A fusion type electric energy meter electricity larceny detection system and method

Abstract

The present application relates to the technical field of electric power metering, in particular to a kind of electric energy meter electricity larceny detection system and method of fusion, including data acquisition module, noise feature extraction module, data fusion processing module, evidence storage module and communication data transceiver module, each module cooperates to constitute complete electricity larceny detection and evidence closed loop.The method comprises: obtaining the original data of electric meter voltage and environmental magnetic field intensity by data acquisition module, and constructing normal noise feature sample set in the initial stage of grid connection;Multi-dimensional noise feature vector is extracted from voltage data, and the deviation degree of real-time feature and normal sample is calculated by Mahalanobis distance, combined with dynamically updated adaptive threshold and magnetic field intensity to judge electricity larceny;Automatic retention electricity larceny related original waveform data, respond to master station data query and evidence upload request;The present application improves the reliability and robustness of magnetic interference type electricity larceny detection, effectively reduces the false alarm and miss rate, and adapts to different operating environments and long-term use scenarios of electric meter.

Description

Technical Field

[0001] This invention belongs to the field of power metering technology, specifically relating to an integrated electricity meter theft detection system and method. Background Technology

[0002] In the field of electricity metering, magnetic interference is a common method of electricity theft. Thieves interfere with the normal meter readings by applying external magnetic fields, causing distorted meter data and resulting in economic losses for power companies. Therefore, developing reliable and accurate detection technology for magnetic interference has become a crucial requirement in the field of electricity meter anti-theft. Currently, the industry mainly relies on two conventional solutions for preventing magnetic interference-based electricity theft:

[0003] (1) Single magnetic field detection technology, the core of which is to install a magnetic sensitive element inside the meter, pre-set a fixed magnetic field strength threshold, and detect the ambient magnetic field strength in real time through the magnetic sensitive element. When the detected value exceeds the preset threshold, the electricity theft alarm is triggered directly. However, this technology has obvious defects. Legitimate strong magnetic field sources in the environment, such as high-power audio equipment, microwave ovens, and motors that are starting, are very likely to trigger false alarms, reduce the reliability of the alarm and increase the workload of maintenance personnel in ineffective investigation. Moreover, electricity thieves can avoid detection by long-term interference with weak magnetic fields or by shielding the magnetic sensitive element, resulting in missed alarms.

[0004] (2) Single electrical quantity analysis technology: By collecting grid voltage and current signals in real time, it analyzes characteristic parameters such as waveform distortion, phase shift or harmonic content, and uses abnormal parameters as the basis for judging electricity theft. However, this technology is not sensitive enough to nonlinear, slowly changing or carefully designed magnetic interference electricity theft behavior. The changes in electrical quantity parameters are weak and difficult to capture, and it is easy to miss the report. At the same time, fluctuations such as normal load switching of the grid will cause temporary changes in electrical quantity parameters. This technology is difficult to distinguish between normal fluctuations and abnormal electricity theft, has poor environmental adaptability, and has a high risk of false alarm.

[0005] In view of this, the present invention is hereby proposed. Summary of the Invention

[0006] In order to solve the above-mentioned technical problems in the prior art, the present invention provides an integrated electricity meter theft detection system and method, which solves the problems of low reliability, high false alarm rate, easy circumvention, or insufficient ability to identify specific electricity theft behaviors that are common in existing magnetic interference-based electricity theft prevention technologies.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows:

[0008] Firstly, an integrated electricity meter theft detection system includes:

[0009] The data acquisition module is used to collect the raw voltage data of the electricity meter and the magnetic field strength data of the environment in which the electricity meter is located;

[0010] A noise feature extraction module, connected to the data acquisition module, is used to extract noise feature vectors from the raw voltage data;

[0011] The data fusion processing module is connected to the noise feature extraction module and the data acquisition module, and is used to perform fusion judgment based on the noise feature vector and magnetic field strength data, and output the electricity theft judgment result.

[0012] The evidence collection and storage module is connected to the data fusion and processing module and is used to save relevant original data as evidence when the electricity theft judgment result is that electricity theft exists.

[0013] The communication data transceiver module is connected to the data fusion processing module and the evidence collection and storage module, and is used to interact with the main station to transmit the electricity theft judgment results and the stored evidence data.

