Anti-electricity theft method and system based on smart meter

Abstract

The present invention discloses a method and system for preventing electricity theft based on a smart meter. The method includes: detecting whether an anti-electricity theft analysis condition is met based on a preset polling anti-electricity theft detection time, an acquired signal indicator light status, multiple spacing values, and multiple connection pressure values, wherein the anti-electricity theft analysis condition includes a first anti-electricity theft analysis condition and a second anti-electricity theft analysis condition; if the first anti-electricity theft analysis condition is met, generating an anti-electricity theft analysis result based on collected electricity usage information; if the second anti-electricity theft analysis condition is met, uploading the electricity usage information, collected location information, and environmental information to a cloud server, so that the cloud server performs an anti-electricity theft analysis based on the electricity usage information, location information, and environmental information using a hybrid model to obtain an anti-electricity theft analysis result, and then sending the anti-electricity theft analysis result to the smart meter; and generating a graded alarm based on the anti-electricity theft analysis result. This application reduces the false alarm rate of anti-electricity theft and improves the accuracy of anti-electricity theft.

Description

Technical Field

[0001] The present invention relates to the technical field of anti-electricity theft, and in particular to an anti-electricity theft method and system based on a smart meter. Background Art

[0002] With the continuous development of society and advancements in technology, electricity has become a fundamental energy source for modern life and production. However, the power system faces numerous challenges in its operation. Electricity theft not only causes economic losses to power companies but also negatively impacts the stable operation of the power system and the rational allocation of power resources. Therefore, how to effectively prevent electricity theft has become a key concern for power companies and researchers.

[0003] Although existing smart meter anti-theft methods have achieved physical protection and basic data monitoring, the following problems still exist: 1. Weak physical tampering protection: For example, relying solely on mechanical locks, single mechanical locks are easy to crack, and there is a lack of effective monitoring of high-tech theft methods (such as magnetic field interference and line tampering); 2. High false alarm rate: Maintenance operations are difficult to distinguish from theft, and manual verification is required; 3. The single data detection dimension makes the anti-theft accuracy low. For example, only detecting current makes it difficult to adapt to the identification of theft in complex electricity usage scenarios. Summary of the Invention

[0004] The embodiments of the present invention provide a method and system for preventing electricity theft based on a smart meter, aiming to solve the problem of low accuracy of existing electricity theft prevention methods.

[0005] In a first aspect, the present invention provides an anti-electricity theft method based on a smart meter. The smart meter includes a door. The edge of the door is provided with multiple sets of ultrasonic distance sensors to detect the distance between the door and the door frame to obtain multiple distance values. The terminal area inside the door is provided with multiple sets of pressure sensors to detect the connection pressure of the terminal to obtain multiple connection pressure values. The method includes:

[0006] Detecting whether anti-electricity theft analysis conditions are met according to a preset polling anti-electricity theft detection time, the acquired signal indicator light status, the plurality of spacing values, and the plurality of connection pressure values, wherein the anti-electricity theft analysis conditions include a first anti-electricity theft analysis condition and a second anti-electricity theft analysis condition;

[0007] If the first anti-electricity theft analysis condition is met, generating an anti-electricity theft analysis result based on the collected electricity usage information;

[0008] If the second anti-electricity theft analysis condition is met, uploading the electricity usage information, the collected location information, and the environmental information to a cloud server, so that the cloud server performs an anti-electricity theft analysis based on the electricity usage information, the location information, and the environmental information using a hybrid model to obtain the anti-electricity theft analysis result, and sending the anti-electricity theft analysis result to the smart meter;

[0009] A graded alarm is issued according to the anti-electricity theft analysis result.

[0010] In a second aspect, the present invention further provides an anti-electricity theft system based on a smart meter, comprising:

[0011] A smart electricity meter includes a door, wherein the edge of the door is provided with multiple sets of ultrasonic distance sensors for detecting the distance between the door and the door frame to obtain multiple distance values, and a terminal area inside the door is provided with multiple sets of pressure sensors for detecting the connection pressure of the terminal to obtain multiple connection pressure values. The smart electricity meter also includes a detection module and a generation module, wherein the detection module is configured to detect whether an anti-theft analysis condition is met based on a preset polling anti-theft detection time, an acquired signal indicator light state, multiple distance values, and multiple connection pressure values; and the generation module is configured to generate an anti-theft analysis result based on the collected electricity usage information.

[0012] a cloud server comprising an anti-electricity theft analysis module and a sending module, wherein the anti-electricity theft analysis module is configured to perform an anti-electricity theft analysis based on the electricity usage information uploaded by the smart meter and the collected location information and environmental information using a hybrid model to obtain the anti-electricity theft analysis result, and the sending module is configured to send the anti-electricity theft analysis result to the smart meter;

[0013] The smart meter further includes an alarm module, which is used to generate graded alarms based on the anti-electricity theft analysis results.

