Fusion self-adaptive deconstruction method, device and equipment for line loss electric quantity of transformer area and medium

Abstract

The invention discloses a fusion self-adaptive deconstruction method, device, equipment and medium for the line loss electric quantity of a transformer area, and relates to the technical field of power systems. Abnormal data correction is carried out on a data type historical electric quantity data set through an abnormal value detection rule, the data quality is guaranteed, quantitative conversion is carried out on a text type abnormal cause label set, and the data quality is improved. And text information is converted into a numerical form which can be calculated by the model, and cross-modal data fusion is realized. And then comprehensive influence of various abnormal causes is incorporated into electric quantity loss calculation, so that the limitation of a traditional method on single factor or independent factor analysis is broken through, and a multi-factor synergistic effect is accurately reflected. Then performing model training on the neural network model through a multi-objective loss function and a dynamic coding mechanism, constructing a transformer area line loss electric quantity deconstruction model, enabling the model to automatically capture a complex coupling relationship between causes and line loss, and combining real-time data to update parameters so as to adapt to a dynamic scene; and the identification precision and generalization capability of the line loss abnormal causes are improved.

Description

Technical Field

[0001] The present application relates to the technical field of power systems, and in particular to a method, device, equipment and medium for fusion adaptive deconstruction of line loss electricity in a substation. Background Art

[0002] In power systems, substation line loss management is a critical component in ensuring the economic operation, security, and stability of the power grid. While the development of smart grids has enabled substation line loss analysis with vast amounts of data, the widespread integration of distributed energy resources has made substation power flows increasingly complex, exposing the shortcomings of traditional line loss management methods.

[0003] In related technologies, line loss analysis often uses numerical power data and relies on fixed rule models to handle anomalies. However, the applicant recognizes that existing methods can only process single numerical data, resulting in incomplete data cleaning and loss of potential related information, which affects the accuracy of analysis. Traditional models treat the causes of line loss anomalies as independent factors, making it difficult to accurately depict line loss changes under the synergy of multiple factors, resulting in deconstruction results that deviate from reality. In addition, the model's lack of adaptive learning and dynamic updating capabilities limits the accuracy of line loss power deconstruction and fails to meet the needs of refined smart grid management. Summary of the Invention

[0004] In view of this, the present application provides a fusion adaptive deconstruction method, device, equipment and medium for substation line loss electricity. The main purpose is to solve the problem that the existing method can only process single numerical data, the traditional model is difficult to accurately depict the line loss changes under the synergistic effect of multiple factors, and the model's adaptive learning and dynamic update capabilities are insufficient.

[0005] According to the first aspect of the present application, a method for fusion adaptive deconstruction of line loss in a substation area is provided, the method comprising:

[0006] Acquire a data-type historical electricity data set and a text-type abnormality cause label set, perform abnormal data correction on the data-type historical electricity data set, and quantify the text-type abnormality cause label set into a multi-dimensional line loss abnormality cause feature vector;

[0007] The modified data-type historical electricity data set is used to calculate the total power loss caused by the superposition of multiple abnormal causes for each abnormal line loss area, as well as the total power supply for the abnormal line loss month for each abnormal line loss area;

[0008] The total power loss due to the superposition of multiple abnormal causes of each abnormal line loss area, the total power supply of each abnormal line loss area in a month, and the multi-dimensional abnormal line loss cause feature vector are input into the neural network model, and the model is trained through a multi-objective loss function and a dynamic coding mechanism to construct a line loss power deconstruction model for the area;

[0009] Collecting the power data of the substation line loss, inputting the power data of the substation line loss into the power deconstruction model of the substation line loss for analysis, and outputting the power loss caused by each abnormal line loss.

[0010] According to a second aspect of the present application, a device for fusion and adaptive deconstruction of line loss in a substation area is provided, the device comprising:

[0011] a data processing module, configured to obtain a data-type historical electricity data set and a text-type abnormality cause label set, perform abnormal data correction on the data-type historical electricity data set, and quantify the text-type abnormality cause label set into a multi-dimensional line loss abnormality cause feature vector;

[0012] A data calculation module is used to calculate the total power loss caused by the superposition of multiple abnormal causes for each abnormal line loss area using the modified data-type historical power data set, and the total power supply of each abnormal line loss area in the month of abnormal line loss;

[0013] An adaptive learning module is configured to input the total power loss due to the superposition of multiple abnormal causes of each abnormal line loss substation, the total monthly power supply of each abnormal line loss substation, and the multi-dimensional abnormal line loss cause feature vector into a neural network model, and perform model training through a multi-objective loss function and a dynamic coding mechanism to construct a substation line loss power deconstruction model;

[0014] The deconstruction analysis module is used to collect the line loss power data of the substation area, input the line loss power data of the substation area into the line loss power deconstruction model of the substation area for analysis, and output the power loss caused by each abnormal line loss.

[0015] According to a third aspect of the present application, a device is provided, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the methods described in the first aspect when executing the computer program.

[0016] According to a fourth aspect of the present application, a medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the methods in the first aspect are implemented.

[0017] By means of the above technical solution, the technical solution provided by the embodiment of the present application has at least the following advantages:

[0018] The present application provides a method, device, equipment and medium for fusion adaptive deconstruction of line loss electricity in a substation. The present application obtains a data-type historical electricity data set and a text-type abnormal cause label set, performs abnormal data correction on the data-type historical electricity data set, and quantifies the text-type abnormal cause label set into a multi-dimensional line loss abnormal cause feature vector. By setting an outlier detection rule, the data-type historical electricity data set is corrected for abnormal data to ensure data quality; the text-type abnormal cause label set is quantitatively converted, the abnormal cause described in text is converted into a multi-dimensional line loss abnormal cause feature vector, and the text information is converted into a numerical form that can be calculated by the model, thereby realizing cross-modal data fusion. The corrected data-type historical electricity data set is then used to calculate the total loss electricity under the superposition of multiple abnormal causes for each line loss abnormal substation, as well as the total monthly power supply of line loss abnormalities for each line loss abnormal substation. By incorporating the comprehensive impact of multiple abnormal causes into the loss electricity calculation, the limitations of traditional methods on single factors or independent analysis of factors are broken, and the synergistic effect of multiple factors is accurately reflected, providing a scientific and quantitative basis for the definition and management of line loss responsibilities. The total power loss due to the superposition of multiple abnormal causes for each abnormal line loss substation, the total monthly power supply for each abnormal line loss substation, and the multi-dimensional line loss cause feature vector are then input into the neural network model. The model is trained using a multi-objective loss function and a dynamic encoding mechanism to construct a substation line loss power deconstruction model. This model automatically captures the complex coupling relationship between causes and line losses. It also updates parameters based on real-time data to adapt to dynamic scenarios, improving the accuracy and generalization of identifying abnormal line loss causes and providing self-evolutionary technical support for smart grid line loss management. In practical applications, the collected substation line loss power data can be input into the substation line loss power deconstruction model for analysis, outputting the power loss associated with each abnormal line loss cause, enabling accurate tracing of the causes of line losses and providing a quantitative basis for line loss management.

[0019] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present application. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0021] Figure 1 A schematic flow chart of a method for fusion and adaptive deconstruction of line loss in a transformer area provided in an embodiment of the present application is shown;

[0022] Figure 2A A schematic flow chart of another method for fusion and adaptive deconstruction of line loss in an area provided in an embodiment of the present application is shown;

[0023] Figure 2B A schematic diagram of the architecture of a method for deconstructing line loss in a transformer area based on electric energy analysis and adaptive learning of uncertainty of abnormal causes provided by an embodiment of the present application is shown;

[0024] Figure 3 A schematic diagram of the structure of a fusion adaptive deconstruction of line loss electricity in a transformer area provided by an embodiment of the present application is shown;

[0025] Figure 4 A schematic diagram of the device structure of a device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0026] In the description of the present application, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like to indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0027] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. Throughout the description of this application, "plurality" means two or more, unless otherwise specifically defined.

[0028] In this application, unless otherwise specified or limited, the terms "mounted," "connected," "connect," "fixed," etc. should be understood broadly. For example, they can refer to fixed connection, detachable connection, or integral connection; mechanical connection or electrical connection; direct connection or indirect connection through an intermediate medium; or internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in this application based on specific circumstances.

