Intelligent anti-theft electric energy metering box

Abstract

The present application relates to the technical field of electricity larceny prevention, in particular to a kind of intelligent electricity larceny prevention electric energy metering box, including the intelligent electricity larceny prevention unit embedded in electric energy metering box, and the intelligent electricity larceny prevention unit includes multidimensional sensing module and edge computing module;Multidimensional sensing module at least includes the first, second current sensor for real-time acquisition electric energy metering box incoming line total loop current vector;Edge computing module is configured to: according to the signal of the first, second current sensor, the difference of the sum of total loop current vector and each branch current vector is calculated in real time, obtain residual current vector, judge whether there is electricity larceny behavior.The physical defense line set by the present application is based on the total-branch circuit current vector difference calculation of independent sensor, and the pre-table shunt is found from the physical level of Kirchhoff's law, and any electricity larcener cannot bypass;By reconstructing the error of multidimensional electrical characteristics through self-encoder model analysis, new type, high-frequency electronic electricity larceny method that has never appeared can be effectively detected.

Description

Technical Field

[0001] This invention relates to the field of anti-theft electricity technology, specifically to an intelligent anti-theft electricity metering box. Background Technology

[0002] Electricity metering is a core component of electricity marketing and settlement, and its accuracy directly impacts the interests of both power supply companies and users. However, electricity theft has long plagued the power industry, causing significant economic losses and safety hazards.

[0003] Currently, preventing electricity theft from electricity metering boxes mainly relies on lead seals, mechanical locks, and regular manual inspections. This traditional, point-based, passive defense model has the following significant drawbacks: Firstly, there are blind spots in perception; traditional anti-theft methods are largely ineffective against new types of electricity theft, such as strong magnetic interference, high-frequency harmonic injection, and concealed current diversion before the meter. Existing technologies, such as the patent application CN121347887A which discloses a novel anti-theft system and application method for smart meter boxes, still primarily rely on data collected by the meter itself for analysis, making it difficult to cope with strong magnetic interference attacks targeting the metering chip itself. Another example is the patent application CN121499873A, which discloses an anti-theft detection system for electricity meter boxes. Although it involves multi-dimensional feature vector construction, its data source is the raw electrical energy data and communication link metadata uploaded by the terminal. The delay in data transmission and processing results in insufficient real-time response capabilities on-site.

[0004] Secondly, the judgment is severely delayed; even if some existing intelligent solutions collect data, they only send it to a remote master station for analysis, and by the time electricity theft is discovered, it has often been going on for months. Patent application CN122017340A discloses an intelligent management electricity metering box and metering method with multiple security protections, proposing to identify electricity theft patterns based on a lightweight AI diagnostic model loaded with structured edge monitoring data. However, its diagnostic results still need to be uploaded for comprehensive judgment, limiting on-site closed-loop processing capabilities. This "data upload-cloud analysis-command feedback" model cannot meet the demand for millisecond-level response to prevent electricity theft. Summary of the Invention

[0005] To overcome the shortcomings of the existing technology, the present invention aims to provide an intelligent anti-theft electricity metering box. It performs Kirchhoff's current law verification in real time through an edge computing module, and inputs the extracted multi-dimensional electrical features into a hybrid model that includes physical rules and self-encoder reconstruction for comprehensive judgment. Once electricity theft is confirmed, closed-loop processing is completed on-site, including waveform recording, image capture, active warning or remote circuit breaker tripping, so as to solve the problems mentioned in the background technology.

[0006] To achieve the above objectives, the present invention provides an intelligent anti-theft electricity metering box, including an intelligent anti-theft unit embedded in the electricity metering box. The intelligent anti-theft unit includes a multi-dimensional sensing module, an edge computing module connected to the multi-dimensional sensing module, and an evidence collection and execution module connected to the edge computing module. The multidimensional sensing module includes at least a first current sensor for real-time acquisition of the total current vector of the incoming line of the power metering box, a second current sensor for real-time acquisition of the current vector of at least one outgoing branch, and a magnetic sensor for detecting the strength of an external abnormal magnetic field. The edge computing module is configured to: calculate the difference between the total loop current vector and the sum of the branch current vectors in real time based on the signals from the first current sensor and the second current sensor to obtain the residual current vector; and determine whether there is any electricity theft based at least on the residual current vector and a preset anomaly detection model. The evidence collection and execution module is used to collect evidence, including at least images or videos, when the edge computing module determines that there is electricity theft, and to execute preset active anti-theft actions.