[0014] Furthermore, the data acquisition module includes a voltage sampling circuit and a magnetic field detection element;

[0015] The voltage sampling circuit collects raw voltage data at a preset sampling rate. After the meter is connected to the grid, within a preset time period, the data acquisition module collects noise feature vectors at a preset time window and caches them in RAM to construct a normal noise feature sample set.

[0016] The magnetic field detection element is used to detect the strength of the ambient magnetic field.

[0017] Furthermore, the noise feature vector is a five-dimensional feature vector, which includes: noise kurtosis, noise variance, and the mean of the three power spectral density bands.

[0018] Furthermore, the noise kurtosis The calculation formula is:

[0019]

[0020] in, For noise kurtosis, The length of the noise sequence. For noise sequence values, The noise mean is... This represents the noise standard deviation.

[0021] Furthermore, the process of obtaining the noise variance is as follows: the fundamental component is extracted from the original voltage data using a fast Fourier transform; the fundamental component is subtracted from the original voltage data to obtain the noise signal; the variance of the noise sequence composed of the noise signal within a preset time window is calculated to obtain the noise variance. The specific calculation formula is as follows:

[0022]

[0023] in, The length of the noise sequence. For noise sequence values, This represents the noise mean.

[0024] Furthermore, the process of obtaining the three power spectral density bands is as follows:

[0025] A fast Fourier transform is performed on the noise signal within a preset time window, and the power spectral density is calculated. The power spectral density is divided into a low-frequency band of 10~100Hz, a mid-frequency band of 100~500Hz, and a high-frequency band of 500~1000Hz. The mean value of the power spectral density of each frequency band is calculated to obtain the mean value of the three power spectral density bands.

[0026] Furthermore, the data fusion processing module includes:

[0027] The Mahalanobis distance calculation unit is used to calculate the Mahalanobis distance between the real-time noise feature vector and the normal noise feature sample set.

[0028] An adaptive threshold calculation unit is used to dynamically update the judgment threshold based on historical Mahalanobis distance data;

[0029] The magnetic field detection unit is used to receive the magnetic field strength data of the magnetic field detection element and determine whether the magnetic field strength exceeds a preset magnetic field threshold of 300mT.

[0030] Furthermore, the formula for calculating the Mahalanobis distance is:

[0031]

[0032] in, The Mahalanobis distance, This is a real-time noise feature vector. This is the mean vector of the set of normal noise feature samples. It is the inverse of the covariance matrix. This represents the matrix transpose operation.

[0033] Furthermore, the adaptive threshold calculation unit calculates the mean of all Mahalanobis distances over the past 3 days, with a preset update cycle of 3 days. and standard deviation A new judgment threshold is calculated based on the mean and standard deviation. The calculation formula is:

[0034] .

[0035] Furthermore, the fusion judgment logic of the data fusion processing module includes:

[0036] When the Mahalanobis distance exceeds the judgment threshold and the magnetic field strength exceeds the preset magnetic field threshold of 300mT, the suspected electricity theft sign is set and a 5-minute timer is started.

[0037] If the Mahalanobis distance continuously exceeds the judgment threshold and the magnetic field strength continuously exceeds the preset magnetic field threshold during the timing period, the electricity theft flag is set and the judgment result of electricity theft is output; otherwise, the electricity theft suspicion flag is cleared.

[0038] The evidence collection and storage module stores relevant raw data including voltage, current and power data within 20 seconds before and 60 seconds after the electricity theft incident. The relevant raw data is transferred from random access memory to data flash memory for storage.

[0039] The communication data transceiver module responds to the master station's instructions and uploads the electricity theft flag and related raw data.

[0040] Secondly, an integrated method for detecting electricity theft from electricity meters includes:

[0041] S1. Data Acquisition: Acquire raw voltage data of the meter and magnetic field strength data of the environment where the meter is located;

[0042] S2. Noise Feature Extraction: Extract noise feature vectors from the raw voltage data;

[0043] S3. Data fusion processing: Based on the noise feature vector and magnetic field strength data, perform fusion judgment and output the electricity theft judgment result;

[0044] S4. Evidence Collection and Preservation: If the electricity theft determination result indicates that electricity theft exists, the relevant original data shall be preserved as evidence.

[0045] S5. Communication Interaction: Interact with the main station to transmit the electricity theft judgment results and the saved evidence data.

