Power distribution network area line loss detection method and system

Abstract

The invention discloses a line loss detection method and system for a power distribution network station area, and relates to the technical field of line loss detection, and the method comprises the steps: collecting the operation data of a station area main power supply transformer and each user branch; performing time synchronization and data cleaning on the acquired operation data to obtain a preliminary line loss result; establishing a line loss feature library; identifying and correcting the initial line loss result by using a machine learning algorithm to generate a line loss detection result; and carrying out transformer area operation state classification, classification result and line loss trend prediction, and issuing grading early warning information. The technical problems that in the prior art, power distribution network area line loss detection lacks full-process coverage, the deviation of the line loss detection result is large, and abnormal conditions are not recognized in time are solved, and the technical effects that full-process management of power distribution network area line loss detection is achieved, and the line loss detection accuracy and the abnormal condition recognition efficiency are improved are achieved.

Description

Technical Field

[0001] This invention relates to the field of line loss detection technology, specifically to a method and system for detecting line loss in distribution network areas. Background Technology

[0002] In distribution network operation and management, transformer substation line loss is a core indicator for measuring power supply efficiency and management level. Its accurate detection and anomaly management directly affect the economic benefits and power supply reliability of the power grid. Currently, traditional transformer substation line loss detection methods mostly rely on manual calculation or single-dimensional data calculation, which has obvious technical limitations: On the one hand, some methods only collect operating data from the main transformer or a few nodes, failing to achieve comprehensive coverage of user branch data, and the data processing stage lacks strict time synchronization and cleaning mechanisms, which can easily lead to distorted line loss calculation results due to data deviations; on the other hand, traditional methods mostly use fixed formulas to estimate theoretical line loss, making it difficult to establish accurate models by combining historical operating patterns, meteorological changes, time-period characteristics, and other dynamic factors. This leads to misjudgments or omissions when identifying line loss anomalies, and the anomaly response is delayed, making it impossible to quickly locate the source of the problem.

[0003] Existing technologies suffer from technical problems such as a lack of full-process coverage in distribution network line loss detection, large deviations in line loss detection results, and untimely identification of abnormal situations. Summary of the Invention

[0004] This application provides a method and system for detecting line loss in distribution network areas, which addresses the technical problems in the prior art of lacking full-process coverage in distribution network area line loss detection, large deviations in line loss detection results, and untimely identification of abnormal situations.

[0005] In view of the above problems, this application provides a method and system for detecting line loss in distribution network areas.

[0006] A first aspect of this application provides a method for detecting line losses in a distribution network area, the method comprising:

[0007] The system collects operational data from the main power supply transformer and each user branch line in the distribution area, including at least voltage, current, and power factor. It performs time synchronization and data cleaning on the collected operational data, calculates the difference between theoretical power supply energy and actual metered energy based on the power balance principle, and obtains preliminary line loss results. It then establishes a line loss feature database by combining historical operational data, meteorological data, and time-period characteristic variables. Based on the line loss feature database, it uses machine learning algorithms to identify and correct the preliminary line loss results, generating line loss detection results. Finally, it classifies the operating status of the distribution area based on the line loss detection results, predicts the classification results and line loss trends, and issues tiered early warning information.

[0008] A second aspect of this application provides a distribution network substation line loss detection system, the system comprising:

[0009] The system comprises the following modules: a data acquisition module for collecting operational data from the main power supply transformer and each user branch in the distribution area, including at least voltage, current, and power factor; a preliminary line loss result acquisition module for synchronizing and cleaning the collected operational data, calculating the difference between theoretical power supply energy and actual metered energy based on the principle of power balance, and obtaining preliminary line loss results; a line loss feature library establishment module for establishing a line loss feature library by combining historical operational data, meteorological data, and time-period characteristic variables; a line loss detection result generation module for identifying and correcting the preliminary line loss results based on the line loss feature library using machine learning algorithms, and generating line loss detection results; and an early warning information release module for classifying the distribution area's operational status based on the line loss detection results, predicting the classification results and line loss trends, and releasing tiered early warning information.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] The system collects operational data from the main power supply transformer and each user branch line in the distribution network area. It performs time synchronization and data cleaning on the collected data, calculates the difference between theoretical power supply energy and actual metered energy based on the power balance principle, and obtains preliminary line loss results. A line loss feature database is established by combining historical operational data, meteorological data, and time-period characteristic variables. Based on this database, machine learning algorithms are used to identify and correct the preliminary line loss results, generating line loss detection results. The system then classifies the operating status of the distribution network area based on the line loss detection results, predicts line loss trends, and issues tiered early warning information. This achieves full-process management of line loss detection in the distribution network area, improving the accuracy of line loss detection and the efficiency of anomaly identification. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This application provides a schematic flowchart of a method for detecting line losses in a distribution network area.

[0014] Figure 2 This is a schematic diagram of a distribution network substation line loss detection system provided in an embodiment of this application.

[0015] Figure labeling: 10 for running data acquisition module, 20 for preliminary line loss result acquisition module, 30 for line loss feature library establishment module, 40 for line loss detection result generation module, and 50 for early warning information release module. Detailed Implementation

[0016] This application provides a method and system for detecting line loss in distribution network areas, which addresses the technical problems in existing technologies such as the lack of full-process coverage in distribution network area line loss detection, large deviations in line loss detection results, and untimely identification of abnormal situations.

