Photovoltaic electricity larceny prevention alarm device

Abstract

The invention relates to the field of intelligent early warning of photovoltaic electricity larceny prevention, and particularly discloses a photovoltaic electricity larceny prevention alarm device, which comprises the following steps of: firstly, acquiring an electricity utilization curve graph of a user to be analyzed in a plurality of preset time periods and transformer area line loss data of a plurality of preset time points in a power system as input data; and then performing deep convolutional coding and analysis on the input data by using a machine learning technology to obtain a classification result, wherein the classification result is used for representing whether the user to be analyzed has an electricity stealing behavior or not. Therefore, real-time early warning of photovoltaic electricity stealing behaviors can be realized, so that benefits of enterprises are effectively protected, the market order is maintained, and healthy development of the photovoltaic industry is promoted.

Description

Technical Field

[0001] The present application relates to the field of intelligent early warning for photovoltaic anti-electricity theft, and more specifically, to a photovoltaic anti-electricity theft alarm device. Background Art

[0002] Currently, the common method to prevent electricity theft is to install electricity meters and data monitoring equipment to monitor actual power generation and output, compare them with recorded power generation, and promptly detect anomalies. However, electricity meters and data monitoring equipment are easily tampered with or bypassed, making it impossible to completely eliminate electricity theft.

[0003] Therefore, we look forward to a photovoltaic anti-electricity theft alarm device that uses machine learning technology to comprehensively monitor and analyze user electricity usage information and substation line loss parameters to achieve real-time early warning of photovoltaic electricity theft, thereby effectively protecting the interests of enterprises, maintaining market order, and promoting the healthy development of the photovoltaic industry. Summary of the Invention

[0004] To address the aforementioned technical issues, the present application is proposed. An embodiment of the present application provides a photovoltaic power theft prevention alarm device that first collects power consumption curves for multiple predetermined time periods for the user to be analyzed, as well as line loss data for multiple predetermined time points in the power system, as input data. Machine learning techniques are then used to perform deep convolutional coding and analysis on these input data to obtain a classification result, which indicates whether the user to be analyzed has engaged in power theft. This facilitates real-time early warning of photovoltaic power theft, thereby effectively protecting the interests of businesses, maintaining market order, and promoting the healthy development of the photovoltaic industry.

[0005] According to this aspect of the present application, there is provided a photovoltaic anti-electricity theft alarm device, comprising:

[0006] Photovoltaic anti-electricity theft data collection module, used to collect the power consumption curves of the users to be analyzed within multiple predetermined time periods and the line loss data of the substations at multiple predetermined time points in the power system;

[0007] A photovoltaic anti-electricity theft feature extraction module is used to extract and analyze the power consumption curves of the user to be analyzed in multiple predetermined time periods and the substation line loss data at multiple predetermined time points in the power system to obtain a user power consumption time series change feature matrix and a substation line loss comprehensive feature matrix;

[0008] A photovoltaic anti-electricity theft feature fusion module is used to perform deep feature integration on the user's electricity consumption time series variation feature matrix and the substation line loss comprehensive feature matrix to obtain a photovoltaic electricity theft feature matrix;

[0009] The photovoltaic anti-electricity theft classification result generating module is used to obtain a classification result based on the photovoltaic electricity theft feature matrix, and the classification result is used to indicate whether the user to be analyzed has engaged in electricity theft behavior.

[0010] In combination with this aspect of the present application, in a photovoltaic anti-electricity theft alarm device of the first aspect of the present application, the photovoltaic anti-electricity theft feature extraction module includes: a user electricity consumption feature extraction unit, which is used to feature encode the electricity consumption curve diagrams within multiple predetermined time periods of the user to be analyzed to obtain the user electricity consumption time series change feature matrix; a substation line loss feature extraction unit, which is used to feature encode the substation line loss data at multiple predetermined time points in the power system to obtain the substation line loss comprehensive feature matrix.

[0011] Compared to existing technologies, the photovoltaic power theft alarm device provided in this application first collects power consumption curves for the user to be analyzed over multiple predetermined time periods, as well as line loss data for the power system at multiple predetermined time points, as input data. It then uses machine learning techniques to perform deep convolutional coding and analysis on this input data to obtain a classification result, which indicates whether the user to be analyzed has engaged in power theft. This helps provide real-time early warning of photovoltaic power theft, effectively protecting the interests of businesses, maintaining market order, and promoting the healthy development of the photovoltaic industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0013] Figure 1 The figure shows a schematic block diagram of a photovoltaic anti-electricity theft alarm device according to an embodiment of the present application.

[0014] Figure 2 The figure shows a schematic block diagram of a photovoltaic anti-electricity theft feature extraction module in a photovoltaic anti-electricity theft alarm device according to an embodiment of the present application.

[0015] Figure 3 The figure shows a schematic block diagram of a user electricity usage feature extraction unit in a photovoltaic anti-electricity theft feature extraction module in a photovoltaic anti-electricity theft alarm device according to an embodiment of the present application.

[0016] Figure 4 The figure shows a schematic block diagram of the station line loss feature extraction unit in the photovoltaic anti-electricity theft feature extraction module in the photovoltaic anti-electricity theft alarm device according to an embodiment of the present application.

