Electricity larceny prevention system for distributed photovoltaic grid-connected power generation system

Abstract

The invention relates to the field of electricity larceny prevention, and particularly discloses an electricity larceny prevention system for a distributed photovoltaic grid-connected power generation system. The method comprises the following steps of: firstly, acquiring solar radiation intensity values at a plurality of preset time points acquired by a sensor, photovoltaic power generation system output power at a plurality of preset time points acquired by a photovoltaic operation monitoring system and historical energy data acquired by a database, and then performing feature extraction and correlation analysis on the three values by using a deep learning technology to obtain the solar radiation intensity values, the photovoltaic power generation system output power and the historical energy data; and finally, whether an electricity stealing alarm of the distributed photovoltaic grid-connected power generation system needs to be sent out is judged through a classifier, so that the electricity stealing behavior is found in time, the safety of the system is improved, and loss is prevented.

Description

Technical Field

[0001] The present application relates to the field of anti-electricity theft, and more specifically, to an anti-electricity theft system for a distributed photovoltaic grid-connected power generation system. Background Art

[0002] Distributed photovoltaic grid connection refers to connecting photovoltaic power generation systems distributed in different locations through the power grid to achieve resource sharing and complementarity, and improve the overall efficiency and stability of the photovoltaic power generation system.

[0003] To promote the adoption of distributed photovoltaic grid-connected power generation systems, these systems enjoy preferential feed-in tariffs. However, some criminals, attracted by these profits, engage in electricity theft. Electricity theft refers to the unpaid withdrawal of electricity from the public grid. It is a long-standing and widespread problem, causing significant losses and negative impacts on the country, businesses, and society. The methods used to steal electricity vary, and anti-theft systems and technical measures are constantly evolving. Some criminals illegally obtain electricity price subsidies through various means, such as tampering with distributed photovoltaic grid-connected power generation systems to falsely inflate photovoltaic power meter readings, or directly withdrawing electricity from the grid and feeding it back to the grid through photovoltaic system equipment. Manual investigation of electricity theft is time-consuming and lacks accurate information on the time and amount of electricity stolen.

[0004] Therefore, an anti-electricity theft system for a distributed photovoltaic grid-connected power generation system is desired. Summary of the Invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an anti-electricity theft system for a distributed photovoltaic grid-connected power generation system. The system first obtains solar radiation intensity values ​​at multiple predetermined time points collected by a sensor, photovoltaic power generation system output power at multiple predetermined time points collected by a photovoltaic operation monitoring system, and historical energy data collected by a database. Then, deep learning technology is used to perform feature extraction and correlation analysis on the three. Finally, a classifier is used to determine whether it is necessary to issue an electric theft alarm for the distributed photovoltaic grid-connected power generation system, thereby timely detecting electric theft, improving system security, and preventing losses.

[0006] According to one aspect of the present application, there is provided an anti-electricity theft system for a distributed photovoltaic grid-connected power generation system, comprising:

[0007] The photovoltaic grid-connected anti-electricity theft data acquisition module is used to obtain the solar radiation intensity values ​​at multiple predetermined time points collected by the sensor, the photovoltaic power generation system output power at multiple predetermined time points collected by the photovoltaic operation monitoring system, and the historical energy data collected by the database;

[0008] a photovoltaic grid-connected power theft prevention data extraction module, configured to extract a photovoltaic grid-connected power generation anti-power theft global feature vector and a historical energy global feature vector from the solar radiation intensity values ​​at multiple predetermined time points collected by the sensor, the photovoltaic power generation system output power at multiple predetermined time points collected by the photovoltaic operation monitoring system, and the historical energy data collected by the database;

[0009] The photovoltaic grid-connected power theft alarm generating module is used to determine whether it is necessary to issue a distributed photovoltaic grid-connected power generation system power theft alarm based on the photovoltaic grid-connected power generation anti-power theft global feature vector and the historical energy global feature vector.

[0010] Compared with the existing technology, the present application provides an anti-electricity theft system for a distributed photovoltaic grid-connected power generation system, which first obtains the solar radiation intensity values ​​at multiple predetermined time points collected by a sensor, the photovoltaic power generation system output power at multiple predetermined time points collected by a photovoltaic operation monitoring system, and historical energy data collected by a database, and then uses deep learning technology to perform feature extraction and correlation analysis on the three. Finally, a classifier is used to determine whether it is necessary to issue an electricity theft alarm for the distributed photovoltaic grid-connected power generation system, thereby timely discovering electricity theft, improving system security, and preventing losses. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] 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.

[0012] Figure 1 4 is a block diagram of an anti-electricity theft system for a distributed photovoltaic grid-connected power generation system according to an embodiment of the present application.

[0013] Figure 2 1 is a block diagram of a photovoltaic grid-connected anti-electricity-theft data extraction module in an anti-electricity-theft system for a distributed photovoltaic grid-connected power generation system according to an embodiment of the present application.

[0014] Figure 3 The present invention is a block diagram of a photovoltaic grid-connected power theft alarm generation module in an anti-power theft system for a distributed photovoltaic grid-connected power generation system according to an embodiment of the present application.

[0015] Figure 4 is a block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0016] 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.

