cluster on the sample power user, as the difference between the characteristic user cluster and the sample user cluster similarity.
  • the similarity calculation module 208 is also set to:
  • the validity detection module 209 is also set to:
  • the stealing behavior detection module 210 is also set to:
  • the target power user in the characteristic user cluster has a power stealing behavior belonging to the power stealing type.
  • the device for detecting electricity stealing behavior provided by the embodiment of the present application can execute the method for detecting electricity stealing behavior provided by the embodiment of the present application, and has corresponding functional modules and beneficial effects for executing the method.
  • FIG. 3 is a schematic structural diagram of a computer device provided in Embodiment 3 of the present application.
  • FIG. 3 shows a block diagram of an exemplary computer device 12 suitable for implementing embodiments of the present application.
  • the computer device 12 shown in FIG. 3 is only one example.
  • computer device 12 takes the form of a general-purpose computing device.
  • Components of computer device 12 may include at least one processor or processing unit 16 , system memory 28 , bus 18 connecting various system components including system memory 28 and processing unit 16 .
  • Bus 18 represents at least one of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using one of several bus structures.
  • bus structures include, for example, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MCA) bus, the Enhanced ISA bus, the Video Electronics Standard Association (VESA) ) Local bus and Peripheral Component Interconnect (PCI) bus.
  • ISA Industry Standard Architecture
  • MCA Micro Channel Architecture
  • VESA Video Electronics Standard Association
  • PCI Peripheral Component Interconnect
  • Computer device 12 includes a variety of computer system readable media. These media can be any available media that can be accessed by computer device 12 and include both volatile and nonvolatile media, removable and non-removable media.
  • System memory 28 may include computer system readable media in the form of volatile memory, such as Random Access Memory (RAM) 30 and/or cache memory 32 .
  • Computer device 12 may include other removable/non-removable, volatile/nonvolatile computer system storage media.
  • storage system 34 may be used to read from and write to non-removable, non-volatile magnetic media (commonly referred to as a "hard drive”).
  • Disk drives for reading and writing to removable non-volatile disks (such as "floppy disks") and for removable non-volatile optical disks (such as Compact Disc-Read Only Memory (CD) -ROM), Digital Video Disc-Read Only Memory (Digital Video Disc-Read Only Memory, DVD-ROM) or other optical media) CD-ROM drive.
  • each drive may be connected to bus 18 via at least one data medium interface.
  • Memory 28 may include at least one program product having a set (eg, at least one) of program modules configured to perform the functions of various embodiments of the present application.
  • a program/utility tool 40 having a set (at least one) of program modules 42 may be stored, for example, in memory 28, such program modules 42 including an operating system, at least one application program, other program modules, and program data, in these examples Each or combination may include implementations of network environments.
  • the program modules 42 generally perform the functions and/or methods of the embodiments described herein.
  • Computer device 12 may also communicate with at least one external device 14 (e.g., a keyboard, pointing device, display 24, etc.), and at least one device that enables a user to interact with 12 A device (eg, network card, modem, etc.) capable of communicating with at least one other computing device. This communication can be performed through an input/output (Input/Output, I/O) interface 22 .
  • the computer device 12 can also communicate with at least one network (such as a local area network (Local Area Network, LAN), a wide area network (Wide Area Network, WAN) and/or a public network such as the Internet) through the network adapter 20. As shown, network adapter 20 communicates with other modules of computer device 12 via bus 18 .
  • the processing unit 16 executes various functional applications and data processing by running the programs stored in the system memory 28 , for example, realizing the method for detecting power theft provided by the embodiment of the present application.
  • Embodiment 4 of the present application also provides a computer-readable storage medium.
  • a computer program is stored on the computer-readable storage medium.
  • the computer program is executed by a processor, each process of the detection method for the above-mentioned stealing behavior can be realized, and the same can be achieved. To avoid repetition, the technical effects will not be repeated here.
  • the computer-readable storage medium may include, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or a combination thereof.
  • Examples of computer readable storage media include: an electrical connection having at least one lead, a portable computer disk, a hard disk, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (such as electronic programmable read-only memory (Electronic Programable Read Only Memory, EPROM) or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or a suitable combination of the above.
  • a computer-readable storage medium may be a tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本申请提供了一种窃电行为的检测方法、装置、计算机设备和存储介质，该方法包括：接收多个电表检测到的涌浪电流，从涌浪电流中提取原始电流特征，对原始电流特征进行主成分分析，以降维为目标电流特征，根据样本电流特征将窃电行为的窃电类型相同的样本用电户划分至同一个样本用户簇，针对每个K值，根据目标电流特征、样本电流特征对目标用电户、样本用电户进行K均值聚类，得到K个目标用户簇，根据样本用电户的分布信息计算样本用户簇与目标用户簇之间的相似度，根据相似度与目标用电户的分布信息检测目标用户簇的有效性，基于目标用户簇有效的检测结果，在目标用户簇中依据样本用电户检测发生窃电行为的目标用电户，保证对目标用电户检测窃电的准确性。

