Electricity stealing behavior identification method and system

Abstract

The invention relates to the technical field of electricity stealing behavior recognition, and discloses an electricity stealing behavior recognition method and system, and the method comprises the steps: collecting electricity consumption behavior data, and converting the data into a frequency domain component; performing frequency domain analysis on the frequency domain component according to the initial frequency band demarcation point to obtain an initial frequency domain feature, and adjusting the initial frequency band demarcation point according to the initial frequency domain feature to obtain a frequency band demarcation point; according to the frequency band demarcation point, performing frequency domain analysis on the frequency domain component to obtain frequency domain features, and integrating the frequency domain features into a frequency domain feature matrix; and inputting the frequency domain feature matrix into the feature extraction model to obtain a convolution feature vector, and performing data processing on the convolution feature vector through an electricity stealing behavior detection algorithm to obtain an electricity stealing probability. According to the invention, by dynamically adjusting the frequency band demarcation point, the accuracy of frequency domain feature extraction is improved, the deep electricity consumption features are extracted through the feature extraction model, the electricity stealing probability is calculated based on the feature values after nonlinear processing, and the accuracy and efficiency of electricity stealing behavior recognition are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electricity theft identification, and in particular to an electricity theft identification method and system. Background Art

[0002] With the widespread application of power systems and the growing demand for electricity, electricity theft has become one of the important challenges faced by power companies. Electricity theft not only directly affects the economic benefits of power companies, but also disrupts the normal balance of electricity supply and demand and affects the operational stability of the power grid. In order to address this problem, power companies and related research institutions have proposed a variety of electricity theft detection methods, including real-time monitoring methods based on smart meters, anomaly detection methods based on user electricity consumption data, and behavior recognition technology based on machine learning. Among them, smart meters are widely used in the detection of electricity theft because they can collect users' electricity consumption data in real time. Anomaly detection based on user electricity consumption data analyzes historical data and identifies abnormal electricity consumption patterns to discover potential electricity theft. In recent years, with the development of artificial intelligence technology, deep learning and data mining methods have been introduced into the field of electricity theft detection. By constructing complex classification models, users' electricity consumption behaviors are pattern recognized and classified, thereby improving the accuracy and response speed of electricity theft detection.

[0003] However, the existing technology still has some shortcomings in application. The rule-based anomaly detection method relies on manually set thresholds and rules, which is difficult to adapt to complex user behaviors and changing electricity consumption patterns, and often leads to misjudgments and missed judgments. The machine learning-based electricity theft detection method is often too simple in feature extraction, fails to make full use of frequency domain information, and is difficult to effectively capture the periodic characteristics of user electricity consumption behavior. In the interactive processing of multi-dimensional features, the existing methods lack sophisticated nonlinear interaction modeling, resulting in the correlation between features being ignored, thereby affecting the accuracy of detection. Traditional electricity theft detection models are usually difficult to maintain efficient computing performance when facing large amounts of real-time data. Although deep learning models have certain advantages in feature recognition, there is still a trade-off between detection accuracy and computational complexity. How to build an efficient electricity theft detection model that can effectively capture frequency domain features and combine deep learning with complex feature interactions is still a difficulty in the current technical field. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a method and system for identifying electricity theft, which can solve the limitations of existing electricity theft identification methods in capturing the dynamic changes and periodic characteristics of electricity usage behavior, thereby achieving the effect of improving the efficiency and accuracy of electricity theft identification.

[0005] In a first aspect, the present invention provides a method for identifying electricity theft, the method comprising: Collecting electricity usage behavior data of the user to be identified, and converting the electricity usage behavior data into frequency domain components through fast Fourier transform; According to the preset initial frequency band demarcation point, the frequency domain component is subjected to frequency domain analysis to obtain initial frequency domain characteristics, and according to the initial frequency domain characteristics, the initial frequency band demarcation point is adjusted to obtain a frequency band demarcation point; According to the frequency band demarcation point, the frequency domain components are subjected to frequency domain analysis to obtain frequency domain features, and the frequency domain features are integrated into a frequency domain feature matrix, wherein the frequency domain features include high frequency energy, low frequency energy and spectrum entropy; The frequency domain feature matrix is ​​input into a preset feature extraction model to obtain a convolution feature vector, and the convolution feature vector is processed by an electricity theft behavior detection algorithm to obtain the electricity theft probability of the user to be identified.

[0006] Further, the step of performing frequency domain analysis on the frequency domain components according to the preset initial frequency band demarcation point to obtain initial frequency domain features, and adjusting the initial frequency band demarcation point according to the initial frequency domain features to obtain the frequency band demarcation point includes: According to a preset initial frequency band demarcation point, extracting high-frequency energy and low-frequency energy from the frequency domain components, and calculating an energy ratio between the high-frequency energy and the low-frequency energy; Calculate the spectral entropy based on the power spectral density of each frequency domain component; The energy ratio is compared with an energy ratio threshold, and the spectrum entropy is compared with a spectrum entropy threshold, and according to the comparison result, the initial frequency band demarcation point is adjusted to obtain a frequency band demarcation point.