[0014] The present invention provides an anti-electricity theft method and system based on a smart electricity meter. The method includes: detecting whether an anti-electricity theft analysis condition is met based on a preset polling anti-electricity theft detection time, an acquired signal indicator light state, a plurality of spacing values, and a plurality of connection pressure values, wherein the anti-electricity theft analysis condition includes a first anti-electricity theft analysis condition and a second anti-electricity theft analysis condition; if the first anti-electricity theft analysis condition is met, generating an anti-electricity theft analysis result based on collected electricity usage information; if the second anti-electricity theft analysis condition is met, uploading the electricity usage information, collected location information, and environmental information to a cloud server, so that the cloud server performs anti-electricity theft analysis based on the electricity usage information, the location information, and the environmental information through a hybrid model to obtain the anti-electricity theft analysis result, and sending the anti-electricity theft analysis result to the smart electricity meter; and issuing a graded alarm based on the anti-electricity theft analysis result. The present application first detects whether the anti-electricity theft analysis conditions are met based on the preset polling anti-electricity theft detection time, signal indicator light status, multiple spacing values, and multiple connection pressure values. Then, when the first anti-electricity theft analysis condition is met, the anti-electricity theft analysis result is generated based on the collected electricity usage information; when the second anti-electricity theft analysis condition is met, the cloud server performs anti-electricity theft analysis through a hybrid model based on the electricity usage information, location information, and environmental information to obtain the anti-electricity theft analysis result; finally, a graded alarm is issued based on the anti-electricity theft analysis result. During the entire anti-electricity theft analysis and alarm process, since whether the anti-electricity theft analysis conditions are met is monitored based on multi-dimensional data, not only is the electricity theft effectively monitored in an all-round manner, but it can also distinguish between maintenance operations and electricity theft behaviors, thereby improving the accuracy of anti-electricity theft; and when different electricity theft analysis conditions are met, the anti-electricity theft analysis results are generated locally on the smart meter or on the cloud server based on multi-dimensional data, thereby reducing the false alarm rate of anti-electricity theft and improving the accuracy of anti-electricity theft. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 A block diagram of an anti-electricity theft system based on a smart meter according to an embodiment of the present invention is shown;

[0017] Figure 2 A schematic diagram of a flow chart of an anti-electricity theft method based on a smart meter according to an embodiment of the present invention is shown;

[0018] Figure 3 A schematic diagram of a sub-process of the anti-electricity theft method based on a smart meter according to an embodiment of the present invention is shown;

[0019] Figure 4 Another sub-flow diagram of the anti-electricity theft method based on a smart meter according to an embodiment of the present invention is shown;

[0020] Figure 5 Another sub-process diagram of the anti-electricity theft method based on a smart meter according to an embodiment of the present invention is shown;

[0021] Figure 6 A schematic diagram of another sub-process of the anti-electricity theft method based on a smart meter according to an embodiment of the present invention is shown;

[0022] Reference numerals:

[0023] 10. Anti-electricity theft system based on smart meter; 11. Smart meter; 111. Box door; 112. Detection module; 113. Generation module; 114. Alarm module; 115. Meter body; 116. Voltage and current acquisition module; 117. Location monitoring module; 118. Environmental monitoring module; 119. Communication module; 12. Cloud server; 121. Anti-electricity theft analysis module; 122. Distribution module. DETAILED DESCRIPTION

[0024] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0025] Directional terms used herein, such as "upper," "lower," "front," "back," "left," "right," "inner," "outer," and "side," refer only to directions in the accompanying drawings. Therefore, these directional terms are intended to illustrate and facilitate understanding of the present invention and are not intended to limit the present invention. Furthermore, in the accompanying drawings, similar or identical structures are denoted by the same reference numerals.

[0026] Embodiments of the present invention provide a smart meter-based electricity theft prevention method and system, addressing the low accuracy of existing electricity theft prevention methods. The method first detects whether an electricity theft analysis condition is met based on a preset polling electricity theft detection time, a signal indicator light status, multiple spacing values, and multiple connection pressure values. Then, when the first electricity theft analysis condition is met, an electricity theft analysis result is generated based on the collected electricity usage information. When the second electricity theft analysis condition is met, a cloud server performs an electricity theft analysis using a hybrid model based on the electricity usage information, location information, and environmental information to obtain an electricity theft analysis result. Finally, a graded alarm is generated based on the electricity theft analysis result. Throughout the entire electricity theft analysis and alarm process, the satisfaction of the electricity theft analysis conditions is monitored based on multi-dimensional data. This not only effectively monitors electricity theft in all aspects, but also distinguishes between maintenance operations and electricity theft, thereby improving the accuracy of electricity theft prevention. Furthermore, when different electricity theft analysis conditions are met, an electricity theft analysis result is generated locally on the smart meter or on a cloud server based on the multi-dimensional data, thereby reducing the false alarm rate and improving the accuracy of electricity theft prevention.