[0046] Compared with existing technologies, the integrated electricity meter theft detection system and method provided by this invention includes a data acquisition module, a noise feature extraction module, a data fusion processing module, an evidence collection and storage module, and a communication data transceiver module. These modules work together to form a complete closed loop for electricity theft detection and evidence collection. The detection method is implemented based on this system: the data acquisition module acquires raw meter voltage data and environmental magnetic field strength, constructing a normal noise feature sample set during the initial grid connection phase; the noise feature extraction module extracts multi-dimensional noise feature vectors from the voltage data; the data fusion processing module calculates the deviation between real-time features and normal samples using Mahalanobis distance, and performs fusion judgment based on dynamically updated adaptive thresholds and magnetic field strength. After a delayed secondary confirmation of persistent abnormalities, electricity theft is determined; the evidence collection and storage module automatically stores raw waveform data related to electricity theft; and the communication module responds to data queries and evidence upload requests from the main station. This system significantly improves the reliability and robustness of magnetic interference-based electricity theft detection, effectively reduces false alarm and false negative rates, and is adaptable to different operating environments and long-term meter usage scenarios. Attached Figure Description

[0047] Figure 1 This is an architecture diagram of an electricity meter theft detection system provided in an embodiment of the present invention;

[0048] Figure 2 A flowchart of an electricity meter theft detection method provided in an embodiment of the present invention;

[0049] Figure 3 This is an architecture diagram of the data fusion processing module provided in an embodiment of the present invention;

[0050] Figure 4 This is a schematic diagram of the fusion judgment logic provided in an embodiment of the present invention. Detailed Implementation

[0051] The technical solution of the present invention will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are not all embodiments of the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.

[0052] It should be noted that, unless otherwise specifically stated, the relative arrangement and numerical expressions of the components and steps described in these embodiments should not be construed as limiting the scope of the invention.

[0053] The following description of exemplary embodiments is merely illustrative and is not intended to limit the invention or its application or use in any way. Techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail herein, but where applicable, such techniques, methods, and apparatus should be considered part of this specification.

[0054] Terminology Explanation:

[0055] Data Flash: A type of non-volatile memory with fast read and write speeds and large capacity, but it needs to be erased page by page (usually 4096 bytes) before writing.

[0056] Random Access Memory (RAM): Used as a high-speed cache for chips, with access speeds much faster than DataFlash, but it is volatile memory and data will be lost when power is off.

[0057] Mahalanobis distance (MD) is a method for measuring the "distance" between multidimensional data points. It takes into account the correlation between dimensions and the covariance structure of the data distribution and is used in fields such as anomaly detection and pattern recognition.

[0058] Fast Fourier Transform (FFT): An efficient algorithm for computing the Discrete Fourier Transform (DFT), converting time-domain signals into frequency-domain signals. It is a core tool in applications such as signal analysis, spectrum detection, and filtering.

[0059] Fundamental component: The sinusoidal component with the lowest frequency and most concentrated energy in a periodic signal. It is the first term (i.e., the first harmonic) among all the harmonics that make up a periodic signal. It is often used to determine the normal state of a system or as a reference for filtering and feature extraction.

[0060] Example 1

[0061] See Figure 1 , Figure 1 This is an architecture diagram of an integrated electricity meter theft detection system proposed in this invention, which may specifically include:

[0062] M1, Data Acquisition Module, is used to collect raw voltage data from the electricity meter and magnetic field strength data of the environment in which the meter is located; specifically including:

[0063] M11, Voltage Sampling Circuit: The voltage sampling circuit collects raw voltage data at a preset sampling rate. After the meter is connected to the grid, within a preset time period, the data acquisition module collects noise feature vectors at a preset time window and caches them in RAM to construct a normal noise feature sample set.

[0064] Specifically, the voltage sampling circuit continuously samples the grid voltage at a fixed sampling rate of 6.4kHz to obtain raw voltage data. Within 3 days after the meter is installed and connected to the grid, the voltage data in each window is preliminarily processed in 50ms intervals to extract noise feature-related data and cache it in RAM (random access memory, used as a high-speed cache, with access speed faster than DataFlash). Finally, a normal noise feature sample set is constructed, which is used for subsequent Mahalanobis distance calculation, adaptive threshold initial value update, and evidence preservation benchmark.

[0065] M12, Magnetic Field Detection Element: The magnetic field detection element is used to detect the strength of the ambient magnetic field. A Hall effect sensor is used to detect the magnetic field strength of the environment in which the meter is located in real time, converting the physical quantity of the magnetic field strength into an electrical signal and transmitting it to the data fusion processing module.