[0007] Furthermore, the step of comparing the energy ratio with an energy ratio threshold, and comparing the spectrum entropy with a spectrum entropy threshold, and adjusting the initial frequency band demarcation point according to the comparison result to obtain the frequency band demarcation point includes: Determine whether the energy ratio is greater than an energy ratio threshold, if so, determine whether the spectrum entropy is greater than a spectrum entropy threshold, if so, reduce the initial frequency band demarcation point, if so, keep the initial frequency band demarcation point unchanged; If it is less than or equal to the energy ratio threshold, determine whether the spectrum entropy is less than or equal to the spectrum entropy threshold. If it is less than or equal to the spectrum entropy threshold, increase the initial frequency band boundary point. If it is greater than the spectrum entropy threshold, keep the initial frequency band boundary point unchanged.

[0008] Furthermore, the high frequency energy is calculated using the following formula: In the formula, Indicates the initial frequency band boundary point, represents the maximum frequency, represents the yth frequency component, represents the power spectral density of the yth frequency component; The low frequency energy is calculated using the following formula: In the formula, Indicates the minimum frequency; The spectrum entropy is calculated using the following formula: In the formula, represents the spectral entropy of the power spectral density of the yth frequency component, Represents the normalized power spectral density of the y-th frequency component.

[0009] Furthermore, the feature extraction model is constructed based on a deep convolutional neural network, and the feature extraction model includes three convolutional layers connected in sequence, and residual connections are set between each convolutional layer.

[0010] Furthermore, the step of performing data processing on the convolution feature vector through an electricity theft behavior detection algorithm to obtain the electricity theft probability of the user to be identified includes: Performing nonlinear feature processing on each eigenvalue in the convolution feature vector, and summing the processed eigenvalues ​​to obtain a nonlinear influence value; According to the processed eigenvalues, the interaction value of each feature pair in the convolution feature vector is calculated, and the interaction values ​​are summed to obtain a nonlinear interaction value; The electricity theft probability of the user to be identified is calculated according to the nonlinear influence value and the nonlinear interaction value.

[0011] Furthermore, the processed eigenvalue is expressed by the following formula: In the formula, x i represents the i-th eigenvalue in the convolution feature vector, and q represents the attenuation factor; The nonlinear influence value is calculated using the following formula: Where n represents the total number of eigenvalues ​​in the convolution feature vector, and c represents the bias constant; The interaction value is calculated using the following formula: In the formula, x j Represents the jth eigenvalue in the convolution feature vector; The nonlinear interaction value is calculated using the following formula: In the formula, represents the adjustment parameter; The probability of electricity theft is calculated using the following formula: In the formula, Represents the Sigmoid function.

[0012] Furthermore, after the step of obtaining the electricity theft probability of the user to be identified, the method further includes: Determining the electricity theft risk interval of the user to be identified according to the electricity theft probability and a preset electricity theft probability threshold; The power theft probability is dynamically adjusted according to the real-time power consumption behavior data of the user to be identified, and a power theft warning reminder is issued according to the changing trend of the corresponding power theft risk interval.

[0013] Furthermore, after the step of performing the electricity theft warning step, the method further includes: The electricity usage behavior data and the electricity theft analysis data are stored and backed up, and an abnormal analysis report is generated according to the electricity theft warning reminder and the electricity theft analysis data, wherein the electricity theft analysis data includes the frequency domain characteristics and the electricity theft probability.

[0014] In a second aspect, the present invention provides a system for identifying electricity theft, the system comprising: A data processing module, used to collect the electricity usage behavior data of the user to be identified, and convert the electricity usage behavior data into frequency domain components through fast Fourier transform; A feature extraction module, configured to perform frequency domain analysis on the frequency domain components according to a preset initial frequency band demarcation point to obtain an initial frequency domain feature, and to adjust the initial frequency band demarcation point according to the initial frequency domain feature to obtain a frequency band demarcation point; According to the frequency band demarcation point, the frequency domain components are subjected to frequency domain analysis to obtain frequency domain features, and the frequency domain features are integrated into a frequency domain feature matrix, wherein the frequency domain features include high frequency energy, low frequency energy and spectrum entropy; The electricity theft identification module is used to input the frequency domain feature matrix into a preset feature extraction model to obtain a convolution feature vector, and perform data processing on the convolution feature vector through an electricity theft behavior detection algorithm to obtain the electricity theft probability of the user to be identified.