[0029] The following describes exemplary embodiments of the present application in more detail with reference to the accompanying drawings. Although exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0030] In the power system, substation line loss management is a key link in ensuring the economic operation, security and stability of the power grid. With the development of smart grids, substation line loss analysis has gained massive data support. However, the widespread access to distributed energy has made substation power flow increasingly complex, and traditional line loss management methods have gradually exposed many shortcomings. From the current state of the art, existing substation line loss power decomposition methods have the following obvious shortcomings:

[0031] In terms of data processing, existing technologies can only process a single type of data. They fail to integrate numerical data (such as power inflow and power loss) with non-numerical data (such as anomaly cause labels), making it difficult to uncover potential connections between different data types. In practice, the causes of line loss anomalies represented by anomaly cause labels (such as equipment aging and power consumption fluctuations) are closely linked to numerical data. However, the lack of an effective fusion mechanism leads to incomplete data cleaning and partial information acquisition, affecting the accuracy of subsequent line loss analysis and easily missing key influencing factors.

[0032] In terms of model construction, existing technologies do not consider the cumulative impact of abnormal causes in substation samples, treating each abnormal cause as an independent factor and failing to analyze the dynamic coupling relationship between it and the line loss rate (such as the combined impact of multiple abnormal causes on power loss). This makes it impossible to accurately quantify the contribution of each abnormal cause when calculating power loss. When making judgments, the factors are isolated and it is difficult to reveal the mechanism of the combined action of multiple factors. This leads to a deviation between the decomposition results of line loss power and the actual situation, affecting the scientific nature of the loss reduction strategy.

[0033] In terms of model learning capabilities, existing technologies are mostly based on fixed rules or simple statistical methods, lacking the ability to adaptively learn from the uncertainty of anomaly causes. The causes of abnormal line loss in substations are uncertain and dynamic (such as new types of anomaly causes), but the existing models cannot update and learn in real time based on new data, making them difficult to adapt to complex and changing line loss analysis scenarios. This leads to insufficient model generalization and the inability to accurately identify the ever-changing causes of line loss anomalies and quantify their impact.

[0034] To solve this problem, this application proposes a fusion adaptive deconstruction method for line loss in substations. First, a data-based historical electricity data set and a text-based abnormal cause label set are obtained, the numerical anomalies are corrected, and the text labels are quantified into feature vectors. Then, based on the corrected data, the total loss electricity of the line loss abnormal substation with multiple causes and the total power supply in the abnormal month are calculated. Then, these data and the feature vectors are input into a neural network, and a deconstruction model is constructed by training with a multi-objective loss function and a dynamic coding mechanism. Finally, the model is used to analyze real-time data and output the loss electricity of each abnormal cause. This solution improves data quality through cross-modal data fusion and cleaning. The comprehensive calculation of multiple causes breaks the limitations of single analysis and reflects synergistic effects. The adaptive neural network model realizes the learning of the uncertainty of abnormal causes, providing self-evolutionary technical support for smart grid line loss management. It has the advantages of reliable data quality, accurate analysis, and adaptability to dynamic scenarios. The executor of this application may be a substation line loss electricity deconstruction system, which relies on the computing power of the server to provide services to users. The server may be an independent server, or it may provide cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), as well as big data and artificial intelligence platforms and other basic cloud computing servers, so as to output the electricity loss caused by each abnormal line loss, realize the accurate deconstruction of the substation line loss electricity, and provide more scientific technical support for substation line loss management.

[0035] The embodiment of the present application provides a method for fusion adaptive deconstruction of line loss in a substation area, such as Figure 1 As shown, the method includes:

[0036] 101. Obtain a data-type historical electricity data set and a text-type abnormality cause label set, perform abnormal data correction on the data-type historical electricity data set, and quantify the text-type abnormality cause label set into a multi-dimensional line loss abnormality cause feature vector.

[0037] In the embodiment of the present application, the data-type historical electricity data set covers the numerical electric energy data (such as voltage, current, electricity, etc.) generated by the operation of the substation, including the monthly power supply and monthly power loss of the normal line loss month in the past year in the line loss abnormal substation, as well as the monthly power supply and monthly power loss of the line loss abnormal month, and the monthly power loss caused by different abnormal causes. The sampling granularity is monthly, which is the basic quantitative basis for line loss calculation. The text-type abnormality cause label set contains at least one line loss abnormality cause label for the line loss abnormal substation. The line loss abnormality cause label indicates the cause of the line loss abnormality in the line loss abnormal month, so that the collected data covers the time dimension and the cause dimension, providing sufficient material for subsequent comprehensive analysis of multiple causes and model learning of complex relationships. Moreover, through the correction of numerical data and quantitative encoding of text labels, the standardized processing of cross-modal data is achieved, the data quality is effectively improved, and the subsequent line loss analysis is based on more reliable data.

[0038] 102. Use the revised data-type historical electricity data set to calculate the total power loss caused by the superposition of multiple abnormal causes for each line loss abnormal area, as well as the total power supply for each line loss abnormal area in the month of line loss abnormality.

[0039] In the embodiment of the present application, the total power loss under the superposition of multiple abnormal causes is calculated, and the comprehensive impact of multiple abnormal causes is taken into consideration. Unlike the traditional method of single analysis or simple superposition of the impact of each cause, the complex scenario of multiple factors intertwined in the operation of the power grid is truly restored, and the actual situation of abnormal line loss caused by the coordination of multiple causes is accurately presented, which provides a scientific basis for clarifying the proportion of each cause in the comprehensive impact when defining the line loss responsibility and formulating a multi-cause collaborative governance strategy when governing.

[0040] 103. The total power loss due to the superposition of multiple abnormal causes of each line loss abnormality substation, the total monthly power supply of each line loss abnormality substation and the multi-dimensional line loss abnormality cause feature vector are input into the neural network model, and the model is trained through a multi-objective loss function and a dynamic coding mechanism to construct a substation line loss power deconstruction model.

[0041] In the embodiment of the present application, the total power loss of multiple abnormal causes superimposed on each other can reflect the line loss results caused by the combined effect of multiple causes. The total power supply in the month of abnormal line loss can provide basic data on the scale of power supply in the substation. The multi-dimensional line loss abnormal cause feature vector is quantified from the text-type abnormal cause label, and the cause described in text is converted into a numerical feature that can be calculated by the model, realizing cross-modal data fusion input. Then, a multi-objective loss function is used to balance multiple training objectives, such as minimizing the single cause loss prediction error, the total loss prediction error, etc., so that the model can take into account both local and overall prediction accuracy during the learning process. In addition, the introduction of a dynamic coding mechanism makes the model adaptive, and it can automatically adjust the coding method and model parameters according to data changes, such as adding new abnormal cause types, to adapt to dynamic line loss scenarios.

[0042] 104. Collect the power data of the substation line loss, input the power data of the substation line loss into the power deconstruction model of the substation line loss for analysis, and output the power loss caused by each abnormal line loss.

[0043] In the embodiment of the present application, data such as the substation line loss and electricity data covers the real-time or latest electric energy monitoring information of the substation, such as numerical data such as voltage, current, and electricity, as well as possible related text-type abnormality cause clues such as descriptions of newly discovered equipment abnormalities in operation and maintenance, providing real-time / latest input for model analysis, so that the substation line loss and electricity deconstruction model outputs the loss of electricity caused by each abnormal cause of line loss, accurately decomposes the electricity loss values ​​caused by different abnormal causes in the substation line loss, and realizes quantitative tracing of the causes of line loss.

[0044] The embodiment of the present application provides a fusion adaptive deconstruction method for line loss electricity in a substation. Compared with the prior art, the embodiment of the present application obtains a data-type historical electricity data set and a text-type abnormal cause label set, performs abnormal data correction on the data-type historical electricity data set, and quantifies the text-type abnormal cause label set into a multi-dimensional line loss abnormal cause feature vector. By setting an outlier detection rule, the data-type historical electricity data set is corrected for abnormal data to ensure data quality; the text-type abnormal cause label set is quantitatively converted, the abnormal cause described in text is converted into a multi-dimensional line loss abnormal cause feature vector, and the text information is converted into a numerical form that can be calculated by the model, thereby realizing cross-modal data fusion. The corrected data-type historical electricity data set is then used to calculate the total loss electricity under the superposition of multiple abnormal causes for each line loss abnormal substation, as well as the total monthly power supply of line loss abnormalities for each line loss abnormal substation. By incorporating the comprehensive impact of multiple abnormal causes into the loss electricity calculation, the limitations of traditional methods on single factors or independent analysis of factors are broken, and the synergistic effect of multiple factors is accurately reflected, providing a scientific and quantitative basis for the definition and management of line loss responsibilities. The total power loss due to the superposition of multiple abnormal causes for each abnormal line loss substation, the total monthly power supply for each abnormal line loss substation, and the multi-dimensional line loss cause feature vector are then input into the neural network model. The model is trained using a multi-objective loss function and a dynamic encoding mechanism to construct a substation line loss power deconstruction model. This model automatically captures the complex coupling relationship between causes and line losses. It also updates parameters based on real-time data to adapt to dynamic scenarios, improving the accuracy and generalization of identifying abnormal line loss causes and providing self-evolutionary technical support for smart grid line loss management. In practical applications, the collected substation line loss power data can be input into the substation line loss power deconstruction model for analysis, outputting the power loss associated with each abnormal line loss cause, enabling accurate tracing of the causes of line losses and providing a quantitative basis for line loss management.