[0017] Figure 5 The figure shows a schematic block diagram of a photovoltaic anti-electricity theft classification result generation module in a photovoltaic anti-electricity theft alarm device according to an embodiment of the present application.

[0018] Figure 6 The figure illustrates a flow chart of a photovoltaic anti-electricity theft alarm method according to an embodiment of the present application. DETAILED DESCRIPTION

[0019] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0020] Exemplary devices

[0021] Figure 1 FIG2 shows a schematic block diagram of a photovoltaic anti-electricity theft alarm device according to an embodiment of the present application. Figure 1 As shown, according to an embodiment of the present application, the photovoltaic anti-electricity theft alarm device 100 includes: a photovoltaic anti-electricity theft data acquisition module 110, which is used to collect electricity consumption curves within multiple predetermined time periods of the user to be analyzed and substation line loss data at multiple predetermined time points in the power system; a photovoltaic anti-electricity theft feature extraction module 120, which is used to extract and analyze the electricity consumption curves within multiple predetermined time periods of the user to be analyzed and the substation line loss data at multiple predetermined time points in the power system to obtain the user's electricity consumption time series change feature matrix and the substation line loss comprehensive feature matrix; a photovoltaic anti-electricity theft feature fusion module 130, which is used to perform deep feature integration on the user's electricity consumption time series change feature matrix and the substation line loss comprehensive feature matrix to obtain a photovoltaic electricity theft feature matrix; a photovoltaic anti-electricity theft classification result generation module 140, which is used to obtain a classification result based on the photovoltaic electricity theft feature matrix, and the classification result is used to indicate whether the user to be analyzed has committed electricity theft.

[0022] Currently, the common method for preventing electricity theft is to install electricity meters and data monitoring equipment to monitor actual power generation and output, compare them with recorded power generation, and promptly detect anomalies. However, electricity meters and data monitoring equipment are easily tampered with or bypassed, making them incapable of completely eliminating electricity theft. Therefore, a photovoltaic anti-theft alarm device is desired. Using machine learning technology, it comprehensively monitors and analyzes user electricity usage information and line loss parameters across the distribution area. This device can provide real-time early warning of photovoltaic electricity theft, effectively protecting business interests, maintaining market order, and promoting the healthy development of the photovoltaic industry.

[0023] In this embodiment of the present application, the photovoltaic anti-electricity theft data collection module 110 is configured to collect electricity usage graphs for the user to be analyzed over multiple predetermined time periods, as well as substation line loss data for multiple predetermined time points in the power system. It should be understood that the electricity usage graph data reflects the user's electricity usage over different time periods, including information such as fluctuations in electricity consumption and peak hours. This data can reveal a user's electricity usage habits and behavioral patterns, which is important for identifying abnormal electricity usage (such as electricity theft). Therefore, to better analyze whether a user is engaging in electricity theft, electricity usage graphs for the user to be analyzed are first collected over multiple predetermined time periods. It should be understood that while electricity usage graphs reflect the user's electricity usage, they lack information about the overall operation of the power system. Based on this, it is considered that abnormal substation line loss data may indicate potential electricity theft. Substation line loss data reflects the flow of electricity in the power system, as well as changes in parameters such as voltage and current, and can help determine system stability and the direction of electricity flow. In other words, substation line loss data contains more comprehensive power system data, providing more comprehensive information for determining the operating status of the power system. Therefore, in order to analyze electricity theft more comprehensively and accurately, it is also necessary to collect line loss data (such as electric energy, voltage, current, etc.) at multiple predetermined time points in the power system.

[0024] In this embodiment of the present application, the photovoltaic anti-electricity theft feature extraction module 120 is configured to perform feature extraction and analysis on the power consumption curves of the user to be analyzed within multiple predetermined time periods and the substation line loss data at multiple predetermined time points in the power system, respectively, to obtain a user power consumption time series variation feature matrix and a substation line loss comprehensive feature matrix. It should be understood that after collecting these input data, feature extraction and analysis tasks are further performed on these input data.

[0025] Specifically, Figure 2 The figure shows a schematic block diagram of a photovoltaic anti-electricity theft feature extraction module in a photovoltaic anti-electricity theft alarm device according to an embodiment of the present application. Figure 2 As shown, the photovoltaic anti-electricity theft feature extraction module 120 includes: a user electricity consumption feature extraction unit 121, which is used to feature encode the electricity consumption curve diagrams of the user to be analyzed in multiple predetermined time periods to obtain the user electricity consumption time series change feature matrix; a substation line loss feature extraction unit 122, which is used to feature encode the substation line loss data at multiple predetermined time points in the power system to obtain the substation line loss comprehensive feature matrix.