[0007] As a further improvement to this technical solution, the anomaly detection model includes a physical rule verification model, which is configured to generate a first electricity theft alarm signal when the amplitude of the residual current vector exceeds a preset dynamic threshold and the duration exceeds a set time.

[0008] As a further improvement to this technical solution, the anomaly detection model also includes an anomaly feature recognition model based on neural networks; the edge computing module is further configured to: extract multiple dimensions of current and voltage of each circuit, and input the multiple dimensions of features into the anomaly feature recognition model to output the confidence level of electricity theft behavior; when the first electricity theft alarm signal is generated and / or the confidence level of electricity theft behavior exceeds the second threshold, it is determined that electricity theft behavior exists.

[0009] As a further improvement to this technical solution, the multiple dimensional features include at least one of the following: fundamental current RMS value, total harmonic distortion rate, specific harmonic content rate, power factor, phase angle, voltage imbalance, and active / reactive power fluctuation mode.

[0010] As a further improvement to this technical solution, the abnormal feature recognition model includes an autoencoder model and / or a one-dimensional convolutional-long short-term memory hybrid model; the autoencoder model is used to identify abnormal states based on the reconstruction error of the input feature vector, and the one-dimensional convolutional-long short-term memory hybrid model is used to identify specific types of electricity theft.

[0011] As a further improvement to this technical solution, the multi-dimensional sensing module also includes a zero-sequence current detection unit and / or an anti-vandalism detection circuit; the anti-vandalism detection circuit is a wire circuit pre-embedded in the interlayer of the power metering box shell, and the edge computing module is connected to the anti-vandalism detection circuit and is used to determine that a power metering box vandalism event has occurred when the wire circuit is broken.

[0012] As a further improvement to this technical solution, the evidence collection and execution module includes an image acquisition device, an audio playback device, a flashing warning device, and a circuit breaker control interface; the active anti-theft action includes controlling the image acquisition device to capture or record video, controlling the audio playback device to emit a warning voice, controlling the flashing warning device to emit a flashing light, and controlling the incoming main circuit breaker to disconnect, or one or more of the following:

[0013] As a further improvement to this technical solution, an independent backup power module is also included, which is used to maintain the operation of all functions of the intelligent anti-theft unit for a predetermined period of time after the main power supply is interrupted.

[0014] As a further improvement to this technical solution, the intelligent anti-theft method for the electricity metering box includes: The first current sensor is used to collect the total current vector of the incoming line circuit in real time, and the second current sensor is used to collect the current vector of each outgoing branch in real time. The edge computing module calculates the difference between the total loop current vector and the sum of the branch current vectors to obtain the residual current vector; Extract multi-dimensional electrical features including residual current characteristics, current harmonic characteristics, and power factor characteristics; The multidimensional electrical features are input into a preset anomaly detection model in the edge computing module. The anomaly detection model includes a physical rule verification unit based on Kirchhoff's current law and an anomaly reconstruction unit based on an autoencoder. The output of the anomaly detection model is used to determine whether there is electricity theft. When electricity theft is detected, the evidence collection and execution module is triggered to collect on-site image evidence and perform proactive handling actions.

[0015] As a further improvement to this technical solution, the step of determining whether electricity theft has occurred based on the output of the anomaly detection model includes: When the magnitude of the residual current vector exceeds the dynamic threshold and persists for a first time period, a preliminary abnormal signal is generated; Simultaneously, the multidimensional electrical features extracted within the preset time window are input into the self-encoder to calculate the reconstruction error. When the reconstruction error exceeds a set value, a second abnormal signal is generated. Based on the combined preliminary abnormal signal and the second abnormal signal, the theft of electricity is confirmed to have occurred.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This intelligent anti-theft electricity metering box uses a physical defense system based on the calculation of the vector difference between the total and branch circuit currents using independent sensors. It detects the current shunting before the meter from the physical level of Kirchhoff's laws, making it impossible for any electricity thief to bypass it.