[0045] Furthermore, as a refinement and extension of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, the embodiment of the present application provides another fusion adaptive deconstruction method of the line loss in the substation area, such as Figure 2A As shown, the method includes:

[0046] 201. Obtain a data-type historical electricity data set and a text-type abnormality cause label set, and perform abnormal data correction on the data-type historical electricity data set.

[0047] In an embodiment of the present application, multiple line loss abnormality areas are determined in a data-type historical electricity data set. For the i-th line loss abnormality area, the power supply and power loss of each line loss normal month of the i-th line loss abnormality area are extracted from the data-type historical electricity data set. Then, an irregular power supply data judgment inequality is obtained. Based on the power supply of each line loss normal month of the i-th line loss abnormality area, the line loss normal month that satisfies the irregular power supply data judgment inequality among the multiple line loss normal months of the i-th line loss abnormality area is used as the target line loss normal month, and at least one target line loss normal month is obtained. The calculation formula is as shown in the following formula 1:

[0048] Formula 1:

[0049]

[0050] Among them, W supply,i,t represents the power supply in the tth month with normal line loss in the ith abnormal line loss area, represents the average monthly power supply of the line loss in the i-th abnormal line loss area, δ supply,i N represents the standard deviation of monthly power supply in the i-th abnormal line loss area with normal line loss. normal,i represents the number of normal line loss months in the i-th abnormal line loss area. The power supply in each target normal line loss month is then replaced by the average power supply in the i-th abnormal line loss area in normal line loss months to obtain the corrected power supply in each target normal line loss month.

[0051] Then, an inequality for judging irregular power loss data is obtained. Based on the power loss of each normal line loss month in the i-th abnormal line loss substation, the normal line loss month that satisfies the inequality for judging irregular power loss data is used as the designated normal line loss month among multiple normal line loss months in the i-th abnormal line loss substation. At least one designated normal line loss month is obtained. The calculation formula is as shown in the following formula 2:

[0052] Formula 2:

[0053]

[0054] Among them, W loss,i,tIt represents the power loss in the tth month with normal line loss in the ith abnormal line loss area. represents the average monthly power loss of the line loss in the i-th abnormal line loss area, δ loss,i It represents the standard deviation of monthly power loss in the normal line loss area of ​​the ith abnormal line loss area, N normal,i represents the number of normal line loss months in the i-th abnormal line loss area. The power loss in each specified normal line loss month is replaced by the average power loss in the normal line loss month in the abnormal line loss area to obtain the corrected power loss in each specified normal line loss month.

[0055] Then, the corrected power supply of at least one target month with normal line loss and the corrected power loss of at least one specified month with normal line loss are used to correct the power supply and power loss of multiple normal line loss months in the i-th abnormal line loss substation in the data-based historical power data set. Using statistical methods, namely the mean + standard deviation + 3δ principle, the abnormal fluctuations in power supply and power loss are quantitatively defined. This is more scientific and operational than subjective judgment, while ensuring that the identified abnormal monthly data truly deviates from the normal pattern, eliminating interference for subsequent analysis. Furthermore, the correction logic is based on the statistical characteristics of the substation's own normal monthly data, that is, replacing abnormal values ​​with their own average values ​​rather than arbitrarily deleting or replacing them with fixed values. This approach not only preserves the historical normal patterns of the substation, but also eliminates the impact of abnormal data on subsequent analysis, ensuring the authenticity and continuity of the data.

[0056] 202. Quantify the text-based anomaly cause label set into a multi-dimensional line loss anomaly cause feature vector.

[0057] In an embodiment of the present application, a plurality of line loss abnormality cause label types are obtained by statistically analyzing the abnormality cause label set of line loss abnormality areas. For the i-th line loss abnormality area, at least one line loss abnormality cause label corresponding to the i-th line loss abnormality area is extracted from the abnormality cause label set of line loss abnormality areas, and vector encoding is performed on the at least one line loss abnormality cause label using the plurality of line loss abnormality cause label types to obtain a line loss abnormality cause feature vector for the i-th line loss abnormality area. The calculation formula is as shown in the following formula 3:

[0058] Formula 3: m i =[m i,1 ,m i,2 ,...,m i,k ,...,m i,K ],

[0059] Among them, m i represents the characteristic vector of the line loss anomaly in the ith line loss anomaly area, m i,k Indicates the existence of the kth line loss anomaly cause label in the i-th line loss anomaly area, m i,k =1 means that the i-th line loss abnormal area has the k-th line loss abnormal cause label, mi,k =0 means that the kth line loss anomaly cause label does not exist in the i-th line loss anomaly substation, and K represents the number of line loss anomaly cause label types. For example, an adaptive coding mechanism for line loss causes in different substations is established, and the set of anomaly causes in different substations is encoded using the hot coding principle to obtain multi-hot vector encodings of line loss anomaly causes in multiple substations. The line loss anomaly cause feature vectors of multiple line loss anomaly substations are then used as multi-dimensional line loss anomaly cause feature vectors. The text-described anomaly cause labels are converted into numerical vectors, breaking the barriers between text data and numerical models, realizing cross-modal fusion of text and numerical data, and turning text information that was originally difficult to be directly processed by the model into computable and learnable features, providing richer data dimensions for subsequent line loss analysis and improving data quality from the perspective of data type expansion.

[0060] 203. For the i-th abnormal line loss substation, the target power supply and target power loss of each normal line loss month in the i-th abnormal line loss substation in the revised data-type historical power data set are used to calculate the average line loss rate of the i-th abnormal line loss substation, and the specified power supply and specified power loss of each abnormal line loss month in the i-th abnormal line loss substation in the revised data-type historical power data set are used to calculate the abnormal line loss rate of each abnormal line loss month in the i-th abnormal line loss substation.

[0061] In the embodiment of the present application, the target power supply and target power loss of each normal line loss month for the i-th abnormal line loss substation are extracted from the corrected data-type historical power data set. The target power supply and target power loss of each normal line loss month for the i-th abnormal line loss substation are used to calculate the average line loss rate of the i-th abnormal line loss substation. The calculation formula is as shown in the following formula 4:

[0062] Formula 4:

[0063] in, represents the average line loss rate of the ith abnormal line loss area, W loss,i,t ′ It represents the target power loss in the tth month with normal line loss in the ith abnormal line loss area, W supply,i,t ′ N represents the target power supply in the tth month with normal line loss in the ith abnormal line loss area. normal,i represents the number of months with normal line loss in the i-th substation with abnormal line loss. By calculating the average line loss rate, a normal line loss benchmark is established for each substation. This benchmark, based on the substation's own historical data, avoids the problem of large differences in line loss rates across substations, which can lead to misjudgment based on a unified standard. This allows for more accurate identification of line loss anomalies, indirectly verifies the effectiveness of data correction, and ensures the quality and accuracy of line loss analysis data.

[0064] Then, the specified power supply and specified power loss of each abnormal line loss month in the i-th abnormal line loss area are extracted from the corrected data-type historical power data set. The abnormal line loss rate of each abnormal line loss month in the i-th abnormal line loss area is calculated using the specified power supply and specified power loss of each abnormal line loss month in the i-th abnormal line loss area. The calculation formula is as follows:

[0065] Formula 5:

[0066] Among them, μ unnormal,i,m W represents the abnormal line loss rate of the mth abnormal line loss month in the i-th abnormal line loss area, unnormal,supply,i,m Indicates the designated power supply in the mth line loss abnormal month in the i-th line loss abnormal area, W unnormal,loss,i,m This represents the specified amount of power lost in the mth month of abnormal line loss in the i-th abnormal line loss area. The abnormal line loss rate directly quantifies the severity of the abnormal line loss. Combined with the text-based abnormality cause label, it can further analyze which combinations of causes lead to a significant increase in the line loss rate and explore the synergistic effects of multiple causes.