[0026] First, feature extraction and analysis operations are performed on the power consumption curves of the user to be analyzed in multiple predetermined time periods. Specifically, Figure 3 The figure shows a schematic block diagram of the user power consumption feature extraction unit in the photovoltaic anti-electricity theft feature extraction module in the photovoltaic anti-electricity theft alarm device according to an embodiment of the present application. Figure 3 As shown, the user electricity consumption feature extraction unit 121 includes: an image pixel enhancement subunit 121-1, which is used to pass the electricity consumption curve graphs of the user to be analyzed in multiple predetermined time periods through a pixel enhancement module based on a generative adversarial network to obtain multiple user electricity consumption pixel enhanced images; a user electricity consumption analysis subunit 121-2, which is used to pass the multiple user electricity consumption pixel enhanced images through an electricity consumption feature extractor based on a deep and shallow feature fusion module to obtain multiple user electricity consumption feature matrices; a user electricity consumption time series feature extraction subunit 121-3, which is used to arrange the multiple user electricity consumption feature matrices into a three-dimensional tensor of user electricity consumption changes according to the time dimension and then pass them through a electricity consumption time series change feature extraction module based on a three-dimensional convolutional neural network model to obtain a user electricity consumption time series change feature map; a feature map pooling subunit 121-4, which is used to perform a pooling operation on the user electricity consumption time series change feature map along the channel dimension to obtain a user electricity consumption time series change feature matrix.

[0027] It should be understood that the original electricity usage graph may contain noise or blur, and contain relatively little information. Direct feature extraction may result in information loss or misunderstanding. Therefore, to extract more effective information and reduce the possibility of information loss, a pixel enhancement module based on a generative adversarial network is used to perform pixel enhancement processing on the electricity usage graphs of the user to be analyzed for multiple predetermined time periods, thereby obtaining multiple pixel-enhanced images of the user's electricity usage.

[0028] Next, considering that the user electricity usage pixel enhanced image contains not only shallow features such as edges and textures, but also higher-level deep features, it is necessary to comprehensively consider the shallow and deep features in the image when analyzing the user electricity usage information. Specifically, in an embodiment of the present application, the multiple user electricity usage pixel enhanced images are respectively passed through an electricity usage feature extractor based on a deep and shallow feature fusion module to obtain multiple user electricity usage feature matrices. In this way, by fusing these features, it is possible to fully utilize information at different levels and improve the efficiency of feature extraction and characterization capabilities.

[0029] In a specific embodiment of the present application, the user electricity consumption analysis subunit 121-2 is used to: extract a shallow feature matrix from the i-th layer of the electricity consumption feature extraction module, where the i-th layer is the first to sixth layers of the electricity consumption feature extraction module; extract a deep feature matrix from the j-th layer of the electricity consumption feature extraction module, where the ratio between the j-th layer and the i-th layer is greater than or equal to 5; and use the shallow and deep feature fusion module to fuse the shallow feature matrix and the deep feature matrix to obtain each user electricity consumption feature matrix in the multiple user electricity consumption feature matrices.

[0030] Furthermore, it should be understood that a user's electricity usage changes over time. Therefore, in order to better preserve the temporal continuity and changing trends of the user's electricity usage characteristics, the multiple user electricity usage feature matrices are first arranged according to the time dimension into a three-dimensional tensor of user electricity usage changes. Then, a power usage time series change feature extraction module based on a three-dimensional convolutional neural network model is used to feature encode the three-dimensional tensor of user electricity usage changes to obtain a user electricity usage time series change feature graph. The three-dimensional convolutional neural network model can simultaneously consider features in both the time and space dimensions, which helps to capture the complex spatiotemporal features in the user electricity usage time series changes.

[0031] In a specific embodiment of the present application, the user electricity consumption timing feature extraction subunit 121-3 is used to: use each layer of the three-dimensional convolutional neural network model to perform the following on the input data in the forward pass of the layer: perform convolution processing based on the three-dimensional convolution kernel on the input data to obtain a convolution feature map; perform pooling processing based on the local feature matrix on the convolution feature map to obtain a pooled feature map; perform nonlinear activation on the pooled feature map to obtain an activation feature map; wherein, the input of the first layer of the three-dimensional convolutional neural network model is the three-dimensional tensor of the user electricity consumption change, and the output of the last layer of the three-dimensional convolutional neural network model is the user electricity consumption timing change feature map.

[0032] Then, considering that high-dimensional feature maps increase the computational burden, reduce the training and inference speed of the model, and affect the efficiency of real-time warning, in order to reduce the computational complexity of the model, the user electricity usage time series change feature map is further subjected to pooling and dimensionality reduction processing. Specifically, in this embodiment of the application, the user electricity usage time series change feature map is pooled along the channel dimension to obtain the user electricity usage time series change feature matrix.

[0033] Then, feature extraction and analysis operations are performed on the line loss data of the substation area at multiple predetermined time points in the power system. Specifically, Figure 4 The figure shows a schematic block diagram of the station area line loss feature extraction unit in the photovoltaic anti-electricity theft feature extraction module in the photovoltaic anti-electricity theft alarm device according to an embodiment of the present application. Figure 4 As shown, the substation line loss feature extraction unit 122 includes: a substation line loss analysis subunit 122-1, which is used to arrange the substation line loss data of multiple predetermined time points in the power system into multiple substation line loss input vectors according to the time dimension and then pass them through the substation line loss time series feature extraction module based on the time encoder to obtain multiple substation line loss feature vectors; a substation line loss comprehensive feature extraction subunit 122-2, which is used to arrange the multiple substation line loss feature vectors into a substation line loss comprehensive input matrix and then pass them through the substation line loss comprehensive feature extraction module based on the feature encoder to obtain a substation line loss comprehensive feature matrix.