[0066] M2, a noise feature extraction module, connected to the data acquisition module, is used to extract noise feature vectors from the raw voltage data; it extracts a five-dimensional noise feature vector from the raw voltage data transmitted by the data acquisition module, the vector including noise kurtosis, noise variance, and the mean power spectral density in the low-frequency band (10~100Hz). ), average power spectral density in the mid-frequency band (100~500Hz) ), average power spectral density in the high-frequency band (500~1000Hz) Specifically, it includes:

[0067] M21, Noise Kurtosis Unit: When noise changes and deviates from the normal distribution range, it can display the sharpness of the noise signal distribution and is exceptionally sensitive to impulse interference. When the magnetic field changes, it generates magnetic pulses, causing impulse interference to line noise; the specific formula is:

[0068]

[0069] in, For noise kurtosis, The length of the noise sequence. For noise sequence values, The noise mean is... This represents the noise standard deviation.

[0070] M22, Noise Variance Unit: Used to reflect the overall fluctuation intensity of the signal; firstly, the raw signal acquired by the data acquisition module... The waveform is fed into an FFT to extract the 50Hz fundamental component. The noise signal can be obtained by subtracting the fundamental voltage component from the original voltage waveform signal. Then, the individual noises within the 50ms window are combined into a noise sequence and the "noise variance" is calculated. ".

[0071] M23, Noise Three-Segment Power Spectrum Density Band Unit: Used to illustrate the characteristics of electromagnetic interference in different frequency bands; performs FFT transformation on a "50ms window noise signal sequence" and calculates its power spectral density. The signal power spectrum is divided into three specific frequency bands: low frequency (10~100Hz), mid frequency (100~500Hz), and high frequency (500~1000Hz), and the average PSD value of each band is calculated. And cache it in RAM.

[0072] The five features are combined into a single five-dimensional noise feature vector, denoted as . Additionally, the five-dimensional feature vector can be replaced with a multi-dimensional vector containing time-domain or frequency-domain features such as skewness, waveform factor, and margin factor; the division of the three power spectral density bands can be adjusted to two or four or more segments, and the frequency band boundaries (such as adjusting the low-frequency band to 5~80Hz) can be adapted according to the actual power grid noise characteristics.

[0073] M3, a data fusion processing module, connected to the noise feature extraction module and the data acquisition module, is used to perform fusion judgment based on the noise feature vector and magnetic field strength data, and output the electricity theft judgment result; see reference Figure 3 Specifically, it includes:

[0074] M31, Mahalanobis distance calculation unit, is used to calculate the Mahalanobis distance between the real-time noise feature vector and the normal noise feature sample set; take the noise feature vector sequence data from the data acquisition module, and for the real-time feature vector x, calculate its Mahalanobis distance with the normal sample set;

[0075] Based on the five-dimensional feature vector mean vector covariance matrix The Mahalanobis distance scalar is calculated using the following formula:

[0076]

[0077] in, The Mahalanobis distance, This represents the feature offset after removing the mean. is the inverse of the covariance matrix; MD is a scalar representing how far away from the normal state; the calculation result is a scalar used to characterize the degree of deviation between real-time features and the normal state.

[0078] In addition, this unit can be replaced by an Euclidean distance calculation unit, which can be used in conjunction with principal component analysis (PCA) for dimensionality reduction calculation; the adaptive threshold calculation unit can use moving average, exponential weighted moving average, or other methods to replace the "mean + N times the standard deviation" calculation method.

[0079] M32, the adaptive threshold calculation unit, is used to dynamically update the judgment threshold based on historical Mahalanobis distance data. It dynamically updates the judgment threshold every 3 days based on the Mahalanobis distance (MD) value during normal operation to adapt to slow aging or environmental changes that may occur during long-term meter operation. It also calculates the average of all MD values ​​over the past 3 days. and standard deviation And calculate the new judgment threshold according to the formula. This enables dynamic updating of thresholds to adapt to aging or environmental changes during long-term operation of the electricity meter.

[0080] M33, Magnetic Field Detection Unit, is used to receive the magnetic field strength data of the magnetic field detection element and determine whether the magnetic field strength exceeds a preset magnetic field threshold of 300mT. It receives the magnetic field strength electrical signal transmitted by the Hall sensor, converts it into a magnetic field strength value, compares it with the preset 300mT magnetic field threshold, and outputs a high-level signal when the magnetic field strength value exceeds 300mT, otherwise outputs a low-level signal.