[0027] In order to better understand the above technical solution, the above technical solution will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0028] See also Figure 1 , Figure 1 The block diagram of the anti-theft system based on smart meter according to the embodiment of the present invention is shown. Figure 1As shown, the anti-theft system 10 based on the smart meter includes a smart meter 11 and a cloud server 12, wherein the smart meter 11 includes a box door 111, and the edge of the box door 111 is provided with multiple groups of ultrasonic distance sensors to collect the distance between the box door 111 and the door frame to obtain multiple distance values, and the terminal area inside the box door 111 is provided with multiple groups of pressure sensors to collect the connection pressure of the terminal to obtain multiple connection pressure values; the smart meter 11 also includes a detection module 112 and a generation module 113, the detection module 112 is used to generate a plurality of connection pressure values according to the preset polling anti-theft detection time, the obtained signal indicator light status, the plurality of distance values and the plurality of connection pressure values. The connection pressure value is detected to determine whether the anti-electricity theft analysis conditions are met; the generation module 113 is used to generate an anti-electricity theft analysis result based on the collected electricity usage information; the cloud server 12 includes an anti-electricity theft analysis module 121 and a sending module 122, the anti-electricity theft analysis module 121 is used to perform anti-electricity theft analysis based on the electricity usage information uploaded by the smart meter 11, the collected location information and environmental information through a hybrid model to obtain the anti-electricity theft analysis result, and the sending module 122 is used to send the anti-electricity theft analysis result to the smart meter 11; the smart meter 11 also includes an alarm module 114, and the alarm module 114 is used to issue a graded alarm based on the anti-electricity theft analysis result. It should be noted that, in this embodiment, there are four groups of ultrasonic distance sensors and six groups of pressure sensors. The four groups of ultrasonic distance sensors are evenly arranged along the edge of the box door 111, and the six groups of pressure sensors are distributed in a ring shape with the terminal block as the center, covering the main force-bearing area of the terminal block plugging and unplugging operations. It can be understood that in other embodiments, the number of groups of ultrasonic distance sensors and pressure sensors can be set according to actual needs. It should also be noted that, in this embodiment, the smart meter 11 also includes a meter body 115, a voltage and current acquisition module 116, a position monitoring module 117, an environmental monitoring module 118 and a communication module 119. The meter body 115 is provided with an electromagnetic shielding layer to prevent interference from high-frequency magnetic fields; the position monitoring module 117 is used to collect the position information; the environmental monitoring module 118 is used to collect the environmental information; the voltage and current acquisition module 116 is respectively connected to the phase line and the neutral line of the smart meter 11 to simultaneously sample the power consumption information of the phase line and the neutral line; the communication module 119 is used to upload the power consumption information, the position information and the environmental information to the cloud server 12, and receive the anti-theft analysis results issued by the cloud server 12. For the sake of simplicity, the method for generating the anti-electricity theft analysis results in the cloud service and the detection method used by the detection module 112 in the smart meter 11, the generation method used by the generation module 113, and the alarm method used by the alarm module 114 will be described in detail in the anti-electricity theft based on the smart meter 11 later, and will not be repeated here.

[0029] The embodiment of the present invention also proposes an anti-theft method based on a smart meter. The anti-theft method based on a smart meter can be used in the anti-theft system based on a smart meter in the above embodiment. The anti-theft system based on a smart meter has been described in detail in the above embodiment. For the sake of brevity, it will not be repeated here. In order to clearly illustrate the workflow of the embodiment of the present invention, the anti-theft method based on a smart meter will be described below in combination with the anti-theft system based on a smart meter in the above embodiment. Figure 2 The anti-electricity theft method based on the smart meter includes steps: S110-S140.

[0030] S110. Detect whether anti-electricity theft analysis conditions are met based on a preset polling anti-electricity theft detection time, the acquired signal indicator light status, the multiple spacing values, and the multiple connection pressure values, wherein the anti-electricity theft analysis conditions include a first anti-electricity theft analysis condition and a second anti-electricity theft analysis condition.

[0031] In this embodiment, the smart meter is provided with a signal indicator light. When the signal indicator light is green, it indicates that the smart meter is operating normally; when it is yellow, it indicates that the smart meter is under maintenance; and when it is red, it indicates that the smart meter is in an alarm state, indicating that electricity theft may occur. It is understandable that when the signal indicator light is green or red, that is, when the signal indicator light is in a normal state or an alarm state, anti-theft monitoring is required. Therefore, the signal indicator light state is obtained, and based on the signal indicator light state, a preset polling anti-theft detection time, multiple spacing values, and multiple connection pressure values, it is detected whether anti-theft analysis conditions are met, wherein the anti-theft analysis conditions include a first anti-theft analysis condition and a second anti-theft analysis condition.