[0034] It should be understood that the substation line loss data includes electric energy, voltage, current, etc., and these data change dynamically in the time dimension. Based on this, in order to better preserve the time series characteristics of the substation line loss data, the substation line loss data at multiple predetermined time points in the power system are first arranged according to the time dimension to obtain multiple substation line loss input vectors. Among them, the multiple substation line loss feature vectors include electric energy input vectors, voltage input vectors, current input vectors, etc. In addition, considering that the time series encoder can effectively model and extract the time series information in the substation line loss data, it helps to capture the time dependence and trend of the line loss data. Therefore, the substation line loss time series feature extraction module based on the time series encoder is further used to perform time series feature encoding on the multiple substation line loss input vectors to obtain multiple substation line loss feature vectors.

[0035] In a specific embodiment of the present application, the substation line loss analysis subunit 122-1 is used to: use the fully connected layer of the time series encoder to perform fully connected encoding on each of the multiple substation line loss input vectors to extract high-dimensional implicit features of the eigenvalues ​​at each position in the substation line loss input vector; and use the one-dimensional convolution layer of the time series encoder to perform one-dimensional convolution encoding on each of the multiple substation line loss input vectors to extract high-dimensional implicit correlation features of the correlation between the eigenvalues ​​at each position in the substation line loss input vector.

[0036] Next, considering that the substation line loss data includes electric energy, voltage, current, etc., a single analysis of characteristic information such as electric energy, voltage, current, etc. may affect the accuracy of the substation line loss feature analysis. Therefore, in order to comprehensively consider the characteristics of different aspects, the characteristic vectors of different types such as electric energy, voltage, current, etc. are first arranged into a two-dimensional matrix according to the dimension of the substation line loss. Specifically, in an embodiment of the present application, the multiple substation line loss feature vectors are arranged into a substation line loss comprehensive input matrix. Then, a substation line loss comprehensive feature extraction module based on a feature encoder is used to feature encode the substation line loss comprehensive input matrix to obtain a substation line loss comprehensive feature matrix. Among them, the feature encoder can map the high-dimensional substation line loss feature matrix to a low-dimensional representation space, extract the key features in the data, reduce the dimension of the data, and retain important information, which is beneficial to the efficiency of model training and reasoning. In this way, through the combination of two-dimensional arrangement and feature encoder, the various feature information of substation line loss can be better integrated, and the important features therein can be extracted to provide more meaningful input for the subsequent early warning model.

[0037] In a specific embodiment of the present application, the substation line loss comprehensive feature extraction subunit 122-2 is used to: use the layers of the feature encoder to perform the following operations on the input data in the forward pass of the layer: use the convolution units of the layers of the feature encoder to perform convolution processing based on a two-dimensional convolution kernel on the input data to obtain a convolution feature map; use the pooling units of the layers of the feature encoder to perform pooling processing on the convolution feature map along the channel dimension to obtain a pooling feature map; and use the activation units of the layers of the feature encoder to perform nonlinear activation on the feature values ​​of each position in the pooling feature map to obtain an activation feature map; wherein the output of the last layer of the feature encoder is the substation line loss comprehensive feature matrix.

[0038] In an embodiment of the present application, the photovoltaic anti-electricity theft feature fusion module 130 is used to perform deep feature integration on the user electricity usage time series variation feature matrix and the substation line loss comprehensive feature matrix to obtain a photovoltaic electricity theft feature matrix. It should be understood that there is a certain correlation between user electricity usage behavior and substation line loss parameters. For example, abnormal user electricity usage behavior may lead to changes in the substation line loss rate, or electricity theft behavior may leave specific patterns in user electricity usage data and substation line loss data. Therefore, in order to more accurately achieve real-time early warning of photovoltaic electricity theft behavior, the correlation characteristics between user electricity usage behavior and substation line loss parameters are further captured. Specifically, in an embodiment of the present application, the user electricity usage time series variation feature matrix and the substation line loss comprehensive feature matrix are deeply integrated to obtain a photovoltaic electricity theft feature matrix.

[0039] In a specific embodiment of the present application, the photovoltaic anti-electricity theft feature fusion module 130 includes: a feature flattening unit, configured to perform feature flattening on the user electricity consumption time series variation feature matrix and the substation line loss comprehensive feature matrix to obtain a user electricity consumption time series variation feature vector and a substation line loss comprehensive feature vector; a feature fusion unit, configured to perform weighted fusion on the user electricity consumption time series variation feature vector and the substation line loss comprehensive feature vector to obtain a photovoltaic electricity theft feature vector; a feature optimization unit, configured to perform feature fine-grained internal architecture optimization based on spectral analysis on the photovoltaic electricity theft feature vector to obtain an optimized photovoltaic electricity theft feature vector; and a feature aggregation unit, configured to perform feature aggregation on the optimized photovoltaic electricity theft feature vector based on feature flattening to obtain the photovoltaic electricity theft feature matrix.

[0040] In particular, in the technical solution of this application, the photovoltaic electricity theft feature vector may suffer from internal information redundancy and noise interference. In high-dimensional data spaces, photovoltaic electricity theft feature vectors often contain a large amount of redundant information, which is not helpful for accurately identifying photovoltaic electricity theft and increases the complexity of subsequent analysis models. Furthermore, the data in the photovoltaic electricity theft feature vector may be contaminated by noise due to factors such as measurement errors and environmental interference. Directly using the photovoltaic electricity theft feature vector increases the risk of model overfitting, reduces generalization ability, and fails to accurately identify photovoltaic electricity theft in different scenarios. Therefore, the photovoltaic electricity theft feature vector is subjected to fine-grained internal architecture optimization based on spectral analysis to obtain an optimized photovoltaic electricity theft feature vector.