[0017] 2. This intelligent anti-theft electricity metering box, through the analysis of reconstruction errors of multi-dimensional electrical characteristics using a self-encoder model, can effectively detect new, high-frequency electronic electricity theft methods that have never appeared before.

[0018] 3. This intelligent anti-theft electricity metering box associates and seals abnormal electrical waveforms with on-site captured images, faces, and vandalism events at the edge end using millisecond-level timestamps, forming a usable and complete local evidence chain.

[0019] 4. This intelligent anti-theft electricity metering box constructs a complete active defense system, from real-time detection and local sound and light deterrence to controlled remote power cut-off, greatly reducing the duration of electricity theft. Attached Figure Description

[0020] The accompanying drawings described herein are for illustrative purposes only.

[0021] Figure 1 This is a block diagram of the intelligent anti-theft electricity metering box module of the present invention; Figure 2 This is a core flowchart of the intelligent anti-theft electricity control method of the present invention. Detailed Implementation

[0022] Under the guidance of this invention, any possible variations of this invention should be considered within the scope of this invention.

[0023] Please see Figures 1-2 As shown, the present invention provides an intelligent anti-theft electricity metering box, including an intelligent anti-theft unit embedded in the electricity metering box. The intelligent anti-theft unit includes a multi-dimensional sensing module, an edge computing module connected to the multi-dimensional sensing module, an evidence collection and execution module connected to the edge computing module, and an independent backup power module. The backup power module is used to maintain the operation of all functions of the intelligent anti-theft unit for a predetermined time after the main power supply is interrupted.

[0024] Specifically, the multi-dimensional sensing module includes at least a first current sensor for real-time acquisition of the current vector of the main inlet circuit of the electricity metering box, a second current sensor for real-time acquisition of the current vector of at least one outlet branch circuit, and a magnetic sensor for detecting the intensity of an external abnormal magnetic field. The multi-dimensional sensing module also includes a zero-sequence current detection unit and / or an anti-vandalism detection circuit. The anti-vandalism detection circuit is a wire circuit pre-embedded in the interlayer of the electricity metering box shell. The edge computing module is connected to the anti-vandalism detection circuit and is used to determine that an electricity metering box vandalism event has occurred when the wire circuit is broken.

[0025] Specifically, the edge computing module is configured to: calculate the difference between the total loop current vector and the sum of the branch current vectors in real time based on the signals from the first current sensor and the second current sensor to obtain the residual current vector; and determine whether there is electricity theft based at least on the residual current vector and a preset anomaly detection model.

[0026] The anomaly detection model includes a physical rule verification model, configured to generate a first electricity theft alarm signal when the amplitude of the residual current vector exceeds a preset dynamic threshold and the duration exceeds a set time. The anomaly detection model also includes a neural network-based anomaly feature recognition model. The edge computing module is further configured to extract multiple dimensions of current and voltage from each circuit and input these multiple dimensions into the anomaly feature recognition model to output a confidence level for electricity theft behavior. When the first electricity theft alarm signal is generated and / or the confidence level for electricity theft behavior exceeds a second threshold, it is determined that electricity theft has occurred.

[0027] Multiple dimensional characteristics include at least one of the following: fundamental current RMS value, total harmonic distortion rate, specific harmonic content, power factor, phase angle, voltage imbalance, and active / reactive power fluctuation mode.

[0028] The abnormal feature recognition model includes an autoencoder model and / or a one-dimensional convolutional-long short-term memory hybrid model; the autoencoder model is used to identify abnormal states based on the reconstruction error of the input feature vector, and the one-dimensional convolutional-long short-term memory hybrid model is used to identify specific types of electricity theft.

[0029] Specifically, the evidence collection and execution module is used to collect evidence, including at least images or videos, when the edge computing module determines that there is electricity theft, and to execute preset proactive anti-theft actions.