[0017] Exemplary Systems

[0018] Figure 1 This is a block diagram of an anti-theft system for a distributed photovoltaic grid-connected power generation system according to an embodiment of the present application. Figure 1 As shown, according to an embodiment of the present application, the anti-electricity theft system 100 for a distributed photovoltaic grid-connected power generation system includes: a photovoltaic grid-connected anti-electricity theft data acquisition module 110, for acquiring solar radiation intensity values ​​at multiple predetermined time points collected by a sensor, photovoltaic power generation system output power at multiple predetermined time points collected by a photovoltaic operation monitoring system, and historical energy data collected by a database; a photovoltaic grid-connected anti-electricity theft data extraction module 120, for extracting a photovoltaic grid-connected power generation anti-electricity theft global feature vector and a historical energy global feature vector from the solar radiation intensity values ​​at multiple predetermined time points collected by the sensor, the photovoltaic power generation system output power at multiple predetermined time points collected by the photovoltaic operation monitoring system, and the historical energy data collected by the database; and a photovoltaic grid-connected power theft alarm generation module 130, for determining whether it is necessary to issue a distributed photovoltaic grid-connected power generation system anti-electricity theft alarm based on the photovoltaic grid-connected power generation anti-electricity theft global feature vector and the historical energy global feature vector.

[0019] In the aforementioned anti-electricity theft system 100 for distributed photovoltaic grid-connected power generation systems, the photovoltaic grid-connected anti-electricity theft data acquisition module 110 is configured to acquire solar radiation intensity values ​​at multiple predetermined time points collected by sensors, photovoltaic power generation system output power at multiple predetermined time points collected by the photovoltaic operation monitoring system, and historical energy data collected by a database. It should be understood that distributed photovoltaic grid-connected systems connect photovoltaic power generation systems distributed across different locations through the power grid to achieve resource sharing and complementarity, thereby improving overall efficiency and stability. These systems enjoy preferential feed-in tariff subsidies to encourage their adoption. However, these incentives have also attracted some criminals who exploit electricity theft to profit illegally. Electricity theft, which involves the unpaid extraction of electricity from the public power grid, is a long-standing and widespread problem, causing serious losses and negative impacts to the country, businesses, and society. The methods used to steal electricity are diverse, and anti-electricity theft systems and technical measures are constantly evolving. Some criminals illegally obtain electricity price subsidies by tampering with photovoltaic power meter readings, or by directly drawing electricity from the grid and feeding it back to the grid through photovoltaic power generation system equipment. Manual investigations for electricity theft are time-consuming and inefficient, and fail to provide accurate information on the time and quantity of electricity theft. Therefore, the technical solution of this application utilizes solar radiation intensity values ​​collected by sensors at multiple predetermined time points, photovoltaic power generation system output power collected by a photovoltaic operation monitoring system at multiple predetermined time points, and historical energy data in a database. Combined with deep learning technology, the system can determine whether electricity theft is occurring in distributed photovoltaic grid-connected power generation systems and issue timely alerts to improve system security and prevent losses.

[0020] Specifically, the acquisition of solar radiation intensity values ​​collected by sensors at multiple predetermined time points, photovoltaic power generation system output power values ​​collected by the photovoltaic operation monitoring system at multiple predetermined time points, and historical energy data collected by the database aims to establish a comprehensive and dynamic information foundation for effectively monitoring and managing the operating status and performance of distributed photovoltaic grid-connected power generation systems. Solar radiation intensity values ​​are a key external environmental factor that directly impacts the power generation efficiency of photovoltaic power generation systems. Therefore, timely and accurate acquisition of this data can help system operators optimize power generation plans and resource utilization. The output power of photovoltaic power generation systems represents the actual power generation capacity of the system. Monitoring this indicator can promptly detect anomalies or faults in system operation and ensure stable system operation. Historical energy data provides a record of the system's past operation and trends. Analysis of this data can reveal system development patterns and potential problems, providing a reference for long-term system operation and optimization. This system can establish a comprehensive operation management system, improve the overall efficiency and reliability of the system, and ensure safe and stable operation.

[0021] In the aforementioned anti-electricity theft system 100 for distributed photovoltaic grid-connected power generation systems, the photovoltaic grid-connected anti-electricity theft data extraction module 120 is configured to extract a photovoltaic grid-connected power generation anti-electricity theft global feature vector and a historical energy global feature vector from the solar radiation intensity values ​​collected by the sensor at multiple predetermined time points, the photovoltaic power generation system output power collected by the photovoltaic operation monitoring system at multiple predetermined time points, and the historical energy data collected by the database. It should be understood that the process of extracting feature vectors from the solar radiation intensity values ​​collected by the sensor, the photovoltaic power generation system output power, and the historical energy data in the database is intended to transform these multidimensional data into representative and information-rich feature representations, enabling the system to more effectively determine whether electricity theft has occurred and respond accordingly. This process, by fully utilizing information from different data sources, can enhance the system's ability to comprehensively understand and monitor the operating status of the photovoltaic grid-connected power generation system. Thus, extracting feature vectors from different data sources helps the system comprehensively understand the operating status of the photovoltaic grid-connected power generation system, improves the system's ability to detect and respond to electricity theft, and ensures the system's safe and stable operation.

[0022] Figure 2 FIG. 1 is a block diagram of a photovoltaic grid-connected anti-electricity theft data extraction module in an anti-electricity theft system for a distributed photovoltaic grid-connected power generation system according to an embodiment of the present application. Figure 2 As shown, in a specific embodiment of the present application, the photovoltaic grid-connected anti-electricity theft data extraction module 120 includes: a solar radiation intensity feature extraction unit 121, used to perform feature extraction on the solar radiation intensity values ​​at multiple predetermined time points collected by the sensor to obtain a solar radiation intensity feature vector; a photovoltaic power generation output power feature extraction unit 122, used to perform feature extraction on the photovoltaic power generation system output power at multiple predetermined time points collected by the photovoltaic operation monitoring system to obtain a photovoltaic power generation output power feature vector; a photovoltaic grid-connected power generation feature aggregation unit 123, used to perform feature aggregation on the solar radiation intensity feature vector and the photovoltaic power generation output power feature vector to obtain the photovoltaic grid-connected power generation anti-electricity theft global feature vector; and a historical energy feature extraction unit 124, used to perform feature extraction on the historical energy data collected by the database to obtain the historical energy global feature vector.