Description

窃电行为的检测方法、装置、计算机设备和存储介质
本公开要求在2021年12月17日提交中国专利局、申请号为202111545671.2的中国专利申请的优先权，以上申请的全部内容通过引用结合在本申请中。
技术领域
本申请涉及电力的技术领域，例如涉及一种窃电行为的检测方法、装置、计算机设备和存储介质。
背景技术
在电力行业，随着经济发展、社会用电量的增大，窃电行为越发频繁，不仅给供电企业带来损失，而且，不安全的用电行为容易造成安全事故。
相关技术中，监测窃电行为的方式主要是考量用电户的用电量，通过自定义规则或者时序模型对用电户的用电量进行分析，从而检测是否发生窃电行为。
用电量是用电户用电的结果，损失的信息较多，比如，同样的用电量，可能是一台空调消耗的，也可能是电饭煲与电视机共同消耗的，还可能是电暖器与电吹风消耗的，这将会导致检测窃电行为的精确度较低。
发明内容
本申请提出了一种窃电行为的检测方法、装置、计算机设备和存储介质，以解决使用用电户的用电量检测窃电行为的精确度较低的问题。
第一方面，本申请实施例提供了一种窃电行为的检测方法，包括：
接收多个电表检测到的涌浪电流，所述电表安装在目标用电户的场所中；
从所述涌浪电流中提取原始电流特征；
对所述原始电流特征进行主成分分析，以降维为目标电流特征；
获取格式与所述目标电流特征相同的样本电流特征，所述样本电流特征关联发生窃电行为的样本用电户；
根据所述样本电流特征将所述窃电行为的窃电类型相同的所述样本用电户划分至同一个样本用户簇；
设置多个K值；
针对每个所述K值，根据所述目标电流特征、所述样本电流特征对所述目标用电户、所述样本用电户进行K均值聚类，得到K个目标用户簇；
根据所述样本用电户的分布信息计算所述样本用户簇与所述目标用户簇之间的相似度；
根据所述相似度与所述目标用电户的分布信息检测所述目标用户簇的有效性；
基于所述目标用户簇有效的检测结果，在所述目标用户簇中依据所述样本用电户检测发生窃电行为的所述目标用电户。
第二方面，本申请实施例还提供了一种窃电行为的检测装置，包括：
涌浪电流接收模块，设置为接收多个电表检测到的涌浪电流，所述电表安装在目标用电户的场所中；
电流特征提取模块，设置为从所述涌浪电流中提取原始电流特征；
电流特征降维模块，设置为对所述原始电流特征进行主成分分析，以降维为目标电流特征；
电流特征获取模块，设置为获取格式与所述目标电流特征相同的样本电流特征，所述样本电流特征关联发生窃电行为的样本用电户；
样本用户簇聚类模块，设置为根据所述样本电流特征将所述窃电行为的窃电类型相同的所述样本用电户划分至同一个样本用户簇；
聚类参数设置模块，设置为设置多个K值；
目标用户簇聚类模块，设置为针对每个所述K值，根据所述目标电流特征、所述样本电流特征对所述目标用电户、所述样本用电户进行K均值聚类，得到K个目标用户簇；
相似度计算模块，设置为根据所述样本用电户的分布信息计算所述样本用户簇与所述目标用户簇之间的相似度；
有效性检测模块，设置为根据所述相似度与所述目标用电户的分布信息检测所述目标用户簇的有效性；
窃电行为检测模块，设置为基于所述目标用户簇有效的检测结果，在所述目标用户簇中依据所述样本用电户检测发生窃电行为的所述目标用电户。
第三方面，本申请实施例还提供了一种计算机设备，所述计算机设备包括：
处理器；
存储器，设置为存储程序，
在所述程序被所述处理器执行时，所述处理器实现如第一方面所述的窃电行为的检测方法。
第四方面，本申请实施例还提供了一种计算机可读存储介质，所述计算机可读存储介质上存储计算机程序，所述计算机程序被处理器执行时实现如第一方面所述的窃电行为的检测方法。
附图说明
图1为本申请实施例一提供的一种窃电行为的检测方法的流程图；
图2为本申请实施例二提供的一种窃电行为的检测装置的结构示意图；
图3为本申请实施例三提供的一种计算机设备的结构示意图。
具体实施方式
下面结合附图和实施例对本申请作说明。可以理解的是，此处所描述的实施例仅仅用于解释本申请。另外还需要说明的是，为了便于描述，附图中仅示出了与本申请相关的部分。
实施例一
图1为本申请实施例一提供的一种窃电行为的检测方法的流程图，本实施例可适用于适用涌浪电流检测窃电行为的情况，该方法可以由窃电行为的检测装置来执行，该窃电行为的检测装置可以由软件和/或硬件实现，可配置在作为服务端的计算机设备中，例如，服务器、工作站、个人电脑等，该方法包括如下步骤：
步骤101、接收多个电表检测到的涌浪电流。
在本实施例中，电表安装在目标用电户的场所中，如住宅、公寓等，电表通常安装有浪涌电流(surge current)的保护器，在该保护器中可以增加元件，用于检测浪涌电流。
其中，浪涌电流也称为接通浪涌或输入浪涌电流，是指电源或电气设备在 接通时消耗的瞬时高输入电流，即，在电源或电气设备接通瞬间，流入电源设备的峰值电流，这是由于为电容器和电感器或变压器充电需要很高的初始电流。
示例性地，接通时，电源中的放电电容器提供低阻抗，当它们从零充电至最大值时，允许大电流流入电路。这些电流可能高达稳态电流的20倍。即使它仅持续约10毫秒，它仍需要30到40个周期才能使电流稳定到正常工作值。如果没有限制，那么大电流除了会在电源线上产生电压骤降之外，还会损坏设备，并导致由同一电源供电的其他设备发生故障。
电表持续地检测每个目标用电户在使用电气设备时产生的涌浪电流，并将该涌浪电路的信息通过物联网等方式传输至服务端。
步骤102、从涌浪电流中提取原始电流特征。
在本实施例中，可以从涌浪电流中提取用于区分正常用电行为与窃电行为的特征，记为原始电流特征。
示例性地，可以将第一时间周期(如1天)划分为多个第二时间周期(如5分钟)。
在每个第二时间周期内生成标记，该标记表示是否接收到涌浪电流，例如，在第二时间周期接收到涌浪电流，则标记为1，在第二时间周期未接收到涌浪电流，则标记为0。
在标记表示接收到涌浪电流的情况下，统计在该第二时间周期内接收涌浪电流的频次。
对涌浪电流分别统计在该第二时间周期内的最大值、最小值、平均值。
在标记表示未接收到涌浪电流的情况下，将在该第二时间周期内接收涌浪电流的频次设置为指定的数值，如0。
将涌浪电流在该第二时间周期内的最大值、最小值、平均值均设置为指定的数值，如0。
将第二时间周期内的标记、频次、最大值、最小值、平均值按照既定的方式排列，组合为原始字符串。
将第一时间周期内的所有原始字符串按照第二时间周期的顺序首位拼接，组成原始电流特征。
当然，上述原始电流特征只是作为示例，在实施本申请实施例时，可以根据实际情况设置其它原始电流特征，例如，对第二时间周期内的涌浪电流累加获得总电流值，对第二时间周期内的涌浪电流计算方差等。另外，除了上述原始电流特征外，本领域技术人员还可以根据实际需要采用其它原始电流特征。
步骤103、对原始电流特征进行主成分分析，以降维为目标电流特征。