[0067] 204. The total power loss due to the superposition of multiple abnormal causes in the i-th line loss abnormal area is calculated using the designated power supply of each line loss abnormal month in the i-th line loss abnormal area, the average line loss rate of the i-th line loss abnormal area, and the abnormal line loss rate of each line loss abnormal month in the i-th line loss abnormal area.

[0068] In the embodiment of the present application, the total power loss in the i-th abnormal line loss area due to the superposition of multiple abnormal causes is calculated using the designated power supply in each abnormal line loss month of the i-th abnormal line loss area, the average line loss rate of the i-th abnormal line loss area, and the abnormal line loss rate of each abnormal line loss month of the i-th abnormal line loss area. The calculation formula is as shown in the following formula 6:

[0069] Formula 6:

[0070]

[0071] Among them, W cause,loss,i,m It represents the total power loss caused by the superposition of multiple abnormal causes in the mth abnormal month of line loss in the i-th abnormal area, W cause,loss,sum,i It represents the total power loss caused by the superposition of multiple abnormal causes in the i-th abnormal line loss area, μ unnormal,i,m W represents the abnormal line loss rate of the mth abnormal line loss month in the i-th abnormal line loss area, unnormal,supply,i,m It represents the designated power supply in the mth month of abnormal line loss in the i-th abnormal line loss area. represents the average line loss rate of the ith abnormal line loss area, N unnormal,iRepresents the number of months with abnormal line losses in the i-th abnormal line loss area. This breaks away from the traditional approach of single-cause or simply cumulative loss calculations. By comparing the line loss rates between abnormal and normal months and combining power supply with the difference in line loss rates, this method more accurately depicts the impact of multiple causes on line losses, clearly presenting the losses resulting from the synergy of multiple causes and providing more realistic quantitative data for line loss analysis and management.

[0072] 205. The total power supply of the line loss abnormality month of the i-th line loss abnormality substation is calculated using the specified power supply of each line loss abnormality month of the i-th line loss abnormality substation.

[0073] In the embodiment of the present application, the total power supply of the line loss abnormal month in the i-th line loss abnormal area is calculated using the specified power supply of each line loss abnormal month in the i-th line loss abnormal area. The calculation formula is as follows:

[0074] Formula 7:

[0075] Among them, W unnormal,supply,i represents the total power supply in the month with abnormal line loss in the ith abnormal line loss area, W unnormal,supply,i,m N represents the designated power supply in the mth month of abnormal line loss in the ith abnormal line loss area. unnormal,i It represents the number of months with abnormal line loss in the i-th abnormal line loss area.

[0076] 206. Determine input data and output data using the total power loss due to the superposition of multiple abnormal causes in each abnormal line loss substation, the total power supply in each abnormal line loss substation per month, and the multi-dimensional abnormal line loss cause feature vector.

[0077] In the embodiment of the present application, the total power loss due to the superposition of multiple abnormal causes in each abnormal line loss area, the total power supply of each abnormal line loss area in a month, and the multi-dimensional abnormal line loss cause feature vector are used as input data, and the calculation formula is as shown in the following formula 8:

[0078] Formula 8: x i =[W unnormal,supply,i ,W ceusa,loss,sum,i ,m i ],

[0079]

[0080] in, represents the input sample of the ith line loss abnormal area in the input data, X represents the input data, W cause,loss,sum,i It represents the total power loss due to the superposition of multiple abnormal causes in the i-th line loss abnormal area, W unnormal,supply,i represents the total power supply in the month with abnormal line loss in the ith abnormal line loss area, m irepresents the characteristic vector of the line loss anomaly cause in the ith line loss anomaly area, N sample Indicates the number of substations experiencing abnormal line losses. By integrating total power supply, total power loss, and causal feature vectors into fixed-format input samples, the neural network model no longer needs to deal with complex data formats and inconsistent dimensions, focusing directly on feature learning and accelerating model convergence. This multi-source data fusion allows the model to simultaneously learn the correlation between numerical power features and text-based causal features, improving its ability to learn the synergistic effects of multiple causes.

[0081] Then, the output data is determined based on the total power loss under the superposition of multiple abnormal causes in each abnormal line loss area and the multi-dimensional line loss abnormal cause characteristic vector. The calculation formula is as follows:

[0082] Formula 9: W cause,i =[W cause,i,1 ,W cause,i,2 ,...,W cause,i,k ,...,W cause,i,K ],

[0083]

[0084] in, represents the output sample of the ith line loss abnormal area in the output data, W cause Indicates output data, W cause,i,k It represents the power loss corresponding to the kth line loss abnormality cause label in the i-th line loss abnormality area, that is, m i,k =0, W cause,i,k Also 0, K represents the number of line loss anomaly cause label types, N sample Indicates the number of areas with abnormal line loss.

[0085] 207. Build a neural network model.

[0086] In an embodiment of the present application, a neural network model is constructed, wherein the fully connected neural network model includes an input layer, a hidden layer, and an output layer. The dimensions of the input layer and the output layer are fixed to handle a variable number of causes. The hidden layer includes multiple fully connected layers (Dense layers), each of which is followed by a ReLU activation function (Rectified Linear Unit) to introduce nonlinearity. The number of hidden layers and the number of neurons are hyperparameters, and the calculation formula is as shown in the following formula 10:

[0087] Formula 10: z (1) =ReLU(W (1) X+b (1) ),

[0088] z (l+1) =ReLU(W (l+1) z(l) +b (l+1) ),

[0089] W cause =W (out) z (out) +b (out) ,

[0090] Among them, z (1) represents the output of the first fully connected layer in the hidden layer, z (l+1) represents the output of the l+1th fully connected layer in the hidden layer, z (l) represents the output of the lth fully connected layer in the hidden layer, ReLU() represents the activation function, ReLU(a)=max(0,a), W (1) Represents the weight matrix of the first fully connected layer in the hidden layer, W (l+1) represents the weight matrix of the l+1th fully connected layer in the hidden layer, X represents the input data, b (1) represents the bias of the first fully connected layer in the hidden layer, b (l+1) represents the bias of the l+1th fully connected layer in the hidden layer, W cause Indicates output data, W (out) represents the weight matrix of the output layer, z (out) represents the output of the hidden layer, b (out) Represents the bias of the output layer. Fully connected layers fully convey the characteristic information of the input data. The ReLU activation function introduces nonlinearity, allowing the model to learn the complex nonlinear relationships between power supply, power loss, and cause labels. Furthermore, by stacking multiple fully connected layers, the model can gradually extract hierarchical features, such as the relationship between a single cause and loss, to deeper features, such as the relationship between multiple causes and loss, strengthening its ability to learn the combined effects of multiple causes. The number of hidden layers and neurons, as hyperparameters, can be flexibly adjusted based on the actual data scale and complexity, allowing the model to find the optimal structure in both simple and complex substation scenarios, improving model generalization.

[0091] 208. Construct a multi-objective loss function.

[0092] In the embodiment of the present application, a multi-objective loss function is constructed, and the calculation formula is as follows:

[0093] Formula 11:

[0094]

[0095]

[0096] Among them, L total Represents the multi-objective loss function, L MSErepresents the mean square error loss function caused by anomalies, L sum represents the sum constraint loss function, N sample represents the number of abnormal line loss areas, λ represents a hyperparameter, K represents the number of label types of abnormal line loss causes, W represents the predicted power loss corresponding to the kth line loss abnormality cause label in the i-th line loss abnormality area, cause,i,k Indicates the power loss corresponding to the kth line loss abnormality cause label in the i-th line loss abnormality area, W unnormal,supply,i,k represents the monthly power supply corresponding to the kth line loss cause label in the i-th line loss anomaly substation. The anomaly cause mean squared error loss function directly measures the error between the model's predicted power loss and the true value for each cause label, driving the model to accurately learn the cause loss prediction and ensuring high accuracy for individual cause loss predictions. The sum constraint loss function introduces physical constraints on monthly power input to prevent the prediction results from deviating from the reasonable range of actual power supply, making the prediction more consistent with the physical laws of grid operation and improving the practicality and credibility of the solution. The hyperparameter λ is then used to balance the weights of the two, allowing the model to find the optimal balance between accurate prediction and satisfying physical constraints, avoiding overfitting and enhancing the model's robustness and generalization capabilities in complex line loss scenarios. This builds a solid training framework for accurately deconstructing line loss causes and supporting intelligent governance decisions.

[0097] 209. The neural network model is trained using input data, output data, and multi-objective loss function to obtain a decomposition model of the substation line loss electricity.

[0098] In an embodiment of the present application, input data, output data, and a multi-objective loss function are used to train a neural network model to obtain a substation line loss power deconstruction model to ensure the accuracy of line loss cause deconstruction in basic scenarios.