[0041] Specifically, the feature optimization unit is used to: first, calculate the global fine-grained autocorrelation pattern matrix of the photovoltaic power theft feature vector, which is expressed as:

[0042]

[0043] M=D1⊙D2

[0044] v i ,v j ∈V

[0045] Where V represents the photovoltaic power theft characteristic vector, v i and v j They represent the i-th and j-th eigenvalues ​​of the photovoltaic electricity theft feature vector, w1, w2, w3 and w4 represent different weight hyperparameters, ⊙ represents matrix dot product, D1 represents the forward weighted matrix of photovoltaic electricity theft features, and D2 represents the reverse weighted matrix of photovoltaic electricity theft features. Represents the value of the (i, j)th position in the forward weighted matrix of photovoltaic power theft characteristics, represents the value of the (i, j)th position of the inverse weighted matrix of photovoltaic power theft characteristics, and M represents the global fine-grained autocorrelation graph matrix.

[0046] Specifically, by constructing a global, fine-grained autocorrelation pattern matrix, we capture the fine-grained global correlations among the dimensions or elements of the photovoltaic power theft feature vector. This transforms the implicit structural information within the feature into a directly analyzable mathematical object, allowing the structural information within the feature vector to be clearly quantified and presented. This provides a clear mathematical representation for in-depth analysis of the feature's inherent structure and lays the foundation for subsequent spectral analysis of the feature.

[0047] Secondly, the global fine-grained autocorrelation pattern matrix is ​​spectrally parsed to obtain a set of fine-grained spectral component encoding vectors of photovoltaic power theft characteristics, which can be expressed as follows:

[0048]

[0049] Where Λ represents a diagonal matrix, λ1 and λ m denote the first and mth eigenvalues ​​of the diagonal matrix, respectively, (·) T represents the transpose of the vector, U represents the set of encoding vectors of the fine-grained spectral components of photovoltaic power theft characteristics, x1, x2 and x m They respectively represent the first, second and mth photovoltaic electricity theft feature fine-grained spectral component encoding vectors in the set of photovoltaic electricity theft feature fine-grained spectral component encoding vectors.

[0050] That is, through spectral analysis, the global fine-grained autocorrelation pattern matrix is ​​decomposed into a set of fine-grained spectral component encoding vectors of photovoltaic power theft characteristics, and a new orthogonal basis guided by the internal structure of the data is constructed to represent different independent change patterns within the characteristics. This achieves effective decoupling of the complex correlation structure within the characteristics, and decomposes the complex structure into independent basic structural patterns sorted by importance, so that each independent structure can be processed more accurately in the subsequent process.

[0051] Next, information constraint compression processing is performed on each photovoltaic power theft feature fine-grained spectral component code vector in the set of photovoltaic power theft feature fine-grained spectral component code vectors to obtain a set of photovoltaic power theft feature fine-grained spectral component modulation code vectors, which is expressed as follows:

[0052]

[0053] Among them, x i represents the i-th photovoltaic power theft feature fine-grained spectral component encoding vector in the set of photovoltaic power theft feature fine-grained spectral component encoding vectors, ||·|| represents the Euclidean norm, y i represents the fine-grained spectral component modulation coding vector of the i-th photovoltaic power theft feature.

[0054] That is, through information constraint compression processing, the higher-order dependencies beyond the second-order statistics in the coding vector of the fine-grained spectral component of photovoltaic power theft characteristics are captured or the discrimination ability is enhanced, and each decoupled structural component is refined and independently nonlinearly modulated to enhance useful information and suppress noise. This breaks the pure linear processing framework and introduces complexity and nonlinear processing capabilities, so that the generated modulation coding vector of the fine-grained spectral component of photovoltaic power theft characteristics can better highlight useful information and reduce noise interference, providing better quality features for subsequent accurate judgment of power theft behavior.

[0055] Then, the internal architecture significant control parameters of each photovoltaic power theft feature fine granularity spectral component modulation coding vector in the set of photovoltaic power theft feature fine granularity spectral component modulation coding vectors are calculated to obtain a set of internal architecture significant control parameters, which can be expressed as follows:

[0056]

[0057] Among them, α and β represent different weight parameters, ||·||1 represents the first norm, ||·||2 represents the second norm, L represents the length of the modulation coding vector of the fine-grained spectral component of the photovoltaic power theft feature, and a i represents y i The corresponding internal architecture significantly regulates the parameters.

[0058] That is, the internal architecture significant control parameters of the modulation coding vector of each fine-grained spectral component of the photovoltaic power theft feature are calculated to quantify the value of its structural information, providing an importance basis based on the current modulation state for subsequent operations such as weighted fusion. The generated internal architecture significant control parameters can dynamically evaluate the importance of each structural component, allowing subsequent processing to perform weighted fusion based on the actual value of each fine-grained spectral component of the photovoltaic power theft feature, thereby improving the accuracy and effectiveness of feature representation.