原始电流特征以向量的形式表示，由于原始电流特征的维度较多，在维持 原始电流特征的主要特征分量的情况下，可以通过PCA(Principal Component Analysis，主成分分析)算法降低原始电流特征的维度，记为目标电流特征，使得目标电流特征的维度降低至个或十的量级，从而降低运算量。
其中，PCA通过线性变换将原始数据(原始电流特征)变换为一组各维度线性无关的表示(目标电流特征)。
在一实施例中，假设有m条n维的原始电流特征，则可以将原始电流特征组合成矩阵，得到n行m列第一电流矩阵X。
将第一电流矩阵X中的每一行数据执行零均值化(减去这一行的均值)。
在执行零均值化完成的情况下，对第一电流矩阵X计算协方差矩阵C，
Figure PCTCN2022136079-appb-000001
其中，X T为X的转置矩阵。
计算协方差矩阵C的特征值与特征向量，以及，按照特征值的大小对特征向量从上到下按行排列，并取前q(q为正整数)行组成新的矩阵，记为第二电流矩阵P。
计算第二电流矩阵P与第一电流矩阵X之间的乘积，获得降维之后的目标电流特征Y，即，Y＝PX。
步骤104、获取格式与目标电流特征相同的样本电流特征。
在本实施例中，可以预先通过其他方式(如人工甄别)筛选发生窃电行为的样本用电户，对该样本用户设置样本电流特征，即，样本电流特征关联发生窃电行为的样本用电户。
其中，构建样本电流特征的方式与构建目标电流特征的方式相同，使得样本电流特征的格式(如维度的数量及含义等)与构建目标电流特征的格式(如维度的数量及含义等)，即，获取安装在样本用电户的场所中多个电表检测到的涌浪电流，从涌浪电流中提取原始电流特征，对原始电流特征进行主成分分析，以降维为样本电流特征。
步骤105、根据样本电流特征将窃电行为的窃电类型相同的样本用电户划分至同一个样本用户簇。
在实际应用中，窃电行为的窃电类型较多，例如，欠压法窃电，欠流法窃电，欠压法窃电有断开零线的方式窃电，欠流法窃电有表内分流窃电和相零对调窃电等，窃电行为通常并非持续进行，而是有选择性地挑选部分时间段进行，使得不同窃电类型的窃电行为会对浪涌电流产生不同的影响。
例如，断开零线窃电是将电路的零线断开，即，电表前面和后面的零线，使得电能表的电压线圈在通电时无法按照电压线路进行回路，进而不同正常工作，这就导致电表对用电户的用电量无法进行计量。
又如，表内分流窃电是在流经电表的相线电流进出孔之间并联一条分流线，使得用电户在用电时，电表的电流线圈没有电流通过或有局部电流通过。
再如，相零对调窃电是将电表中相线和零线的出入口进行对调，在断电后，将连接电表的零线改用其他零线或接地线，电表没有电流通过。
在本实施例中，可以调用K均值(K-Means)、谱聚类(Spectral Clustering)，层次聚类(Hierarchical Clustering)等聚类算法，使用样本电流特征对样本用电户进行聚类，将样本用电户划分至不同的样本用户簇中，这些样本用户簇可以由运维人员登录客户端人工检查是否准确聚类，如果未准确聚类，则运维人员在客户端向服务端发出调整指令，通知服务端在一样本用户簇中增加样本用电户、删除样本用电户，如果准确聚类，则可以认定同一个样本用户簇中样本用电户的窃电行为的窃电类型相同。
步骤106、设置多个K值。
步骤107、针对每个K值，根据目标电流特征、样本电流特征对目标用电户、样本用电户进行K均值聚类，得到K个目标用户簇。
在本实施例中，使用K均值(K-Means)算法对目标用电户进行聚类，K(K为正整数)值表示聚类的簇的数量，K值的选取对K均值聚类的结果(簇中包含的目标用电户)影响明显，不同K值得到的结果不一样，如果手动设置K值，人工观察K均值聚类的结果的优劣，操作繁琐，效率较低。
为此，本实施例可以设置多个合理的K值，如3-100，针对一个K值执行一次K均值聚类，得到K个目标用户簇，由于K均值聚类的计算简单，可以保证多次K均值聚类的正常执行，将样本用电户、样本电流特征、样本用户簇作为先验知识，以先验知识作为监督，连同目标电流特征、样本电流特征对目标用电户、样本用电户进行K均值聚类，得到K个目标用户簇，从不同K值对应的目标用户簇中选择合适的目标用户簇。
在本申请的一个实施例中，步骤107可以包括如下步骤：
步骤1071、针对每个K值，初始化K个目标用户簇。
针对每个K值执行K均值聚类时，可以初始化K个目标用户簇，每个目标用户簇中具有中心，该中心初始可以是随机设置的，也可以是选择彼此距离尽可能远的K个点作为中心，还可以是先对数据(目标电流特征、样本电流特征)用层次聚类算法或者Canopy算法进行聚类，得到K个簇之后，从每个簇中选择一个点，该点可以是该类簇的中心点，或者是距离类簇中心点最近的那个点等。
步骤1072、计算组合电流特征与中心之间的距离。
在每轮训练中，将组合用电户视为一个点，使用其组合电流特征计算该点与中心之间的距离，如欧氏距离、余弦距离等。
其中，组合用电户包括目标用电户、样本用电户，组合电流特征包括目标电流特征、样本电流特征。
步骤1073、将组合用电户划入距离最小的目标用户簇中。
针对既定的组合用电户，可以将该组合用电户与每个目标用户簇的中心之间的距离进行比较，选择距离最小的目标用户簇作为该组合用电户归属的目标用户簇，从而将组合用电户划入距离最小的目标用户簇中。
步骤1074、在每个目标用户簇中，计算组合用电户的组合电流特征计算平均值，以更新目标用户簇的中心。
将每个组合用电户重新划分至目标用户簇中之后，每个目标用户簇包含了多个组合用电户，对多个组合用电户的组合电流特征计算平均值，作为该目标用户簇新的中心。
步骤1078、判断中心在更新时的变化幅度是否小于或等于第一阈值；基于中心在更新时的变化幅度小于或等于第一阈值的判断结果，执行步骤1079，基于中心在更新时的变化幅度大于第一阈值的判断结果，返回执行步骤1072。
步骤1079、确定目标用户簇收敛。
针对同一个目标用户簇，可以计算更新前的中心与更新后的中心之间的差异，作为更新时的变化幅度，将更新时的变化幅度与预设的第一阈值进行比较。
如果更新时的变化幅度小于或等于第一阈值，表示中心更新的变化较小，可以确认目标用户簇收敛，K均值聚类完成。
如果更新时的变化幅度大于第一阈值，表示中心更新的变化较大，可以确认目标用户簇未收敛，进入下一轮训练，重新执行步骤1072-步骤1077，直至目标用户簇收敛。
步骤108、根据样本用电户的分布信息计算样本用户簇与目标用户簇之间的相似度。
样本用电户作为先验知识，具有既定的窃电类型，因此，可以对比样本用电户在目标用户簇中的分布信息与样本用电户在样本用户簇中的分布信息，从而计算样本用户簇与目标用户簇之间的相似度，该相似度用于表征对样本用电户聚类的相似程度。
在本申请的一个实施例中，步骤108可以包括如下步骤：
步骤1081、在目标用户簇中统计属于各个窃电类型的样本用电户的总数量。
在K均值聚类中，样本用电户划分在不同的目标用户簇中，划分目标用户簇的情况可能与划分样本用户簇的情况可能并不一致，同一个目标用户簇中，可以包含至少一个窃电类型的样本用电户。
在本实施例中，针对每个目标用户簇中，统计该目标用户簇中不同窃电类型的样本用电户的总数量。