[0015] The present invention provides a method and system for identifying electricity theft. The present invention can improve the accuracy of frequency domain feature extraction through a dynamic adjustment mechanism of the frequency band demarcation point, capture deeper electricity consumption characteristics in the frequency domain characteristics through a deep convolutional neural network, and calculate the probability of electricity theft based on the characteristic values ​​after nonlinear processing, which can improve the accuracy and efficiency of electricity theft identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flow chart of a method for identifying electricity theft in an embodiment of the present invention; Figure 2 It is a structural diagram of the electricity theft identification system in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] See also Figure 1 The first embodiment of the present invention provides a method for identifying electricity theft, which includes steps S10 to S40: Step S10, collecting the electricity usage behavior data of the user to be identified, and converting the electricity usage behavior data into frequency domain components through fast Fourier transform; Step S20, performing frequency domain analysis on the frequency domain components according to the preset initial frequency band demarcation point to obtain initial frequency domain features, and adjusting the initial frequency band demarcation point according to the initial frequency domain features to obtain a frequency band demarcation point; Step S30, performing frequency domain analysis on the frequency domain components according to the frequency band demarcation point to obtain frequency domain features, and integrating the frequency domain features into a frequency domain feature matrix, wherein the frequency domain features include high frequency energy, low frequency energy and spectrum entropy; Step S40: input the frequency domain feature matrix into a preset feature extraction model to obtain a convolution feature vector, and perform data processing on the convolution feature vector through an electricity theft behavior detection algorithm to obtain the electricity theft probability of the user to be identified.

[0019] The present invention determines whether the user has engaged in electricity theft by analyzing the user's electricity usage behavior data. Specifically, the electricity usage behavior data of each user is first collected. This can be done by installing a smart meter at the power entrance of each user and setting the collection frequency and collection cycle of the smart meter to collect data. The electricity usage behavior data includes the current, voltage and power at different time points within the collection cycle. By installing a smart meter at the power entrance of each user and setting a reasonable collection frequency and collection cycle, the user's electricity usage behavior data can be accurately collected, including the changes in current, voltage and power at different time points. This refined data collection method can not only monitor the user's electricity usage in real time, but also provide high-precision basic data for subsequent abnormal detection and electricity theft behavior identification, significantly improving the system's ability to analyze complex electricity usage patterns and the accuracy of identifying abnormal behaviors.

[0020] Then, the collected electricity usage behavior data is preprocessed, and the preprocessing here includes using a statistical anomaly detection algorithm to identify and remove noise and erroneous electricity usage behavior data, and the preprocessed electricity usage behavior data is sorted according to timestamps. Preferably, Z-Score can be used to detect electricity usage behavior data that obviously deviates from the normal range, thereby removing abnormal and greatly fluctuating noise data, using linear interpolation to complete the missing electricity usage behavior data, and standardizing the completed electricity usage behavior data. Of course, other data preprocessing methods can also be used to remove noise and complete missing values ​​for electricity usage behavior data, and there are no excessive restrictions here.

[0021] In this embodiment, data integrity and consistency are ensured through data preprocessing, which can improve the accuracy and reliability of subsequent analysis and model prediction. Sorting the data by timestamp further ensures the continuity and effectiveness of the time series analysis, which helps to more accurately capture the user's electricity consumption behavior patterns.

[0022] For the preprocessed electricity consumption behavior data, the present invention uses a frequency domain analysis method to extract the frequency domain features of the electricity consumption behavior data. Specifically, the electricity consumption behavior data is first converted into frequency domain components by fast Fourier transform. Fast Fourier transform is a conventional signal processing tool for obtaining frequency information from the time domain. For the obtained frequency domain components, based on the preset initial frequency band demarcation point, the initial frequency domain features of the frequency domain components are extracted, and the initial frequency domain features are analyzed. The initial frequency band demarcation point is dynamically adjusted according to the analysis result, and the frequency domain features are re-extracted according to the adjusted frequency band demarcation point, so as to improve the accuracy of feature extraction. The specific adjustment steps include: According to a preset initial frequency band demarcation point, extracting high-frequency energy and low-frequency energy from the frequency domain components, and calculating an energy ratio between the high-frequency energy and the low-frequency energy; Calculate the spectral entropy based on the power spectral density of each frequency domain component; The energy ratio is compared with an energy ratio threshold, and the spectrum entropy is compared with a spectrum entropy threshold, and according to the comparison result, the initial frequency band demarcation point is adjusted to obtain a frequency band demarcation point.

[0023] In this embodiment, firstly, the high-frequency energy and the low-frequency energy are calculated according to the obtained frequency domain components and the set initial frequency band demarcation point, and the formulas are: in, Indicates the initial frequency band demarcation point, which is used to divide high-frequency energy and low-frequency energy. The frequency band demarcation point is adaptively adjusted according to changes in user behavior. Indicates the maximum frequency of the current spectrum analysis. Indicates the minimum frequency of the current spectrum analysis. represents the yth frequency component after fast Fourier transform, represents the power spectral density of the yth frequency component, that is, the power spectral density of the frequency component is represented by square modulus of the frequency component. Indicates the increment for each frequency bin in the integral calculation.

[0024] According to the obtained high-frequency energy and low-frequency energy, the energy ratio of high-frequency energy to low-frequency energy is calculated. The formula is: Then calculate the total energy of the entire frequency domain between the minimum frequency and the maximum frequency in the current spectrum analysis, that is, the sum of the power spectrum densities of all frequency domain components, normalize the power spectrum density of each frequency domain component, and calculate the spectrum entropy of each power spectrum density. The formula is: In the formula, represents the spectral entropy of the power spectral density of the yth frequency component, is the normalized power spectral density of the yth frequency component.