[0023] It should be understood that by extracting features from solar radiation intensity values, the system can better understand the characteristics and patterns of solar radiation, thereby providing a more accurate information foundation for system operation and management. Feature extraction can help the system identify abnormal changes or electricity theft, allowing timely measures to safeguard the normal operation of the grid-connected photovoltaic power generation system. During feature extraction, statistical features of solar radiation intensity, such as mean, variance, maximum, and minimum values, can be considered to reflect the overall intensity and range of solar radiation. Furthermore, frequency domain features, such as power spectral density or spectrum characteristics, can be extracted to reveal periodic variations or frequency information in solar radiation. Time series features are also crucial here, including trend analysis and periodicity analysis, helping the system understand the long-term trends and periodic patterns of solar radiation. Therefore, feature extraction from solar radiation intensity values ​​is a crucial step in optimizing system monitoring and management capabilities, helping to improve system safety, stability, and efficiency, and achieving sustainable operation and development of photovoltaic power generation systems.

[0024] Furthermore, feature extraction is performed on the PV system output power at multiple predetermined time points, collected by the PV operation monitoring system. This aims to extract key features from the output power data, enabling the system to more accurately analyze the PV system's operating status, power generation performance, and potential anomalies, thereby determining whether electricity theft is occurring. By extracting the PV output power feature vector, the system can transform complex output power data into a more representative and informative feature representation, effectively supporting system monitoring and management.

[0025] Furthermore, feature aggregation is performed on the solar radiation intensity feature vector and the photovoltaic power output power feature vector. This aims to comprehensively utilize the characteristic information of solar radiation and photovoltaic power output to more comprehensively describe the operating status and performance characteristics of the photovoltaic grid-connected power generation system, allowing the system to more accurately and comprehensively determine whether electricity theft occurs. Feature aggregation organically combines information on solar radiation and power output, providing the system with a more comprehensive and multi-dimensional feature representation, which helps the system conduct more in-depth analysis and identification of the operating status of the photovoltaic power generation system. During the feature aggregation process, various methods can be used to integrate the solar radiation intensity feature vector and the photovoltaic power output feature vector, such as simple feature concatenation, weighted averaging, and feature fusion. Through feature aggregation, the system can comprehensively consider information on solar radiation and photovoltaic power output, better reflect the overall operating status and characteristics of the photovoltaic power generation system, and provide the system with a more comprehensive and accurate feature description.

[0026] Specifically, feature extraction is performed on historical energy data collected from the database. This aims to extract key features from historical energy data to reveal the historical operation, energy utilization, and potential problems of the energy system, providing the system with a more comprehensive representation of historical data features. By extracting global feature vectors of historical energy, the system can transform historical energy data into a more representative and informative feature form, providing richer information support for system analysis and decision-making.

[0027] In a specific embodiment of the present application, the solar radiation intensity feature extraction unit 121 includes: constructing the solar radiation intensity values ​​at multiple predetermined time points collected by the sensor into a solar radiation intensity input vector; and passing the solar radiation intensity input vector through a solar radiation intensity feature encoder based on a convolutional neural network to obtain the solar radiation intensity feature vector.

[0028] It should be understood that when constructing the solar radiation intensity input vector, the solar radiation intensity values ​​at multiple predetermined time points are usually arranged in chronological order and combined into a vector. This vector format enables the system to obtain solar radiation data at multiple time points at once, facilitating the system's analysis and prediction of the changing trends and periodicity of solar radiation. By constructing the solar radiation intensity input vector, the system can better understand the changing patterns of solar radiation and provide more accurate and timely information support for the operation and management of the system. The construction of the solar radiation intensity input vector also helps the system conduct a holistic analysis of solar radiation data and improve the system's utilization efficiency of solar energy resources. By integrating solar radiation data from multiple time points into a single vector, the system can more comprehensively consider the temporal and spatial variation characteristics of solar radiation, providing a more scientific basis for the system's optimized scheduling and operation.

[0029] Furthermore, the main purpose of converting the solar radiation intensity input vector into a solar radiation intensity feature vector through a convolutional neural network-based solar radiation intensity feature encoder is to extract and learn high-level feature representations from solar radiation data, thereby more effectively characterizing the complex patterns and variations of solar radiation. Through the solar radiation intensity feature encoder, the solar radiation intensity input vector is processed through a series of convolutional layers, pooling layers, and activation functions, gradually extracting and extracting local and global features from the solar radiation data. The convolutional layers effectively capture the spatial correlation in the solar radiation data, the pooling layers help reduce the data dimension and retain important information, and the activation functions introduce nonlinear factors to enhance the model's expressive power. Through these processes, the solar radiation intensity feature encoder transforms the original solar radiation intensity input vector into a more representative and advanced solar radiation intensity feature vector, enabling the system to better understand the inherent structure and characteristics of solar radiation data, realize automatic feature extraction and representation learning of solar radiation data, avoid the tedious process of manually designing features, and improve the system's efficiency and accuracy in processing solar radiation data. Specifically, each layer of the solar radiation intensity feature encoder based on the convolutional neural network is used to perform convolution processing, mean pooling processing based on the local feature matrix, and nonlinear activation processing on the input data in the forward pass of the layer so that the last layer of the solar radiation intensity feature encoder based on the convolutional neural network outputs the solar radiation intensity feature vector, wherein the input of the solar radiation intensity feature encoder based on the convolutional neural network is the solar radiation intensity input vector.