[0081] See Figure 4 The fusion judgment logic of this module is as follows: When the Mahalanobis distance output by the Mahalanobis distance calculation unit exceeds the current judgment threshold of the adaptive threshold calculation unit, and the magnetic field detection unit outputs a high level, the "suspected electricity theft" flag is set, and a 5-minute timer is started. During the 5-minute timer, the Mahalanobis distance and magnetic field strength are continuously monitored. If both always meet the condition of "Madalanobis distance exceeding the threshold + magnetic field strength exceeding 300mT", the "electricity theft" flag is set, and the judgment result of electricity theft is output. If the above conditions are not met at any time during the timer, the "suspected electricity theft" flag is cleared, and no electricity theft judgment result is output.

[0082] M4, the evidence collection and storage module, is connected to the data fusion processing module and is used to save relevant original data as evidence when the electricity theft judgment result is that electricity theft exists;

[0083] Specifically, when the "electricity theft" flag is received from the data fusion processing module, the module is triggered to start. It reads the original voltage, current and power data from RAM within 20 seconds before and 60 seconds after the electricity theft event, and transfers this data to DataFlash (data flash memory, non-volatile memory with fast read and write speed and large capacity, erased in 4096-byte pages before writing) for storage to ensure that the data is not lost after power failure, forming a complete chain of evidence of electricity theft.

[0084] M5, a communication data transceiver module, is connected to the data fusion processing module and the evidence collection and storage module, and is used to interact with the main station to transmit the electricity theft judgment results and the stored evidence data.

[0085] Specifically, the module establishes a connection with the power master station using wired or wireless communication. The master station periodically sends electricity theft mark query commands to the module, and the module responds to the commands and reports the current electricity theft mark status. When the master station sends an evidence data upload command, the module reads the original voltage, current, and power data stored in DataFlash and uploads it to the master station according to the preset communication protocol for subsequent verification by power operation and maintenance personnel.

[0086] Example 2

[0087] See Figure 2 , Figure 2 This is a flowchart of a fusion-based electricity meter theft detection method proposed in this invention, which may specifically include:

[0088] S1. Data Acquisition: Acquire raw voltage data from the meter and magnetic field strength data of the environment in which the meter is located; specifically including:

[0089] S11. The voltage sampling circuit continuously samples the grid voltage at a sampling rate of 6.4kHz to obtain the raw voltage data.

[0090] S12. Within 3 days after the electricity meter is connected to the power grid, the raw voltage data in each window is preprocessed in 50ms time windows, noise feature related data is extracted and cached in RAM, and a normal noise feature sample set is constructed.

[0091] S13. The magnetic field strength of the environment where the meter is located is detected in real time by the Hall sensor, and the physical quantity of magnetic field strength is converted into an electrical signal and output.

[0092] S2. Noise Feature Extraction: Extracting noise feature vectors from the raw voltage data; specifically including:

[0093] S21. Noise signal acquisition: Extract the 50Hz fundamental component from the raw voltage data acquired in step S11 through FFT transformation, and subtract this fundamental component from the raw voltage data to obtain the noise signal.

[0094] S22. Noise Kurtosis Calculation: The noise signal obtained in step S21 is divided into noise sequences in 50ms windows. The kurtosis of the noise sequence within each window is calculated based on the formula:

[0095]

[0096] in, For noise kurtosis, The length of the noise sequence. For noise sequence values, The noise mean is... This represents the noise standard deviation.

[0097] S23. Noise Variance Calculation: For the noise signal obtained in step S21, divide the noise sequence into 50ms windows, and calculate the variance of the noise sequence within each window based on the formula:

[0098]

[0099] in, The length of the noise sequence. For noise sequence values, This represents the noise mean.

[0100] S24. Calculation of three power spectral density bands: For the noise signal obtained in step S21, divide the noise sequence into 50ms windows, perform FFT transformation on the noise sequence in each window and calculate the power spectral density; divide the power spectral density into three frequency bands: 10~100Hz, 100~500Hz and 500~1000Hz, and calculate the mean value of the power spectral density of each frequency band respectively.

[0101] S25. Feature Vector Combination: Combine the noise kurtosis calculated in step S22, the noise variance calculated in step S23, and the mean power spectral density of the three frequency bands calculated in step S24 to form a five-dimensional noise feature vector.