步骤1082、将总数量最大的目标用户簇设置为归属窃电类型的特征用户簇。
针对既定的窃电类型，该窃电类型下的样本用电户可能划分至不同的目标 用户簇中，将不同的目标用户簇中该窃电类型下的样本用电户的总数量进行比较，从而选择总数量最大的目标用户簇，标记该窃电类型，记为表征该窃电类型的特征用户簇。
步骤1083、针对指定的窃电类型，计算特征用户簇与样本用户簇之间在样本用电户上的重合程度，作为特征用户簇与样本用户簇之间的相似度。
针对既定的窃电类型，样本用电户是既定的，在先聚类为样本用户簇、在后聚类为特征用户簇，此时，可计算特征用户簇与样本用户簇之间在样本用电户上的重合程度，作为特征用户簇与样本用户簇之间的相似度。
在一实施例中，一方面，统计特征用户簇中所有样本用电户的第一子数量X1，该样本用电户并不区分属于何种窃电类型，以及，统计特征用户簇中属于指定的窃电类型的样本用电户的第二子数量X2，另一方面，统计样本用户簇中属于指定的窃电类型的样本用电户的第三子数量X3。
将第一子数量X1依次加上第三子数量X3并减去第二子数量X2，得到目标子数量X4，即，X4＝X1+X3-X2。
计算第二子数量与目标子数量之间的比值X2/X4，作为特征用户簇与样本用户簇之间的相似度。
步骤109、根据相似度与目标用电户的分布信息检测目标用户簇的有效性。
样本用户簇作为先验知识，聚类的性能得到检验，可以以样本用户簇作为参考，评估样本用户簇与目标用户簇聚之间在样本用电户分布上的相似度，从而检验目标用户簇聚类的性能。
此外，目标用电户中发生窃电行为与正常用电行为的存在一定规律的，这些规律以分布信息呈现，可以检验目标用户簇聚类的性能。
因此，本实施例可以结合样本用户簇与目标用户簇聚之间在样本用电户分布上的相似度、目标用电户的分布信息检测目标用户簇的有效性，即，检测检测目标用户簇是有效还是无效。
在一实施例中，在确定特征用户簇之后，可以统计特征用户簇中的目标用电户占所有目标用电户的占比。
分别将相似度与预设的第二阈值进行比较、将占比与预设的第三阈值进行比较。
在相似度大于或等于第二阈值、且占比小于或等于第三阈值的情况下，确定目标用户簇有效。
在相似度小于第二阈值、或占比大于第三阈值的情况下，确定目标用户簇无效。
步骤110、基于目标用户簇有效的检测结果，在目标用户簇中依据样本用电户检测发生窃电行为的目标用电户。
针对有效的聚类结果(K个目标用户簇)，可以以样本用户簇作为参考，在目标用户簇中依据样本用电户检测发生窃电行为的目标用电户。
在一实施例中，在确定特征用户簇之后，可以确定特征用户簇中的目标用电户发生归属特征用户簇对应窃电类型的窃电行为。
在本实施例中，接收多个电表检测到的涌浪电流，电表安装在目标用电户的场所中，从涌浪电流中提取原始电流特征，对原始电流特征进行主成分分析，以降维为目标电流特征，获取格式与目标电流特征相同的样本电流特征，样本电流特征关联发生窃电行为的样本用电户，根据样本电流特征将窃电行为的窃电类型相同的样本用电户划分至同一个样本用户簇，设置多个K值，针对每个K值，根据目标电流特征、样本电流特征对目标用电户、样本用电户进行K均值聚类，得到K个目标用户簇，根据样本用电户的分布信息计算样本用户簇与目标用户簇之间的相似度，根据相似度与目标用电户的分布信息检测目标用户簇的有效性，基于目标用户簇有效的检测结果，在目标用户簇中依据样本用电户检测发生窃电行为的目标用电户，涌浪电流是用电户用电的过程，可以在一定程度上体现用电户的用电行为本身，保留了较为丰富的信息量，可提高特征的质量，从而可以提高聚类的准确性，再者，以样本用电户及其样本用户簇作为先验知识，对K均值聚类的结果进行检验，不仅可以保证K均值聚类的结果的准确性，从而保证对目标用电户检测窃电行为的准确性，而且避免手动设置K值时人工观察K均值聚类的性能，大大减少了工作量，提高了检测的效率，降低了检测的成本，此外，K均值聚类操作简单、计算量少，占用的资源少，可以满足对大量用电户检测窃电行为。
需要说明的是，对于方法实施例，为了简单描述，故将其都表述为一系列的动作组合，但是本领域技术人员应该知悉，本申请实施例并不受所描述的动作顺序的限制，因为依据本申请实施例，一些步骤可以采用其他顺序或者同时进行。其次，本领域技术人员也应该知悉，说明书中所描述的实施例均属于优选实施例，所涉及的动作并不一定是本申请实施例所必须的。
实施例二
图2为本申请实施例二提供的一种窃电行为的检测装置的结构框图，该装置可以包括如下模块：
涌浪电流接收模块201，设置为接收多个电表检测到的涌浪电流，所述电表安装在目标用电户的场所中；
电流特征提取模块202，设置为从所述涌浪电流中提取原始电流特征；
电流特征降维模块203，设置为对所述原始电流特征进行主成分分析，以降维为目标电流特征；
电流特征获取模块204，设置为获取格式与所述目标电流特征相同的样本电流特征，所述样本电流特征关联发生窃电行为的样本用电户；
样本用户簇聚类模块205，设置为根据所述样本电流特征将所述窃电行为的窃电类型相同的所述样本用电户划分至同一个样本用户簇；
聚类参数设置模块206，设置为设置多个K值；
目标用户簇聚类模块207，设置为针对每个所述K值，根据所述目标电流特征、所述样本电流特征对所述目标用电户、所述样本用电户进行K均值聚类，得到K个目标用户簇；
相似度计算模块208，设置为根据所述样本用电户的分布信息计算所述样本用户簇与所述目标用户簇之间的相似度；
有效性检测模块209，设置为根据所述相似度与所述目标用电户的分布信息检测所述目标用户簇的有效性；
窃电行为检测模块210，设置为基于所述目标用户簇有效的检测结果，在所述目标用户簇中依据所述样本用电户检测发生窃电行为的所述目标用电户。
在本申请的一个实施例中，所述电流特征提取模块202还设置为：
将第一时间周期划分为多个第二时间周期；
在每个所述第二时间周期内生成标记，所述标记表示是否接收到所述涌浪电流；
在所述标记表示接收到所述涌浪电流的情况下，统计接收所述涌浪电流的频次；
对所述涌浪电流分别统计最大值、最小值、平均值；
在所述标记表示未接收到所述涌浪电流的情况下，将接收所述涌浪电流的频次设置为指定的数值；
将所述涌浪电流的最大值、最小值、平均值均设置为指定的数值；
将所述第二时间周期内的所述标记、所述频次、所述最大值、所述最小值、所述平均值组合为原始字符串；
将所述第一时间周期内的所有所述原始字符串组成原始电流特征。
在本申请的一个实施例中，所述电流特征降维模块203还设置为：
将所述原始电流特征组合成第一电流矩阵；
将所述第一电流矩阵中的每一行数据执行零均值化；
在执行所述零均值化完成的情况下，对所述第一电流矩阵计算协方差矩阵；
计算所述协方差矩阵的特征值与特征向量；
按照所述特征值的大小对所述特征向量从上到下按行排列，并取前q行组成 第二电流矩阵；
计算所述第二电流矩阵与所述第一电流矩阵之间的乘积，获得降维之后的目标电流特征。
在本申请的一个实施例中，所述目标用户簇聚类模块207还设置为：