[0059] Next, a normalization constraint process is performed on the set of internal architecture significant control parameters to obtain a set of internal architecture significant modulation weighting parameters, which is expressed as:

[0060] w i =Softmax(a i )

[0061] Among them, Softmax(·) represents the normalization function, w i Indicates a i The corresponding internal architecture significantly modulates the weighting parameters.

[0062] Specifically, a method similar to Softmax normalization is employed, where the sharpness of the parameters is significantly controlled through the internal architecture. This allows for normalized constraint processing to yield a set of significantly modulated weighted parameters within the internal architecture, ensuring a stable and controllable fusion process. This ensures that when the modulation coding vectors of the fine-grained spectral components of the photovoltaic power theft characteristics are subsequently analyzed, each fine-grained spectral component of the photovoltaic power theft characteristics can be properly integrated into the fusion, avoiding analysis bias caused by unreasonable parameters and improving the reliability of power theft judgments.

[0063] Finally, based on the set of significant modulation weighting parameters of the internal architecture, a fine-grained aggregation operation is performed on the set of fine-grained spectral component modulation coding vectors of the photovoltaic power theft characteristics to obtain an optimized photovoltaic power theft feature vector, which is expressed as follows:

[0064]

[0065] Among them, V ' Represents the optimized photovoltaic power theft feature vector.

[0066] That is, the internal architecture significant modulation weighting parameters are used to finely aggregate the modulation coding vectors of the fine-grained spectral components of the photovoltaic electricity theft characteristics, and an optimized photovoltaic electricity theft feature vector is obtained that concentrates key structural information and improves the characterization capability. The structural components that are significant and contain key information are strengthened, and the secondary or noise components are suppressed, so as to more accurately judge the photovoltaic electricity theft behavior, making the optimized photovoltaic electricity theft feature vector more robust and discriminative, and providing better quality features for downstream machine learning tasks.

[0067] In the embodiment of the present application, the photovoltaic anti-electricity theft classification result generation module 140 is used to obtain a classification result based on the photovoltaic electricity theft feature matrix, and the classification result is used to indicate whether the user to be analyzed has engaged in electricity theft. Specifically, Figure 5 FIG2 is a schematic block diagram of a photovoltaic anti-electricity theft alarm device according to an embodiment of the present application. Figure 5 As shown, the photovoltaic anti-electricity theft classification result generation module 140 includes: a photovoltaic electricity theft feature analysis unit 141, which is used to pass the photovoltaic electricity theft feature matrix through a photovoltaic electricity theft feature extraction module based on a convolutional neural network model to obtain a photovoltaic electricity theft classification feature matrix; and a photovoltaic electricity theft feature classification unit 142, which is used to pass the photovoltaic electricity theft classification feature matrix through a classifier to obtain the classification result.

[0068] It's understandable that feature fusion alone may not extract high-level, abstract features from the data, nor fully exploit the underlying patterns and regularities within the data. Therefore, to better extract higher-level, more abstract feature information, a photovoltaic electricity theft feature extraction module based on a convolutional neural network model performs deep convolution encoding on the photovoltaic electricity theft feature matrix to generate a photovoltaic electricity theft classification feature matrix. By stacking convolutional and pooling layers, the convolutional neural network model extracts abstract features from the data layer by layer, from low-level features to high-level features, helping to discover deeper patterns in the data.

[0069] Next, considering that the classifier can quickly and efficiently classify data, thereby achieving real-time warnings, timely detection of electricity theft is crucial to protecting corporate interests and maintaining market order. Therefore, the classifier is used to perform feature classification on the photovoltaic electricity theft classification feature matrix to obtain a classification result. The classification result is used to indicate whether the user to be analyzed has engaged in electricity theft.

[0070] In a specific embodiment of the present application, the cropping positioning feature classification unit 142 is used to: use the classifier to process the photovoltaic power theft classification feature matrix according to the following formula to obtain the classification result; wherein the formula is: O = softmax{(W c ,B c)|Project(M)}, where Project(M) represents the projection of the photovoltaic power theft classification feature matrix into a vector, W c is the weight matrix, B c represents a bias vector, softmax represents a normalized exponential function, and O represents the classification result.

[0071] It's worth noting that in addition to using a classifier to classify the PV power theft classification feature matrix, anomaly detection algorithms can also be used to detect power theft in PV systems. Anomaly detection algorithms can help identify data points that deviate from normal behavior patterns, thereby detecting potential power theft. The following is a possible implementation process for this approach: 1. Data Collection and Preprocessing: Collect user power usage information and substation line loss parameter data from the PV system and perform data preprocessing, including missing value handling and data standardization. 2. Feature Extraction: Extract features from the data to construct a PV power theft feature matrix, which is used to describe various characteristics of the PV system. 3. Anomaly Detection Algorithm Selection: Select an appropriate anomaly detection algorithm, such as those based on statistical methods (such as Z-score and boxplots), clustering algorithms (such as K-means and DBSCAN), or density-based algorithms (such as LOF and Isolation Forest). 4. Model Training: Use the selected anomaly detection algorithm to train the PV power theft feature matrix to learn normal behavior patterns. 5. Anomaly Detection: Input new data into the trained anomaly detection model to detect whether there are any anomalous data points. Abnormal data points may indicate potential electricity theft. 6. Threshold Setting and Output: Based on the output of the anomaly detection algorithm and the set threshold, determine whether electricity theft has occurred. When the degree of anomaly exceeds the threshold, a theft warning is issued. 7. Feedback and Improvement: Based on actual conditions, the anomaly detection algorithm is continuously optimized to improve the accuracy and robustness of electricity theft detection.