[0032] In one embodiment, such as this embodiment, Figure 3 As shown, step S110 specifically includes steps S111-S114:

[0033] S111: If the acquired signal indicator light state is not a preset indicator light state, determining the preset polling anti-electricity theft detection time, the plurality of spacing values, and the plurality of connection pressure values;

[0034] S112: If the preset polling anti-electricity theft detection time is reached, determining that the first anti-electricity theft analysis condition is met;

[0035] S113: If at least one of the plurality of spacing values is greater than a preset reference spacing value, it is determined that the second anti-electricity theft analysis condition is satisfied;

[0036] S114: If at least one of the plurality of connection pressure values is greater than a preset upper connection pressure limit or at least one of the plurality of connection pressure values is less than a preset lower connection pressure limit, it is determined that the second anti-electricity theft analysis condition is satisfied.

[0037] In this embodiment, the preset indicator light status is the maintenance status, and the signal indicator light status is not the preset indicator light status, which means that the signal indicator light status is not the maintenance status. It can be understood that at this time, the signal indicator light status is normal status or alarm status, that is, the signal indicator light is green or red. When the preset indicator light state is not in the maintenance state, the preset polling anti-electricity theft detection time, multiple spacing values and multiple connection pressure values are judged; if the preset polling anti-electricity theft detection time is reached, it indicates that the fixed anti-electricity theft detection time is reached, and only a simple anti-electricity theft detection of the smart meter needs to be performed locally, then it is determined that the first anti-electricity theft analysis condition is met; if at least one of the multiple spacing values is greater than the preset reference spacing value, it indicates that the box door may be abnormally opened, and a more accurate anti-electricity theft detection analysis needs to be performed through the cloud server, then it is determined that the second anti-electricity theft analysis condition is met; if at least one of the multiple connection pressure values is greater than the preset connection upper limit pressure value or at least one of the connection pressure values is less than the preset connection lower limit pressure value, it indicates that there may be abnormal line operation or circuit short circuit, and a more accurate anti-electricity theft detection analysis needs to be performed through the cloud server, then it is determined that the second anti-electricity theft analysis condition is met. It should be noted that in this embodiment,

[0038] The preset reference spacing value, the preset upper connection pressure value, and the preset lower connection pressure value can all be set according to actual needs and are not specifically limited here. It should also be noted that in this embodiment, by determining the status of the signal indicator light, it is possible to clearly distinguish between maintenance operations and electricity theft.

[0039] S120: If the first anti-electricity theft analysis condition is met, generate an anti-electricity theft analysis result based on the collected electricity usage information.

[0040] In this embodiment, the electricity usage information includes current data and voltage data. Generating an anti-theft analysis result based on the collected electricity usage information includes: calculating current balance based on the current data, and calculating power factor data based on the current data and voltage data using a power calculation formula; and generating the anti-theft analysis result based on the current balance and power factor data. It should be noted that in this embodiment, the power calculation formula is P=UIcosφ, where P is power, U is the voltage data, I is the current data, and cosφ is the power factor data. Current balance is the difference between the phase line voltage data and the neutral line voltage data. It should also be noted that in this embodiment, if the current balance is greater than a preset current balance or the power factor data exhibits an abnormal jump, the anti-theft analysis result is set as suspected theft. If the current balance is not greater than the preset current balance or the power factor data does not exhibit an abnormal jump, the anti-theft analysis result is set as normal electricity usage.

[0041] S130. If the second anti-electricity theft analysis condition is met, the electricity usage information, the collected location information, and the environmental information are uploaded to a cloud server, so that the cloud server performs an anti-electricity theft analysis based on the electricity usage information, the location information, and the environmental information through a hybrid model to obtain the anti-electricity theft analysis result, and sends the anti-electricity theft analysis result to the smart meter.

[0042] In this embodiment, the environmental information includes temperature, humidity and light intensity, such as Figure 4As shown, step S130 specifically includes steps S131-S136: S131, the cloud server preprocesses the current data and the voltage data to obtain current processed data and voltage processed data, and performs feature enhancement on the current processed data and the voltage processed data to obtain current enhanced data and voltage enhanced data; S132, calculates power factor data according to the current data and the voltage data through a power calculation formula, and performs Fourier transform and normalization on the power factor data in sequence to obtain power normalized features; S133, performs geo-hashing on the location information to generate a geo-location code; S134, normalizes the temperature and humidity to obtain normalized temperature and humidity; S135, performs logarithmic transformation on the light intensity to obtain transformed light intensity; S136, performs anti-electricity theft analysis through the hybrid model according to the current enhanced data, the voltage enhanced data, the power normalized features, the geo-location code, the normalized temperature and humidity, and the transformed light intensity to obtain the anti-electricity theft analysis result. It should be noted that in this embodiment, the preprocessing includes noise filtering, missing value repair, and normalization. Specifically, noise filtering uses a Butterworth bandpass filter to attenuate high-frequency harmonics and low-frequency drift while retaining valid signal components; missing value repair uses cubic spline interpolation, filling in the gaps using curve fitting of adjacent time data; and feature enhancement uses a sliding window. The normalization performed on the current and voltage data in the preprocessing, as well as the normalization performed on the temperature and humidity data, are both standardization processes, while the normalization performed on the power frequency domain features is minimum and maximum normalization. It should also be noted that, in this embodiment, the specific process of calculating the power factor data according to the current data and the voltage data through the power calculation formula is as described above and will not be repeated here; the power factor data is sequentially Fourier transformed and normalized to obtain a power normalized feature, specifically, the power factor data is Fourier transformed to obtain a power frequency domain feature, and the power frequency domain feature is normalized to obtain a power normalized feature; the location information is geo-hashed, which not only compresses the data dimension but also retains spatial proximity.