针对每个所述K值，初始化K个目标用户簇，每个所述目标用户簇中具有中心；
计算组合电流特征与所述中心之间的距离，所述组合电流特征包括所述目标电流特征、所述样本电流特征；
将组合用电户划入所述距离最小的所述目标用户簇中，所述组合用电户包括所述目标用电户、所述样本用电户；
在每个所述目标用户簇中，计算所述组合用电户的所述组合电流特征计算平均值，以更新所述目标用户簇的中心；
判断所述中心在更新时的变化幅度是否小于或等于第一阈值；基于所述中心在更新时的变化幅度小于或等于第一阈值的判断结果，确定所述目标用户簇收敛；基于所述中心在更新时的变化幅度大于第一阈值的判断结果，返回执行所述计算组合电流特征与所述中心之间的距离。
在本申请的一个实施例中，所述相似度计算模块208还设置为：
在所述目标用户簇中统计属于各个所述窃电类型的所述样本用电户的总数量；
将所述总数量最大的所述目标用户簇设置为归属所述窃电类型的特征用户簇；
针对指定的所述窃电类型，计算所述特征用户簇与所述样本用户簇之间在所述样本用电户上的重合程度，作为所述特征用户簇与所述样本用户簇之间的相似度。
在本申请的一个实施例中，所述相似度计算模块208还设置为：
统计所述特征用户簇中所有所述样本用电户的第一子数量；
统计所述特征用户簇中属于指定的所述窃电类型的所述样本用电户的第二子数量；
统计所述样本用户簇中属于指定的所述窃电类型的所述样本用电户的第三子数量；
将所述第一子数量依次加上所述第三子数量并减去所述第二子数量，得到目标子数量；
计算所述第二子数量与所述目标子数量之间的比值，作为所述特征用户簇与所述样本用户簇之间的相似度。
在本申请的一个实施例中，所述有效性检测模块209还设置为：
统计所述特征用户簇中的所述目标用电户占所有所述目标用电户的占比；
在所述相似度大于或等于第二阈值、且所述占比小于或等于第三阈值的情况下，确定所述目标用户簇的有效；
在本申请的一个实施例中，所述窃电行为检测模块210还设置为：
确定所述特征用户簇中的所述目标用电户发生归属所述窃电类型的窃电行为。
本申请实施例所提供的窃电行为的检测装置可执行本申请实施例所提供的窃电行为的检测方法，具备执行方法相应的功能模块和有益效果。
实施例三
图3为本申请实施例三提供的一种计算机设备的结构示意图。图3示出了适于用来实现本申请实施方式的示例性计算机设备12的框图。图3显示的计算机设备12仅仅是一个示例。
如图3所示，计算机设备12以通用计算设备的形式表现。计算机设备12的组件可以包括：至少一个处理器或者处理单元16，系统存储器28，连接不同系统组件(包括系统存储器28和处理单元16)的总线18。
总线18表示几类总线结构中的至少一种，包括存储器总线或者存储器控制器，外围总线，图形加速端口，处理器或者使用多种总线结构中的总线结构的局域总线。举例来说，这些体系结构包括工业标准体系结构(Industry Standard Architecture，ISA)总线，微通道体系结构(Micro Channel Architecture，MCA)总线，增强型ISA总线、视频电子标准协会(Video Electronic Standard Association，VESA)局域总线以及外围组件互连(Peripheral Component Interconnect，PCI)总线。
计算机设备12包括多种计算机系统可读介质。这些介质可以是任何能够被计算机设备12访问的可用介质，包括易失性和非易失性介质，可移动的和不可移动的介质。
系统存储器28可以包括易失性存储器形式的计算机系统可读介质，例如随机存取存储器(Random Access Memory，RAM)30和/或高速缓存存储器32。计算机设备12可以包括其它可移动/不可移动的、易失性/非易失性计算机系统存储介质。仅作为举例，存储系统34可以用于读写不可移动的、非易失性磁介质(通常称为“硬盘驱动器”)。可以提供用于对可移动非易失性磁盘(例如“软盘”)读写的磁盘驱动器，以及对可移动非易失性光盘(例如便携式紧凑磁盘只读存储器(Compact Disc-Read Only Memory，CD-ROM)，数字通用光盘只读存储器(Digital Video Disc-Read Only Memory，DVD-ROM)或者其它光介质)读写 的光盘驱动器。在这些情况下，每个驱动器可以通过至少一个数据介质接口与总线18相连。存储器28可以包括至少一个程序产品，该程序产品具有一组(例如至少一个)程序模块，这些程序模块被配置以执行本申请各实施例的功能。
具有一组(至少一个)程序模块42的程序/实用工具40，可以存储在例如存储器28中，这样的程序模块42包括操作系统、至少一个应用程序、其它程序模块以及程序数据，这些示例中的每一个或组合中可能包括网络环境的实现。程序模块42通常执行本申请所描述的实施例中的功能和/或方法。
计算机设备12也可以与至少一个外部设备14(例如键盘、指向设备、显示器24等)通信，还可与至少一个使得用户能与该计算机设备12交互的设备通信，和/或与使得该计算机设备12能与至少一个其它计算设备进行通信的设备(例如网卡，调制解调器等)通信。这种通信可以通过输入/输出(Input/Output，I/O)接口22进行。并且，计算机设备12还可以通过网络适配器20与至少一个网络(例如局域网(Local Area Network，LAN)，广域网(Wide Area Network，WAN)和/或公共网络，例如因特网)通信。如图所示，网络适配器20通过总线18与计算机设备12的其它模块通信。应当明白，可以结合计算机设备12使用其它硬件和/或软件模块，包括：微代码、设备驱动器、冗余处理单元、外部磁盘驱动阵列、独立磁盘冗余阵列(Redundant Array of Inexpensive Disks，RAID)系统、磁带驱动器以及数据备份存储系统等。
处理单元16通过运行存储在系统存储器28中的程序，从而执行各种功能应用以及数据处理，例如实现本申请实施例所提供的窃电行为的检测方法。
实施例四
本申请实施例四还提供一种计算机可读存储介质，计算机可读存储介质上存储有计算机程序，该计算机程序被处理器执行时实现上述窃电行为的检测方法的各个过程，且能达到相同的技术效果，为避免重复，这里不再赘述。
其中，计算机可读存储介质例如可以包括电、磁、光、电磁、红外线、或半导体的系统、装置或器件，或者以上的组合。计算机可读存储介质的示例包括：具有至少一个导线的电连接、便携式计算机磁盘、硬盘、随机存取存储器(RAM)、只读存储器(Read Only Memory，ROM)、可擦式可编程只读存储器(如电子可编程只读存储器(Electronic Programable Read Only Memory，EPROM)或闪存)、光纤、便携式紧凑磁盘只读存储器(CD-ROM)、光存储器件、磁存储器件、或者上述的合适的组合。在本文件中，计算机可读存储介质可以是包含或存储程序的有形介质，该程序可以被指令执行系统、装置或者器件使用或者与其结合使用。