[0072] In summary, the photovoltaic anti-electricity theft alarm device according to the embodiment of the present application is described. It first collects electricity consumption curves for multiple predetermined time periods of the user to be analyzed, as well as line loss data for multiple predetermined time points in the power system. It then uses machine learning technology to perform deep convolutional coding and analysis on this input data to obtain a classification result. The classification result is used to indicate whether the user to be analyzed has engaged in electricity theft. This helps to achieve real-time early warning of photovoltaic electricity theft, thereby effectively protecting the interests of enterprises, maintaining market order, and promoting the healthy development of the photovoltaic industry.

[0073] As described above, the photovoltaic anti-electricity theft alarm device 100 according to the embodiments of the present application can be implemented in various wireless terminals, such as a server for photovoltaic anti-electricity theft alarms. In one example, the photovoltaic anti-electricity theft alarm device 100 according to the embodiments of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the photovoltaic anti-electricity theft alarm device 100 can be a software module in the operating system of the wireless terminal, or it can be an application developed specifically for the wireless terminal. Of course, the photovoltaic anti-electricity theft alarm device 100 can also be one of the many hardware modules of the wireless terminal.

[0074] Alternatively, in another example, the photovoltaic anti-electricity theft alarm device 100 and the wireless terminal may be separate devices, and the photovoltaic anti-electricity theft alarm device 100 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0075] Exemplary Methods

[0076] Figure 6 The figure shows a flow chart of the photovoltaic anti-electricity theft alarm method according to an embodiment of the present application. Figure 6 As shown, the photovoltaic anti-electricity theft alarm method according to the embodiment of the present application includes: S1, collecting electricity consumption curves of the user to be analyzed in multiple predetermined time periods and substation line loss data at multiple predetermined time points in the power system; S2, performing feature extraction and analysis on the electricity consumption curves of the user to be analyzed in multiple predetermined time periods and the substation line loss data at multiple predetermined time points in the power system to obtain the user's electricity consumption time series change feature matrix and the substation line loss comprehensive feature matrix; S3, performing deep feature integration on the user's electricity consumption time series change feature matrix and the substation line loss comprehensive feature matrix to obtain a photovoltaic electricity theft feature matrix; S4, obtaining a classification result based on the photovoltaic electricity theft feature matrix, and the classification result is used to indicate whether the user to be analyzed has engaged in electricity theft behavior.

[0077] Here, those skilled in the art will appreciate that the specific functions and operations of each step in the above photovoltaic anti-theft alarm method have been described in detail in the referenced examples. Figure 1 The photovoltaic anti-electricity theft alarm device has been described in detail, and therefore, its repeated description will be omitted.

[0078] In the several embodiments provided herein, it should be understood that the disclosed systems, devices, apparatuses, and methods may be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the module division is merely a logical functional division, and actual implementation may employ other division methods.

[0079] In addition, the functional modules in the various embodiments of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0080] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0081] Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the claims are intended to be embraced therein. Any reference to a figure in a claim should not be construed as limiting the claim to which it relates.

[0082] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a device claim may also be implemented by a single unit or device through software or hardware. Second-order terms are used to indicate names and do not imply any particular order.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit of the technical solutions of the present invention.

Claims

1. A photovoltaic anti-electricity theft alarm device, characterized in that: include: Photovoltaic anti-electricity theft data collection module, used to collect the power consumption curves of the users to be analyzed within multiple predetermined time periods and the line loss data of the substations at multiple predetermined time points in the power system; A photovoltaic anti-electricity theft feature extraction module is used to extract and analyze the power consumption curves of the user to be analyzed in multiple predetermined time periods and the substation line loss data at multiple predetermined time points in the power system to obtain a user power consumption time series change feature matrix and a substation line loss comprehensive feature matrix; A photovoltaic anti-electricity theft feature fusion module is used to perform deep feature integration on the user's electricity consumption time series variation feature matrix and the substation line loss comprehensive feature matrix to obtain a photovoltaic electricity theft feature matrix; The photovoltaic anti-electricity theft classification result generating module is used to obtain a classification result based on the photovoltaic electricity theft feature matrix, and the classification result is used to indicate whether the user to be analyzed has engaged in electricity theft behavior.

2. The photovoltaic anti-electricity theft alarm device according to claim 1, characterized in that: The photovoltaic anti-electricity theft feature extraction module includes: A user electricity consumption feature extraction unit is used to perform feature coding on the electricity consumption curves of the user to be analyzed in multiple predetermined time periods to obtain a characteristic matrix of the user's electricity consumption time series variation; The substation line loss feature extraction unit is used to perform feature encoding on the substation line loss data at multiple predetermined time points in the power system to obtain the substation line loss comprehensive feature matrix.