[0043] Further, if Figure 5As shown, step S136 specifically includes steps S1361-S1365: S1361, determining behavioral characteristics based on the current enhancement data, the voltage enhancement data, the power normalization characteristics, the geographic location code, the normalized temperature and humidity, and the transformed light intensity; S1362, splicing the current enhancement data, the voltage enhancement data, and the power normalization characteristics based on time to generate multidimensional time series characteristics; S1363, splicing the geographic location code, the normalized temperature and humidity, and the transformed light intensity based on time to generate multidimensional spatial environment characteristics; S1364, using the principal component analysis method to reduce the dimensions of the multidimensional time series characteristics and the multidimensional spatial environment characteristics to obtain reduced dimension time series characteristics and reduced dimension spatial environment characteristics; S1365, inputting the behavioral characteristics, the reduced dimension time series characteristics, and the reduced dimension spatial environment characteristics into the hybrid model to perform anti-electricity theft analysis to obtain the anti-electricity theft analysis result. It should be noted that, in this embodiment, the principal component analysis method is an unsupervised statistical method used to extract key features from high-dimensional data. It converts the original variables into a few uncorrelated principal components through linear transformation while retaining the maximum variance information in the data.

[0044] Furthermore, step S1361 specifically includes: calculating the current balance and the nighttime load proportion based on the current enhancement data and the voltage enhancement data, and marking the load abnormality jump point based on the power normalization feature; calculating the correlation between the current enhancement data and the normalized temperature and humidity to obtain the temperature-humidity-current correlation, and marking the light abnormality jump point based on the transformed light intensity; determining the user's area based on the geographic location code, and determining the regional load deviation based on the average power consumption of other users in the user's area; characterizing the behavioral characteristics based on the current balance, the nighttime load proportion, the load abnormality jump point, the temperature-humidity-current correlation, the light abnormality jump point and the regional load deviation. It should be noted that in this embodiment, the nighttime load percentage is calculated based on the current enhancement data and the voltage enhancement data, using a time-based approach. Specifically, based on a 24-hour day, the total load is first calculated, followed by the nighttime load. Finally, the nighttime load percentage is calculated based on the total load and the nighttime load. Load anomaly transition points are marked based on the power normalization feature. For example, if the power normalization feature suddenly drops from 0.9 to 0.5, the load at that moment is marked as a load anomaly transition point. The correlation between the current enhancement data and the normalized temperature and humidity is calculated using an existing correlation algorithm. Light anomaly transition points are marked based on the changing light intensity. It is understood that a sudden increase in the changing light intensity at night, indicating that the door may have been opened, is marked as a light anomaly transition point. The regional load deviation is determined based on the average electricity usage of other users in the user's area. It is understood that users in a certain area generally have lower nighttime loads. If the nighttime load of individual users suddenly increases without reasonable cause, it may indicate electricity theft. It should also be noted that in this embodiment, the calculation of current balance based on the current enhancement data is as described above and will not be repeated here. The reason why the current balance, the nighttime load proportion, the abnormal load jump point, the temperature-humidity-current correlation, the abnormal illumination jump point and the regional load deviation are used to characterize the behavioral characteristics is that the current balance characterizes whether the line integrity is abnormal, the nighttime load proportion can identify the user's irregular electricity usage behavior, the abnormal load jump point can capture the sudden change pattern of electricity usage, the temperature-humidity-current correlation can characterize the relationship between the environment and electricity usage, the abnormal illumination jump point can monitor physical operation traces, and the regional load deviation can locate abnormal individuals in a group.