Claims (10)

  1. 一种窃电行为的检测方法，包括：
    接收多个电表检测到的涌浪电流，所述电表安装在目标用电户的场所中；
    从所述涌浪电流中提取原始电流特征；
    对所述原始电流特征进行主成分分析，以降维为目标电流特征；
    获取格式与所述目标电流特征相同的样本电流特征，所述样本电流特征关联发生窃电行为的样本用电户；
    根据所述样本电流特征将所述窃电行为的窃电类型相同的所述样本用电户划分至同一个样本用户簇；
    设置多个K值；
    针对每个所述K值，根据所述目标电流特征、所述样本电流特征对所述目标用电户、所述样本用电户进行K均值聚类，得到K个目标用户簇；
    根据所述样本用电户的分布信息计算所述样本用户簇与所述目标用户簇之间的相似度；
    根据所述相似度与所述目标用电户的分布信息检测所述目标用户簇的有效性；
    基于所述目标用户簇有效的检测结果，在所述目标用户簇中依据所述样本用电户检测发生窃电行为的所述目标用电户。
  2. 根据权利要求1所述的方法，其中，所述从所述涌浪电流中提取原始电流特征，包括：
    将第一时间周期划分为多个第二时间周期；
    在每个所述第二时间周期内生成标记，所述标记表示是否接收到所述涌浪电流；
    在所述标记表示接收到所述涌浪电流的情况下，统计接收所述涌浪电流的频次；
    对所述涌浪电流分别统计最大值、最小值、平均值；
    在所述标记表示未接收到所述涌浪电流的情况下，将接收所述涌浪电流的频次设置为指定的数值；
    将所述涌浪电流的最大值、最小值、平均值均设置为指定的数值；
    将所述第二时间周期内的所述标记、所述频次、所述最大值、所述最小值、所述平均值组合为原始字符串；
    将所述第一时间周期内的所有所述原始字符串组成原始电流特征。
  3. 根据权利要求1所述的方法，其中，所述对所述原始电流特征进行主成分分析，以降维为目标电流特征，包括：
    将所述原始电流特征组合成第一电流矩阵；
    将所述第一电流矩阵中的每一行数据执行零均值化；
    在执行所述零均值化完成的情况下，对所述第一电流矩阵计算协方差矩阵；
    计算所述协方差矩阵的特征值与特征向量；
    按照所述特征值的大小对所述特征向量从上到下按行排列，并取前q行组成第二电流矩阵；
    计算所述第二电流矩阵与所述第一电流矩阵之间的乘积，获得降维之后的 目标电流特征。
  4. 根据权利要求1所述的方法，其中，所述针对每个所述K值，根据所述目标电流特征、所述样本电流特征对所述目标用电户、所述样本用电户进行K均值聚类，得到K个目标用户簇，包括：
    针对每个所述K值，初始化K个目标用户簇，每个所述目标用户簇中具有中心；
    计算组合电流特征与所述中心之间的距离，所述组合电流特征包括所述目标电流特征、所述样本电流特征；
    将组合用电户划入所述距离最小的所述目标用户簇中，所述组合用电户包括所述目标用电户、所述样本用电户；
    在每个所述目标用户簇中，计算所述组合用电户的所述组合电流特征计算平均值，以更新所述目标用户簇的中心；
    判断所述中心在更新时的变化幅度是否小于或等于第一阈值；基于所述中心在更新时的变化幅度小于或等于所述第一阈值的判断结果，确定所述目标用户簇收敛；基于所述中心在更新时的变化幅度大于所述第一阈值的判断结果，返回执行所述计算组合电流特征与所述中心之间的距离。
  5. 根据权利要求1-4中任一项所述的方法，其中，所述根据所述样本用电户的分布信息计算所述样本用户簇与所述目标用户簇之间的相似度，包括：
    在所述目标用户簇中统计属于各个所述窃电类型的所述样本用电户的总数量；
    将所述总数量最大的所述目标用户簇设置为归属所述窃电类型的特征用户 簇；
    针对指定的所述窃电类型，计算所述特征用户簇与所述样本用户簇之间在所述样本用电户上的重合程度，作为所述特征用户簇与所述样本用户簇之间的相似度。
  6. 根据权利要求5所述的方法，其中，所述针对指定的所述窃电类型，计算所述特征用户簇与所述样本用户簇之间在所述样本用电户上的重合程度，作为所述特征用户簇与所述样本用户簇之间的相似度，包括：
    统计所述特征用户簇中所有所述样本用电户的第一子数量；
    统计所述特征用户簇中属于指定的所述窃电类型的所述样本用电户的第二子数量；
    统计所述样本用户簇中属于指定的所述窃电类型的所述样本用电户的第三子数量；
    将所述第一子数量依次加上所述第三子数量并减去所述第二子数量，得到目标子数量；
    计算所述第二子数量与所述目标子数量之间的比值，作为所述特征用户簇与所述样本用户簇之间的相似度。
  7. 根据权利要求5所述的方法，其中，所述根据所述相似度与所述目标用电户的分布信息检测所述目标用户簇的有效性，包括：
    统计所述特征用户簇中的所述目标用电户占所有所述目标用电户的占比；
    在所述相似度大于或等于第二阈值、且所述占比小于或等于第三阈值的情况下，确定所述目标用户簇的有效；
    所述在所述目标用户簇中依据所述样本用电户检测发生窃电行为的所述目标用电户，包括：
    确定所述特征用户簇中的所述目标用电户发生归属所述窃电类型的窃电行为。
  8. 一种窃电行为的检测装置，包括：
    涌浪电流接收模块，设置为接收多个电表检测到的涌浪电流，所述电表安装在目标用电户的场所中；
    电流特征提取模块，设置为从所述涌浪电流中提取原始电流特征；
    电流特征降维模块，设置为对所述原始电流特征进行主成分分析，以降维为目标电流特征；
    电流特征获取模块，设置为获取格式与所述目标电流特征相同的样本电流特征，所述样本电流特征关联发生窃电行为的样本用电户；
    样本用户簇聚类模块，设置为根据所述样本电流特征将所述窃电行为的窃电类型相同的所述样本用电户划分至同一个样本用户簇；
    聚类参数设置模块，设置为设置多个K值；
    目标用户簇聚类模块，设置为针对每个所述K值，根据所述目标电流特征、所述样本电流特征对所述目标用电户、所述样本用电户进行K均值聚类，得到K个目标用户簇；
    相似度计算模块，设置为根据所述样本用电户的分布信息计算所述样本用户簇与所述目标用户簇之间的相似度；
    有效性检测模块，设置为根据所述相似度与所述目标用电户的分布信息检 测所述目标用户簇的有效性；
    窃电行为检测模块，设置为基于所述目标用户簇有效的检测结果，在所述目标用户簇中依据所述样本用电户检测发生窃电行为的所述目标用电户。
  9. 一种计算机设备，所述计算机设备包括：
    处理器；
    存储器，设置为存储程序，
    在所述程序被所述处理器执行时，所述处理器实现如权利要求1-7中任一项所述的窃电行为的检测方法。
  10. 一种计算机可读存储介质，所述计算机可读存储介质上存储计算机程序，所述计算机程序被处理器执行时实现如权利要求1-7中任一项所述的窃电行为的检测方法。
PCT/CN2022/136079 2021-12-17 2022-12-02 窃电行为的检测方法、装置、计算机设备和存储介质 WO2023109527A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111545671.2A CN113933556B (zh) 2021-12-17 2021-12-17 一种窃电行为的检测方法、装置、计算机设备和存储介质
CN202111545671.2 2021-12-17

