Electricity stealing identification method, device and equipment based on multi-dimensional power data fusion
By using a multi-dimensional power data fusion method, combined with the physical characteristics of electricity theft and weighted processing, accurate identification of electricity theft behavior is achieved, solving the problem of high false negative rate in existing technologies and improving the accuracy and adaptability of electricity theft detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 国网河北省电力有限公司营销服务中心
- Filing Date
- 2026-01-15
- Publication Date
- 2026-05-15
AI Technical Summary
Existing methods for detecting electricity theft are ill-suited to the diverse and covert methods of electricity theft, resulting in a high rate of missed detections and an inability to accurately capture potential electricity theft activities.
By acquiring multi-dimensional power data of target users, combining it with the characteristics of electricity theft mechanisms, performing weighted processing and using a dedicated identification model, and integrating the probability scores of each electricity theft mechanism for judgment, comprehensive coverage and accurate identification of different electricity theft mechanisms can be achieved.
It significantly reduces the false alarm rate and the false alarm rate, improves the accuracy of electricity theft identification, adapts to the dynamic changes of different user groups and new electricity theft methods, and provides more reliable anti-electricity theft technology support.
Smart Images
Figure CN122045663A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power technology, and in particular to a method, apparatus and equipment for identifying electricity theft based on multi-dimensional power data fusion. Background Technology
[0002] Electricity theft not only seriously disrupts the stable order of the electricity market and causes significant economic losses, but also poses a huge hidden danger to electricity users and public safety.
[0003] With the continuous development of technology, electricity theft methods are rapidly upgrading towards intelligence, diversification, and concealment: from traditional physical methods such as undercurrent method, undervoltage method, phase shifting method, and differential expansion method that change the structure of electricity meters or lines, they are gradually evolving into high-tech electricity theft modes such as strong magnetic interference, high-power wireless interference, and remote control, which greatly increases the difficulty and challenge for power companies in combating electricity theft.
[0004] Current anti-electricity theft measures mostly rely on electricity information collection systems to obtain real-time data such as user electricity consumption, maximum line loss rate, and transformer area line loss rate. Common detection paths are mainly divided into three categories: First, methods based on power grid state analysis, which calculate electrical quantity deviations such as power balance, voltage and current residuals, and line impedance changes in the distribution network, and combine them with state estimation, Kalman filtering, and other algorithms to determine whether there are non-technical energy loss anomalies caused by electricity theft; Second, methods based on machine learning, which use algorithms such as neural networks and temporal convolutional networks to mine the temporal patterns and group differences in user electricity consumption behavior and identify abnormal features that deviate from the normal pattern; Third, hardware-based methods, which develop anti-tamper smart meters, feeder terminal devices, remote inspection sensors, and other equipment to physically block traditional electricity theft operations such as short-circuit current and unauthorized meter tampering, or to capture traces of electricity theft such as strong magnetic interference and poor voltage circuit contact.
[0005] However, with the increasing variety of electricity theft methods, existing detection methods are difficult to adapt to and are prone to missed detections. Therefore, a new method is urgently needed to accurately detect potential electricity theft and provide efficient support for anti-electricity theft efforts. Summary of the Invention
[0006] This invention provides a method, apparatus, and device for identifying electricity theft based on multi-dimensional power data fusion, in order to solve the problem of missed detection in existing electricity theft detection methods.
[0007] In a first aspect, embodiments of the present invention provide a method for identifying electricity theft based on multi-dimensional power data fusion, comprising: Obtain multi-dimensional power data of the target user during a preset time period; Based on the preset feature dimensions corresponding to different electricity theft mechanisms, at least one dimension of power data corresponding to each electricity theft mechanism is extracted from the multi-dimensional power data; For each theft mechanism, the power data of at least one dimension is weighted and input into the identification model corresponding to the theft mechanism to obtain the probability score of the theft mechanism. The electricity theft identification result is determined based on the probability scores of all electricity theft mechanisms.
[0008] In one possible implementation, the weighting of at least one dimension of power data for each electricity theft mechanism includes: Obtain the preset first weight for each feature dimension; Based on the power data for each feature dimension, determine the second weight for each feature dimension; The first and second weights of each feature dimension are fused to obtain the comprehensive weight of each feature dimension, and a weighted feature vector is obtained based on the comprehensive weight.
[0009] In one possible implementation, obtaining the preset first weight for each feature dimension includes: Cluster all users based on electricity consumption behavior data, and determine the group to which the target user belongs based on the clustering results; Within the target user's group, the discriminative power of each feature dimension is calculated using the entropy weight method, and the first weight of each feature dimension is determined based on the discriminative power.
[0010] In one possible implementation, determining the second weight for each feature dimension based on the power data for each feature dimension includes: Obtain the historical power feature baseline for each feature dimension; Calculate the deviation between the power data and the historical power characteristic baseline; Based on the deviation, a second weight is determined for each feature dimension.
[0011] In one possible implementation, determining the electricity theft identification result based on the probability scores of all electricity theft mechanisms includes: If the probability score of any electricity theft mechanism is greater than or equal to the first preset threshold, then it is determined that an electricity theft event has occurred using that mechanism. If the probability scores of all electricity theft mechanisms are less than the first preset threshold, then the comprehensive electricity theft probability is determined according to the preset scenario weights corresponding to each electricity theft mechanism; if the comprehensive electricity theft probability is greater than the second preset threshold, then it is determined that an electricity theft event has occurred. Wherein, the second preset threshold is less than the first preset threshold.
[0012] In one possible implementation, before extracting at least one dimension of power data corresponding to each theft mechanism from the multi-dimensional power data according to the preset feature dimensions corresponding to different theft mechanisms, the method further includes: By analyzing the relationship between the tampering parameters of each electricity theft mechanism and the changes in power data in each dimension, the correlation coefficient between each electricity theft mechanism and each feature dimension is determined. Based on the correlation coefficient, the feature dimension corresponding to each theft mechanism is determined.
[0013] In one possible implementation, the feature dimensions include: peak active power, average active power, valley active power, total reactive power, total active power, and power factor.
[0014] In one possible implementation, the recognition model is a pre-trained neural network model.
[0015] Secondly, embodiments of the present invention provide a power theft detection device based on multi-dimensional power data fusion, comprising: The acquisition module is used to acquire multi-dimensional power data of the target user; The extraction module is used to extract at least one dimension of power data corresponding to each theft mechanism from the multi-dimensional power data based on the preset feature dimensions corresponding to different theft mechanisms. The calculation module is used to weight the power data of at least one dimension for each theft mechanism and input it into the identification model corresponding to the theft mechanism to obtain the probability score of the theft mechanism. The determination module is used to determine the electricity theft identification result based on the probability scores of all electricity theft mechanisms.
[0016] Thirdly, embodiments of the present invention provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect or any possible implementation thereof.
[0017] The beneficial effects of this invention compared to the prior art are as follows: In this embodiment of the invention, by accurately acquiring multi-dimensional power data of target users, comprehensive coverage of abnormal characteristics of different electricity theft mechanisms is achieved, avoiding the limitation that a single power parameter is insufficient to capture complex electricity theft behaviors. It extracts corresponding feature dimensions according to preset electricity theft mechanisms, combines weighted processing to highlight the contribution of core abnormal features and weaken redundant information interference, and then adapts a dedicated recognition model to the characteristics of each electricity theft mechanism, significantly improving the accuracy of single-mechanism electricity theft identification. Finally, based on the probability scores of all electricity theft mechanisms, a fusion judgment is made, significantly reducing the false negative and false positive rates. Simultaneously, it adapts to the differences in electricity consumption among different user groups such as industrial and residential users, as well as the dynamic changes in new electricity theft methods, enhancing the generalization ability and engineering practicality of the method, and providing more reliable and efficient technical support for power inspection. Attached Figure Description
[0018] Figure 1 This is a flowchart illustrating the implementation of the electricity theft identification method based on multi-dimensional power data fusion provided in this embodiment of the invention. Figure 2 This is a schematic diagram of the structure of the electricity theft identification device based on multi-dimensional power data fusion provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation
[0019] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.
[0020] To illustrate the technical solution described in this invention, specific embodiments are described below.
[0021] See Figure 1 The document illustrates a flowchart of the implementation of the electricity theft identification method based on multi-dimensional power data fusion provided in an embodiment of the present invention, which is described in detail below: Step S101: Obtain multi-dimensional power data of the target user during a preset time period.
[0022] Here, the preset time period is a collection window defined in advance to capture the time characteristics of electricity theft. It needs to be flexibly set according to the duration of electricity theft, such as one day, one week, one month, etc.
[0023] Multi-dimensional power data mainly comes from smart meters, advanced metering systems, or electricity information collection systems. These systems have already achieved automated collection and storage of user power data, requiring no additional hardware. After the raw data is collected, simple preprocessing is required to avoid interference from invalid data, including: missing value imputation, outlier removal, and unit unification.
[0024] For example, multi-dimensional power data includes, but is not limited to: (1) Total active power Definition: The total power (unit: kW) of the electrical energy actually consumed by the user within a preset time period is the core basis for electricity billing.
[0025] Data acquisition logic: The smart meter collects the instantaneous active power in real time, accumulates and calculates the total power according to the preset time period, and then converts it into the average total power (or directly collects the cumulative total active power within the time period).
[0026] Physical meaning: Electricity theft (such as undercurrent or undervoltage) will directly cause its abnormal decrease, which is a common sensitive parameter for all electricity theft mechanisms.
[0027] (2) Peak active power Definition: The maximum instantaneous active power within a preset time period (unit: kW).
[0028] Data collection logic: Iterate through all instantaneous active power data within the time period and take the maximum value.
[0029] Physical meaning: It reflects the load intensity of users' peak electricity consumption. When there is insufficient current or insufficient voltage, the peak value will be significantly lower than the normal electricity consumption baseline (e.g., the peak power of industrial users drops sharply from 50kW to 20kW).
[0030] (3) Peak active power Definition: The minimum instantaneous active power within a preset time period (unit: kW, usually ≥0).
[0031] Data collection logic: Iterate through all instantaneous active power data within the time period and take the minimum value (removing abnormal zero / negative values).
[0032] Physical meaning: It reflects the load intensity during the off-peak hours of a user's electricity consumption. When using the differential method or strong magnetic field to steal electricity, the off-peak value will fluctuate irregularly (e.g., the normal off-peak value is stable at 5kW, but after electricity theft, it fluctuates between 0-3kW).
[0033] (4) Average active power Definition: The average value of active power within a preset time period (unit: kW).
[0034] Data collection logic: Total active power ÷ duration of the time period (e.g., total power in 24 hours ÷ 24), or directly calculate the arithmetic mean of all instantaneous active power within the time period.
[0035] Physical meaning: It reflects the average load level of users' electricity consumption. When there is undervoltage or electricity theft by differential method, the average value will be consistently low or the fluctuation coefficient will increase.
[0036] (5) Total reactive power Definition: The total reactive power (unit: kVar) that maintains equipment operation during a preset time period without actual energy consumption during user electricity use.
[0037] Data Acquisition Logic: Consistent with the total active power acquisition logic, the smart meter collects instantaneous reactive power and then accumulates and converts it.
[0038] Physical significance: The core sensitive parameter of the phase shifting method of electricity theft. The phase misalignment will cause its abnormal increase (e.g., the normal total reactive power = 10kVar, which rises to 30kVar after electricity theft).
[0039] (6) Power factor Definition: The ratio of active power to apparent power, ranging from 0 to 1 (the closer to 1, the higher the power efficiency).
[0040] Data acquisition logic: Calculate based on instantaneous values, and then take the average power factor over the time period.
[0041] Physical significance: Key parameters for phase-shifting methods and strong magnetic field electricity theft will decrease significantly during electricity theft.
[0042] Step S102: Based on the preset feature dimensions corresponding to different electricity theft mechanisms, extract at least one dimension of power data corresponding to each electricity theft mechanism from the multi-dimensional power data.
[0043] This embodiment targets common electricity theft methods (such as undercurrent method, undervoltage method, phase shift method, differential expansion method, meterless method, and strong magnetic method). By testing and analyzing the relationship between the tampering parameters of each electricity theft mechanism and the changes in power data in each dimension, the correlation coefficient between each electricity theft mechanism and each feature dimension is determined. Based on the correlation coefficient, the feature dimension corresponding to each electricity theft mechanism is determined (for each electricity theft mechanism, the top N feature dimensions with the largest correlation coefficient are selected).
[0044] For example, the feature dimensions corresponding to each mechanism of electricity theft are shown in Table 1: Table 1
[0045] Step S103: For each theft mechanism, weight the power data of at least one dimension and input it into the identification model corresponding to the theft mechanism to obtain the probability score of the theft mechanism.
[0046] This embodiment focuses on the most sensitive power parameter for each physical mechanism of electricity theft. By weighting and amplifying the contribution of this parameter to electricity theft characteristics, a model specifically adapted to the mechanism is used to calculate the probability score of a user's electricity theft under that mechanism. Here, the identification model can be a pre-trained neural network model or other algorithm models, and this application does not limit it.
[0047] In one possible implementation, for each electricity theft mechanism, its power data in at least one dimension is weighted, including: Step S1031: Obtain the preset first weight for each feature dimension.
[0048] First, cluster all users based on their electricity consumption behavior data. Then, determine the target user's group based on the clustering results. For example, using the k-means clustering algorithm, users can be divided into typical groups such as industrial, residential, and commercial.
[0049] Then, within the target user's group, the discriminative power (dispersion) of each feature dimension is calculated using the entropy weight method, and the first weight of each feature dimension is determined based on the discriminative power. For example, for at least one dimension of power data mentioned above, based on the discriminative power within the group, features with high discriminative power are assigned high initial weights, and the remaining features are assigned weights in decreasing order of discriminative power, with a total weight of 1.
[0050] Step S1032: Determine the second weight of each feature dimension based on the power data of each feature dimension.
[0051] Obtain the historical power characteristic baseline for each feature dimension; calculate the deviation between the power data and the historical power characteristic baseline; based on the deviation, determine the second weight for each feature dimension. This step is based on the individual user's electricity consumption habits, calculating the degree to which the real-time power characteristics deviate from their normal state, and converting this deviation into an individual-specific second weight.
[0052] The historical power characteristic baseline serves as a normal reference standard for a user's power characteristic dimension under normal electricity consumption conditions. Deviation is a quantified value of how far a user's current (or recent) power characteristic value deviates from its historical baseline: a larger deviation indicates a more abnormal current performance in that characteristic dimension, and is more likely to be caused by electricity theft. The second weight is the individual-level feature importance allocation: the higher the deviation of a feature dimension (the more abnormal the current performance), the greater the contribution of that feature to identifying whether the user is stealing electricity, and therefore it is assigned a higher weight; conversely, features with low deviation (consistent with normal habits) have lower weights (to avoid interfering with the identification results). The essence of this method is to create a profile of the user's normal electricity consumption, then examine the degree to which the current electricity consumption data deviates from this profile, and finally convert the degree of deviation into feature weights. This allows the model to focus more on the user's most abnormal power characteristics when identifying electricity theft, thereby improving the accuracy of individual electricity theft identification and avoiding misjudging normal fluctuations consistent with the user's habits as electricity theft.