[0045] Furthermore, the hybrid model includes a cross-domain attention model and a linear regression model, and step S1361 specifically includes: splicing the behavioral features, the reduced dimensionality time series features, and the reduced dimensionality space environment features to obtain a spliced feature vector; inputting the spliced feature vector into the cross-domain attention model to analyze the importance of different features to electricity theft, and assigning corresponding weight values to the behavioral features, the reduced dimensionality time series features, and the reduced dimensionality space environment features to obtain behavioral weight values, time series weight values, and spatial environment weight values; weighting the behavioral features, the reduced dimensionality time series features, and the reduced dimensionality space environment features according to the behavioral weight values, time series weight values, and spatial environment weight values to obtain weighted features; inputting the weighted features into the linear regression model to output the electricity theft probability, and generating the anti-electricity theft analysis results according to the electricity theft probability. It should be noted that in this embodiment, the cross-domain attention model includes a multi-head attention mechanism, that is, through multiple independent attention heads, to learn the feature associations of different subspaces in the spliced feature vector separately, thereby enhancing the model's ability to capture complex relationships; the cross-domain attention model is built based on a deep learning framework (such as Transformer or BERT) and is obtained through a multi-stage training strategy and specific task target optimization, where the multi-stage training strategy includes a pre-training stage and a fine-tuning stage. Specific task target optimization refers to designing and optimizing the objective function (loss function) for a specific downstream task (such as classification, generation, etc.) during the model training process, so that the model parameters are adjusted to the state that is most suitable for the task. Its core is to convert a general pre-trained model into a task-specific model, such as the cross-domain attention model in this embodiment, by adjusting the loss function, model structure or training strategy to maximize the task performance indicator. It is understandable that when the cross-domain attention model is performing weight allocation, the spatial environment weight value is minimized, and the temporal weight value and behavior weight value are not limited. It should also be noted that, in this embodiment, generating the anti-theft analysis result based on the electricity theft probability specifically includes: if the electricity theft probability is within a first electricity theft probability range (i.e., if the electricity theft probability is less than a first preset electricity theft probability), setting the anti-theft analysis result to normal electricity usage; if the electricity theft probability is within a second electricity theft probability range (i.e., if the first preset electricity theft probability is less than or equal to the electricity theft probability and less than the second preset electricity theft probability), setting the anti-theft analysis result to suspected electricity theft; and if the electricity theft probability is within a third electricity theft probability range (i.e., if the electricity theft probability is greater than or equal to the second preset electricity theft probability), setting the anti-theft analysis result to confirmed electricity theft. In this embodiment, the first and second preset electricity theft probabilities are 0.5 and 0.8, respectively. In other embodiments, the first and second preset electricity theft probabilities can be set based on actual needs.

[0046] S140: issuing graded alarms based on the anti-electricity theft analysis results.

[0047] In this embodiment, if Figure 6 As shown, step S140 specifically includes steps S141-S142: S141, if the anti-electricity theft analysis result is suspected electricity theft, triggering a primary alarm to control the buzzer to alarm, and controlling the camera to shoot the smart meter to obtain multiple images; S142, if the anti-electricity theft analysis result is confirmed electricity theft, triggering a high-level alarm to cut off the power supply. It should be noted that, in this embodiment, after triggering a primary alarm to control a buzzer to sound an alarm and controlling a camera to capture a plurality of images of the smart meter, the process further includes: uploading the plurality of images to the cloud server, so that the cloud server detects the appearance of the smart meter using the computer vision algorithm in the hybrid model based on the plurality of images to obtain a detection result, wherein the computer vision algorithm includes a traditional computer vision algorithm and a computer vision algorithm based on deep learning; and the detection of the appearance of the smart meter specifically involves checking whether the appearance of the smart meter is abnormal, for example, whether there are pry marks or whether the wires are exposed; if the detection result is that the appearance is abnormal, that is, there are pry marks or the wires are exposed, then the step of triggering a high-level alarm to cut off the power supply is executed. It should also be noted that, in this embodiment, detecting the appearance of the smart meter using the computer vision algorithm in the hybrid model can further improve the accuracy of anti-theft electricity detection.

[0048] Furthermore, after the step of triggering a high-level alarm to cut off the power supply, it may also include: uploading the anti-electricity theft analysis results, multiple spacing values, multiple connection pressure values, the signal indicator light status, the power usage information, the location information and the environmental information to the blockchain evidence storage platform to ensure that they cannot be tampered with.

[0049] In summary, in this embodiment, whether the anti-theft analysis conditions are met is first detected based on multi-dimensional data (preset polling anti-theft detection time, signal indicator status, multiple spacing values, and multiple connection pressure values). Then, when the first anti-theft analysis condition is met, the generation module in the smart meter generates an anti-theft analysis result based on the collected electricity usage information; when the second anti-theft analysis condition is met, the communication module uploads the electricity usage information, location information, and environmental information to the cloud server, so that the cloud server performs anti-theft analysis based on the electricity usage information, location information, and environmental information through a hybrid model to obtain an anti-theft analysis result. , the hybrid model includes a cross-domain attention model and a linear regression model; finally, a graded alarm is given according to the anti-electricity theft analysis results. In the entire anti-electricity theft analysis and alarm process, whether the anti-electricity theft analysis conditions are met is monitored based on multi-dimensional data, which not only effectively monitors the electricity theft in an all-round way, but also distinguishes between maintenance operations and electricity theft behaviors, thereby improving the accuracy of anti-electricity theft; and when different electricity theft analysis conditions are met, the anti-electricity theft analysis results are generated locally on the smart meter or in the cloud server based on multi-dimensional data, which not only improves the real-time nature and efficiency of anti-electricity theft, but also reduces the false alarm rate of anti-electricity theft and improves the accuracy of anti-electricity theft.