Publications (1)

Publication Number Publication Date
WO2023109527A1 true WO2023109527A1 (zh) 2023-06-22

Family

ID=79289163

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/136079 WO2023109527A1 (zh) 2021-12-17 2022-12-02 窃电行为的检测方法、装置、计算机设备和存储介质

Country Status (2)

Country Link
CN (1) CN113933556B (zh)
WO (1) WO2023109527A1 (zh)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116701947A (zh) * 2023-08-02 2023-09-05 成都汉度科技有限公司 一种窃电行为检测方法及系统
CN116777121A (zh) * 2023-08-18 2023-09-19 武汉振铭科技发展有限公司 基于大数据的非法用电的稽查方法、存储介质及电子设备
CN116910596A (zh) * 2023-07-26 2023-10-20 江苏方天电力技术有限公司 一种基于改进dbscan聚类的用户窃电分析方法、设备及存储介质
CN117495109A (zh) * 2023-12-29 2024-02-02 国网山东省电力公司禹城市供电公司 一种基于深井网络的窃电用户识别系统
CN117853270A (zh) * 2024-01-15 2024-04-09 国网浙江省电力有限公司杭州供电公司 一种基于大数据分析的反窃电监控系统及方法
CN117929823A (zh) * 2024-03-21 2024-04-26 江苏西欧电子有限公司 一种基于物联网实现安全用电防窃电的智能型电能表

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113933556B (zh) * 2021-12-17 2022-04-08 广东电网有限责任公司东莞供电局 一种窃电行为的检测方法、装置、计算机设备和存储介质

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103942453A (zh) * 2014-05-07 2014-07-23 华北电力大学 一种针对非技术性损失的智能用电异常检测方法
WO2015015365A1 (en) * 2013-07-31 2015-02-05 Manohar Alfred An electricity theft deterrent system and a method thereof
CN107085153A (zh) * 2017-04-27 2017-08-22 深圳市欧瑞博电子有限公司 一种用电异常火灾预警方法及系统
CN109146705A (zh) * 2018-07-02 2019-01-04 昆明理工大学 一种用电特征指标降维与极限学习机算法进行窃电检测的方法
CN109615004A (zh) * 2018-12-07 2019-04-12 江苏瑞中数据股份有限公司 一种多源数据融合的防窃电预警方法
CN113128596A (zh) * 2021-04-21 2021-07-16 广东电网有限责任公司汕尾供电局 一种窃电检测方法、装置及计算机可读存储介质
CN113933556A (zh) * 2021-12-17 2022-01-14 广东电网有限责任公司东莞供电局 一种窃电行为的检测方法、装置、计算机设备和存储介质