[0053] Step S1033: The first weight and the second weight of each feature dimension are fused to obtain the comprehensive weight of each feature dimension, and the weighted feature vector is obtained based on the comprehensive weight.
[0054] Here, the fusion coefficient can be set empirically. After weighting the power data in at least one dimension, a weighted power feature vector is formed. Essentially, this is a new combination of features created by assigning importance weights to the original power data, allowing the model to focus more on power features sensitive to electricity theft. Ultimately, the set of multiple weighted features constitutes the input data that the model can directly identify. After weighting, abnormal power features caused by electricity theft are amplified, while normal fluctuations are suppressed, allowing the model to more accurately identify electricity theft.
[0055] Step S104: Determine the electricity theft identification result based on the probability scores of all electricity theft mechanisms.
[0056] In this embodiment, if the probability score of any electricity theft mechanism is greater than or equal to a first preset threshold, an electricity theft event is determined to have occurred using that mechanism. If the probability scores of all electricity theft mechanisms are less than the first preset threshold, a comprehensive electricity theft probability is determined based on the preset scenario weights corresponding to each mechanism. If the comprehensive electricity theft probability is greater than a second preset threshold, an electricity theft event is determined to have occurred. The second preset threshold is less than the first preset threshold. The second threshold covers cases where multiple mechanisms are slightly suspected (each individual score is not high enough, but the combined score is significant) or cases of compound electricity theft (using two mechanisms simultaneously, where individual scores are not high but the combined score is sufficient), thus avoiding missed detections.
[0057] In this embodiment of the invention, by accurately acquiring multi-dimensional power data of target users, comprehensive coverage of abnormal characteristics of different electricity theft mechanisms is achieved, avoiding the limitation that a single power parameter is insufficient to capture complex electricity theft behaviors. It extracts corresponding feature dimensions according to preset electricity theft mechanisms, combines weighted processing to highlight the contribution of core abnormal features and weaken redundant information interference, and then adapts a dedicated recognition model to the characteristics of each electricity theft mechanism, significantly improving the accuracy of single-mechanism electricity theft identification. Finally, based on the probability scores of all electricity theft mechanisms, a fusion judgment is made, significantly reducing the false negative and false positive rates. Simultaneously, it adapts to the differences in electricity consumption among different user groups such as industrial and residential users, as well as the dynamic changes in new electricity theft methods, enhancing the generalization ability and engineering practicality of the method, and providing more reliable and efficient technical support for power inspection.
[0058] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
[0059] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.
[0060] Figure 2 The diagram shows a schematic of the structure of the electricity theft identification device based on multi-dimensional power data fusion provided in an embodiment of the present invention. For ease of explanation, only the parts related to the embodiment of the present invention are shown.
[0061] like Figure 2 As shown, the electricity theft detection device 2 based on multi-dimensional power data fusion includes: Module 21 is used to acquire multi-dimensional power data of the target user; Extraction module 22 is used to extract at least one dimension of power data corresponding to each theft mechanism from the multi-dimensional power data according to the preset feature dimensions corresponding to different theft mechanisms. The calculation module 23 is used to weight the power data of at least one dimension of each theft mechanism and input it into the identification model corresponding to the theft mechanism to obtain the probability score of the theft mechanism. The determination module 24 is used to determine the electricity theft identification result based on the probability scores of all electricity theft mechanisms.
[0062] In one possible implementation, the computing module 23 is used for: Obtain the preset first weight for each feature dimension; Based on the power data for each feature dimension, determine the second weight for each feature dimension; The first and second weights of each feature dimension are fused to obtain the comprehensive weight of each feature dimension, and a weighted feature vector is obtained based on the comprehensive weight.
[0063] In one possible implementation, the computing module 23 is used for: Cluster all users based on electricity consumption behavior data, and determine the group to which the target user belongs based on the clustering results; Within the target user's group, the discriminative power of each feature dimension is calculated using the entropy weight method, and the first weight of each feature dimension is determined based on the discriminative power.