[0050] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and such modifications or substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A method for preventing electricity theft based on a smart meter, characterized in that: The smart meter includes a door, wherein the edge of the door is provided with multiple sets of ultrasonic distance sensors to collect the distance between the door and the door frame to obtain multiple distance values; the terminal area inside the door is provided with multiple sets of pressure sensors to collect the connection pressure of the terminal to obtain multiple connection pressure values. The method includes: Detecting whether an anti-electricity theft analysis condition is met according to a preset polling anti-electricity theft detection time, the obtained signal indicator light state, a plurality of the spacing values, and a plurality of the connection pressure values, wherein the anti-electricity theft analysis condition includes a first anti-electricity theft analysis condition and a second anti-electricity theft analysis condition, and a signal indicator light is provided on the smart meter; If the first anti-electricity theft analysis condition is met, generating an anti-electricity theft analysis result based on the collected electricity usage information of the smart meter; If the second anti-electricity theft analysis condition is met, uploading the electricity usage information and the collected location information and environmental information of the smart meter to a cloud server, so that the cloud server performs an anti-electricity theft analysis based on the electricity usage information, the location information, and the environmental information using a hybrid model to obtain the anti-electricity theft analysis result, and sending the anti-electricity theft analysis result to the smart meter; Producing graded alarms based on the anti-electricity theft analysis results; The detecting whether the anti-electricity theft analysis condition is met according to the preset polling anti-electricity theft detection time, the acquired signal indicator light status, the plurality of the spacing values, and the plurality of the connection pressure values includes: If the obtained signal indicator light state is not the preset indicator light state, judging the preset polling anti-electricity theft detection time, the plurality of the spacing values, and the plurality of the connection pressure values; If the preset polling anti-electricity theft detection time is reached, it is determined that the first anti-electricity theft analysis condition is met; If at least one of the plurality of spacing values is greater than a preset reference spacing value, it is determined that the second anti-electricity theft analysis condition is met; If at least one of the plurality of connection pressure values is greater than a preset upper connection pressure limit or at least one of the plurality of connection pressure values is less than a preset lower connection pressure limit, it is determined that the second anti-electricity theft analysis condition is met.

2. The method according to claim 1, characterized in that The electricity usage information includes current data and voltage data. The generating of the anti-electricity theft analysis result based on the collected electricity usage information includes: Calculating a current balance degree according to the current data, and calculating a power factor data according to the current data and the voltage data using a power calculation formula; The anti-electricity theft analysis result is generated according to the current balance and power factor data.

3. The method according to claim 2, characterized in that The environmental information includes temperature, humidity and light intensity; The cloud server performs anti-electricity theft analysis based on the electricity usage information, the location information, and the environmental information through a hybrid model to obtain the anti-electricity theft analysis result, including: The cloud server preprocesses the current data and the voltage data to obtain current processed data and voltage processed data, and performs feature enhancement on the current processed data and the voltage processed data to obtain current enhanced data and voltage enhanced data; Calculating power factor data according to the current data and the voltage data using a power calculation formula, and performing Fourier transform and normalization processing on the power factor data in sequence to obtain a power normalization feature; Performing geohashing on the location information to generate a geolocation code; Normalizing the temperature and humidity to obtain normalized temperature and humidity; Performing logarithmic transformation on the light intensity to obtain transformed light intensity; The anti-electricity theft analysis result is obtained by performing anti-electricity theft analysis through the hybrid model according to the current enhancement data, the voltage enhancement data, the power normalization feature, the geographic location code, the normalized temperature and humidity, and the transformed light intensity.

4. The method according to claim 3, characterized in that The anti-electricity theft analysis result is obtained by performing the anti-electricity theft analysis through the hybrid model according to the current enhancement data, the voltage enhancement data, the power normalization feature, the geographic location code, the normalized temperature and humidity, and the transformed light intensity, including: determining a behavior feature based on the current enhancement data, the voltage enhancement data, the power normalization feature, the geographic location code, the normalized temperature and humidity, and the transformed light intensity; Splicing the current enhancement data, the voltage enhancement data, and the power normalization feature based on time to generate a multi-dimensional time series feature; The geographic location code, the normalized temperature and humidity, and the transformed light intensity are spliced based on time to generate a multi-dimensional spatial environmental feature; Using a principal component analysis method to reduce the dimensions of the multidimensional time series features and the multidimensional spatial environment features to obtain reduced-dimensional time series features and reduced-dimensional spatial environment features; The behavioral features, the dimension-reduced time series features, and the dimension-reduced space environment features are input into the hybrid model to perform anti-electricity theft analysis to obtain the anti-electricity theft analysis result.