[0030] In a specific embodiment of the present application, the photovoltaic power generation output power feature extraction unit 122 includes: arranging the photovoltaic power generation system output power at multiple predetermined time points collected by the photovoltaic operation monitoring system into a photovoltaic power generation output power input vector; and passing the photovoltaic power generation output power input vector through a photovoltaic power generation output power feature encoder based on a convolutional neural network to obtain the photovoltaic power generation output power feature vector.

[0031] It should be understood that when constructing the photovoltaic power generation output power input vector, the output power of the photovoltaic power generation system at multiple predetermined time points is typically arranged in chronological order and combined into a single vector. This vector format enables the system to obtain photovoltaic power generation data at multiple time points at once, facilitating the analysis and prediction of the changing trends and volatility of the photovoltaic power generation system's output power. By constructing the photovoltaic power generation output power input vector, the system can better understand the power generation efficiency, stability, and potential problems of the photovoltaic power generation system. By integrating the photovoltaic power generation system output power data at multiple time points into a single vector, the system can more comprehensively consider the operating status and performance of the photovoltaic power generation system, providing a more scientific basis for the system's optimized scheduling and operation.

[0032] Furthermore, the main purpose of converting the photovoltaic power generation output power input vector into a photovoltaic power generation output power feature vector by using a photovoltaic power generation output power feature encoder based on a convolutional neural network is to extract and learn high-level feature representations in photovoltaic power generation data, thereby better describing the operating status and performance characteristics of the photovoltaic power generation system. Here, the convolutional neural network, as a deep learning model suitable for processing data with spatial structure, has good feature extraction and representation learning capabilities when processing time series data. Among them, the generation of the photovoltaic power generation output power feature vector enables the system to better understand the inherent characteristics and change patterns of photovoltaic power generation data, providing the system with richer and more useful information. Specifically, the photovoltaic power generation output power input vector is fully connected encoded using the fully connected layer of the photovoltaic power generation output power feature encoder based on the convolutional neural network to extract the high-dimensional implicit features of the eigenvalues ​​of each position in the photovoltaic power generation output power input vector; and the photovoltaic power generation output power input vector is one-dimensionally encoded using the one-dimensional convolutional layer of the photovoltaic power generation output power feature encoder based on the convolutional neural network to extract the high-dimensional implicit correlation features of the correlation between the eigenvalues ​​of each position in the photovoltaic power generation output power input vector.

[0033] In a specific embodiment of the present application, the photovoltaic grid-connected power generation feature aggregation unit 123 includes: differentiating the solar radiation intensity feature vector and the photovoltaic power generation output power feature vector to obtain a photovoltaic grid-connected power generation differential feature vector; and passing the photovoltaic grid-connected power generation differential feature matrix through a photovoltaic grid-connected power generation feature differential convolutional neural network to obtain the photovoltaic grid-connected power generation anti-electricity theft global feature vector.

[0034] It should be understood that differential operations can help the system identify the correlation and temporal relationship between solar radiation intensity and photovoltaic power output, improving the system's monitoring and analysis capabilities for grid-connected photovoltaic power generation systems. Here, by performing a differential operation on the solar radiation intensity eigenvector and the photovoltaic power output eigenvector, the system can obtain a photovoltaic grid-connected power generation differential eigenvector, which contains information on the variation and trend between solar radiation intensity and photovoltaic power output. This differential eigenvector can more accurately reflect the impact of changes in solar radiation intensity on photovoltaic power output, helping the system to more promptly detect abnormal conditions and performance fluctuations in the photovoltaic power generation system, thereby improving the system's real-time monitoring and early warning capabilities. By analyzing the variation patterns and trends in the differential eigenvector, the system can gain a deeper understanding of the dynamic relationship between solar radiation intensity and photovoltaic power output, providing a more accurate and effective reference for system optimization, scheduling, and operational management.

[0035] Furthermore, the PV grid-connected power generation feature differential convolutional neural network combines the temporal information of differential features with the feature extraction capabilities of convolutional neural networks, effectively learning global feature representations within the PV grid-connected power generation system. The generation of global feature vectors for PV grid-connected power generation anti-theft enables the system to gain a deeper understanding of the overall characteristics and changing patterns of PV grid-connected power generation data, providing the system with a more comprehensive and advanced information representation. Through the PV grid-connected power generation feature differential convolutional neural network, the system can automatically extract and learn representations of PV grid-connected power generation data, improving the system's monitoring and analysis capabilities for the PV grid-connected power generation system and further enhancing its safety and efficiency. Specifically, each layer of the photovoltaic grid-connected power generation feature differential convolutional neural network is used to perform the following operations on the input data in the forward pass of the layer: convolution processing is performed on the input data based on the convolution kernel to generate a convolution feature map; global mean pooling processing is performed on the convolution feature map based on the feature matrix to generate a pooled feature map; and nonlinear activation is performed on the eigenvalues ​​of each position in the pooled feature map to generate an activation feature addition map; wherein the output of the last layer of the photovoltaic grid-connected power generation feature differential convolutional neural network is the photovoltaic grid-connected power generation anti-electricity theft global feature vector, the input from the second layer to the last layer of the photovoltaic grid-connected power generation feature differential convolutional neural network is the output of the previous layer, and the input of the photovoltaic grid-connected power generation feature differential convolutional neural network is the photovoltaic grid-connected power generation differential feature matrix.

[0036] In a specific embodiment of the present application, the historical energy feature extraction unit 124 includes: passing the historical energy data collected from the database through a converter-based historical energy data context semantic encoder to obtain multiple historical energy feature vectors; and cascading the multiple historical energy feature vectors to obtain the historical energy global feature vector.