[0099] When a new line loss anomaly cause label type is detected based on the dynamic coding mechanism, the new line loss anomaly cause label type is used to expand the dimension of the multidimensional line loss anomaly cause feature vector. Then, the dimension of the expanded multidimensional line loss anomaly cause feature vector is used to adjust the model structure of the neural network model to obtain an updated neural network model. Subsequently, training data containing the new line loss anomaly cause label type is obtained, and the updated neural network model is trained using the training data to obtain updated model parameters. Finally, the updated model parameters are used to adjust the substation line loss power deconstruction model. There is no need to retrain the entire model, only the parts associated with the new causes are updated, and new scenarios can be quickly adapted. Whether it is a new line loss cause caused by power grid equipment upgrades, changes in operation and maintenance modes, or environmental factors, the model can be quickly adapted through dynamic updates, providing technical support for the full life cycle line loss management of smart grids.

[0100] 210. Collect the power data of line loss in the substation area, input the power data of line loss in the substation area into the power deconstruction model of line loss in the substation area for analysis, and output the power loss caused by each abnormal reason of line loss.

[0101] In the embodiment of the present application, for a certain substation that needs to analyze the line loss power deconstruction object, the substation line loss power data is collected, and the substation line loss power data x=[W unnormal,supply ,W cause,loss,sum ,m] Input the area line loss power deconstruction model for analysis and output the power loss W for each abnormal line loss reason cause =[W cause,1 ,W cause,2 ,...,W cause,i ,...,W cause,K ], to achieve accurate quantitative tracing of the causes of line loss.

[0102] Based on the above process, the schematic diagram of the architecture of a method for deconstructing line loss in a transformer area based on electric energy analysis and adaptive learning of uncertainty of abnormal causes proposed in an embodiment of the present application is as follows:

[0103] like Figure 2B As shown, first enter step 1 to collect data. It is necessary to obtain the operating data of the abnormal line loss area and the normal area data for comparative analysis. At the same time, collect text-type abnormal cause labels to reserve basic materials for subsequent analysis; then enter step 2 data cleaning. On the one hand, perform routine processing on the collected data. On the other hand, use statistical or algorithmic means to accurately identify and correct irregular data to ensure data quality and lay a solid foundation for subsequent calculations and modeling; then enter step 3 loss power calculation. First, based on the cleaned data, calculate the average line loss rate of the abnormal line loss area during the normal operating period and the line loss rate in the abnormal month, and then Through comparative analysis, the total power loss generated in the substation due to the superposition of multiple abnormal causes is calculated, and the synergistic impact of multiple factors is quantified. Then, step 4 is entered to build an adaptive learning model. First, the text-based abnormal cause label is encoded, and an input containing abnormal cause vector, power loss and other information is constructed. Relying on the fully connected neural network architecture, the model learns the complex relationship between abnormal causes and power loss. After model training and multi-objective loss function optimization, a model that can accurately deconstruct the causes of line losses is finally generated, and the power loss results corresponding to each abnormal cause are output, completing the entire line loss analysis process and achieving a closed loop from data collection to accurate tracing of the cause.

[0104] The embodiment of the present application provides a fusion adaptive deconstruction method for line loss electricity in a substation. Compared with the prior art, the embodiment of the present application obtains a data-type historical electricity data set and a text-type abnormal cause label set, performs abnormal data correction on the data-type historical electricity data set, and quantifies the text-type abnormal cause label set into a multi-dimensional line loss abnormal cause feature vector. By setting an outlier detection rule, the data-type historical electricity data set is corrected for abnormal data to ensure data quality; the text-type abnormal cause label set is quantitatively converted, the abnormal cause described in text is converted into a multi-dimensional line loss abnormal cause feature vector, and the text information is converted into a numerical form that can be calculated by the model, thereby realizing cross-modal data fusion. The corrected data-type historical electricity data set is then used to calculate the total loss electricity under the superposition of multiple abnormal causes for each line loss abnormal substation, as well as the total monthly power supply of line loss abnormalities for each line loss abnormal substation. By incorporating the comprehensive impact of multiple abnormal causes into the loss electricity calculation, the limitations of traditional methods on single factors or independent analysis of factors are broken, and the synergistic effect of multiple factors is accurately reflected, providing a scientific and quantitative basis for the definition and management of line loss responsibilities. The total power loss due to the superposition of multiple abnormal causes for each abnormal line loss substation, the total monthly power supply for each abnormal line loss substation, and the multi-dimensional line loss cause feature vector are then input into the neural network model. The model is trained using a multi-objective loss function and a dynamic encoding mechanism to construct a substation line loss power deconstruction model. This model automatically captures the complex coupling relationship between causes and line losses. It also updates parameters based on real-time data to adapt to dynamic scenarios, improving the accuracy and generalization of identifying abnormal line loss causes and providing self-evolutionary technical support for smart grid line loss management. In practical applications, the collected substation line loss power data can be input into the substation line loss power deconstruction model for analysis, outputting the power loss associated with each abnormal line loss cause, enabling accurate tracing of the causes of line losses and providing a quantitative basis for line loss management.

[0105] Further, as Figure 1 The specific implementation of the method, the embodiment of the present application provides a fusion adaptive deconstruction device for the line loss electricity in the substation area, such as Figure 3 As shown, the device includes: a data processing module 301, a data calculation module 302, an adaptive learning module 303 and a deconstruction analysis module 304.

[0106] The data processing module 301 is configured to obtain a data-based historical electricity data set and a text-based abnormality cause label set, perform abnormal data correction on the data-based historical electricity data set, and quantify the text-based abnormality cause label set into a multi-dimensional line loss abnormality cause feature vector;

[0107] The data calculation module 302 is configured to calculate the total power loss caused by the superposition of multiple abnormal causes for each abnormal line loss area using the modified data-type historical power data set, and the total power supply for each abnormal line loss area in the month of abnormal line loss;

[0108] The adaptive learning module 303 is configured to input the total power loss due to the superposition of multiple abnormal causes for each abnormal line loss substation, the total monthly power supply of each abnormal line loss substation, and the multi-dimensional abnormal line loss cause feature vector into a neural network model, and perform model training through a multi-objective loss function and a dynamic coding mechanism to construct a substation line loss power deconstruction model;

[0109] The deconstruction analysis module 304 is used to collect the substation line loss power data, input the substation line loss power data into the substation line loss power deconstruction model for analysis, and output the power loss caused by each line loss abnormality.

[0110] In a specific application scenario, the data processing module 301 is used to determine multiple line loss abnormality substations in the data-type historical electricity data set; for the i-th line loss abnormality substation, extract the power supply and power loss of each line loss normal month of the i-th line loss abnormality substation in the data-type historical electricity data set; obtain an irregular power supply data judgment inequality, and based on the power supply of each line loss normal month of the i-th line loss abnormality substation, take the line loss normal month that meets the irregular power supply data judgment inequality among the multiple line loss normal months of the i-th line loss abnormality substation as the target line loss normal month, and obtain at least one target line loss normal month.

[0111]

[0112] Among them, W supply,i,t represents the power supply in the tth month with normal line loss in the i-th abnormal line loss area, represents the average monthly power supply of the line loss in the i-th abnormal line loss area, δ supply,i N represents the standard deviation of the monthly power supply with normal line loss in the i-th abnormal line loss area, normal,i represents the number of normal line loss months in the i-th abnormal line loss substation; replaces the power supply of each target normal line loss month with the average power supply of the i-th abnormal line loss substation in the normal line loss month to obtain the corrected power supply of each target normal line loss month; obtains an irregular power loss data judgment inequality, and based on the power loss of each normal line loss month in the i-th abnormal line loss substation, takes the normal line loss month that meets the irregular power loss data judgment inequality among multiple normal line loss months in the i-th abnormal line loss substation as a designated normal line loss month, to obtain at least one designated normal line loss month.

[0113]

[0114] Among them, W loss,i,t It represents the power loss in the tth month with normal line loss in the i-th abnormal line loss area, represents the average monthly power loss of the line loss in the i-th abnormal line loss area, δ loss,i It represents the standard deviation of monthly power loss in the normal line loss area of ​​the i-th abnormal line loss area, N normal,i Represents the number of normal line loss months in the i-th line loss abnormal substation; replaces the lost electricity in each of the specified normal line loss months with the average value of the lost electricity in the normal line loss months in the line loss abnormal substation to obtain the corrected lost electricity in each of the specified normal line loss months; uses the corrected power supply of at least one target normal line loss month and the corrected loss electricity of at least one specified normal line loss month to correct the power supply and loss electricity of multiple normal line loss months in the i-th line loss abnormal substation in the data-type historical electricity data set.