5. The method according to claim 4, characterized in that The determining of the behavior feature according to the current enhancement data, the voltage enhancement data, the power normalization feature, the geographic location code, the normalized temperature and humidity, and the transformed light intensity includes: Calculating the current balance degree according to the current enhancement data, calculating the nighttime load proportion according to the current enhancement data and the voltage enhancement data, and marking the load abnormality jump point according to the power normalization feature; Calculating the correlation between the current enhancement data and the normalized temperature and humidity to obtain a temperature-humidity-current correlation, and marking an abnormal illumination jump point according to the transformed illumination intensity; Determine the region where the user is located according to the geographic location code, and determine the regional load deviation according to the average power consumption of other users in the region where the user is located; The behavior characteristics are characterized according to the current balance, the nighttime load proportion, the load abnormality jump point, the temperature-humidity-current correlation, the illumination abnormality jump point, and the regional load deviation.

6. The method according to claim 4, characterized in that The hybrid model includes a cross-domain attention model and a linear regression model, and the behavioral features, the reduced-dimensionality time series features, and the reduced-dimensionality space environment features are input into the hybrid model to perform anti-electricity theft analysis to obtain the anti-electricity theft analysis result, including: Splicing the behavior feature, the reduced-dimensionality temporal feature, and the reduced-dimensionality spatial environment feature to obtain a spliced feature vector; Inputting the concatenated feature vector into the cross-domain attention model to analyze the importance of different features to electricity theft, and assigning corresponding weights to the behavioral features, the reduced-dimensionality temporal features, and the reduced-dimensionality spatial environment features to obtain behavioral weight values, temporal weight values, and spatial environment weight values; The behavior feature, the reduced-dimensionality time series feature, and the reduced-dimensionality space environment feature are weighted according to the behavior weight value, the time series weight value, and the space environment weight value to obtain a weighted feature; The weighted features are input into the linear regression model to output the probability of electricity theft, and the anti-electricity theft analysis result is generated according to the probability of electricity theft.

7. The method according to any one of claims 1 to 6, characterized in that The step of issuing a graded alarm based on the anti-electricity theft analysis result includes: If the anti-electricity theft analysis result is suspected electricity theft, a primary alarm is triggered to control a buzzer to sound an alarm, and a camera is controlled to capture multiple images of the smart meter; If the anti-electricity theft analysis result confirms electricity theft, a high-level alarm is triggered to cut off the power supply.

8. The method according to claim 7, characterized in that After triggering the primary alarm to control the buzzer to sound an alarm and controlling the camera to photograph the smart meter to obtain a plurality of photographed images, the method further includes: Uploading the plurality of captured images to the cloud server, so that the cloud server detects the appearance of the smart meter according to the plurality of captured images using the computer vision algorithm in the hybrid model to obtain a detection result; If the detection result is that the appearance is abnormal, the step of triggering a high-level alarm to cut off the power supply is executed. 9.An anti-theft system based on smart meters, characterized in that: include: A smart electricity meter, comprising a door, wherein the edge of the door is provided with a plurality of ultrasonic distance sensors for collecting the distance between the door and the door frame to obtain a plurality of distance values, and a plurality of pressure sensors are provided in the terminal area inside the door to collect the connection pressure of the terminal to obtain a plurality of connection pressure values; the smart electricity meter is provided with a signal indicator light and further comprises a detection module and a generation module, the detection module being configured to detect whether an anti-electricity theft analysis condition is satisfied based on a preset polling anti-electricity theft detection time, an acquired signal indicator light state, a plurality of the distance values, and a plurality of the connection pressure values, wherein the anti-electricity theft analysis condition comprises a first anti-electricity theft analysis condition and a second anti-electricity theft analysis condition; the generation module being configured to generate an anti-electricity theft analysis result based on the collected electricity usage information of the smart electricity meter; a cloud server comprising an anti-electricity theft analysis module and a sending module, wherein the anti-electricity theft analysis module is configured to perform an anti-electricity theft analysis based on the electricity usage information uploaded by the smart meter and the collected location information and environmental information of the smart meter using a hybrid model to obtain the anti-electricity theft analysis result, and the sending module is configured to send the anti-electricity theft analysis result to the smart meter; The smart meter further includes an alarm module, which is used to generate graded alarms based on the anti-electricity theft analysis results. The detecting whether the anti-electricity theft analysis condition is met according to the preset polling anti-electricity theft detection time, the acquired signal indicator light status, the plurality of the spacing values, and the plurality of the connection pressure values includes: If the obtained signal indicator light state is not the preset indicator light state, judging the preset polling anti-electricity theft detection time, the plurality of the spacing values, and the plurality of the connection pressure values; If the preset polling anti-electricity theft detection time is reached, it is determined that the first anti-electricity theft analysis condition is met; If at least one of the plurality of spacing values is greater than a preset reference spacing value, it is determined that the second anti-electricity theft analysis condition is met; If at least one of the plurality of connection pressure values is greater than a preset upper connection pressure limit or at least one of the plurality of connection pressure values is less than a preset lower connection pressure limit, it is determined that the second anti-electricity theft analysis condition is met.

Images