[0037] It should be understood that the converter-based historical energy data context semantic encoder combines the powerful feature extraction and representation learning capabilities of the converter model to transform historical energy data into feature vector representations with semantic information. The historical energy data context semantic encoder first encodes the historical energy data using the converter model, converting the data sequence into a high-dimensional semantic representation. The converter model effectively captures the contextual information and semantic associations in the energy data sequence, thereby extracting the underlying features and patterns in the data. This encoding method effectively expresses and encodes the time series characteristics of the historical energy data and the correlations between the data. The encoder converts the historical energy data into multiple historical energy feature vectors, each of which corresponds to a semantic representation of a different time period or feature dimension in the data sequence. These feature vectors contain important features and patterns in the historical energy data and can reflect the operating status and performance characteristics of the energy system at different time periods. Specifically, the historical energy data collected by the database is segmented to obtain a word sequence; the embedding layer of the converter-based historical energy data context semantic encoder is used to map each word in the word sequence into a word embedding vector to obtain a sequence of word embedding vectors; and the converter-based Bert model of the converter-based historical energy data context semantic encoder is used to perform global context semantic encoding on the sequence of word embedding vectors to obtain multiple historical energy feature vectors.

[0038] Furthermore, the purpose of concatenating multiple historical energy feature vectors to obtain a global historical energy feature vector is to comprehensively utilize the information contained in each feature vector, thereby more comprehensively describing and characterizing the characteristics and changing patterns of the entire historical energy data. By concatenating multiple feature vectors to form a larger global feature vector, different aspects of information can be integrated, allowing the system to more accurately capture the overall characteristics and trends of historical energy data, improving the system's understanding and analysis of the historical operation of the energy system. The process of concatenating multiple historical energy feature vectors can be viewed as a process of fusing and integrating the information of each feature vector. Each historical energy feature vector may capture different aspects of the data, such as energy consumption, energy production, and energy utilization efficiency. By concatenating these feature vectors, this diverse information can be organically combined to form a more comprehensive and comprehensive global historical energy feature vector.

[0039] In the aforementioned anti-theft system 100 for a distributed photovoltaic grid-connected power generation system, the photovoltaic grid-connected power theft alarm generation module 130 is configured to determine whether to issue a distributed photovoltaic grid-connected power generation system power theft alarm based on the photovoltaic grid-connected power generation anti-theft global feature vector and the historical energy global feature vector. As a distributed energy system, photovoltaic grid-connected power generation systems are susceptible to power theft, making timely alarm issuance crucial for preventing losses and maintaining system operation. Using the photovoltaic grid-connected power generation anti-theft global feature vector, the system can monitor and analyze the system's current operating status in real time, including key parameters such as photovoltaic power generation output power, grid access, and system load. These feature vectors reflect the system's real-time operating status and help the system identify abnormal behavior and signs of power theft. The historical energy global feature vector provides comprehensive features of the system's historical operating data, including information such as energy consumption patterns, power generation efficiency trends, and load variation patterns. By comprehensively considering the photovoltaic grid-connected power generation anti-theft global feature vector and the historical energy global feature vector, the system can compare the current operating status with historical data characteristics to detect abnormalities and power theft. If the system detects that the current characteristics match the electricity theft characteristics in historical data or an abnormal situation occurs, the system can automatically issue an electricity theft alarm and notify relevant personnel to conduct further investigation and processing to protect the safety and stable operation of the system.

[0040] Figure 3 FIG. 1 is a block diagram of a photovoltaic grid-connected power theft alarm generation module in an anti-theft power system for a distributed photovoltaic grid-connected power generation system according to an embodiment of the present application. Figure 3 As shown, in a specific embodiment of the present application, the photovoltaic grid-connected electricity theft alarm generation module 130 includes: an electricity theft judgment feature fusion unit 131, which is used to fuse the photovoltaic grid-connected power generation anti-electricity theft global feature vector and the historical energy global feature vector to obtain an electricity theft judgment classification feature vector; an electricity theft judgment feature optimization unit 132, which is used to perform feature fine internal structure optimization based on inherent decomposition on the electricity theft judgment classification feature vector to obtain an optimized electricity theft judgment classification feature vector; and an electricity theft alarm early warning judgment unit 133, which is used to pass the optimized electricity theft judgment classification feature vector through a classifier to obtain a classification result, and the classification result is used to determine whether it is necessary to issue an electricity theft alarm for the distributed photovoltaic grid-connected power generation system.

[0041] It should be understood that the purpose of fusing the global PV grid-connected power generation anti-theft feature vector and the historical energy global feature vector to generate the theft classification feature vector is to comprehensively utilize information from the current system state and historical data to improve the accuracy and reliability of theft identification. By fusing these two types of global feature vectors, the system can more comprehensively assess system operation, identify potential theft, and take appropriate measures to ensure safe and stable system operation. This combined feature vector can more comprehensively describe system operation, identify potential theft, and provide timely alerts and decision support to system operators.

[0042] In particular, in the technical solution of this application, the electricity theft classification feature vector is key for determining whether electricity theft has occurred in a distributed photovoltaic grid-connected power generation system. Improving the convergence of the electricity theft classification feature vector during the classification process by the classifier is crucial to ensuring that the system accurately and efficiently identifies electricity theft. Specifically, convergence refers to the ability of the classification feature vector to gradually stabilize and approach the optimal solution during training. If the electricity theft classification feature vector can quickly and accurately converge to the correct position in the class probability label space, the classifier can more accurately distinguish between normal and abnormal (electricity theft) behavior. In practical applications, misclassification (misclassifying normal behavior as electricity theft or missing electricity theft) can lead to unnecessary economic losses and loss of reputation. Improving convergence can reduce the occurrence of these misclassifications. Convergence also affects the speed of the training process. If the electricity theft classification feature vector can converge quickly, the entire training process can be shortened, thereby accelerating the deployment and application of the model. A model with good convergence generally also has better generalization ability, meaning that the model not only performs well on the training data but also adapts to new, unseen data. Overfitting refers to a situation where a model performs well on training data but poorly on new data. Improving the convergence of the electricity theft classification feature vector helps find a more generalized feature representation, thereby reducing the risk of overfitting. Based on this, the electricity theft classification feature vector is optimized using a feature microstructure optimization method based on inherent decomposition to obtain an optimized electricity theft classification feature vector.