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN203689675U (zh) * 2014-01-20 2014-07-02 国家电网公司 低压客户用电信息采集终端远程接控装置
CN106682079B (zh) * 2016-11-21 2020-06-05 云南电网有限责任公司电力科学研究院 一种基于聚类分析的用户用电行为检测方法
CN107169145B (zh) * 2017-06-19 2019-11-26 武汉大学 一种基于聚类算法的用户窃电严重等级检测的方法
CN207096324U (zh) * 2017-09-12 2018-03-13 国网四川省电力公司成都供电公司 一种电能计量监测用防窃电电路
CN108896804A (zh) * 2018-04-28 2018-11-27 广东电网有限责任公司 窃电检测方法、装置及电子设备
CN110288039B (zh) * 2019-06-29 2022-09-30 河南工业大学 基于用户用电负荷特征窃电检测方法
CN111340065B (zh) * 2020-02-08 2022-08-09 南京邮电大学 一种基于复杂用户行为分析的用户负荷窃电模型挖掘系统及方法
CN111525697B (zh) * 2020-05-09 2022-10-25 西安交通大学 基于电流监测与线路拓扑分析的中低压配电网防窃电方法和系统

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2015015365A1 (en) * 2013-07-31 2015-02-05 Manohar Alfred An electricity theft deterrent system and a method thereof
CN103942453A (zh) * 2014-05-07 2014-07-23 华北电力大学 一种针对非技术性损失的智能用电异常检测方法
CN107085153A (zh) * 2017-04-27 2017-08-22 深圳市欧瑞博电子有限公司 一种用电异常火灾预警方法及系统
CN109146705A (zh) * 2018-07-02 2019-01-04 昆明理工大学 一种用电特征指标降维与极限学习机算法进行窃电检测的方法
CN109615004A (zh) * 2018-12-07 2019-04-12 江苏瑞中数据股份有限公司 一种多源数据融合的防窃电预警方法
CN113128596A (zh) * 2021-04-21 2021-07-16 广东电网有限责任公司汕尾供电局 一种窃电检测方法、装置及计算机可读存储介质
CN113933556A (zh) * 2021-12-17 2022-01-14 广东电网有限责任公司东莞供电局 一种窃电行为的检测方法、装置、计算机设备和存储介质

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116910596A (zh) * 2023-07-26 2023-10-20 江苏方天电力技术有限公司 一种基于改进dbscan聚类的用户窃电分析方法、设备及存储介质
CN116701947A (zh) * 2023-08-02 2023-09-05 成都汉度科技有限公司 一种窃电行为检测方法及系统
CN116701947B (zh) * 2023-08-02 2023-11-03 成都汉度科技有限公司 一种窃电行为检测方法及系统
CN116777121A (zh) * 2023-08-18 2023-09-19 武汉振铭科技发展有限公司 基于大数据的非法用电的稽查方法、存储介质及电子设备
CN116777121B (zh) * 2023-08-18 2023-11-03 武汉振铭科技发展有限公司 基于大数据的非法用电的稽查方法、存储介质及电子设备
CN117495109A (zh) * 2023-12-29 2024-02-02 国网山东省电力公司禹城市供电公司 一种基于深井网络的窃电用户识别系统
CN117495109B (zh) * 2023-12-29 2024-03-22 国网山东省电力公司禹城市供电公司 一种基于神经网络的窃电用户识别系统
CN117853270A (zh) * 2024-01-15 2024-04-09 国网浙江省电力有限公司杭州供电公司 一种基于大数据分析的反窃电监控系统及方法
CN117929823A (zh) * 2024-03-21 2024-04-26 江苏西欧电子有限公司 一种基于物联网实现安全用电防窃电的智能型电能表
CN117929823B (zh) * 2024-03-21 2024-06-07 江苏西欧电子有限公司 一种基于物联网实现安全用电防窃电的智能型电能表

Also Published As

Publication number Publication date
CN113933556B (zh) 2022-04-08
CN113933556A (zh) 2022-01-14

Similar Documents

Publication Publication Date Title
WO2023109527A1 (zh) 窃电行为的检测方法、装置、计算机设备和存储介质
CN109740641A (zh) 一种基于主成分分析和k-means聚类的台区用户识别方法
CN109856299A (zh) 一种变压器在线监测差异化阈值动态设置方法、系统
EP3968040A1 (en) Abnormal electricity use recognition method and device, and computer readable storage medium
CN107328974B (zh) 一种窃电识别方法及装置
CN113125903A (zh) 线损异常检测方法、装置、设备及计算机可读存储介质
Therrien et al. Assessment of measurement-based phase identification methods
CN109934301B (zh) 一种电力负荷聚类分析方法、装置和设备
CN113189418B (zh) 一种基于电压数据的拓扑关系识别方法
CN110569876A (zh) 非侵入式负荷识别方法、装置及计算设备
Babu et al. Detecting unusual customer consumption profiles in power distribution systems—APSPDCL
CN109583679A (zh) 一种多算法融合的窃电疑似度分析方法
CN102324007A (zh) 基于数据挖掘的异常检测方法
CN109885597A (zh) 基于机器学习的用户分群处理方法、装置及电子终端
CN112381351A (zh) 一种基于奇异谱分析的用电行为变化检测方法及系统
KR102472573B1 (ko) 전기 이용 고객 데이터 군집 분석 방법
CN114091783A (zh) 一种企业用电预警方法、装置、计算机设备和存储介质
CN112801135B (zh) 一种基于特征量相关性的发电厂厂用电系统故障选线方法及设备
CN110610212A (zh) 一种配电网的变压器的故障分类方法和故障分类装置
CN116760175A (zh) 非侵入式负荷识别方法、装置、终端及存储介质
Frank et al. Extracting operating modes from building electrical load data
CN115081933B (zh) 一种基于改进谱聚类的低压用户拓扑构建方法和系统
CN110212503B (zh) 一种适用于小电阻接地系统的高阻接地保护方法
Saini et al. Pq events classification and detection-a survey
Pan et al. Study on intelligent anti–electricity stealing early-warning technology based on convolutional neural networks

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22906277

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE