Electricity theft detection method, device, equipment, storage medium and program product

Abstract

The present application relates to a method, device, equipment, storage medium and program product for detecting electricity theft, and belongs to the technical field of electricity theft detection. The electricity theft detection method includes: obtaining an enhanced electricity theft sample, the enhanced electricity theft sample includes multiple electricity usage information and the electricity theft type corresponding to the electricity usage information; calculating the separation degree of the electricity theft type in the enhanced electricity theft sample, the separation degree is used to indicate the difficulty of distinguishing the electricity theft type from other electricity theft types, and using the enhanced electricity theft sample and the separation degree of the electricity theft type to obtain a trained classification model for each electricity theft type; using the trained classification model to classify the input electricity theft suspicion data, and determine the electricity theft type and user information of the electricity theft suspicion data. The electricity theft detection method provided by the present application can improve the efficiency of electricity theft detection.

Description

Technical Field

[0001] The present invention relates to the technical field of electricity theft detection, and in particular to an electricity theft detection method, device, equipment, storage medium and program product. Background Art

[0002] With the rapid development of social economy, the demand for electricity is increasing. Although the huge demand for electricity has promoted the steady development of the power industry, it has also increased the occurrence of electricity theft. Electricity theft will seriously affect the profits of enterprises and harm the development of the national economy. Therefore, electricity theft detection is of great significance to the profits of power grid companies and the safety and stability of power grid operation.

[0003] Traditional methods of detecting electricity theft mostly use grassroots power workers' inspections, on-site inspections and evidence collection to investigate electricity theft. However, this detection method is prone to users' non-cooperation, blocking power workers, and deliberately disrupting the scene during implementation. Therefore, there is a problem of low efficiency in electricity theft detection. Summary of the invention

[0004] Based on this, it is necessary to provide a method, device, equipment, storage medium and program product for electricity theft detection in response to the above technical problems, which can improve the problem of low efficiency of electricity theft detection.

[0005] A first aspect of an embodiment of the present application provides a method for detecting electricity theft, the method comprising: obtaining an enhanced electricity theft sample, the enhanced electricity theft sample comprising a plurality of electricity usage information and electricity theft types corresponding to the electricity usage information; calculating a degree of separation of electricity theft types in the enhanced electricity theft sample, the degree of separation being used to indicate the difficulty of distinguishing an electricity theft type from other electricity theft types, and obtaining a trained classification model for each electricity theft type using the enhanced electricity theft sample and the degree of separation of the electricity theft type; and classifying input suspected electricity theft data using the trained classification model, and determining the electricity theft type and user information of the suspected electricity theft data.

[0006] In one of the embodiments, obtaining an enhanced electricity theft sample includes: obtaining a real electricity theft sample, using the real electricity theft sample to train a generative adversarial model, the generative adversarial module includes a generator and a discriminator, the generator is used to generate a first augmented electricity theft sample according to random noise, and the discriminator is used to discriminate between the real electricity theft sample and the first augmented electricity theft sample; using the real electricity theft sample and the generator in the generative adversarial model obtained after training to generate an enhanced electricity theft sample.

[0007] In one of the embodiments, the method further includes: extracting features from real electricity theft samples to obtain feature vectors of the real electricity theft samples; correspondingly, generating enhanced electricity theft samples using a generator in a generative adversarial model obtained after training, including: generating enhanced electricity theft samples using the feature vectors of the real electricity theft samples and the generator in the generative adversarial model obtained after training.

[0008] In one of the embodiments, the separation degree of the enhanced electricity theft samples and the electricity theft types is used to obtain a trained classification model for each electricity theft type, including: classifying the enhanced electricity theft samples according to the separation degree to obtain multiple enhanced electricity theft sample subsets; and training multiple preset classification models respectively according to the enhanced electricity theft sample subsets to obtain a trained classification model for each electricity theft type.

[0009] In one of the embodiments, the classification model includes a decision plane. After the input suspected electricity theft data is classified using the classification model, the method further includes: calculating the distance between the suspected electricity theft data and the decision plane of the corresponding classification model. If the distance is less than a preset threshold, the suspected electricity theft data is secondary classified using the KNN algorithm, and the electricity theft type and user information of the suspected electricity theft data are determined.

[0010] In one embodiment, calculating the separation degree of the electricity theft types in the enhanced electricity theft sample includes: calculating the separation degree of the electricity theft types in the enhanced electricity theft sample using a projection vector method.

[0011] A second aspect of an embodiment of the present application provides a power theft detection device, the device comprising:

[0012] An acquisition module, used to acquire an enhanced electricity theft sample, the enhanced electricity theft sample including a plurality of electricity usage information and the electricity theft type corresponding to the electricity usage information;

[0013] A calculation module, used to calculate the separation degree of the electricity theft type in the enhanced electricity theft sample, the separation degree is used to indicate the difficulty of distinguishing the electricity theft type from other electricity theft types, and to obtain a trained classification model for each electricity theft type by using the enhanced electricity theft sample and the separation degree of the electricity theft type;

[0014] The processing module is used to classify the input suspected electricity theft data using the classification model, and determine the electricity theft type and user information of the suspected electricity theft data.

[0015] A third aspect of an embodiment of the present application provides a computer device, which includes: a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the steps of any method of the first aspect of the embodiment of the present application are implemented.

[0016] A fourth aspect of an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of any method of the first aspect of the embodiment of the present application are implemented.

[0017] A fifth aspect of an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of any method of the first aspect of the embodiment of the present application.

[0018] The beneficial effects brought by the technical solution provided by the embodiment of the present application include at least:

[0019] The electricity theft detection method provided in the embodiment of the present application obtains an enhanced electricity theft sample, wherein the enhanced electricity theft sample includes a plurality of electricity usage information and the electricity theft type corresponding to the electricity usage information, and then calculates the separation degree of the electricity theft type in the enhanced electricity theft sample, the separation degree is used to indicate the difficulty of distinguishing the electricity theft type from other electricity theft types, and uses the enhanced electricity theft sample and the separation degree of the electricity theft type to obtain a trained classification model for each electricity theft type, and finally uses the trained classification model to classify the input electricity theft suspicion data, and determine the electricity theft type and user information of the electricity theft suspicion data. The electricity theft detection method provided in the embodiment of the present application can improve the efficiency of electricity theft detection and the accuracy of electricity theft detection because it calculates the separation degree of the electricity theft type, and then classifies the input electricity theft suspicion data according to the corresponding trained classification model, and determines the electricity theft type and user information of the electricity theft suspicion data. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A flowchart of a method for detecting electricity theft provided in an embodiment of the present application;

[0021] Figure 2 A flowchart of another electricity theft detection method provided in an embodiment of the present application;

[0022] Figure 3 A schematic diagram of the loss function of the discriminator provided in an embodiment of the present application;

[0023] Figure 4 A schematic diagram of a loss function of a generator provided in an embodiment of the present application;

[0024] Figure 5 A schematic diagram of 10 rounds of sample training provided in an embodiment of the present application;

[0025] Figure 6 A schematic diagram of 30 rounds of sample training provided in an embodiment of the present application;

[0026] Figure 7 A schematic diagram of 100 rounds of sample training provided in an embodiment of the present application;

[0027] Figure 8 A schematic diagram for comparing the accuracy of electricity theft detection provided in an embodiment of the present application;

[0028] Fig. 9 A schematic diagram for comparing false detection rates of electricity theft detection provided in an embodiment of the present application;

[0029] Fig.10 A structural diagram of an electricity theft detection device is provided for an embodiment of the present application;

[0030] Fig.11 A schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0031] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0032] With the rapid development of social economy, the demand for electricity is increasing. Although the huge demand for electricity has promoted the steady development of the power industry, it has also increased the occurrence of electricity theft. Electricity theft will seriously affect the profits of enterprises and harm the development of the national economy. Therefore, electricity theft detection is of great significance to the profits of power grid companies and the safety and stability of power grid operation.

[0033] Traditional methods of detecting electricity theft mostly use grassroots power workers' inspections, on-site inspections and evidence collection to investigate electricity theft. However, this detection method is prone to users' non-cooperation, blocking power workers, and deliberately disrupting the scene during implementation. Therefore, there is a problem of low efficiency in electricity theft detection.

[0034] Based on this, it is necessary to provide a method for detecting electricity theft in order to solve the above technical problems, so as to improve the efficiency of detecting electricity theft.

[0035] In the electricity theft detection method, an enhanced electricity theft sample is obtained, wherein the enhanced electricity theft sample includes a plurality of electricity usage information and the electricity theft type corresponding to the electricity usage information, and then the separation degree of the electricity theft type in the enhanced electricity theft sample is calculated, the separation degree is used to indicate the difficulty of distinguishing the electricity theft type from other electricity theft types, and the enhanced electricity theft sample and the separation degree of the electricity theft type are used to obtain a trained classification model for each electricity theft type, and finally the trained classification model is used to classify the input electricity theft suspicion data, and determine the electricity theft type and user information of the electricity theft suspicion data. The electricity theft detection method provided in the embodiment of the present application can improve the efficiency and accuracy of electricity theft detection because it calculates the separation degree of the electricity theft type, and then classifies the input electricity theft suspicion data according to the corresponding trained classification model, and determines the electricity theft type and user information of the electricity theft suspicion data.

[0036] It should be noted that the execution subject of the electricity theft detection method provided in the embodiment of the present application may be an electricity theft detection device, which may be implemented as part or all of a computer device through software, hardware, or a combination of software and hardware. In the following method embodiments, the execution subject is a computer device as an example for explanation.

[0037] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0038] Please refer to Figure 1 , which shows a flow chart of a method for detecting electricity theft provided by an embodiment of the present application, and the method for detecting electricity theft can be applied to the computer device mentioned above. Figure 1 As shown, the electricity theft detection method may include the following steps:

[0039] Step 101: Obtain an enhanced electricity theft sample.

[0040] Among them, the enhanced electricity theft sample includes multiple electricity usage information and the electricity theft type corresponding to the electricity usage information. The electricity usage information includes: user identification information, electricity usage information of the user, current information or voltage information, etc. The electricity theft type is determined based on the analysis and judgment of the electricity usage information, current information or voltage information.

[0041] Step 102: Calculate the separation degree of electricity theft types in the enhanced electricity theft samples, and use the enhanced electricity theft samples and the separation degree of electricity theft types to obtain a trained classification model for each electricity theft type.

[0042] The degree of separation is used to indicate the difficulty of distinguishing one type of electricity theft from other types of electricity theft.

[0043] In practical applications, the projection vector method is used to calculate the degree of separation between classes. The types of leaf nodes, i.e. different types of electricity theft, are determined according to the degree of separation. SVM is trained at the decision points of the decision tree for classification, and multiple classification models of support vector machines based on decision trees are obtained.

[0044] Step 103: Classify the input suspected electricity theft data using the trained classification model, and determine the electricity theft type and user information of the suspected electricity theft data.

[0045] The electricity theft detection method provided in the embodiment of the present application obtains an enhanced electricity theft sample, wherein the enhanced electricity theft sample includes a plurality of electricity usage information and the electricity theft type corresponding to the electricity usage information, and then calculates the separation degree of the electricity theft type in the enhanced electricity theft sample, the separation degree is used to indicate the difficulty of distinguishing the electricity theft type from other electricity theft types, and uses the enhanced electricity theft sample and the separation degree of the electricity theft type to obtain a trained classification model for each electricity theft type, and finally uses the trained classification model to classify the input electricity theft suspicion data, and determine the electricity theft type and user information of the electricity theft suspicion data. The electricity theft detection method provided in the embodiment of the present application can improve the efficiency of electricity theft detection and the accuracy of electricity theft detection because it calculates the separation degree of the electricity theft type, and then classifies the input electricity theft suspicion data according to the corresponding trained classification model, and determines the electricity theft type and user information of the electricity theft suspicion data.

[0046] In one embodiment, obtaining an enhanced electricity theft sample includes: obtaining a real electricity theft sample, using the real electricity theft sample to train a generative adversarial model, the generative adversarial model includes a generator and a discriminator, the generator is used to generate a first augmented electricity theft sample according to random noise, and the discriminator is used to discriminate between the real electricity theft sample and the first augmented electricity theft sample;

[0047] Enhanced electricity theft samples are generated using real electricity theft samples and the generator in the generative adversarial model obtained after training.

[0048] In practical applications, the training process of the generative adversarial model includes: initializing the parameters of the VAE encoder, the generator and discriminator networks of the WGAN, and the real electricity theft samples are passed through the VAE encoder to obtain a low-dimensional latent vector Z carrying the main features of the input information source; when training the discriminator network, the obtained latent vector Z and the noise distribution are input into the generator to generate electricity theft sample data, the generator G is fixed, and then the constraint conditions are used as the optimization target, the loss value of the discriminator is calculated, and the network parameters are updated using the Adam optimizer, and the discriminator is trained to distinguish between true and false as much as possible. When training the generator network, after cyclically updating the discriminator k times, the discriminator network weights are fixed, the generator network loss value is calculated in the same way, and the network parameters are updated using Adam. The generator G is updated once, so that the discriminator cannot distinguish between true and false as much as possible. After multiple update iterations, the adversarial training between the discriminator and the generator reaches a Nash equilibrium, and the discriminator cannot determine whether the data comes from real data or data generated by the generator.

[0049] In one of the embodiments, the method further includes: extracting features from real electricity theft samples to obtain feature vectors of the real electricity theft samples; correspondingly, generating enhanced electricity theft samples using a generator in a generative adversarial model obtained after training, including: generating enhanced electricity theft samples using the feature vectors of the real electricity theft samples and the generator in the generative adversarial model obtained after training.

[0050] In practical applications, an n-dimensional real electricity theft sample is input, and the m-dimensional sample mean and m-dimensional sample variance are obtained through the VAE encoding network; ε is sampled on N(0,1), and the latent variable is calculated using the equation z=ε×σ+μ, so that the latent variable satisfies z~N(μ,σ 2 );The generated factors and Input into the generator; the generating factor generates a sample set through the generating network. The generated sample set is a false label. The sample and the real sample with the true label in the original training set are used as the training set of the discriminant network. The positive discriminant network is trained as a binary classifier, and the probability value close to 1 is output for the real sample, and the probability value close to 0 is output for the generated sample. The trained parameters of the positive discriminant network are transferred to the negative discriminant network, that is, the relevant information of the positive network is transferred to the negative network. The discriminant network of the negative has the same ability to distinguish true from false as the positive discriminant network, but the discriminant network outputs a probability value close to 0 for the real sample and a probability value close to 1 for the generated sample. The reason for the confrontation between the two is that there is a contradiction between the two discriminant networks. The negative discriminant network is trained using a mixed sample set consisting of the sample set generated by the generative network and the real samples in the training set, and the parameters of the generative network are updated to guide the generative network to generate more "real" samples, so that the positive discriminant network cannot distinguish whether the sample is a real sample or a generated sample, until both the positive discriminant network and the negative discriminant network cannot distinguish the true or false of the sample, and the output judgment probability value is close to 0.5; finally, it is judged whether the authenticity constraint and similarity constraint are met; if they are met, the corresponding enhanced electricity theft sample is obtained, otherwise the data is discarded and the above training process is repeated.

[0051] Among them, the network of the data generation model based on VAE-WGAN consists of three parts, namely the VAE encoder, the GAN generator and the discriminator. The objective functions of each network part are as follows:

[0052] The objective function of the encoder of the data generation model is:

[0053] max[γ1·E q(z∣x) [logp(X|Z)]-D KL (q(z|x)‖p(z))]

[0054]

[0055] The objective function of the generator in the data generation model is:

[0056]

[0057] The objective function of the discriminator in the data generation model is:

[0058]

[0059] Where: p data is the true distribution of the original sample; x is the data that obeys the distribution of the original sample; Represents the generated sample with original sample information; according to Generated sample data; z represents the generated factor of the input with original features; q(z|x) is the model of the encoder; γ1 and γ2 are hyperparameters.

[0060] The constraints of the VAE-WGAN-based data generation model are:

[0061] The authenticity constraint is to ensure that the generated data is close to the actual situation, and the authenticity loss L r Defined as:

[0062] L r =W(G(z);x); where: G(z) represents the generated data of the generator; W(G(z);x) represents the Wasserstein distance between the generated data and the real sample.

[0063] The Wasserstein distance is defined as:

[0064]

[0065] Where: ∏(p z ,p data ) is p z and p data is the set of joint distributions γ of marginal distributions; W(G(z); x) represents the infimum of γ(x,z) expectation, which means p z Fitting p data The distance x needs to be moved to z.

[0066] The generated data should be as similar as possible to the actual electricity theft data, so the similarity loss is defined as: L s =||G(z),I||2, where I represents the real sample, and the bi-norm is used to measure the similarity of the two matrices.

[0067] Therefore, the final constraints for data generation are (real and similar):

[0068] In one embodiment, a trained classification model for each electricity theft type is obtained by enhancing the separation of electricity theft samples and electricity theft types, including:

[0069] Classify the enhanced electricity theft samples according to the separation degree to obtain multiple enhanced electricity theft sample subsets;

[0070] The preset multiple classification models are trained respectively according to the enhanced electricity theft sample subset to obtain a trained classification model for each electricity theft type.

[0071] First, feature extraction is performed on real electricity theft samples to obtain feature vectors of real electricity theft samples, and then enhanced electricity theft samples are generated based on the feature vectors and the trained generator, which can improve the generation efficiency of enhanced electricity theft samples.

[0072] In one of the embodiments, the classification model includes a decision plane. After the input suspected electricity theft data is classified using the classification model, the method further includes: calculating the distance between the suspected electricity theft data and the decision plane of the corresponding classification model. If the distance is less than a preset threshold, the suspected electricity theft data is secondary classified using the KNN algorithm, and the electricity theft type and user information of the suspected electricity theft data are determined.

[0073] Among them, the projection vector method can be used to calculate the separation degree of electricity theft types in the enhanced electricity theft samples.

[0074] The classification module is a classification model of support vector machine (SVM). SVM is trained for classification at the decision points of the decision tree to obtain a multi-classification model of support vector machine based on the decision tree, the decision plane of SVM, and SVM is used for classification. The distance between the suspected electricity theft user and the decision plane is calculated and it is determined whether the distance is greater than a given threshold. If not, KNN is used for secondary correction based on the determined support vector to obtain the number and type of the suspected electricity theft user. If so, the number and type of the suspected electricity theft user are obtained, and the accuracy, false detection rate and other model evaluation indicators are used for the obtained number and type of the suspected electricity theft user.

[0075] The training of the classification model includes: Hierarchical classifiers will produce error accumulation, so it is necessary to construct a partial binary decision tree with high accuracy and small classification error, which will avoid the impact of the misclassification of the upper nodes on the classification of the lower nodes, so the type of leaf nodes is determined according to the size of the separation degree. The separation degree is calculated by projection vector.

[0076] For a training set of N categories of samples, let the vector set of the i-th electricity theft category be Among them l i is the number of samples of the i-th sample, then the class center of this class of samples is:

[0077]

[0078] Given a set of electricity theft samples X i ={x1,x2,…,x ni} and X j ={y1,y2,…,y nj}, calculate X i With X j The distance between the cluster centers in the feature space H:

[0079]

[0080] Calculate X i With X j The maximum projection distance in the feature space H and And the number of samples in the intersection class domain space sum i and sum j :

[0081]

[0082]

[0083] when X i and X j The inter-class separation measure is:

[0084] when X i and X j The inter-class separation measure is:

[0085]

[0086] Electricity Theft Sample Category X i Separation test with other types of electricity theft samples Sep i The calculation formula is:

[0087]

[0088] Sep i The larger the value, the easier it is to separate the electricity theft method. If the separability values ​​of the electricity theft methods are all greater than 0, it means that this type of method does not intersect with other electricity theft methods; if it is less than 0, select Sep i The electricity theft method represented by the maximum value is separated.

[0089] The construction method of the decision tree in the multi-class classification method of the support vector machine based on the decision tree is to separate one class from the rest of the classes at each layer, and continue in the same way until all classes have been classified. Finally, the leaf nodes represent the results of various types of electricity theft. The classification process is:

[0090] 1) SVM1, the first decision point of the decision tree, separates the first category of electricity theft features from the second category, the third category, ..., the Nth category and normal samples, and constructs classifier SVM1;

[0091] 2) The SVMi of the i-th decision point of the decision tree separates the i-th category of electricity theft features from the i+1-th category, the i+2-th category, …, the N-th category and normal samples, and constructs a classifier SVMi;

[0092] 3) The last decision point SVMn of the decision tree separates the n-th type of electricity theft features from normal samples and constructs the classifier SVMn.

[0093] Classification strategy of the electricity theft detection model

[0094] Given (x i ,f i )(i=1,2,…,N) are training samples in the electricity theft sample set, where x i is the electricity theft characteristic index corresponding to the electricity theft sample, f i ∈{-1,+1} is x i The label of f i = +1 indicates that the user has stolen electricity, M is the number of training samples, and the support vector machine model can be expressed as:

[0095]

[0096] Where: C is the slack variable ξ i The penalty coefficient.

[0097] The sample data of electricity theft is disorganized and belongs to the situation of nonlinear indivisibility. The kernel function K(x i ,x j )= <x i ,x j >Instead of <x i ,x j >, the kernel function can map the nonlinear indivisible situation of the support vector machine in the low-dimensional space to the linearly separable situation in the high-dimensional space, construct a linear decision boundary, and divide the samples to be tested into corresponding classes. Using the Lagrangian function, the optimization problem is simplified to the form of a quadratic programming problem:

[0098]

[0099] Where: K(·) is the kernel function.

[0100] This paper uses Gaussian radial basis function, and the kernel function can be written as:

[0101]

[0102] By solving the above problem, we can obtain the optimal classification hyperplane with a boundary

[0103]

[0104] The corresponding decision function for nonlinear classification in the input space based on SVM is:

[0105]

[0106] Where: w is the normal vector of the decision boundary; b is the regularization parameter in the m-dimensional space. (where b∈ξ m ).

[0107] {(s1,f1),(s2,f2),…,(s n ,s n )},x i ∈R n , support vector near the SVM decision plane, y i is the label of the data sample set. The electricity theft sample is a nonlinear sample. When performing KNN classification, the distance between the sample to be identified and each support vector in the feature space is calculated, and the class to which the sample to be classified belongs is determined by the distance. Therefore, the distance calculation formula is as follows:

[0108] d(f,f i )=||φ(x)-φ(f i )|| 2 =k(s,s)-2k(s,x i )+k(x i ,x i )

[0109] Where: The kernel function selects RBF, where x represents the support vector of the SVM decision plane, and f i Represents the sample to be corrected on the decision plane. If there is x i to {s1,s2,…s k Any distance of these k samples is smaller than that of sample x i The distance between the remaining nk samples is called sample x. i If the attribute of most samples among these k samples is f i , then the sample x i The property of f i .

[0110] D(x) represents the distance from sample f to the decision plane, and its calculation formula is:

[0111]

[0112] The ideal threshold d should be able to distinguish as many misclassified positive samples as possible, and to judge as few original negative samples as possible, so as to ensure that the least negative samples are used for secondary correction by the KNN algorithm. Therefore, the calculation formula of the threshold d is:

[0113]

[0114] If the value of D(x) is within the range of [0, d], it means that the sample is near the decision plane. We believe that the sample may be a misjudged sample. We use the support vector of the decision plane as a reference and use KNN for secondary classification to determine the category of the sample. If the value of D(x) is greater than d, SVM can accurately identify the category of the sample. In this step, the input is the test sample within the range of [0, d] and the support vector obtained after training SVM, and the category of the sample to be tested is output.

[0115] See also Figure 2 , which shows another electricity theft detection method provided by an embodiment of the present application, the method comprising the following steps:

[0116] Step 201: obtain a real electricity theft sample, and use the real electricity theft sample to train a generative adversarial model, where the generative adversarial model includes a generator and a discriminator;

[0117] Step 202: Generate enhanced electricity theft samples using real electricity theft samples and the generator in the generative adversarial model obtained after training.

[0118] Step 203: extract features from real electricity theft samples to obtain feature vectors of real electricity theft samples;

[0119] Step 204: Generate an enhanced electricity theft sample using the feature vector of the real electricity theft sample and the generator in the generative adversarial model obtained after training.

[0120] Step 205: Calculate the separation degree of the electricity theft types in the enhanced electricity theft samples by using the projection vector method.

[0121] Step 206: Obtain a trained classification model for each electricity theft type by using the enhanced electricity theft samples and the separation degree of the electricity theft types;

[0122] Step 207: Classify the enhanced electricity theft samples according to the separation degree to obtain multiple enhanced electricity theft sample subsets;

[0123] Step 208: Train the preset multiple classification models respectively according to the enhanced electricity theft sample subset to obtain a trained classification model for each electricity theft type.

[0124] Step 209: Calculate the distance between the suspected electricity theft data and the decision plane of the corresponding classification model. If the distance is less than a preset threshold, use the KNN algorithm to perform secondary classification on the suspected electricity theft data.

[0125] Step 210: Determine the electricity theft type and user information of the suspected electricity theft data.

[0126] Among them, the real electricity theft samples in the embodiment of the present application are real data from a certain area. The data have been desensitized. According to the traditional electricity theft detection method, a variety of different electricity theft methods are obtained. The various different electricity theft methods are analyzed and summarized, and a total of six different electricity theft methods are obtained. The distribution of the data is shown in Table 1. It can be found that the number of normal users is more than 50 times the number of abnormal users, and there is a serious class imbalance problem between normal samples and abnormal samples. This shows that in the problem of electricity theft detection, corresponding strategies need to be considered to solve the problem of class imbalance.

[0127] Table 1 Distribution of electricity stealing users

[0128] Data Description Original Dataset Number of electricity theft users / person 1098 Number of users of electricity theft type 1 / person 196 Number of users of electricity theft type 2 / person 94 Number of users of electricity theft type 3 / person 128 Number of users of electricity theft type 4 / person 146 Number of users of electricity theft type 5 / person 289 Number of users of electricity theft type 6 / person 245 Ratio of normal users to abnormal users 50.56

[0129] Figure 3 The loss function of the discriminator provided in the embodiment of the present application, Figure 4 The loss function of the generator provided in the embodiment of the present application. From the convergence characteristics of the loss function of the discriminator, it can be seen that the loss function of the GAN discriminator is oscillating and the convergence effect is not good. WGAN and VAE-WGAN have better convergence characteristics, but the method provided in the embodiment of the present application converges earlier than WGAN. From the convergence characteristics of the loss function of the generator, it can be seen that the convergence characteristics of the GAN generator are relatively good at the beginning, but it will oscillate after a certain number of iterations. VAE, WGAN and VAE-WGAN all have good convergence characteristics, but the method provided in the embodiment of the present application has a faster convergence speed than the other two methods. It is found through the loss function curves of the generator and the discriminator that the training process of GAN is unstable, while WAGN and VAE-WGAN have good stability, but the algorithm proposed by the method provided in the embodiment of the present application has a faster convergence speed and better convergence characteristics than WGAN.

[0130] Figure 5 A schematic diagram of 10 rounds of sample training provided in an embodiment of the present application, Figure 6 A schematic diagram of 30 rounds of sample training provided in an embodiment of the present application, Figure 7 The schematic diagram of 100 rounds of sample training provided in the embodiment of the present application shows that the time series generated after 10 rounds of training is quite different from the actual load series of the user; the gap between the generated sequence and the actual sequence after 30 rounds of training is narrowed, and the generated samples are close to the distribution of the real electricity theft samples; when the training reaches 100 rounds, the generated samples are very close to the real electricity theft curve, and the curve at this time basically coincides with the real curve. However, it can be found that the values ​​of the generated samples are not completely consistent with the original samples, and the two maintain similar fluctuation patterns, which ensures the difference between the synthesized electricity theft samples and the real samples, increases the diversity of electricity theft samples, and improves the generalization ability of the electricity theft detection model.

[0131] Figure 8 A schematic diagram comparing the accuracy of electricity theft detection provided in the embodiments of the present application, Fig. 9 A schematic diagram comparing the false detection rate of electricity theft detection provided by the embodiment of the present application. By comparison, it can be found that the use of a data generation algorithm to balance the number of normal samples and abnormal samples improves the accuracy of the detection algorithm, but the more accurate the generated samples are, the more the accuracy is improved. Under the same data generation model, the algorithm proposed in this article has a higher accuracy rate, and under the same detection model, the data generation model proposed in this article has better generation accuracy.

[0132] The electricity theft detection method provided in the embodiment of the present application obtains an enhanced electricity theft sample, wherein the enhanced electricity theft sample includes a plurality of electricity usage information and the electricity theft type corresponding to the electricity usage information, and then calculates the separation degree of the electricity theft type in the enhanced electricity theft sample, the separation degree is used to indicate the difficulty of distinguishing the electricity theft type from other electricity theft types, and uses the enhanced electricity theft sample and the separation degree of the electricity theft type to obtain a trained classification model for each electricity theft type, and finally uses the trained classification model to classify the input electricity theft suspicion data, and determine the electricity theft type and user information of the electricity theft suspicion data. The electricity theft detection method provided in the embodiment of the present application can improve the efficiency of electricity theft detection and the accuracy of electricity theft detection because it calculates the separation degree of the electricity theft type, and then classifies the input electricity theft suspicion data according to the corresponding trained classification model, and determines the electricity theft type and user information of the electricity theft suspicion data.

[0133] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0134] like Fig.10 As shown, an embodiment of the present application provides a power theft detection device, which includes: an acquisition module 11, a calculation module 12 and a processing module 13.

[0135] Among them, the acquisition module 11 is used to obtain enhanced electricity theft samples, which include multiple electricity usage information and electricity theft types corresponding to the electricity usage information; the calculation module 12 is used to calculate the separation degree of the electricity theft types in the enhanced electricity theft samples, which is used to indicate the difficulty of distinguishing the electricity theft type from other electricity theft types, and use the enhanced electricity theft samples and the separation degree of the electricity theft type to obtain a trained classification model for each electricity theft type; the processing module 13 is used to use the classification model to classify the input suspected electricity theft data, and determine the electricity theft type and user information of the suspected electricity theft data.

[0136] In one of the embodiments, the acquisition module 11 is specifically used to: obtain real electricity theft samples, use the real electricity theft samples to train a generative adversarial model, the generative adversarial model includes a generator and a discriminator, the generator is used to generate a first expanded electricity theft sample according to random noise, and the discriminator is used to discriminate between the real electricity theft sample and the first expanded electricity theft sample; use the real electricity theft sample and the generator in the generative adversarial model obtained after training to generate an enhanced electricity theft sample.

[0137] In one embodiment, the device further includes: an extraction module 14, which is used to extract features from real electricity theft samples to obtain feature vectors of the real electricity theft samples; correspondingly, the acquisition module 11 is also used to generate enhanced electricity theft samples using the feature vectors of the real electricity theft samples and the generator in the generative adversarial model obtained after training.

[0138] In one embodiment, the processing module 13 is specifically used to obtain a trained classification model for each electricity theft type using the enhanced electricity theft samples and the degree of separation of the electricity theft types, including: classifying the enhanced electricity theft samples according to the degree of separation to obtain multiple enhanced electricity theft sample subsets; and training multiple preset classification models respectively according to the enhanced electricity theft sample subsets to obtain a trained classification model for each electricity theft type.

[0139] In one embodiment, the classification model includes a decision plane, and the device also includes: a classification module 15, which is used to calculate the distance between the suspected electricity theft data and the decision plane of the corresponding classification model. If the distance is less than a preset threshold, the KNN algorithm is used to perform secondary classification on the suspected electricity theft data, and determine the electricity theft type and user information of the suspected electricity theft data.

[0140] In one embodiment, the calculation module 12 is specifically used to: calculate the separation degree of the electricity theft types in the enhanced electricity theft sample by using the projection vector method.

[0141] The electricity theft detection device provided in the embodiment of the present application can implement the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0142] Each module in the above-mentioned electricity theft detection device can be implemented in whole or in part by software, hardware or a combination thereof. Each of 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 corresponding operations of each of the above modules.

[0143] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig.11 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a server resource protection method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0144] Those skilled in the art will understand that Fig.11 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0145] In one embodiment of the present application, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the following steps when executing the computer program: splitting application programs of various applications in the current device according to functions, and encapsulating the split application programs to obtain multiple function programs; receiving an acquisition request sent by a third-party device, wherein the acquisition request includes identification information of a target function program, and determining the target function program according to the identification information of the target function program; and sending the determined target function program to the third-party device, so that the third-party device embeds the received target function program into a preset program architecture to generate a target application.

[0146] In one embodiment of the present application, the processor further implements the following steps when executing the computer program: obtaining an enhanced electricity theft sample, the enhanced electricity theft sample including a plurality of electricity usage information and the electricity theft type corresponding to the electricity usage information; calculating the separation degree of the electricity theft type in the enhanced electricity theft sample, the magnitude of the separation degree is used to indicate the difficulty of distinguishing the electricity theft type from other electricity theft types, and using the enhanced electricity theft sample and the separation degree of the electricity theft type to obtain a trained classification model for each electricity theft type; using the trained classification model to classify the input suspected electricity theft data, and determine the electricity theft type and user information of the suspected electricity theft data.

[0147] In one embodiment of the present application, the processor further implements the following steps when executing the computer program: obtaining a real electricity theft sample, using the real electricity theft sample to train a generative adversarial model, the generative adversarial model including a generator and a discriminator, the generator is used to generate a first augmented electricity theft sample based on random noise, and the discriminator is used to discriminate between the real electricity theft sample and the first augmented electricity theft sample; using the real electricity theft sample and the generator in the generative adversarial model obtained after training to generate an enhanced electricity theft sample.