[0025] As an indicator of signal complexity, spectrum entropy can help determine whether the user's power consumption behavior is complex peak behavior or stable basic load behavior. Then, based on the energy ratio of high-frequency energy and low-frequency energy and the spectrum entropy of each power spectrum density, the energy ratio change threshold and spectrum entropy threshold are set to adjust the frequency band boundary point and dynamically adjust the high-frequency and low-frequency frequency band ranges. The specific adjustment steps include: Determine whether the energy ratio is greater than an energy ratio threshold, if so, determine whether the spectrum entropy is greater than a spectrum entropy threshold, if so, reduce the initial frequency band demarcation point, if so, keep the initial frequency band demarcation point unchanged; If it is less than or equal to the energy ratio threshold, determine whether the spectrum entropy is less than or equal to the spectrum entropy threshold. If it is less than or equal to the spectrum entropy threshold, increase the initial frequency band boundary point. If it is greater than the spectrum entropy threshold, keep the initial frequency band boundary point unchanged.

[0026] In this embodiment, the energy ratio is first compared with the energy ratio threshold. When the energy ratio is greater than the energy ratio change threshold, it means that the energy proportion of the high-frequency part is too high, and there may be abnormal spike behavior. It is necessary to expand the analysis of the high frequency, and then determine whether the spectrum entropy of each power spectrum density is greater than or equal to the spectrum entropy threshold. If the spectrum entropy is greater than or equal to the spectrum entropy threshold, the value of the initial frequency band dividing point is lowered, otherwise the initial frequency band dividing point remains unchanged; when the energy ratio is less than or equal to the energy ratio change threshold, it means that the low-frequency part is dominant, indicating normal daily electricity consumption behavior, and it is necessary to expand the analysis of the low frequency, then determine whether the spectrum entropy of each power spectrum density is less than or equal to the spectrum entropy threshold. If the spectrum entropy is less than or equal to the spectrum entropy threshold, the value of the initial frequency band dividing point is increased, otherwise the initial frequency band dividing point remains unchanged.

[0027] Based on the dynamically adjusted frequency band boundary point, the frequency domain features of the frequency domain components, namely high-frequency energy, low-frequency energy and spectrum entropy, are re-extracted, and the extracted frequency domain features are integrated into the frequency domain feature matrix F, which is expressed as follows: in, They represent the high-frequency energy, low-frequency energy and spectrum entropy of the user at time point t respectively.

[0028] High-frequency energy can reflect the user's peak power consumption behavior. For example, the startup of high-power equipment or power theft is usually manifested as an increase in the energy of high-frequency signals. Low-frequency energy usually corresponds to the user's daily basic load power consumption behavior. If the proportion of low-frequency energy is significantly reduced relative to high-frequency energy, it may be a signal of abnormal behavior; Spectral entropy is used to reflect the complexity of the signal. High spectral entropy means that the signal distribution is more complex and may contain spike behavior and irregular load. Low spectral entropy means that the signal is more stable and usually represents the basic load.

[0029] In this embodiment, the frequency band demarcation point is dynamically adjusted to improve the accuracy and flexibility of electricity theft identification. Different from the fixed frequency band division scheme in the prior art, this embodiment monitors the user's electricity consumption behavior in real time, combines the change trend of high-frequency energy, low-frequency energy, high-low frequency energy ratio and spectrum entropy, and dynamically adjusts the high-frequency and low-frequency demarcation points. When the high-frequency energy ratio exceeds the threshold, the frequency band demarcation point is automatically lowered to expand the analysis of the high-frequency part to capture peak load and electricity theft behavior. When the low-frequency energy dominates, the demarcation point is automatically increased to focus on daily basic load behavior, reduce excessive calculation of the high-frequency part, and save computing resources. The introduction of spectrum entropy realizes the judgment of the complexity and stability of the signal. High spectrum entropy usually reflects complex peak behavior, while low spectrum entropy corresponds to stable basic load. This adaptive adjustment mechanism ensures that the method provided by the present invention can flexibly adapt to the electricity consumption patterns of different users. Especially when the electricity consumption behavior is complex or changes rapidly, the dynamic adjustment mechanism can significantly improve the accuracy of feature extraction, effectively reduce misjudgment and missed judgment, and enhance the sensitivity and accuracy of abnormality detection.