[0043] Furthermore, the electricity theft judgment classification feature vector is optimized based on inherent decomposition to obtain an optimized electricity theft judgment classification feature vector, including: first, calculating the overall subtle autocorrelation topological matrix of the electricity theft judgment classification feature vector, which is expressed as:

[0044]

[0045] M=D1⊙D2

[0046] v i ,v j ∈V

[0047] Among them, V represents the classification feature vector of electricity theft judgment, v i and v j They represent the i-th and j-th eigenvalues ​​of the electricity theft judgment classification feature vector, w1, w2, w3 and w4 represent different weight hyperparameters, ⊙ represents matrix dot product, D1 represents the forward weight matrix of the electricity theft judgment classification feature, and D2 represents the reverse weight matrix of the electricity theft judgment classification feature. Represents the value of the (i, j)th position of the forward weighted matrix of the electricity theft judgment classification feature, It represents the value of the (i, j)th position of the inverse weighted matrix of the classification feature of electricity theft judgment, and M represents the overall subtle autocorrelation topology matrix.

[0048] Specifically, by quantifying the subtle correlations between the components of the electricity theft classification feature vector at the time series and dimensional levels, the previously implicit structural information between features is converted into a directly analyzable mathematical expression, thus providing a basis for revealing the abnormal correlation patterns of features that may be caused by electricity theft. This explicitly captures the subtle temporal and spatial dependency structures of each dimension in the electricity theft classification feature vector, laying the data foundation for subsequent mining of the deep internal structure of features through operations such as inherent decomposition. This enables the system to more accurately identify abnormal feature combinations consistent with electricity theft from the correlation patterns between features, thereby improving the reliability of electricity theft judgments.

[0049] Secondly, the overall fine autocorrelation topological matrix is ​​inherently decomposed to obtain a set of fine intrinsic component coding vectors of the electricity theft judgment classification feature, which is expressed as follows:

[0050]

[0051] 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 intrinsic components of the classification features of electricity theft judgment, x1, x2 and x m They respectively represent the first, second and mth electricity theft judgment classification feature fine intrinsic component coding vectors in the set of electricity theft judgment classification feature fine intrinsic component coding vectors.

[0052] Specifically, by utilizing inherent decomposition, the complex correlation structure contained within the overall subtle autocorrelation topological matrix is ​​broken down into a series of independent, importance-ordered, subtle eigencomponent encoding vectors of the electricity theft classification features. This forms an orthogonal basis guided by the intrinsic structure of the data, thereby isolating distinct independent patterns of feature variation. This transforms the internally coupled correlation structure of the features into independently analyzable basic patterns, providing clear structural units for subsequent dimensionality compression. This allows the system to focus on the feature variation patterns that are critical to electricity theft identification, eliminating redundant information interference and more accurately capturing feature correlation anomalies caused by electricity theft.

[0053] Then, dimension compression is performed on each of the fine intrinsic component coding vectors of the electricity theft judgment classification feature in the set of the electricity theft judgment classification feature fine intrinsic component coding vectors to obtain a set of electricity theft judgment classification feature fine intrinsic component modulation coding vectors, which is expressed as follows:

[0054]

[0055] Among them, x i represents the i-th fine intrinsic component coding vector of the electricity theft judgment classification feature in the set of fine intrinsic component coding vectors, ||·|| represents the Euclidean norm, y i Represents the modulation coding vector of the subtle intrinsic component of the i-th electricity theft judgment classification feature.

[0056] Specifically, through targeted dimensionality reduction, while retaining key feature variation patterns, redundant dimensions are eliminated, enhancing the ability of the encoding vectors of the fine intrinsic components of the electricity theft classification features to characterize electricity theft anomalies. This allows the compressed modulation encoding vectors of the fine intrinsic components of the electricity theft classification features to both reflect the independent variation patterns of the feature structure and highlight the core differential information related to electricity theft at a lower dimension. This reduces feature dimensionality without losing key information for identifying electricity theft, reduces computational redundancy, and improves model processing efficiency, enabling subsequent weight allocation and feature fusion to more accurately focus on core feature patterns that are strongly correlated with electricity theft.

[0057] Next, the internal structure important modulation factor of each electricity theft judgment classification feature fine intrinsic component modulation code vector in the set of the electricity theft judgment classification feature fine intrinsic component modulation code vector is calculated to obtain a set of internal structure important modulation factors, which can be expressed as follows:

[0058]

[0059] 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 intrinsic component of the classification feature of electricity theft judgment, a irepresents y i The corresponding important modulation factors of the internal structure.

[0060] That is, the scalar value of the modulation coding vector of the subtle intrinsic component of each electricity theft judgment classification feature is calculated to quantify its importance in characterizing the electricity theft characteristics, provide a dynamic weight basis based on the current modulation state for subsequent weighted fusion, and realize adaptive feature reconstruction, so that the system can dynamically adjust the weight of each feature component in the electricity theft judgment, more accurately focus on the features that play a key role in the electricity theft judgment, and improve the system's recognition ability and judgment accuracy of complex electricity theft behaviors.