[0115] In a specific application scenario, the data processing module 301 is used to obtain a plurality of line loss abnormality cause label types from the line loss abnormality area abnormality cause label set; for the i-th line loss abnormality area, extract at least one line loss abnormality cause label corresponding to the i-th line loss abnormality area from the line loss abnormality area abnormality cause label set, and use the plurality of line loss abnormality cause label types to perform vector encoding on the at least one line loss abnormality cause label to obtain a line loss abnormality cause feature vector for the i-th line loss abnormality area.

[0116] m i =[m i,1 ,m i,2 ,...,m i,k ,...,m i,K ],

[0117] Among them, m i represents the line loss anomaly cause characteristic vector of the i-th line loss anomaly area, m i,k Indicates the existence of the kth line loss abnormality cause label in the i-th line loss abnormal area, m i,k =1 indicates that the i-th abnormal line loss station area has the k-th abnormal line loss cause label, m i,k =0 indicates that the kth line loss abnormality cause label does not exist in the i-th line loss abnormality station, and K indicates the number of line loss abnormality cause label types; the line loss abnormality cause feature vectors of multiple line loss abnormality stations are used as the multi-dimensional line loss abnormality cause feature vector.

[0118] In a specific application scenario, the data calculation module 302 is used to calculate the average line loss rate of the i-th abnormal line loss substation using the target power supply and target power loss of each normal line loss month in the i-th abnormal line loss substation in the revised data-type historical power data set, and calculate the abnormal line loss rate of each abnormal line loss month in the i-th abnormal line loss substation using the specified power supply and specified power loss of each abnormal line loss month in the i-th abnormal line loss substation in the revised data-type historical power data set; calculate the total power loss of the i-th abnormal line loss substation under the superposition of multiple abnormal causes using the specified power supply of each abnormal line loss month in the i-th abnormal line loss substation, the average line loss rate of the i-th abnormal line loss substation, and the abnormal line loss rate of each abnormal line loss month in the i-th abnormal line loss substation.

[0119]

[0120] Among them, W cause,loss,i,m It represents the total power loss caused by the superposition of multiple abnormal causes of line loss in the mth line loss abnormal month in the i-th line loss abnormal area, W cause,loss,sum,i It represents the total power loss caused by the superposition of multiple abnormal causes in the i-th abnormal line loss area, μ unnormal,i,m W represents the abnormal line loss rate of the mth abnormal line loss month in the i-th abnormal line loss area, unnormal,supply,i,m represents the designated power supply in the mth line loss abnormal month of the i-th line loss abnormal area, represents the average line loss rate of the i-th abnormal line loss area, N unnormal,i represents the number of abnormal line loss months in the i-th abnormal line loss area; the total power supply of the i-th abnormal line loss area in the abnormal line loss month is calculated using the specified power supply of each abnormal line loss month in the i-th abnormal line loss area,

[0121]

[0122] Among them, W unnormal,supply,i represents the total power supply in the month with abnormal line loss in the i-th abnormal line loss area, W unnormal,supply,i,m N represents the designated power supply in the mth line loss abnormal month in the i-th line loss abnormal area. unnormal,i It represents the number of months with abnormal line loss in the i-th abnormal line loss area.

[0123] In a specific application scenario, the data calculation module 302 is used to extract the target power supply and target power loss of the i-th abnormal line loss substation in each normal line loss month from the corrected data-type historical power data set, and calculate the average line loss rate of the i-th abnormal line loss substation using the target power supply and target power loss of the i-th abnormal line loss substation in each normal line loss month.

[0124]

[0125] in, represents the average line loss rate of the i-th abnormal line loss area, W loss,i,t ′ It represents the target power loss in the tth month with normal line loss in the i-th abnormal line loss area, W supply,i,t ′ N represents the target power supply in the tth month with normal line loss in the i-th abnormal line loss area. normal,i represents the number of normal line loss months in the i-th abnormal line loss area; extracting the specified power supply and specified power loss of each abnormal line loss month in the i-th abnormal line loss area from the corrected data-type historical power data set, and calculating the abnormal line loss rate of each abnormal line loss month in the i-th abnormal line loss area using the specified power supply and specified power loss of each abnormal line loss month in the i-th abnormal line loss area,

[0126]

[0127] Among them, μ unnormal,i,m W represents the abnormal line loss rate of the mth abnormal line loss month in the i-th abnormal line loss area, unnormal,supply,i,m It represents the designated power supply in the mth line loss abnormal month of the i-th line loss abnormal area, W unnormal,loss,i,m It represents the designated power loss in the mth line loss abnormal month in the i-th line loss abnormal area.

[0128] In a specific application scenario, the adaptive learning module 303 is used to take the total power loss due to the superposition of multiple abnormal causes of each abnormal line loss area, the total power supply of the abnormal line loss month of each abnormal line loss area, and the multi-dimensional line loss abnormal cause feature vector as input data.

[0129] x i =[W unnormal,supply,i ,W cause,loss,sum,i ,m i ],

[0130]

[0131] Among them, x i represents the input sample of the ith line loss abnormal area in the input data, X represents the input data, and W cause,loss,sum,i It represents the total power loss caused by the superposition of multiple abnormal causes in the i-th abnormal line loss area, W unnormal,supply,i represents the total power supply in the month with abnormal line loss in the i-th abnormal line loss area, m i represents the line loss anomaly cause characteristic vector of the i-th line loss anomaly area, N sampleIndicates the number of abnormal line loss areas; output data is determined based on the total power loss under the superposition of multiple abnormal causes of each abnormal line loss area and the multi-dimensional line loss abnormal cause feature vector,

[0132] W cause,i =[W cause,i,1 ,W cause,i,2 ,...,W cause,i,k ,...,W cause,i,K ],

[0133]

[0134] Among them, W cause,i represents the output sample of the ith line loss abnormal area in the output data, W cause Represents the output data, W cause,i,k represents the power loss corresponding to the kth line loss abnormality cause label in the i-th line loss abnormality area, K represents the number of line loss abnormality cause label types, N sample Indicates the number of abnormal line loss areas; constructs the neural network model, which includes an input layer, a hidden layer and an output layer, and the hidden layer includes multiple fully connected layers,

[0135] z (1) =ReLU(W (1) X+b (1) ),

[0136] z (l+1) =ReLU(W (l+1) z (l) +b (l+1) ),

[0137] W cause =W (out) z (out) +b (out) ,

[0138] Among them, z (1) represents the output of the first fully connected layer in the hidden layer, z (k+1) represents the output of the l+1th fully connected layer in the hidden layer, z (l) Represents the output of the lth fully connected layer in the hidden layer, ReLU() represents the activation function, W (1) Represents the weight matrix of the first fully connected layer in the hidden layer, W (l+1) represents the weight matrix of the l+1th fully connected layer in the hidden layer, X represents the input data, b (1) represents the bias of the first fully connected layer in the hidden layer, b (l+1) represents the bias of the l+1th fully connected layer in the hidden layer, W causeRepresents the output data, W (out) represents the weight matrix of the output layer, z (out) represents the output of the hidden layer, b (out) Represents the bias of the output layer; constructs the multi-objective loss function,

[0139]

[0140]

[0141] Among them, L total Represents the multi-objective loss function, L MSE represents the mean square error loss function caused by anomalies, L sum represents the sum constraint loss function, N sample represents the number of abnormal line loss areas, λ represents a hyperparameter, K represents the number of label types of abnormal line loss causes, W represents the predicted power loss corresponding to the kth line loss abnormality cause label in the i-th line loss abnormality area, cause,i,k W represents the power loss corresponding to the kth line loss abnormality cause label in the i-th line loss abnormal area, unnormal,supply,i,k Represents the monthly power supply corresponding to the kth line loss abnormality cause label of the i-th line loss abnormality substation; uses the input data, the output data, and the multi-objective loss function to train the neural network model to obtain the line loss power deconstruction model of the substation.

[0142] In a specific application scenario, the adaptive learning module 303 is used to, when a new line loss abnormality cause label type is detected based on the dynamic coding mechanism, use the new line loss abnormality cause label type to expand the dimension of the multi-dimensional line loss abnormality cause feature vector; use the dimension of the expanded multi-dimensional line loss abnormality cause feature vector to adjust the model structure of the neural network model to obtain an updated neural network model; obtain training data containing the new line loss abnormality cause label type, use the training data to train the updated neural network model to obtain updated model parameters; use the updated model parameters to adjust the substation line loss power deconstruction model.