[0030] Based on the above-mentioned frequency domain feature matrix, the present invention performs convolution feature extraction on the frequency domain feature matrix through a feature extraction model constructed by a deep convolutional neural network. Specifically, the feature extraction model consists of three sequentially connected convolutional layers. Preferably, each layer uses a 3×3 convolution kernel, a step size of 1, a filling method of "same", and the number of convolution kernels in each layer is set to 32; the local features in the frequency domain feature matrix F are extracted by the first layer of convolution kernels, the output of the first layer of convolution is input to the second layer of convolution kernels to extract deep features, and the output of the second layer of convolution is input to the There are three convolutional layers, and residual connections are added between each convolution layer. The input frequency domain feature matrix F is added to the output of the first convolution layer to obtain the residual output of the first layer. The residual output of the first layer is added to the output of the second convolution layer to obtain the residual output of the second layer, and the residual output of the second layer is added to the output of the third convolution layer to obtain the residual output of the third layer. After the three-layer convolution operation and residual connection, the residual output of the third layer is converted into a one-dimensional feature vector through global average pooling, and the one-dimensional feature vector is used as the final output convolution feature vector.

[0031] For the extracted convolution feature vector, in a preferred embodiment, the present invention provides an electricity theft behavior detection algorithm to process the convolution feature vector, thereby obtaining the probability of electricity theft of the user, and the specific processing steps include: Performing nonlinear feature processing on each eigenvalue in the convolution feature vector, and summing the processed eigenvalues ​​to obtain a nonlinear influence value; According to the processed eigenvalues, the interaction value of each feature pair in the convolution feature vector is calculated, and the interaction values ​​are summed to obtain a nonlinear interaction value; The electricity theft probability of the user to be identified is calculated according to the nonlinear influence value and the nonlinear interaction value.

[0032] In this embodiment, nonlinear feature processing is performed on each eigenvalue in the convolution feature vector. The nonlinear feature processing includes logarithmic smoothing processing, exponential decay nonlinear processing and polynomial combination. The eigenvalue after nonlinear feature processing can be expressed as: Among them, θ(x i ) represents the i-th eigenvalue in the convolution feature vector after the nonlinear feature function, x i represents the i-th eigenvalue in the convolution feature vector, and q represents the attenuation factor.

[0033] In the nonlinear feature processing of this embodiment, It represents the natural logarithm transformation of the eigenvalue, which is used for smoothing to reduce the impact of extreme values ​​on subsequent calculations. 0.1 is the smoothing coefficient to prevent the denominator from having a zero value and ensure calculation stability. Represents an exponential decay term. By adjusting the decay factor, it enhances the slight differences in the features and avoids the influence of too large or too small eigenvalues ​​on the calculation. The square term representing the eigenvalue is used as a nonlinear enhancement term to improve the sensitivity of the feature to small fluctuations. Through the combination of the above three nonlinear feature processing, the feature recognition ability is effectively improved in terms of smoothing, small value sensitivity and nonlinear enhancement, so that each feature is better represented at different scales, thereby providing high-quality input for subsequent feature interactions.

[0034] Sum all eigenvalues ​​after nonlinear feature processing and add the bias constant c to obtain the nonlinear influence value A of all eigenvalues. The formula is: Where n represents the total number of eigenvalues ​​in the convolution feature vector, and c represents the bias constant, which is used to adjust the output value range of the numerator.

[0035] Based on the eigenvalues ​​after nonlinear feature processing, the interaction value of each feature pair in the convolution feature vector is calculated. The formula is: In the formula, represents the interaction value between the i-th eigenvalue and the j-th eigenvalue in the convolution feature vector, Represents the product of two eigenvalues ​​after nonlinear processing, which is used to introduce the linear combination of two features. Represents the absolute difference between two eigenvalues ​​and is used to measure the similarity between two features. It is used as a smoothing denominator to prevent the difference from being too small, resulting in unstable calculations.

[0036] The above interaction formula enables the association between each pair of features to be captured more carefully. This design makes the interaction value not limited to simple multiplication or addition, but can reflect the nonlinear and differential relationship between feature pairs, which is particularly suitable for detecting multi-dimensional electricity theft.

[0037] By summing the interaction values ​​of all feature pairs, the nonlinear interaction value B between all feature pairs is calculated, and the formula is: in, It represents the adjustment parameter, which is used to control the weight of the interaction value as the denominator in the overall calculation in the subsequent calculation of the probability of electricity theft, and n represents the total number of eigenvalues ​​in the convolution feature vector.

[0038] After the nonlinear influence value and nonlinear interaction value are calculated through the above steps, the user's power theft probability P can be calculated based on these two values. The formula is: In the formula, P represents the user's probability of stealing electricity, Represents the Sigmoid function, which is used to convert values ​​into the range of [0,1].