[0061] Then, the set of the important modulation factors of the inner structure is normalized to obtain a set of important modulation weight factors of the inner structure, which is expressed as follows:

[0062] w i =Softmax(a i )

[0063] Among them, Softmax(·) represents the normalization function, w i Indicates a i The corresponding internal structure important modulation weight factor.

[0064] That is, through standardization, the important modulation factors of the internal structure are converted into important modulation weight factors of the internal structure that meet certain specifications. This ensures that the sum of these important modulation weight factors is 1 and, at an appropriate scale, avoids problems such as gradient explosion or vanishing, making the feature fusion process more stable and controllable while preventing over-reliance on a small number of feature components. In this way, the system can more stably perform weighted fusion of the modulation code vectors of the subtle intrinsic components of different electricity theft classification features, thereby improving the accuracy and reliability of the overall electricity theft judgment.

[0065] Finally, based on the set of important modulation weight factors of the internal structure, the set of fine intrinsic component modulation coding vectors of the electricity theft judgment classification feature is finely fused to obtain the optimized electricity theft judgment classification feature vector, which is expressed as follows:

[0066]

[0067] Wherein, V' represents the optimized classification feature vector for electricity theft judgment.

[0068] Specifically, based on the importance of the fine-grained intrinsic component modulation coding vectors of each electricity theft classification feature, multiple fine-grained intrinsic component modulation coding vectors are fused into an optimized electricity theft classification feature vector. This condenses the key information of the original electricity theft classification feature and enhances its representational capabilities. This makes the optimized electricity theft classification feature vector more robust and discriminative than the condensed original electricity theft classification feature vector, enhancing key information and suppressing secondary or noisy information. This provides strong support for subsequent accurate judgment of whether distributed photovoltaic grid-connected power generation systems have committed electricity theft.

[0069] Furthermore, by combining optimized classification feature vectors for electricity theft detection with a classifier, the system can intelligently identify and classify electricity theft, improving system security and reliability. The classifier effectively filters and identifies electricity theft, reducing the burden of manual monitoring while improving the accuracy and efficiency of theft detection.

[0070] In summary, the embodiment of the present application first obtains the solar radiation intensity values ​​at multiple predetermined time points collected by the sensor, the output power of the photovoltaic power generation system at multiple predetermined time points collected by the photovoltaic operation monitoring system, and the historical energy data collected by the database, and then uses deep learning technology to perform feature extraction and correlation analysis on the three. Finally, a classifier is used to determine whether it is necessary to issue a distributed photovoltaic grid-connected power generation system electricity theft alarm, so as to timely detect electricity theft behavior, improve the safety of the system, and prevent losses.

[0071] As described above, the anti-electricity theft system 100 for a distributed photovoltaic grid-connected power generation system according to an embodiment of the present application can be implemented in various terminal devices. In one example, the anti-electricity theft system 100 for a distributed photovoltaic grid-connected power generation system can be integrated into a terminal device as a software module and / or a hardware module. For example, the anti-electricity theft system 100 for a distributed photovoltaic grid-connected power generation system can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device. Of course, the anti-electricity theft system 100 for a distributed photovoltaic grid-connected power generation system can also be one of the many hardware modules of the terminal device.

[0072] Alternatively, in another example, the anti-electricity theft system 100 for a distributed photovoltaic grid-connected power generation system and the terminal device may also be separate devices, and the anti-electricity theft system 100 for a distributed photovoltaic grid-connected power generation system may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0073] Exemplary electronic devices

[0074] Below, reference Figure 4 To describe the electronic device according to the embodiment of the present application.

[0075] like Figure 4 As shown, the electronic device 10 includes an input device 11, an input interface 12, a central processing unit 13, a memory 14, an output interface 15, an output device 16, and a bus 17. The input interface 12, the central processing unit 13, the memory 14, and the output interface 15 are interconnected via the bus 17, and the input device 11 and the output device 16 are connected to the bus 17 via the input interface 12 and the output interface 15, respectively, and are further connected to other components of the electronic device 10.

[0076] Specifically, the input device 11 receives input information from the outside and transmits the input information to the central processing unit 13 through the input interface 12; the central processing unit 13 processes the input information based on the computer-executable instructions stored in the memory 14 to generate output information, stores the output information temporarily or permanently in the memory 14, and then transmits the output information to the output device 16 through the output interface 15; the output device 16 outputs the output information to the outside of the electronic device 10 for user use.

[0077] In one embodiment, Figure 4 The electronic device 10 shown can be implemented as a network device, which may include: a memory configured to store programs; a processor configured to run the programs stored in the memory to execute any one of the anti-electricity theft systems for distributed photovoltaic grid-connected power generation systems described in the above embodiments.

[0078] According to an embodiment of the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, an embodiment of the present application includes a computer program product comprising a computer program tangibly embodied on a machine-readable medium, the computer program comprising program code for executing the methods shown in the flowcharts. In such an embodiment, the computer program can be downloaded and installed from a network and / or installed from a removable storage medium.

[0079] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In hardware implementations, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium). As is well known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable, and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those skilled in the art that communication media typically embodies computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0080] It is understood that the above embodiments are merely exemplary embodiments for illustrating the principles of the present application, and the present application is not limited thereto. Those skilled in the art may make various modifications and improvements without departing from the spirit and substance of the present application, and such modifications and improvements are also considered to be within the scope of protection of the present application.