[0148] In one embodiment of the present application, the processor further implements the following steps when executing the computer program: extracting features from real electricity theft samples to obtain feature vectors of the real electricity theft samples; and generating enhanced electricity theft samples using the feature vectors of the real electricity theft samples and the generator in the generative adversarial model obtained after training.

[0149] In one embodiment of the present application, the processor further implements the following steps when executing the computer program: classifying the enhanced electricity theft samples according to the degree of separation to obtain multiple enhanced electricity theft sample subsets; and training multiple preset classification models respectively according to the enhanced electricity theft sample subsets to obtain a trained classification model for each electricity theft type.

[0150] In one embodiment of the present application, the processor further implements the following steps when executing the computer program: calculating the distance between the suspected electricity theft data and the decision plane of the corresponding classification model; if the distance is less than a preset threshold, performing secondary classification on the suspected electricity theft data using the KNN algorithm, and determining the electricity theft type and user information of the suspected electricity theft data.

[0151] In one embodiment of the present application, when the processor executes the computer program, the following steps are further implemented: using a projection vector method to calculate the separation degree of the electricity theft types in the enhanced electricity theft sample.

[0152] The computer device provided in the embodiment of the present application has similar implementation principles and technical effects to those of the above-mentioned method embodiment, and will not be described in detail here.

[0153] In one embodiment of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: application programs of each application in the current device are split according to function, and the split application programs are packaged to obtain multiple function programs; an acquisition request sent by a third-party device is received, the acquisition request includes identification information of the target function program, and the target function program is determined according to the identification information of the target function program; the determined target function program is sent to the third-party device, so that the third-party device embeds the received target function program into a preset program architecture to generate a target application.

[0154] In one embodiment of the present application, when the computer program is executed by a processor, the following steps are implemented: obtaining an enhanced electricity theft sample, the enhanced electricity theft sample including a plurality of electricity usage information and the electricity theft type corresponding to the electricity usage information; calculating the separation degree of the electricity theft type in the enhanced electricity theft sample, the magnitude of the separation degree is used to indicate the difficulty of distinguishing the electricity theft type from other electricity theft types, and using the enhanced electricity theft sample and the separation degree of the electricity theft type to obtain a trained classification model for each electricity theft type; using the trained classification model to classify the input suspected electricity theft data, and determine the electricity theft type and user information of the suspected electricity theft data.

[0155] In one embodiment of the present application, when the computer program is executed by a processor, the following steps are implemented: obtaining a real electricity theft sample, using the real electricity theft sample to train a generative adversarial model, the generative adversarial model includes a generator and a discriminator, the generator is used to generate a first augmented electricity theft sample according to random noise, and the discriminator is used to discriminate between the real electricity theft sample and the first augmented electricity theft sample; using the real electricity theft sample and the generator in the generative adversarial model obtained after training to generate an enhanced electricity theft sample.

[0156] In one embodiment of the present application, when the computer program is executed by a processor, the following steps are implemented: feature extraction is performed on real electricity theft samples to obtain feature vectors of the real electricity theft samples; and enhanced electricity theft samples are generated using the feature vectors of the real electricity theft samples and the generator in the generative adversarial model obtained after training.

[0157] In one embodiment of the present application, when the computer program is executed by a processor, the following steps are implemented: classifying the enhanced electricity theft samples according to the degree of separation to obtain multiple enhanced electricity theft sample subsets; and training multiple preset classification models respectively according to the enhanced electricity theft sample subsets to obtain a trained classification model for each electricity theft type.

[0158] In one embodiment of the present application, when the computer program is executed by a processor, the following steps are implemented: calculating the distance between the suspected electricity theft data and the decision plane of the corresponding classification model; if the distance is less than a preset threshold, performing secondary classification on the suspected electricity theft data using the KNN algorithm, and determining the electricity theft type and user information of the suspected electricity theft data.

[0159] In one embodiment of the present application, when the computer program is executed by a processor, the following steps are implemented: using a projection vector method to calculate the separation degree of the electricity theft types in the enhanced electricity theft sample.

[0160] The computer-readable storage medium provided in this embodiment has similar implementation principles and technical effects to those of the above method embodiments, and will not be described in detail here.

[0161] In one embodiment of the present application, a computer program product is provided, including a computer program, which implements the following steps when executed by a processor: splitting application programs of each application in the current device according to function, and encapsulating the split application programs to obtain multiple function programs; receiving an acquisition request sent by a third-party device, the acquisition request including identification information of the target function program, and determining the target function program according to the identification information of the target function program; sending the determined target function program to the third-party device, so that the third-party device embeds the received target function program into a preset program architecture to generate a target application.

[0162] In one embodiment of the present application, when the computer program is executed by a processor, the following steps are implemented: obtaining an enhanced electricity theft sample, the enhanced electricity theft sample including a plurality of electricity usage information and the electricity theft type corresponding to the electricity usage information; calculating the separation degree of the electricity theft type in the enhanced electricity theft sample, the magnitude of the separation degree is used to indicate the difficulty of distinguishing the electricity theft type from other electricity theft types, and using the enhanced electricity theft sample and the separation degree of the electricity theft type to obtain a trained classification model for each electricity theft type; using the trained classification model to classify the input suspected electricity theft data, and determine the electricity theft type and user information of the suspected electricity theft data.

[0163] In one embodiment of the present application, when the computer program is executed by a processor, the following steps are implemented: obtaining a real electricity theft sample, using the real electricity theft sample to train a generative adversarial model, the generative adversarial model includes a generator and a discriminator, the generator is used to generate a first augmented electricity theft sample according to random noise, and the discriminator is used to discriminate between the real electricity theft sample and the first augmented electricity theft sample; using the real electricity theft sample and the generator in the generative adversarial model obtained after training to generate an enhanced electricity theft sample.