[0039] The present invention proposes a method for identifying electricity theft. In the process of identifying electricity theft, a deep convolutional neural network is used to perform multi-layer convolution feature extraction on a frequency domain feature matrix. The three convolutional layers are set with the same convolution kernel size and step size to ensure the coherence and consistency of feature extraction. The residual connection between each convolutional layer further enhances the learning ability of the model. By retaining the input feature information and enhancing the depth of feature extraction layer by layer, the loss of information in the convolution operation is minimized. This network structure design improves the feature retention and transmission capabilities of the model, so that more detailed and deep-level electricity usage characteristics in the user's electricity usage behavior can be captured in the detection of electricity theft, especially more sensitive to small abnormal changes in the data. After the convolution feature vector is generated, each eigenvalue is further optimized by three nonlinear processing methods: logarithmic smoothing, exponential decay and polynomial combination, so that the feature not only has the advantages of extreme value smoothing processing, reducing the influence of extreme data values ​​on the results, but also is more sensitive to subtle differences through exponential decay, making tiny abnormal usage The electrical features can be effectively identified, and the polynomial enhancement term further improves the ability to identify feature fluctuations, making each feature more expressive at different scales, which is conducive to identifying abnormal power usage behavior in a complex power usage environment. The feature pair interaction value constructed based on the convolution feature vector can carefully capture the nonlinear and differential relationship between the feature values, avoiding the shortcomings of simple weighting or multiplication methods. Through the composite calculation in the interaction formula, the complex correlation relationship between the feature pairs is more comprehensively expressed, so that even when there is a weak correlation between the features, it can be identified and reflected. This multi-level interaction design of the present invention ensures that it can cope with the multi-dimensional feature dependence that may exist in the power theft behavior, thereby further improving the accuracy of power theft identification; the power theft probability is calculated through a comprehensive power theft behavior detection algorithm, and it is converted into the range of [0,1], so that the detection result can intuitively present the user's risk level, which is convenient for subsequent real-time response and graded processing, and provides strong technical support for realizing efficient and accurate power theft behavior identification and processing.

[0040] In a preferred embodiment, after obtaining the electricity theft probability of the user to be identified, the present invention divides the risk interval according to the electricity theft probability, dynamically adjusts the electricity theft probability of the user and issues an early warning according to the risk interval. The specific steps include: Determining the electricity theft risk interval of the user to be identified according to the electricity theft probability and a preset electricity theft probability threshold; The power theft probability is dynamically adjusted according to the real-time power consumption behavior data of the user to be identified, and a power theft warning reminder is issued according to the changing trend of the corresponding power theft risk interval.

[0041] In this embodiment, firstly, according to the calculated power theft probability P' of each user and the preset power theft probability threshold R 1 and R2 , to determine the user's electricity theft risk interval: like , then it is determined that the user's electricity theft probability is in a low-risk range; like , then the user's electricity theft probability is judged to be in the medium risk range; like , it is determined that the user's probability of stealing electricity is in a high-risk range.

[0042] Then, the user's electricity theft probability is continuously adjusted according to the real-time electricity usage behavior data, and corresponding measures are taken when the change in the electricity theft probability causes the risk range to change. Preferably, if the user's electricity theft probability remains in the low-risk range, no early warning measures are taken and normal monitoring is maintained. If the user's electricity theft probability rises to the medium-risk range, a first-level early warning is triggered, the user's electricity usage behavior changes are automatically recorded, and a reminder notification is sent to the user.

[0043] If the user's probability of electricity theft remains in the medium-risk range, the changes in the user's electricity usage behavior will be automatically recorded, and a notification will be sent to the user to remind him that his behavior is risky to avoid escalating to high-risk. If the user's probability of electricity theft drops to the low-risk range, the first-level warning will be lifted and monitoring will continue. If the user's probability of electricity theft rises to the high-risk range, the second-level warning will be triggered, and power management personnel will conduct on-site inspections and take automated restriction measures, including limiting the user's electricity power.

[0044] If the user's probability of electricity theft remains in the high-risk range, the second-level warning status will be maintained and compulsory measures will be added, including power outages and reporting to higher-level regulatory authorities. If the user's probability of electricity theft drops to the medium-risk range, the warning will be downgraded to a first-level warning, automatically recording changes in the user's electricity usage behavior and sending a reminder notification to the user.

[0045] In this embodiment, accurate monitoring and timely warning of electricity theft are achieved by dynamically adjusting the user's probability of electricity theft and dividing the risk intervals. According to the changes in the user's probability of electricity theft, graded warnings and corresponding measures are taken to effectively reduce the risk of electricity theft. Normal monitoring is maintained in the low-risk interval to avoid unnecessary interference. Reminders and records are made in the medium-risk interval to prevent risk escalation. Mandatory measures are taken in the high-risk interval, such as on-site inspections and power outages, to ensure electricity safety. At the same time, the situation is reported to higher-level regulatory authorities, thereby improving the efficiency and safety of electricity management.

[0046] In another preferred embodiment, the present invention stores and backs up all data and generates an abnormal warning analysis report, and the specific steps include: The electricity usage behavior data and the electricity theft analysis data are stored and backed up, and an abnormal analysis report is generated according to the electricity theft warning reminder and the electricity theft analysis data, wherein the electricity theft analysis data includes the frequency domain characteristics and the electricity theft probability.

[0047] In this embodiment, a distributed database system is adopted, and a compression algorithm is used to compress and store all data. The data here include the user's electricity usage behavior data and the electricity theft analysis data generated in the electricity theft identification analysis, such as frequency domain characteristics and electricity theft probability data. At the same time, the database is backed up regularly, and an automatic feedback mechanism is established. After the user's abnormal electricity theft behavior is detected and an early warning is triggered, an abnormal analysis report is automatically generated. The report content includes the time of abnormal occurrence, user ID, and detected feature values, and the abnormal analysis report is sent to relevant management personnel.