[0102] S3. Data Fusion Processing: Based on the noise feature vector and magnetic field strength data, a fusion judgment is performed to output the electricity theft judgment result; specifically including:

[0103] S31. Mahalanobis distance calculation: Obtain the real-time five-dimensional noise feature vector x formed in step S25, combine it with the normal noise feature sample set constructed in step S12, and calculate the distance based on the formula... Calculate the Mahalanobis distance between the real-time feature vector and the normal sample set;

[0104] S32. Adaptive Threshold Update: With a 3-day update cycle, at the beginning of each update cycle, extract all Mahalanobis distance values ​​calculated in step S31 over the past 3 days and calculate their average. and standard deviation Based on the formula Calculate the new judgment threshold and update the currently used judgment threshold;

[0105] S33. Magnetic field strength judgment: Convert the magnetic field strength electrical signal output in step S13 into a magnetic field strength value, compare it with the preset magnetic field threshold of 300mT, and determine whether it exceeds the threshold.

[0106] S34. Fusion Judgment: If the Mahalanobis distance calculated in step S31 exceeds the current judgment threshold in step S32, and the magnetic field strength judged in step S33 exceeds 300mT, then the "suspected electricity theft" flag is set and a 5-minute timer is started; during the timer period, steps S31 to S33 are repeated continuously. If the conditions of "Madalanobis distance exceeding the threshold + magnetic field strength exceeding 300mT" are always met, then the "electricity theft" flag is set and the judgment result of electricity theft is output; if the above conditions are not met at any time during the timer period, then the "suspected electricity theft" flag is cleared and no electricity theft judgment result is output.

[0107] S4. Evidence Collection and Preservation: If the electricity theft determination result indicates that electricity theft exists, the relevant original data shall be preserved as evidence; specifically including:

[0108] S41. When step S34 outputs the judgment result of the existence of electricity theft, read the original voltage data, current data and power data within 20 seconds before the occurrence of the electricity theft event and 60 seconds after the occurrence from the RAM.

[0109] S42. Transfer the data read in step S41 to DataFlash for storage. Before storage, perform an erasure process according to the requirements of DataFlash (4096 bytes per page) to ensure that the data is stored for a long time without being lost.

[0110] S5. Communication Interaction: Interacting with the main station to transmit the electricity theft judgment results and stored evidence data. Specifically, this includes:

[0111] S51, Electricity Theft Flag Query Response: Receives the electricity theft flag query command sent by the power master station and provides feedback on the current status of the "electricity theft" flag (set or not set).

[0112] S52. Evidence Data Upload: Receive the evidence data upload instruction sent by the power master station, read the original data stored in step S42 from DataFlash, and upload it to the master station according to the preset communication protocol.

[0113] In summary, the present invention has the following advantages:

[0114] 1. High detection reliability and significantly reduced false alarm rate: By using dual judgment conditions of magnetic field anomaly and signal noise characteristic anomaly, false alarms caused by legitimate strong magnetic field sources are effectively eliminated; compared with single electrical quantity analysis technology, the calculation of noise feature vector is controllable, and the fusion judgment mechanism improves the ability to resist normal power grid disturbances.

[0115] 2. Accurate electricity theft identification and complete evidence chain: With the help of a delayed secondary confirmation mechanism, only continuous electricity theft behavior is recorded to avoid accidental triggering due to momentary interference; the original waveform data related to the electricity theft event is automatically saved to realize the integration of detection and evidence collection and form a complete evidence chain;

[0116] 3. Excellent adaptability and robustness: It adopts Mahalanobis distance to measure noise changes, which can automatically consider the correlation between features and make the calculation more scientific; the adaptive threshold design can adapt to different installation locations, power grid environments and the aging of the meter after long-term operation, ensuring long-term effective operation.

[0117] The above specific 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 examples, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A fusion-based electricity meter theft detection system, characterized in that, include: The data acquisition module is used to collect the raw voltage data of the electricity meter and the magnetic field strength data of the environment in which the electricity meter is located; A noise feature extraction module, connected to the data acquisition module, is used to extract noise feature vectors from the raw voltage data; The noise feature vector is a five-dimensional feature vector, which includes: noise kurtosis, noise variance, and the mean of three power spectral density bands; the process of obtaining the three power spectral density bands is as follows: Perform a Fast Fourier Transform on the noise signal within a preset time window and calculate the power spectral density, then divide the power spectral density into... low frequency band mid-frequency band and In the high-frequency band, the mean value of the power spectral density of each frequency band is calculated to obtain the mean value of the three power spectral density bands; The data fusion processing module is connected to the noise feature extraction module and the data acquisition module, and is used to perform fusion judgment based on the noise feature vector and magnetic field strength data, and output the electricity theft judgment result. The evidence collection and storage module is connected to the data fusion and processing module and is used to save relevant original data as evidence when the electricity theft judgment result is that electricity theft exists. The communication data transceiver module is connected to the data fusion processing module and the evidence collection and storage module, and is used to interact with the main station to transmit the electricity theft judgment results and the stored evidence data.