[0030] The evidence collection and execution module includes an image acquisition device, an audio playback device, a flashing warning device, and a circuit breaker control interface; the active anti-theft actions include controlling the image acquisition device to capture or record video, controlling the audio playback device to issue a warning voice, controlling the flashing warning device to emit flashing light, and controlling the main incoming circuit breaker to disconnect one or more of the following:

[0031] When the intelligent anti-theft electricity metering box of this invention is in operation, it collects the current vector of the incoming main circuit in real time through a first current sensor and the current vector of each outgoing branch in real time through a second current sensor. The edge computing module calculates the difference between the total circuit current vector and the sum of the current vectors of each branch to obtain the residual current vector. It extracts multi-dimensional electrical features, including residual current characteristics, current harmonic characteristics, and power factor characteristics. The multi-dimensional electrical features are input into a preset anomaly detection model in the edge computing module. The anomaly detection model includes a physical rule verification unit based on Kirchhoff's current law and an anomaly reconstruction unit based on an autoencoder. The output of the anomaly detection model determines whether electricity theft has occurred. When the amplitude of the residual current vector exceeds the dynamic threshold and persists for a first time period, a preliminary anomaly signal is generated. At the same time, the multi-dimensional electrical features extracted within the preset time window are input into the autoencoder to calculate the reconstruction error. When the reconstruction error exceeds a set value, a second anomaly signal is generated. The preliminary anomaly signal and the second anomaly signal are combined to confirm that electricity theft has occurred. When electricity theft is determined to exist, the evidence collection and execution module is triggered to collect on-site image evidence and perform proactive handling actions.

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, a detailed description is provided below in conjunction with the accompanying drawings.

[0033] Example - Apparatus; The metering box of this invention has an intelligent anti-theft unit independently embedded in a standard box. This unit operates completely in parallel with the electricity meter.

[0034] (a) Multidimensional sensing module The multidimensional sensing module specifically includes: Current sensor: The main incoming circuit uses a three-axis high-precision Rogowski coil, and each outgoing branch uses a magnetoresistive sensor array, all of which can perform synchronous AC sampling at a frequency of 12.8kHz to fully capture the current vector from the fundamental wave to the 63rd harmonic.

[0035] Magnetic sensors: Multiple thin-film tunnel magnetoresistive (TMR) sensors are arranged at the door and the partition of the metering chamber to monitor external strong magnetic interference in real time. The dynamic threshold can be set.

[0036] Anti-vandalism detection circuit: Thin wires are pre-embedded in the non-metallic outer shell of the enclosure to form a continuous circuit. Once the enclosure is drilled or cut, causing the wires to break, an open circuit signal is immediately triggered.

[0037] Environmental perception module: Composed of a low-light wide dynamic range camera and an infrared pyroelectric sensor (PIR), used to detect the approach of people and acquire images.

[0038] (ii) Edge computing module The edge computing module is the core, and it adopts an industrial-grade AI core board with an NPU acceleration unit.

[0039] Kirchhoff's Current Law Real-Time Verification: A Fast Fourier Transform (FFT) is performed on the incoming and outgoing currents every 10 milliseconds to obtain the vector values ​​of the fundamental and harmonic frequencies. Then, the complex amplitude of the difference between the total incoming current vector and the sum of all outgoing current vectors is precisely calculated; this is the residual current vector. Under normal operating conditions, this difference is extremely small, reflecting only line-to-ground capacitance leakage. When the vector amplitude suddenly increases beyond a preset dynamic threshold, such as 50mA, and persists for more than 1 second, a high-confidence "pre-meter shunting" preliminary alarm is directly generated.

[0040] Hybrid model anomaly detection: An analysis window is formed every 15 minutes to extract the RMS current value, total harmonic distortion (THD), 3rd / 5th / 7th harmonic content, power factor, voltage imbalance, active power fluctuation mode, etc. of each loop, forming a multi-dimensional feature vector. This vector is input into two models in parallel. Expert rule model: Quickly classifies known electricity theft methods such as undervoltage, undercurrent, and phase shifting; Lightweight autoencoder reconstruction model: Trained using historical normal feature data, the reconstruction error of the model output spikes dramatically when new electricity theft methods cause abnormal feature vector distribution.