Claims

1. An anti-electricity theft system for a distributed photovoltaic grid-connected power generation system, characterized in that: include: The photovoltaic grid-connected anti-electricity theft data acquisition module is used to obtain the solar radiation intensity values ​​at multiple predetermined time points collected by the sensor, the photovoltaic power generation system output power at multiple predetermined time points collected by the photovoltaic operation monitoring system, and the historical energy data collected by the database; a photovoltaic grid-connected power theft prevention data extraction module, configured to extract a photovoltaic grid-connected power generation anti-power theft global feature vector and a historical energy global feature vector from the solar radiation intensity values ​​at multiple predetermined time points collected by the sensor, the photovoltaic power generation system output power at multiple predetermined time points collected by the photovoltaic operation monitoring system, and the historical energy data collected by the database; The photovoltaic grid-connected power theft alarm generating module is used to determine whether it is necessary to issue a distributed photovoltaic grid-connected power generation system power theft alarm based on the photovoltaic grid-connected power generation anti-power theft global feature vector and the historical energy global feature vector.

2. The anti-electricity theft system for a distributed photovoltaic grid-connected power generation system according to claim 1, characterized in that: The photovoltaic grid-connected anti-electricity theft data extraction module includes: a solar radiation intensity feature extraction unit, configured to extract features from the solar radiation intensity values ​​at the plurality of predetermined time points collected by the sensor to obtain a solar radiation intensity feature vector; a photovoltaic power generation output power feature extraction unit, configured to extract features of the photovoltaic power generation system output power at a plurality of predetermined time points collected by the photovoltaic operation monitoring system to obtain a photovoltaic power generation output power feature vector; a photovoltaic grid-connected power generation feature aggregation unit, configured to perform feature aggregation on the solar radiation intensity feature vector and the photovoltaic power generation output power feature vector to obtain a photovoltaic grid-connected power generation anti-electricity theft global feature vector; The historical energy feature extraction unit is used to extract features from the historical energy data collected from the database to obtain the historical energy global feature vector.

3. The anti-electricity theft system for distributed photovoltaic grid-connected power generation system according to claim 2, characterized in that: The solar radiation intensity feature extraction unit includes: constructing the solar radiation intensity values ​​at the plurality of predetermined time points collected by the sensor into a solar radiation intensity input vector; The solar radiation intensity input vector is passed through a solar radiation intensity feature encoder based on a convolutional neural network to obtain the solar radiation intensity feature vector.

4. The anti-electricity theft system for a distributed photovoltaic grid-connected power generation system according to claim 3, characterized in that: The photovoltaic power generation output power feature extraction unit includes: Arranging the photovoltaic power generation system output powers at a plurality of predetermined time points collected by the photovoltaic operation monitoring system into a photovoltaic power generation output power input vector; The photovoltaic power generation output power input vector is passed through a photovoltaic power generation output power feature encoder based on a convolutional neural network to obtain the photovoltaic power generation output power feature vector.

5. The anti-electricity theft system for distributed photovoltaic grid-connected power generation system according to claim 4, characterized in that: The photovoltaic grid-connected power generation characteristic aggregation unit includes: Differentiating the solar radiation intensity characteristic vector and the photovoltaic power generation output power characteristic vector to obtain a photovoltaic grid-connected power generation differential characteristic vector; The photovoltaic grid-connected power generation differential feature matrix is ​​passed through a photovoltaic grid-connected power generation feature differential convolutional neural network to obtain the photovoltaic grid-connected power generation anti-electricity theft global feature vector.

6. The anti-electricity theft system for a distributed photovoltaic grid-connected power generation system according to claim 5, characterized in that: The historical energy feature extraction unit includes: The historical energy data collected from the database is passed through a converter-based historical energy data context semantic encoder to obtain a plurality of historical energy feature vectors; The multiple historical energy feature vectors are cascaded to obtain the historical energy global feature vector.

7. The anti-electricity theft system for a distributed photovoltaic grid-connected power generation system according to claim 6, characterized in that: The photovoltaic grid-connected electricity theft alarm generation module includes: an electricity theft judgment feature fusion unit, configured to fuse the photovoltaic grid-connected power generation anti-electricity theft global feature vector and the historical energy global feature vector to obtain an electricity theft judgment classification feature vector; an electricity theft judgment feature optimization unit, configured to perform feature fine internal structure optimization on the electricity theft judgment classification feature vector based on inherent decomposition to obtain an optimized electricity theft judgment classification feature vector; The electricity theft alarm early warning judgment unit is used to pass the optimized electricity theft judgment classification feature vector through a classifier to obtain a classification result, and the classification result is used to judge whether it is necessary to issue a distributed photovoltaic grid-connected power generation system electricity theft alarm.

8. The anti-electricity theft system for a distributed photovoltaic grid-connected power generation system according to claim 7, characterized in that: The electricity theft judgment feature optimization unit is used to: Calculate the overall subtle autocorrelation topological matrix of the electricity theft judgment classification feature vector; Performing intrinsic decomposition on the overall fine autocorrelation topological matrix to obtain a set of fine intrinsic component coding vectors of electricity theft judgment classification features; Performing dimension compression on each electricity theft judgment classification feature fine intrinsic component coding vector in the set of electricity theft judgment classification feature fine intrinsic component coding vectors to obtain a set of electricity theft judgment classification feature fine intrinsic component modulation coding vectors; Calculating the internal structure important modulation factor of each electricity theft judgment classification feature fine intrinsic component modulation code vector in the set of the electricity theft judgment classification feature fine intrinsic component modulation code vector to obtain a set of internal structure important modulation factors; Normalizing the set of the internal structure important modulation factors to obtain a set of internal structure important modulation weight factors; Based on the set of important modulation weight factors of the internal structure, the set of fine intrinsic component modulation coding vectors of the electricity theft judgment classification feature is finely fused to obtain an optimized electricity theft judgment classification feature vector.

Images