[0048] This embodiment adopts a distributed database system and a compression algorithm to efficiently compress, store and regularly back up all data to ensure data security and scalability. The established automatic feedback mechanism can generate an abnormal analysis report containing detailed information in a timely manner after detecting abnormal electricity theft behavior of users and triggering an early warning, and automatically send it to relevant managers. This process not only improves the efficiency of data management and ensures the stability and reliability of power system operation, but also achieves rapid response and accurate analysis, enhances the monitoring and management capabilities of the power system, and helps to take quick measures to prevent further losses.

[0049] See also Figure 2 Based on the same inventive concept, a second embodiment of the present invention provides a system for identifying electricity theft, including: The data processing module 10 is used to collect the electricity usage behavior data of the user to be identified, and convert the electricity usage behavior data into frequency domain components through fast Fourier transform; The feature extraction module 20 is used to perform frequency domain analysis on the frequency domain components according to the preset initial frequency band demarcation point to obtain initial frequency domain features, and adjust the initial frequency band demarcation point according to the initial frequency domain features to obtain the frequency band demarcation point; According to the frequency band demarcation point, the frequency domain components are subjected to frequency domain analysis to obtain frequency domain features, and the frequency domain features are integrated into a frequency domain feature matrix, wherein the frequency domain features include high frequency energy, low frequency energy and spectrum entropy; The electricity theft identification module 30 is used to input the frequency domain feature matrix into a preset feature extraction model to obtain a convolution feature vector, and perform data processing on the convolution feature vector through an electricity theft behavior detection algorithm to obtain the electricity theft probability of the user to be identified.

[0050] The technical features and technical effects of the electricity theft identification system proposed in the embodiment of the present invention are the same as those of the method proposed in the embodiment of the present invention, and are not described in detail here. Each module in the above-mentioned electricity theft identification system can be implemented in whole or in part by software, hardware, and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0051] In summary, an embodiment of the present invention proposes a method and system for identifying electricity theft. The method collects electricity usage behavior data of a user to be identified, and converts the electricity usage behavior data into frequency domain components through fast Fourier transform; according to a preset initial frequency band demarcation point, frequency domain analysis is performed on the frequency domain components to obtain initial frequency domain features, and according to the initial frequency domain features, the initial frequency band demarcation point is adjusted to obtain a frequency band demarcation point; according to the frequency band demarcation point, frequency domain analysis is performed on the frequency domain components to obtain frequency domain features, and the frequency domain features are integrated into a frequency domain feature matrix, wherein the frequency domain features include high-frequency energy, low-frequency energy and spectrum entropy; the frequency domain feature matrix is ​​input into a preset feature extraction model to obtain a convolution feature vector, and the convolution feature vector is data processed through an electricity theft behavior detection algorithm to obtain the electricity theft probability of the user to be identified. The present invention improves the accuracy of frequency domain feature extraction through the dynamic adjustment mechanism of frequency band demarcation points, extracts deeper power consumption characteristics in the frequency domain characteristics through a deep convolutional neural network, and calculates the probability of electricity theft based on the characteristic values ​​after nonlinear processing, thereby improving the accuracy and efficiency of electricity theft behavior identification.

[0052] Each embodiment in this specification is described in a progressive manner, and the same or similar parts of each embodiment can be directly referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. It should be noted that the technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above-mentioned embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0053] The above-mentioned embodiments only express several preferred implementation modes of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in the technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be based on the protection scope of the claims.

Claims

1. A method for identifying electricity theft, characterized in that: include: Collecting electricity usage behavior data of the user to be identified, and converting the electricity usage behavior data into frequency domain components through fast Fourier transform; According to the preset initial frequency band demarcation point, the frequency domain component is subjected to frequency domain analysis to obtain initial frequency domain characteristics, and according to the initial frequency domain characteristics, the initial frequency band demarcation point is adjusted to obtain a frequency band demarcation point; According to the frequency band demarcation point, the frequency domain components are subjected to frequency domain analysis to obtain frequency domain features, and the frequency domain features are integrated into a frequency domain feature matrix, wherein the frequency domain features include high frequency energy, low frequency energy and spectrum entropy; The frequency domain feature matrix is ​​input into a preset feature extraction model to obtain a convolution feature vector, and the convolution feature vector is processed by an electricity theft behavior detection algorithm to obtain the electricity theft probability of the user to be identified.

2. The method for identifying electricity theft according to claim 1, characterized in that: The step of performing frequency domain analysis on the frequency domain components according to the preset initial frequency band demarcation point to obtain initial frequency domain features, and adjusting the initial frequency band demarcation point according to the initial frequency domain features to obtain the frequency band demarcation point comprises: According to a preset initial frequency band demarcation point, extracting high-frequency energy and low-frequency energy from the frequency domain components, and calculating an energy ratio between the high-frequency energy and the low-frequency energy; Calculate the spectral entropy based on the power spectral density of each frequency domain component; The energy ratio is compared with an energy ratio threshold, and the spectrum entropy is compared with a spectrum entropy threshold, and according to the comparison result, the initial frequency band demarcation point is adjusted to obtain a frequency band demarcation point.