2. The integrated electricity meter theft detection system according to claim 1, characterized in that, The data acquisition module includes a voltage sampling circuit and a magnetic field detection element; The voltage sampling circuit collects raw voltage data at a preset sampling rate. After the meter is connected to the grid, within a preset time period, the data acquisition module collects noise feature vectors at a preset time window and caches them in RAM to construct a normal noise feature sample set. The magnetic field detection element is used to detect the strength of the ambient magnetic field.

3. The integrated electricity meter theft detection system according to claim 1, characterized in that, The formula for calculating the noise kurtosis is: in, For noise kurtosis, The length of the noise sequence. For noise sequence values, The noise mean. This represents the standard deviation of noise.

4. The integrated electricity meter theft detection system according to claim 1, characterized in that, The noise variance acquisition process is as follows: the fundamental component is extracted from the original voltage data using a fast Fourier transform; the fundamental component is subtracted from the original voltage data to obtain the noise signal; the variance of the noise sequence composed of the noise signal within a preset time window is calculated to obtain the noise variance. The specific calculation formula is as follows: in, The length of the noise sequence. For noise sequence values, This represents the noise mean.

5. The integrated electricity meter theft detection system according to claim 2, characterized in that, The data fusion processing module includes: The Mahalanobis distance calculation unit is used to calculate the Mahalanobis distance between the real-time noise feature vector and the normal noise feature sample set. An adaptive threshold calculation unit is used to dynamically update the judgment threshold based on historical Mahalanobis distance data; The magnetic field detection unit is used to receive the magnetic field strength data of the magnetic field detection element and determine whether the magnetic field strength exceeds a preset magnetic field threshold of 300mT.

6. The integrated electricity meter theft detection system according to claim 5, characterized in that, The formula for calculating the Mahalanobis distance is: in, The Mahalanobis distance, This is a real-time noise feature vector. This is the mean vector of the set of normal noise feature samples. It is the inverse of the covariance matrix. This represents the matrix transpose operation.

7. The integrated electricity meter theft detection system according to claim 5, characterized in that, The adaptive threshold calculation unit calculates the mean of all Mahalanobis distances over the past 3 days, with a preset update cycle of 3 days. and standard deviation A new judgment threshold is calculated based on the mean and standard deviation. The calculation formula is: 。 8. The integrated electricity meter theft detection system according to claim 5, characterized in that, The fusion judgment logic of the data fusion processing module includes: When the Mahalanobis distance exceeds the judgment threshold and the magnetic field strength exceeds the preset magnetic field threshold of 300mT, the suspected electricity theft sign is set and a 5-minute timer is started. If the Mahalanobis distance continuously exceeds the judgment threshold and the magnetic field strength continuously exceeds the preset magnetic field threshold during the timing period, the electricity theft flag is set and the judgment result of electricity theft is output; otherwise, the electricity theft suspicion flag is cleared. The evidence collection and storage module stores relevant raw data including voltage, current and power data within 20 seconds before and 60 seconds after the electricity theft incident. The relevant raw data is transferred from random access memory to data flash memory for storage. The communication data transceiver module responds to the master station's instructions and uploads the electricity theft flag and related raw data.

9. A fusion-based method for detecting electricity theft from electricity meters, characterized in that, The integrated electricity meter theft detection system according to any one of claims 1-8 includes: S1. Data Acquisition: Acquire raw voltage data of the meter and magnetic field strength data of the environment where the meter is located; S2. Noise Feature Extraction: Extract noise feature vectors from the raw voltage data; S3. Data fusion processing: Based on the noise feature vector and magnetic field strength data, perform fusion judgment and output the electricity theft judgment result; S4. Evidence Collection and Preservation: If the electricity theft determination result indicates that electricity theft exists, the relevant original data shall be preserved as evidence. S5. Communication Interaction: Interact with the main station to transmit the electricity theft judgment results and the saved evidence data.

Images