[0064] In one possible implementation, the computing module 23 is used for: Obtain the historical power feature baseline for each feature dimension; Calculate the deviation between the power data and the historical power characteristic baseline; Based on the deviation, a second weight is determined for each feature dimension.
[0065] In one possible implementation, the determining module 24 is used to: If the probability score of any electricity theft mechanism is greater than or equal to the first preset threshold, then it is determined that an electricity theft event has occurred using that mechanism. If the probability scores of all electricity theft mechanisms are less than the first preset threshold, then the comprehensive electricity theft probability is determined according to the preset scenario weights corresponding to each electricity theft mechanism; if the comprehensive electricity theft probability is greater than the second preset threshold, then it is determined that an electricity theft event has occurred. Wherein, the second preset threshold is less than the first preset threshold.
[0066] In one possible implementation, before extracting at least one dimension of power data corresponding to each theft mechanism from the multi-dimensional power data according to the preset feature dimensions corresponding to different theft mechanisms, the extraction module 22 is further configured to: By analyzing the relationship between the tampering parameters of each electricity theft mechanism and the changes in power data in each dimension, the correlation coefficient between each electricity theft mechanism and each feature dimension is determined. Based on the correlation coefficient, the feature dimension corresponding to each theft mechanism is determined.
[0067] In one possible implementation, the feature dimensions include: peak active power, average active power, valley active power, total reactive power, total active power, and power factor.
[0068] In one possible implementation, the recognition model is a pre-trained neural network model.
[0069] In this embodiment of the invention, by accurately acquiring multi-dimensional power data of target users, comprehensive coverage of abnormal characteristics of different electricity theft mechanisms is achieved, avoiding the limitation that a single power parameter is insufficient to capture complex electricity theft behaviors. It extracts corresponding feature dimensions according to preset electricity theft mechanisms, combines weighted processing to highlight the contribution of core abnormal features and weaken redundant information interference, and then adapts a dedicated recognition model to the characteristics of each electricity theft mechanism, significantly improving the accuracy of single-mechanism electricity theft identification. Finally, based on the probability scores of all electricity theft mechanisms, a fusion judgment is made, significantly reducing the false negative and false positive rates. Simultaneously, it adapts to the differences in electricity consumption among different user groups such as industrial and residential users, as well as the dynamic changes in new electricity theft methods, enhancing the generalization ability and engineering practicality of the method, and providing more reliable and efficient technical support for power inspection.
[0070] Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. For example... Figure 3 As shown, the electronic device 3 in this embodiment includes a processor 30 and a memory 31. The memory 31 stores a computer program 32. When the processor 30 executes the computer program 32, it implements the steps in the various method embodiments described above. Alternatively, when the processor 30 executes the computer program 32, it implements the functions of each module in the various device embodiments described above.
[0071] For example, computer program 32 may be divided into one or more modules / units, which are stored in memory 31 and executed by processor 30 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 32 in electronic device 3.
[0072] Electronic device 3 may include, but is not limited to, processor 30 and memory 31. Those skilled in the art will understand that... Figure 3 This is merely an example of electronic device 3 and does not constitute a limitation on electronic device 3. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device 3 may also include input / output devices, network access devices, buses, etc.
[0073] For the sake of simplicity and clarity, only the above-described functional modules / units are used as examples. In practical applications, the functions described above can be assigned to different functional modules / units as needed. These modules / units can be implemented in hardware, software, or a combination of both.
[0074] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not detailed or described in a particular embodiment can be referred to in the relevant descriptions of other embodiments. Unless otherwise specified or in conflict with logic, the terminology and / or descriptions between different embodiments are consistent and can be referenced interchangeably. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.
[0075] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.