3. The method for identifying electricity theft according to claim 2, characterized in that: The step of comparing the energy ratio with an energy ratio threshold, and comparing the spectrum entropy with a spectrum entropy threshold, and adjusting the initial frequency band demarcation point according to the comparison result to obtain the frequency band demarcation point comprises: Determine whether the energy ratio is greater than an energy ratio threshold, if so, determine whether the spectrum entropy is greater than a spectrum entropy threshold, if so, reduce the initial frequency band demarcation point, if so, keep the initial frequency band demarcation point unchanged; If it is less than or equal to the energy ratio threshold, determine whether the spectrum entropy is less than or equal to the spectrum entropy threshold. If it is less than or equal to the spectrum entropy threshold, increase the initial frequency band boundary point. If it is greater than the spectrum entropy threshold, keep the initial frequency band boundary point unchanged.

4. The method for identifying electricity theft according to claim 2, characterized in that: The high frequency energy is calculated using the following formula: In the formula, Indicates the initial frequency band boundary point, represents the maximum frequency, represents the yth frequency component, represents the power spectral density of the yth frequency component; The low frequency energy is calculated using the following formula: In the formula, Indicates the minimum frequency; The spectrum entropy is calculated using the following formula: In the formula, represents the spectral entropy of the power spectral density of the yth frequency component, Represents the normalized power spectral density of the y-th frequency component.

5. The method for identifying electricity theft according to claim 1, characterized in that: The feature extraction model is constructed based on a deep convolutional neural network. The feature extraction model includes three convolutional layers connected in sequence, and residual connections are set between each convolutional layer.

6. The method for identifying electricity theft according to claim 1, characterized in that: The step of performing data processing on the convolution feature vector through an electricity theft behavior detection algorithm to obtain the electricity theft probability of the user to be identified comprises: Performing nonlinear feature processing on each eigenvalue in the convolution feature vector, and summing the processed eigenvalues ​​to obtain a nonlinear influence value; According to the processed eigenvalues, the interaction value of each feature pair in the convolution feature vector is calculated, and the interaction values ​​are summed to obtain a nonlinear interaction value; The electricity theft probability of the user to be identified is calculated according to the nonlinear influence value and the nonlinear interaction value.

7. The method for identifying electricity theft according to claim 6, characterized in that: The following formula is used to express the processed eigenvalue: In the formula, x i represents the i-th eigenvalue in the convolution feature vector, and q represents the attenuation factor; The nonlinear influence value is calculated using the following formula: Where n represents the total number of eigenvalues ​​in the convolution feature vector, and c represents the bias constant; The interaction value is calculated using the following formula: In the formula, x j Represents the jth eigenvalue in the convolution feature vector; The nonlinear interaction value is calculated using the following formula: In the formula, represents the adjustment parameter; The probability of electricity theft is calculated using the following formula: In the formula, Represents the Sigmoid function.

8. The method for identifying electricity theft according to claim 1, characterized in that: After the step of obtaining the electricity theft probability of the user to be identified, the method further includes: Determining the electricity theft risk interval of the user to be identified according to the electricity theft probability and a preset electricity theft probability threshold; The power theft probability is dynamically adjusted according to the real-time power consumption behavior data of the user to be identified, and a power theft warning reminder is issued according to the changing trend of the corresponding power theft risk interval.

9. The method for identifying electricity theft according to claim 8, characterized in that: After the step of providing an early warning reminder for electricity theft, the method further includes: The electricity usage behavior data and the electricity theft analysis data are stored and backed up, and an abnormal analysis report is generated according to the electricity theft warning reminder and the electricity theft analysis data, wherein the electricity theft analysis data includes the frequency domain characteristics and the electricity theft probability.

10. A system for identifying electricity theft, characterized in that: include: A data processing module, used to collect the electricity usage behavior data of the user to be identified, and convert the electricity usage behavior data into frequency domain components through fast Fourier transform; A feature extraction module, configured to perform frequency domain analysis on the frequency domain components according to a preset initial frequency band demarcation point to obtain an initial frequency domain feature, and to adjust the initial frequency band demarcation point according to the initial frequency domain feature to obtain a frequency band demarcation point; According to the frequency band demarcation point, the frequency domain components are subjected to frequency domain analysis to obtain frequency domain features, and the frequency domain features are integrated into a frequency domain feature matrix, wherein the frequency domain features include high frequency energy, low frequency energy and spectrum entropy; The electricity theft identification module is used to input the frequency domain feature matrix into a preset feature extraction model to obtain a convolution feature vector, and perform data processing on the convolution feature vector through an electricity theft behavior detection algorithm to obtain the electricity theft probability of the user to be identified.

Images