[0041] When the reconstruction error of the autoencoder model exceeds the threshold, or the expert rule model outputs a known type with a confidence level exceeding 95%, and confirms this by performing an "OR" logic combination with the Kirchhoff law verification alarm, the edge computing module determines that electricity theft has occurred.

[0042] (III) Evidence Collection and Execution Module Upon receiving an alarm signal, the evidence collection and execution module automatically executes the following linkage strategy: Evidence preservation: Immediately capture the original high-frequency waveforms of the entire circuit for 10 seconds before and after the anomaly occurs, store them locally, and then package and upload them. Simultaneously, if the infrared sensor is triggered at this time, immediately start the camera to continuously take 10 photos and a 10-second short video, and run a lightweight face detection algorithm to select and capture faces in the scene. All digital evidence is stamped with a trusted timestamp.

[0043] Active deterrence: The pre-installed voice module inside the box activates a warning: "Abnormal power usage has been recorded in its entirety. Please stop immediately," and drives a high-brightness LED strobe light to flash as a deterrent.

[0044] Remote handling: For high-risk behaviors configured to be power-off, the edge computing module can drive the incoming main magnetic holding circuit breaker to open and set it to a local locked state. It can only be closed by an authorized manager on-site with an authorized terminal.

[0045] (iv) Backup power module The backup power module consists of a supercapacitor and a lithium battery pack, ensuring that the entire intelligent anti-theft unit can run at full power for 15 minutes after the main power supply is cut off, and can fully report the power outage event and the waveform data at the last moment.

[0046] Method Implementation Examples The control method flow of the present invention is as follows: Step 1: Independent of the electricity meter, collect high-frequency current vector signals, magnetic field, enclosure integrity, and environmental status signals in real time and synchronously from the main incoming circuit and each outgoing branch of the enclosure.

[0047] Step 2: The edge computing module calculates the difference between the total loop current vector sum and the current vector sum of each branch to obtain the residual current vector, and then performs Kirchhoff's law verification.

[0048] Step 3: Extract multi-dimensional electrical characteristics such as harmonics and power of each circuit according to the time window to form a feature vector.

[0049] Step 4: Input the feature vector into the pre-built hybrid anomaly detection model within the edge module. This model includes a physical rule-based verification unit and an autoencoder-based anomaly reconstruction unit. The outputs of both are combined to determine whether electricity theft has occurred and the type of theft.

[0050] Step 5: Once the confidence level is deemed sufficient, the edge computing module automatically triggers on-site evidence association and sealing, active audible and visual warnings, and decides whether to execute remote controlled power cut-off based on preset strategies, thereby forming a closed-loop anti-electricity theft prevention mechanism.

[0051] The above embodiments are only for illustrating the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent changes or modifications made in accordance with the spirit and essence of the present invention should be covered within the scope of protection of the present invention.

Claims

1. An intelligent anti-theft electricity metering box, characterized in that: It includes an intelligent anti-theft unit embedded in an electricity metering box. The intelligent anti-theft unit includes a multi-dimensional sensing module, an edge computing module connected to the multi-dimensional sensing module, and an evidence collection and execution module connected to the edge computing module. The multidimensional sensing module includes at least a first current sensor for real-time acquisition of the total current vector of the incoming line of the power metering box, a second current sensor for real-time acquisition of the current vector of at least one outgoing branch, and a magnetic sensor for detecting the strength of an external abnormal magnetic field. The edge computing module is configured to: calculate the difference between the total loop current vector and the sum of the branch current vectors in real time based on the signals from the first current sensor and the second current sensor to obtain the residual current vector; and determine whether there is any electricity theft based at least on the residual current vector and a preset anomaly detection model. The evidence collection and execution module is used to collect evidence, including at least images or videos, when the edge computing module determines that there is electricity theft, and to execute preset active anti-theft actions.