[0164] In one embodiment of the present application, when the computer program is executed by a processor, the following steps are implemented: feature extraction is performed on real electricity theft samples to obtain feature vectors of the real electricity theft samples; and enhanced electricity theft samples are generated using the feature vectors of the real electricity theft samples and the generator in the generative adversarial model obtained after training.

[0165] In one embodiment of the present application, when the computer program is executed by a processor, the following steps are implemented: classifying the enhanced electricity theft samples according to the degree of separation to obtain multiple enhanced electricity theft sample subsets; and training multiple preset classification models respectively according to the enhanced electricity theft sample subsets to obtain a trained classification model for each electricity theft type.

[0166] In one embodiment of the present application, when the computer program is executed by a processor, the following steps are implemented: calculating the distance between the suspected electricity theft data and the decision plane of the corresponding classification model; if the distance is less than a preset threshold, performing secondary classification on the suspected electricity theft data using the KNN algorithm, and determining the electricity theft type and user information of the suspected electricity theft data.

[0167] In one embodiment of the present application, when the computer program is executed by a processor, the following steps are implemented: using a projection vector method to calculate the separation degree of the electricity theft types in the enhanced electricity theft sample.

[0168] The implementation principle and technical effects of the computer program product provided in this embodiment are similar to those of the above method embodiments and will not be described in detail here.

[0169] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited thereto.

[0170] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are 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.

[0171] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for detecting electricity theft, characterized in that: The method comprises: Acquire an enhanced electricity theft sample, wherein the enhanced electricity theft sample includes a plurality of electricity usage information and an electricity theft type corresponding to the electricity usage information; Calculating the separation degree of the electricity theft types in the enhanced electricity theft samples, wherein the magnitude of the separation degree is used to indicate the difficulty of distinguishing any electricity theft type from other electricity theft types, and obtaining a trained classification model for each electricity theft type using the enhanced electricity theft samples and the separation degree of the electricity theft types; Using the trained classification model to classify the input suspected electricity theft data, and determine the electricity theft type and user information of the suspected electricity theft data; The calculating the separation degree of the electricity theft types in the enhanced electricity theft sample comprises: Calculating the separation degree of electricity theft types in the enhanced electricity theft sample by using a projection vector method; The method of obtaining a trained classification model for each electricity theft type by using the enhanced electricity theft samples and the separation degree of the electricity theft types includes: The types of leaf nodes, ie, different types of electricity theft, are determined according to the degree of separation of the electricity theft types. SVM is trained at the decision points of the decision tree for classification, and multiple classification models of support vector machines based on the decision tree are obtained.

2. The method according to claim 1, characterized in that The step of obtaining an enhanced electricity stealing sample includes: Acquire a real electricity theft sample, and use the real electricity theft sample to train a generative adversarial model, wherein the generative adversarial model includes a generator and a discriminator, wherein the generator is used to generate a first expanded electricity theft sample according to random noise, and the discriminator is used to discriminate between the real electricity theft sample and the first expanded electricity theft sample; The enhanced electricity theft sample is generated by using the real electricity theft sample and the generator in the generative adversarial model obtained after training.

3. The method according to claim 2, characterized in that The method further comprises: extracting features from the real electricity theft sample to obtain a feature vector of the real electricity theft sample; Correspondingly, the generating the enhanced electricity theft sample by using the generator in the generative adversarial model obtained after training includes: generating the enhanced electricity theft sample by using the feature vector of the real electricity theft sample and the generator in the generative adversarial model obtained after training.

4. The method according to claim 1, characterized in that The method of obtaining a trained classification model for each electricity theft type by using the enhanced electricity theft samples and the separation degree of the electricity theft types includes: Classifying the enhanced electricity theft samples according to the separation degree to obtain a plurality of enhanced electricity theft sample subsets; A plurality of preset classification models are trained respectively according to the enhanced electricity theft sample subset to obtain a trained classification model for each electricity theft type.

5. The method according to claim 4, characterized in that The classification model includes a decision plane. After the classification model is used to classify the input suspected electricity theft data, the method further includes: The distance between the suspected electricity theft data and the corresponding decision plane of the classification model is calculated. If the distance is less than a preset threshold, the suspected electricity theft data is secondary classified using a KNN algorithm, and the electricity theft type and user information of the suspected electricity theft data are determined.

6. An electricity theft detection device, characterized in that: The device comprises: An acquisition module, used for acquiring an enhanced electricity theft sample, wherein the enhanced electricity theft sample includes a plurality of electricity usage information and the electricity theft type corresponding to the electricity usage information; a calculation module, configured to calculate the separation degree of the electricity theft types in the enhanced electricity theft samples, wherein the separation degree is used to indicate the difficulty of distinguishing the electricity theft type from other electricity theft types, and to obtain a trained classification model for each electricity theft type by using the enhanced electricity theft samples and the separation degree of the electricity theft types; A processing module, used to classify the input suspected electricity theft data using the classification model, and determine the electricity theft type and user information of the suspected electricity theft data; The calculation module is used to calculate the separation degree of the electricity theft types in the enhanced electricity theft samples by using the projection vector method; determine the type of leaf node according to the size of the separation degree of the electricity theft types, that is, different electricity theft types, train SVM at the decision point of the decision tree for classification, and obtain multiple classification models of support vector machines based on the decision tree.

7. The device according to claim 6, characterized in that The acquisition module is specifically used to acquire a real electricity theft sample, and use the real electricity theft sample to train a generative adversarial model, wherein the generative adversarial model includes a generator and a discriminator, wherein the generator is used to generate a first expanded electricity theft sample according to random noise, and the discriminator is used to discriminate between the real electricity theft sample and the first expanded electricity theft sample; The enhanced electricity theft sample is generated by using the real electricity theft sample and the generator in the generative adversarial model obtained after training.

8. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 5 are implemented.

9. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images