[0143] The embodiment of the present application provides a device. Compared with the prior art, the embodiment of the present application obtains a data-type historical electricity data set and a text-type abnormal cause label set, performs abnormal data correction on the data-type historical electricity data set, and quantifies the text-type abnormal cause label set into a multi-dimensional line loss abnormal cause feature vector. The abnormal value detection rule is set to perform abnormal data correction on the data-type historical electricity data set to ensure data quality; the text-type abnormal cause label set is quantitatively converted to convert the abnormal cause described in text into a multi-dimensional line loss abnormal cause feature vector, and the text information is converted into a numerical form that can be calculated by the model to achieve cross-modal data fusion. The corrected data-type historical electricity data set is then used to calculate the total power loss under the superposition of multiple abnormal causes for each line loss abnormal area, as well as the total monthly power supply of line loss abnormalities for each line loss abnormal area. By incorporating the comprehensive impact of multiple abnormal causes into the loss power calculation, the limitations of traditional methods on single factors or independent analysis of factors are broken, and the synergistic effect of multiple factors is accurately reflected, providing a scientific and quantitative basis for the definition and management of line loss responsibilities. The total power loss due to the superposition of multiple abnormal causes for each abnormal line loss substation, the total monthly power supply for each abnormal line loss substation, and the multi-dimensional line loss cause feature vector are then input into the neural network model. The model is trained using a multi-objective loss function and a dynamic encoding mechanism to construct a substation line loss power deconstruction model. This model automatically captures the complex coupling relationship between causes and line losses. It also updates parameters based on real-time data to adapt to dynamic scenarios, improving the accuracy and generalization of identifying abnormal line loss causes and providing self-evolutionary technical support for smart grid line loss management. In practical applications, the collected substation line loss power data can be input into the substation line loss power deconstruction model for analysis, outputting the power loss associated with each abnormal line loss cause, enabling accurate tracing of the causes of line losses and providing a quantitative basis for line loss management.

[0144] It should be noted that for other corresponding descriptions of the functional units involved in the fusion adaptive deconstruction device for the line loss power of the substation provided in the embodiment of the present application, please refer to Figure 1 and Figures 2A to 2B The corresponding description in will not be repeated here.

[0145] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0146] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0147] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

[0148] In an exemplary embodiment, see Figure 4 A device is also provided, comprising a bus, a processor, a memory, and a communication interface. The device may also include an input / output interface and a display device, wherein the various functional units can communicate with each other via the bus. The memory stores a computer program, and the processor is configured to execute the program stored in the memory and implement the method for fusion and adaptive deconstruction of substation line loss electricity in the above-mentioned embodiment.

[0149] A medium stores a computer program, which, when executed by a processor, implements the steps of the method for fusion adaptive deconstruction of line loss electricity in a substation area.

[0150] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present application can be implemented through hardware or by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present application.

[0151] Those skilled in the art will understand that the accompanying drawings are only schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily required to implement the present application.

[0152] Those skilled in the art will appreciate that the modules in the devices in the implementation scenario can be distributed in the devices of the implementation scenario according to the implementation scenario description, or can be modified accordingly and located in one or more devices different from the implementation scenario. The modules in the above implementation scenario can be combined into one module or further split into multiple submodules.

[0153] The above application serial numbers are for description only and do not represent the advantages or disadvantages of the implementation scenarios.

[0154] The above disclosure only describes several specific implementation scenarios of the present application. However, the present application is not limited thereto, and any changes that can be conceived by those skilled in the art should fall within the scope of protection of the present application.

Claims

1. A fusion adaptive deconstruction method for line loss in a substation, characterized by: include: Acquire a data-type historical electricity data set and a text-type abnormality cause label set, perform abnormal data correction on the data-type historical electricity data set, and quantify the text-type abnormality cause label set into a multi-dimensional line loss abnormality cause feature vector; The modified data-type historical electricity data set is used to calculate the total power loss caused by the superposition of multiple abnormal causes for each abnormal line loss area, as well as the total power supply for the abnormal line loss month for each abnormal line loss area; The total power loss due to the superposition of multiple abnormal causes of each abnormal line loss area, the total power supply of each abnormal line loss area in a month, and the multi-dimensional abnormal line loss cause feature vector are input into the neural network model, and the model is trained through a multi-objective loss function and a dynamic coding mechanism to construct a line loss power deconstruction model for the area; Collecting the power data of the substation line loss, inputting the power data of the substation line loss into the power deconstruction model of the substation line loss for analysis, and outputting the power loss caused by each abnormal line loss.

2. The method according to claim 1, characterized in that The performing abnormal data correction on the data type historical electricity data set includes: Determining a plurality of abnormal line loss areas in the data-type historical electricity data set; For the i-th abnormal line loss substation, extract the power supply and power loss of each normal line loss month for the i-th abnormal line loss substation from the data-type historical power data set; Obtaining an irregular power supply data judgment inequality, and based on the power supply of each normal line loss month in the i-th abnormal line loss substation, taking the normal line loss month among multiple normal line loss months in the i-th abnormal line loss substation that satisfies the irregular power supply data judgment inequality as a target normal line loss month, to obtain at least one target normal line loss month. Among them, W supply,i,t represents the power supply in the tth month with normal line loss in the i-th abnormal line loss area, represents the average monthly power supply of the line loss in the i-th abnormal line loss area, δ supplu,i N represents the standard deviation of the monthly power supply with normal line loss in the i-th abnormal line loss area, normal,i represents the number of months with normal line loss in the i-th abnormal line loss area; Replacing the power supply of each target normal line loss month with the average power supply of the normal line loss month of the i-th abnormal line loss substation to obtain the corrected power supply of each target normal line loss month; Obtaining an inequality for determining irregular power loss data; and, based on the power loss of each normal line loss month in the i-th abnormal line loss substation, taking a normal line loss month among multiple normal line loss months in the i-th abnormal line loss substation that satisfies the inequality for determining irregular power loss data as a designated normal line loss month, thereby obtaining at least one designated normal line loss month. Among them, W loss,i,t It represents the power loss in the tth month with normal line loss in the i-th abnormal line loss area, represents the average monthly power loss of the line loss in the i-th abnormal line loss area, δ loss,i It represents the standard deviation of monthly power loss in the normal line loss area of ​​the i-th abnormal line loss area, N normal,i represents the number of months with normal line loss in the i-th abnormal line loss area; Replacing the power loss in each of the specified normal line loss months with the average power loss in the normal line loss months in the abnormal line loss substation to obtain the corrected power loss in each of the specified normal line loss months; The corrected power supply of at least one target normal line loss month and the corrected power loss of at least one specified normal line loss month are used to correct the power supply and power loss of multiple normal line loss months of the i-th line loss abnormal substation in the data-type historical power data set.

3. The method according to claim 1, characterized in that The step of quantifying the abnormal cause label set of the abnormal line loss area into a multi-dimensional abnormal line loss cause feature vector includes: A plurality of line loss abnormality cause label types are obtained by counting the abnormality cause label set of the line loss abnormality area; For the i-th abnormal line loss substation, at least one abnormal line loss cause label corresponding to the i-th abnormal line loss substation is extracted from the abnormal line loss cause label set, and the at least one abnormal line loss cause label is vector-encoded using the multiple abnormal line loss cause label types to obtain a abnormal line loss cause feature vector for the i-th abnormal line loss substation. m i =[m i,1 ,m i,2 ,...,m i,k ,...,m i,K ], Among them, m i represents the line loss anomaly cause characteristic vector of the i-th line loss anomaly area, m i,k Indicates the existence of the kth line loss abnormality cause label in the i-th line loss abnormal area, m i,k =1 indicates that the i-th abnormal line loss station area has the k-th abnormal line loss cause label, m i,k =0 indicates that the kth line loss abnormality cause label does not exist in the i-th line loss abnormality station area, and K indicates the number of line loss abnormality cause label types; The line loss anomaly cause feature vectors of multiple line loss anomaly areas are used as the multi-dimensional line loss anomaly cause feature vector.