3. The photovoltaic anti-electricity theft alarm device according to claim 2, characterized in that: The user electricity consumption feature extraction unit includes: An image pixel enhancement subunit, configured to pass the electricity consumption curve graphs of the user to be analyzed within multiple predetermined time periods through a pixel enhancement module based on a generative adversarial network to obtain multiple user electricity consumption pixel enhanced images; A user electricity consumption analysis subunit, configured to pass the plurality of user electricity consumption pixel enhanced images through an electricity consumption feature extractor based on a deep and shallow feature fusion module to obtain a plurality of user electricity consumption feature matrices; A user electricity usage time series feature extraction subunit, configured to arrange the plurality of user electricity usage feature matrices according to the time dimension into a three-dimensional tensor of user electricity usage changes, and then pass the result through a power usage time series change feature extraction module based on a three-dimensional convolutional neural network model to obtain a user electricity usage time series change feature graph; The feature graph pooling subunit is used to perform a pooling operation on the user electricity usage time series change feature graph along the channel dimension to obtain a user electricity usage time series change feature matrix.

4. The photovoltaic anti-electricity theft alarm device according to claim 3, characterized in that: The user power consumption analysis subunit is used to: Extracting a shallow feature matrix from the i-th layer of the power consumption feature extraction module, wherein the i-th layer is the first to sixth layers of the power consumption feature extraction module; Extracting a deep feature matrix from the jth layer of the power consumption feature extraction module, wherein the ratio between the jth layer and the ith layer is greater than or equal to 5; as well as The shallow feature fusion module is used to fuse the shallow feature matrix and the deep feature matrix to obtain each user power usage feature matrix in the multiple user power usage feature matrices.

5. The photovoltaic anti-electricity theft alarm device according to claim 4, characterized in that: The station area line loss feature extraction unit includes: a substation line loss analysis subunit, configured to arrange the substation line loss data of a plurality of predetermined time points in the power system into a plurality of substation line loss input vectors according to a time dimension, and then pass the plurality of substation line loss time series feature extraction modules based on a time series encoder to obtain a plurality of substation line loss feature vectors; The substation line loss comprehensive feature extraction subunit is used to arrange the multiple substation line loss feature vectors into a substation line loss comprehensive input matrix and then pass it through the substation line loss comprehensive feature extraction module based on the feature encoder to obtain the substation line loss comprehensive feature matrix.

6. The photovoltaic anti-electricity theft alarm device according to claim 5, characterized in that: The substation line loss analysis subunit is used to: Using the fully connected layer of the temporal encoder to perform fully connected encoding on each of the plurality of station area line loss input vectors to extract high-dimensional implicit features of the feature values ​​at each position in the station area line loss input vector; as well as The one-dimensional convolution layer of the temporal encoder is used to perform one-dimensional convolution encoding on each of the multiple station line loss input vectors to extract high-dimensional implicit correlation features of the correlation between the feature values ​​of each position in the station line loss input vector.

7. The photovoltaic anti-electricity theft alarm device according to claim 6, characterized in that: The photovoltaic anti-electricity theft feature fusion module includes: A feature flattening unit, configured to flatten the user power consumption time series variation feature matrix and the substation line loss comprehensive feature matrix to obtain a user power consumption time series variation feature vector and a substation line loss comprehensive feature vector; A feature fusion unit, configured to perform weighted fusion on the user's electricity consumption time series variation feature vector and the substation line loss comprehensive feature vector to obtain a photovoltaic power theft feature vector; a feature optimization unit, configured to perform feature fine-grained internal architecture optimization on the photovoltaic power theft feature vector based on spectrum analysis to obtain an optimized photovoltaic power theft feature vector; A feature aggregation unit is configured to perform feature aggregation on the optimized photovoltaic power theft feature vector based on feature flattening to obtain the photovoltaic power theft feature matrix.

8. The photovoltaic anti-electricity theft alarm device according to claim 7, characterized in that: The feature optimization unit is used to: Calculating a global fine-grained autocorrelation pattern matrix of the photovoltaic power theft feature vector; Performing spectral analysis on the global fine-grained autocorrelation pattern matrix to obtain a set of photovoltaic power theft feature fine-grained spectral component encoding vectors; Performing information constraint compression processing on each photovoltaic power theft feature fine-grained spectral component code vector in the set of photovoltaic power theft feature fine-grained spectral component code vectors to obtain a set of photovoltaic power theft feature fine-grained spectral component modulation code vectors; Calculating the internal architecture significant control parameters of each photovoltaic power theft feature fine granularity spectral component modulation coding vector in the set of photovoltaic power theft feature fine granularity spectral component modulation coding vectors to obtain a set of internal architecture significant control parameters; Performing normalization constraint processing on the set of internal architecture significant control parameters to obtain a set of internal architecture significant modulation weighting parameters; Based on the set of significant modulation weighting parameters of the internal architecture, a fine-grained aggregation operation is performed on the set of fine-grained spectral component modulation coding vectors of the photovoltaic power theft characteristics to obtain an optimized photovoltaic power theft feature vector.

9. The photovoltaic anti-electricity theft alarm device according to claim 8, characterized in that: The photovoltaic anti-electricity theft classification result generation module includes: a photovoltaic electricity theft feature analysis unit, configured to pass the photovoltaic electricity theft feature matrix through a photovoltaic electricity theft feature extraction module based on a convolutional neural network model to obtain a photovoltaic electricity theft classification feature matrix; The photovoltaic electricity theft feature classification unit is configured to pass the photovoltaic electricity theft classification feature matrix through a classifier to obtain the classification result.

Images