Claims
1. A method for identifying electricity theft based on multi-dimensional power data fusion, characterized in that, include: Obtain multi-dimensional power data of the target user during a preset time period; Based on the preset feature dimensions corresponding to different electricity theft mechanisms, at least one dimension of power data corresponding to each electricity theft mechanism is extracted from the multi-dimensional power data; For each theft mechanism, the power data of at least one dimension is weighted and input into the identification model corresponding to the theft mechanism to obtain the probability score of the theft mechanism. The electricity theft identification result is determined based on the probability scores of all electricity theft mechanisms.
2. The electricity theft identification method based on multi-dimensional power data fusion according to claim 1, characterized in that, For each electricity theft mechanism, the power data of at least one dimension is weighted, including: Obtain the preset first weight for each feature dimension; Based on the power data for each feature dimension, determine the second weight for each feature dimension; The first and second weights of each feature dimension are fused to obtain the comprehensive weight of each feature dimension, and a weighted feature vector is obtained based on the comprehensive weight.
3. The electricity theft identification method based on multi-dimensional power data fusion according to claim 2, characterized in that, The step of obtaining the preset first weight for each feature dimension includes: Cluster all users based on electricity consumption behavior data, and determine the group to which the target user belongs based on the clustering results; Within the target user's group, the discriminative power of each feature dimension is calculated using the entropy weight method, and the first weight of each feature dimension is determined based on the discriminative power.
4. The electricity theft identification method based on multi-dimensional power data fusion according to claim 1, characterized in that, The step of determining the second weight for each feature dimension based on the power data for each feature dimension includes: Obtain the historical power feature baseline for each feature dimension; Calculate the deviation between the power data and the historical power characteristic baseline; Based on the deviation, a second weight is determined for each feature dimension.
5. The electricity theft identification method based on multi-dimensional power data fusion according to any one of claims 1 to 4, characterized in that, The determination of the electricity theft identification result based on the probability scores of all electricity theft mechanisms includes: If the probability score of any electricity theft mechanism is greater than or equal to the first preset threshold, then it is determined that an electricity theft event has occurred using that mechanism. If the probability scores of all electricity theft mechanisms are less than the first preset threshold, then the comprehensive electricity theft probability is determined according to the preset scenario weights corresponding to each electricity theft mechanism; if the comprehensive electricity theft probability is greater than the second preset threshold, then it is determined that an electricity theft event has occurred. Wherein, the second preset threshold is less than the first preset threshold.
6. The electricity theft identification method based on multi-dimensional power data fusion according to any one of claims 1 to 4, characterized in that, Before extracting at least one dimension of power data corresponding to each theft mechanism from the multi-dimensional power data according to the preset feature dimensions corresponding to different theft mechanisms, the method further includes: By analyzing the relationship between the tampering parameters of each electricity theft mechanism and the changes in power data in each dimension, the correlation coefficient between each electricity theft mechanism and each feature dimension is determined. Based on the correlation coefficient, the feature dimension corresponding to each theft mechanism is determined.
7. The electricity theft identification method based on multi-dimensional power data fusion according to any one of claims 1 to 4, characterized in that, The characteristic dimensions include: peak active power, average active power, valley active power, total reactive power, total active power, and power factor.
8. The electricity theft identification method based on multi-dimensional power data fusion according to any one of claims 1 to 4, characterized in that, The recognition model is a pre-trained neural network model.
9. A device for identifying electricity theft based on multi-dimensional power data fusion, characterized in that, include: The acquisition module is used to acquire multi-dimensional power data of the target user during a preset time period; The extraction module is used to extract at least one dimension of power data corresponding to each theft mechanism from the multi-dimensional power data based on the preset feature dimensions corresponding to different theft mechanisms. The calculation module is used to weight the power data of at least one dimension for each theft mechanism and input it into the identification model corresponding to the theft mechanism to obtain the probability score of the theft mechanism. The determination module is used to determine the electricity theft identification result based on the probability scores of all electricity theft mechanisms.
10. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 8.