2. The intelligent anti-theft electricity metering box according to claim 1, characterized in that: The anomaly detection model includes a physical rule verification model, which is configured to generate a first electricity theft alarm signal when the amplitude of the residual current vector exceeds a preset dynamic threshold and the duration exceeds a set time.

3. The intelligent anti-theft electricity metering box according to claim 2, characterized in that: The anomaly detection model also includes an anomaly feature recognition model based on neural networks; the edge computing module is further configured to: extract multiple dimensions of current and voltage of each circuit, and input the multiple dimensions of features into the anomaly feature recognition model to output the confidence level of electricity theft behavior; when the first electricity theft alarm signal is generated and / or the confidence level of electricity theft behavior exceeds a second threshold, it is determined that electricity theft behavior exists.

4. The intelligent anti-theft electricity metering box according to claim 3, characterized in that: The multiple dimensional characteristics include at least one of the following: fundamental current RMS value, total harmonic distortion rate, specific harmonic content, power factor, phase angle, voltage imbalance, and active / reactive power fluctuation mode.

5. The intelligent anti-theft electricity metering box according to claim 4, characterized in that: The abnormal feature recognition model includes an autoencoder model and / or a one-dimensional convolutional-long short-term memory hybrid model; the autoencoder model is used to identify abnormal states based on the reconstruction error of the input feature vector, and the one-dimensional convolutional-long short-term memory hybrid model is used to identify specific types of electricity theft.

6. The intelligent anti-theft electricity metering box according to claim 5, characterized in that: The multi-dimensional sensing module also includes a zero-sequence current detection unit and / or an anti-vandalism detection circuit; the anti-vandalism detection circuit is a wire circuit pre-embedded in the interlayer of the power metering box shell, and the edge computing module is connected to the anti-vandalism detection circuit and is used to determine that a power metering box vandalism event has occurred when the wire circuit is broken.

7. The intelligent anti-theft electricity metering box according to claim 6, characterized in that: The evidence collection and execution module includes an image acquisition device, an audio playback device, a flashing warning device, and a circuit breaker control interface; the active anti-theft action includes controlling the image acquisition device to capture or record video, controlling the audio playback device to emit a warning voice, controlling the flashing warning device to emit a flashing light, and controlling the incoming main circuit breaker to disconnect, or one or more of the following:

8. The intelligent anti-theft electricity metering box according to claim 7, characterized in that: It also includes an independent backup power module, which is used to maintain the operation of all functions of the intelligent anti-theft unit for a predetermined period of time after the main power supply is interrupted.

9. The intelligent anti-theft electricity metering box according to claim 8, characterized in that, The intelligent anti-theft methods for this electricity metering box include: The first current sensor is used to collect the total current vector of the incoming line circuit in real time, and the second current sensor is used to collect the current vector of each outgoing branch in real time. The edge computing module calculates the difference between the total loop current vector and the sum of the current vectors of each branch to obtain the residual current vector; Extract multi-dimensional electrical features including residual current characteristics, current harmonic characteristics, and power factor characteristics; The multidimensional electrical features are input into a preset anomaly detection model in the edge computing module. The anomaly detection model includes a physical rule verification unit based on Kirchhoff's current law and an anomaly reconstruction unit based on an autoencoder. The output of the anomaly detection model is used to determine whether there is electricity theft. When electricity theft is detected, the evidence collection and execution module is triggered to collect on-site image evidence and perform proactive handling actions.

10. The intelligent anti-theft electricity metering box according to claim 9, characterized in that: The step of determining whether electricity theft has occurred based on the output of the anomaly detection model includes: When the magnitude of the residual current vector exceeds the dynamic threshold and persists for a first time period, a preliminary abnormal signal is generated; Simultaneously, the multidimensional electrical features extracted within the preset time window are input into the self-encoder to calculate the reconstruction error. When the reconstruction error exceeds a set value, a second abnormal signal is generated. Based on the combined preliminary abnormal signal and the second abnormal signal, the theft of electricity is confirmed to have occurred.

Images