4. The method according to claim 1, wherein The modified data-type historical electricity data set is used to calculate the total power loss caused by the superposition of multiple abnormal causes in each abnormal line loss area, and the total power supply in the abnormal line loss month for each abnormal line loss area, including: For the i-th abnormal line loss substation, the target power supply and target power loss of each normal line loss month for the i-th abnormal line loss substation in the revised data-type historical power data set are used to calculate the average line loss rate of the i-th abnormal line loss substation, and the specified power supply and specified power loss of each abnormal line loss month for the i-th abnormal line loss substation in the revised data-type historical power data set are used to calculate the abnormal line loss rate of each abnormal line loss month for the i-th abnormal line loss substation; The total power loss of the i-th abnormal line loss substation under the superposition of multiple abnormal causes is calculated using the designated power supply of each abnormal line loss month in the i-th abnormal line loss substation, the average line loss rate of the i-th abnormal line loss substation, and the abnormal line loss rate of each abnormal line loss month in the i-th abnormal line loss substation. Among them, W cause,loss,i,m It represents the total power loss caused by the superposition of multiple abnormal causes of line loss in the mth line loss abnormal month in the i-th line loss abnormal area, W cause,loss,sum,i It represents the total power loss caused by the superposition of multiple abnormal causes in the i-th abnormal line loss area, μ unnormal,i,m W represents the abnormal line loss rate of the mth abnormal line loss month in the i-th abnormal line loss area, unnormal,supply,i,m represents the designated power supply in the mth line loss abnormal month of the i-th line loss abnormal area, represents the average line loss rate of the i-th abnormal line loss area, N unnormal,i represents the number of months with abnormal line loss in the i-th abnormal line loss area; The total power supply of the line loss abnormal month of the i-th line loss abnormal area is calculated by using the designated power supply of each line loss abnormal month of the i-th line loss abnormal area. Among them, W unnormal,supply,i represents the total power supply in the month with abnormal line loss in the i-th abnormal line loss area, W unnormal,supply,i,m N represents the designated power supply in the mth line loss abnormal month in the i-th line loss abnormal area. unnormal,i It represents the number of months with abnormal line loss in the i-th abnormal line loss area.

5. The method according to claim 4, characterized in that The method of calculating the average line loss rate of the i-th abnormal line loss area by using the target power supply and target power loss of each normal line loss month for the i-th abnormal line loss area in the revised data-type historical power data set, and calculating the abnormal line loss rate of each abnormal line loss month for the i-th abnormal line loss area by using the specified power supply and specified power loss of each abnormal line loss month for the i-th abnormal line loss area in the revised data-type historical power data set includes: Extract the target power supply and target power loss of each month with normal line loss in the i-th abnormal line loss area from the modified data-type historical power data set, and calculate the average line loss rate of the i-th abnormal line loss area using the target power supply and target power loss of each month with normal line loss in the i-th abnormal line loss area. in, represents the average line loss rate of the i-th abnormal line loss area, W loss,i,t ′ It represents the target power loss in the tth month with normal line loss in the i-th abnormal line loss area, W supply,i,t ′ N represents the target power supply in the tth month with normal line loss in the i-th abnormal line loss area. normal,i represents the number of months with normal line loss in the i-th abnormal line loss area; Extract the designated power supply and designated power loss of each abnormal line loss month in the i-th abnormal line loss area from the modified data-type historical power data set, and calculate the abnormal line loss rate of each abnormal line loss month in the i-th abnormal line loss area using the designated power supply and designated power loss of each abnormal line loss month in the i-th abnormal line loss area. Among them, μ unnormal,i,m W represents the abnormal line loss rate of the mth abnormal line loss month in the i-th abnormal line loss area, unnormal,supply,i,m It represents the designated power supply in the mth line loss abnormal month of the i-th line loss abnormal area, W unnormaa,loss,i,m It represents the designated power loss in the mth line loss abnormal month in the i-th line loss abnormal area.

6. The method according to claim 1, wherein The total power loss due to the superposition of multiple abnormal causes of each abnormal line loss area, the total monthly power supply of each abnormal line loss area, and the multi-dimensional abnormal line loss cause feature vector are input into the neural network model, and the model is trained through a multi-objective loss function and a dynamic coding mechanism to construct a line loss power deconstruction model of the area, including: The total power loss due to the superposition of multiple abnormal causes of each abnormal line loss area, the total power supply of each abnormal line loss area in a month and the multi-dimensional abnormal line loss cause feature vector are used as input data. x i =[W unnormal,supply,i ,W cause,loss,sum,i ,m i ], Among them, x i represents the input sample of the ith line loss abnormal area in the input data, X represents the input data, and W cause,loss,sum,i It represents the total power loss caused by the superposition of multiple abnormal causes in the i-th abnormal line loss area, W unnormal,supply,i represents the total power supply in the month with abnormal line loss in the i-th abnormal line loss area, m i represents the line loss anomaly cause characteristic vector of the i-th line loss anomaly area, N sample Indicates the number of areas with abnormal line loss; Determine output data based on the total power loss caused by the superposition of multiple abnormal causes in each abnormal line loss area and the multi-dimensional line loss abnormal cause feature vector. IN cause,i =[W cause,i,1 ,IN cause,i,2 ,...,IN cause,i,k ,...,IN cause,i,K ], Among them, W cause,i represents the output sample of the ith line loss abnormal area in the output data, W cause Represents the output data, W cause,i,k represents the power loss corresponding to the kth line loss abnormality cause label in the i-th line loss abnormality area, K represents the number of line loss abnormality cause label types, N sample Indicates the number of areas with abnormal line loss; Constructing the neural network model, the neural network model includes an input layer, a hidden layer and an output layer, the hidden layer includes multiple fully connected layers, With (1) =ReLU(W (1) X+b (1) ), With (l+1) =ReLU(W (l+1) With (l) +b (l+1) ), W cause =W (out) z (out) +b (out) , Among them, z (1) represents the output of the first fully connected layer in the hidden layer, z (l+1) represents the output of the l+1th fully connected layer in the hidden layer, z (l) Represents the output of the lth fully connected layer in the hidden layer, ReLU() represents the activation function, W (1) Represents the weight matrix of the first fully connected layer in the hidden layer, W (l+1) represents the weight matrix of the l+1th fully connected layer in the hidden layer, X represents the input data, b (1) represents the bias of the first fully connected layer in the hidden layer, b (l+1) represents the bias of the l+1th fully connected layer in the hidden layer, W cause Represents the output data, W (out) represents the weight matrix of the output layer, z (out) represents the output of the hidden layer, b (out) represents the bias of the output layer; Construct the multi-objective loss function, Among them, L total Represents the multi-objective loss function, L MSE represents the mean square error loss function caused by anomalies, L sum represents the sum constraint loss function, N sample represents the number of abnormal line loss areas, λ represents a hyperparameter, K represents the number of label types of abnormal line loss causes, W represents the predicted power loss corresponding to the kth line loss abnormality cause label in the i-th line loss abnormality area, cause,i,k W represents the power loss corresponding to the kth line loss abnormality cause label in the i-th line loss abnormal area, unnormal,supply,i,k Indicates the monthly power supply corresponding to the kth line loss abnormality cause label in the i-th line loss abnormality area; The neural network model is trained using the input data, the output data, and the multi-objective loss function to obtain the substation line loss power deconstruction model.

7. The method according to claim 6, characterized in that The method further comprises: When a new line loss anomaly cause label type is detected based on the dynamic coding mechanism, the dimension of the multi-dimensional line loss anomaly cause feature vector is expanded using the new line loss anomaly cause label type; The model structure of the neural network model is adjusted using the dimension of the expanded multi-dimensional line loss anomaly cause feature vector to obtain an updated neural network model; Acquire training data containing the newly added line loss anomaly cause label type, and use the training data to train the updated neural network model to obtain updated model parameters; The updated model parameters are used to adjust the substation line loss power deconstruction model.

8. A fusion adaptive deconstruction device for line loss electricity in a substation area, characterized by: include: a data processing module, configured to obtain a data-type historical electricity data set and a text-type abnormality cause label set, perform abnormal data correction on the data-type historical electricity data set, and quantify the text-type abnormality cause label set into a multi-dimensional line loss abnormality cause feature vector; A data calculation module is used to calculate the total power loss caused by the superposition of multiple abnormal causes for each abnormal line loss area using the modified data-type historical power data set, and the total power supply of each abnormal line loss area in the month of abnormal line loss; An adaptive learning module is configured to input the total power loss due to the superposition of multiple abnormal causes of each abnormal line loss substation, the total monthly power supply of each abnormal line loss substation, and the multi-dimensional abnormal line loss cause feature vector into a neural network model, and perform model training through a multi-objective loss function and a dynamic coding mechanism to construct a substation line loss power deconstruction model; The deconstruction analysis module is used to collect the line loss power data of the substation area, input the line loss power data of the substation area into the line loss power deconstruction model of the substation area for analysis, and output the power loss caused by each abnormal line loss.

9. A device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images