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0018] Example 1, as Figure 1 As shown, this application provides a method for detecting line losses in a distribution network area, the method comprising:

[0019] Step S100: Collect the operating data of the main power supply transformer in the distribution area and each user branch, including at least voltage, current and power factor.

[0020] Specifically, by deploying operational data acquisition equipment in the distribution network area, comprehensive operational data is collected for the core power supply nodes of the area, namely the main power supply transformer and the end-user power consumption nodes, namely each user branch. The collected data covers at least three core electrical parameters: voltage, current, and power factor. Among them, the operational data of the main power supply transformer of the area needs to reflect the overall power output status of the power supply end, and the operational data of each user branch needs to accurately correspond to the power consumption of the end-user side. The two types of data together constitute the basic data of the entire power supply-consumption link required for line loss calculation. This provides original data support for subsequent line loss calculation based on the power balance principle and line loss distribution analysis combined with topology structure, ensuring the data integrity and effectiveness of the subsequent line loss detection process.

[0021] Step S200: Synchronize the collected operating data with time and clean the data. Calculate the difference between the theoretical power supply energy and the actual metered energy based on the principle of power balance to obtain preliminary line loss results.

[0022] Specifically, time synchronization processing is performed on the collected operational data of the main power supply transformer and each user branch in the distribution area. Based on the distribution network topology, the distribution area is abstracted into a tree network composed of nodes and edges. The root node is the main power supply transformer in the distribution area, the intermediate nodes are the branch points at each level, and the leaf nodes are the user branches. Then, based on the transmission path between nodes, transmission time difference analysis is performed to determine the node transmission synchronization window. The time synchronization relationship of each data collection node is aligned according to the node transmission synchronization window to establish the time alignment relationship of each topology. Next, the collected operational data is cleaned, and anomalies are removed by filtering out those exceeding the reasonable range. Invalid data is filled in to ensure data quality. Then, based on the power flow and branching relationship represented by the distribution network topology, the power supply balance relationship between the main power supply transformer of the distribution area and the user branch is established. At the same time, multiple line loss monitoring nodes are configured based on the distribution network topology. The multiple line loss monitoring nodes cover the key transmission path nodes of the main power supply transformer of the distribution area and each user branch. According to the power supply balance relationship, the energy difference is calculated for the multiple line loss monitoring nodes in sequence to obtain the difference distribution. Then, the line loss results of multiple nodes are standardized according to the difference distribution, and finally, multi-level preliminary line loss results are obtained.

[0023] Step S300: Combine historical operational data, meteorological data, and time-period characteristic variables to establish a line loss characteristic database.

[0024] Specifically, line loss relationships were fitted to three types of data: historical operational data, meteorological data, and time-specific variables, establishing line loss impact relationships across various dimensions. Subsequently, based on these line loss impact relationships and the distribution network topology, multiple influencing variable features were extracted, including electrical condition features, load behavior features, network topology features, and spatiotemporal environmental features. Electrical condition features encompass voltage deviation rate, three-phase imbalance, and total harmonic distortion rate at transformer substation outlets and key nodes at each level. Load behavior features include daily load factor and peak-valley difference rate calculated based on user branch data, as well as features related to typical user... The similarity of the curve shape of the power mode, the network topology features including the power supply radius, line equivalent impedance and the hierarchical depth of the nodes in the tree network calculated based on the topology, and the spatiotemporal environmental features including the temperature and humidity impact coefficient generated by converting meteorological data, as well as the date type factor and holiday mode label generated by converting time period feature variables; then, the extracted four types of impact variable features are quantified and mapped with the corresponding line loss impact, and a mapping relationship between impact variable features and line loss quantification coefficients is constructed; finally, based on this mapping relationship, the corresponding data time and topology nodes are associated and indexed to complete the establishment of the line loss feature library.

[0025] Step S400: Based on the line loss feature library, use machine learning algorithms to identify and correct the preliminary line loss results, and generate line loss detection results.

[0026] Specifically, based on the established line loss feature library, machine learning algorithms are used to mine the correlation between influencing variables and line loss in the feature library, learn the variation patterns of theoretical line loss, and then construct a benchmark line loss model that can reflect the normal line loss level. The preliminary line loss results are then input into the benchmark line loss model, which determines whether the preliminary line loss results fall within the normal theoretical line loss range, identifying and marking abnormal theoretical line losses that exceed the normal range. Simultaneously, the deviation rate of each level of line loss is calculated based on the corresponding benchmark values ​​output by the benchmark line loss model. Finally, the calculated deviation rate is compared with the pre-stored abnormal correlation patterns in the preset abnormal feature library. Yes, specific abnormal line loss patterns and classifications are identified. Abnormal line loss patterns include suspected electricity theft losses, equipment failure losses, and data quality anomalies. For the identified abnormal line loss patterns, source tracing and location are carried out by combining electrical correlation analysis and distribution network topology. The level with the highest abnormal contribution in the preliminary line loss results is determined. Within the abnormal level, the correlation coefficient between the current of each user branch and the total residual current of the system is calculated, or the equivalent loss power of each line section is calculated. The object with the highest correlation coefficient or the largest equivalent loss power is identified as a high-probability abnormal source and its location information is output. Finally, the normal line loss judgment results, abnormal line loss pattern classifications, and abnormal source location information are integrated to generate complete line loss detection results.

[0027] Step S500: Classify the operating status of the transformer area based on the line loss detection results, predict the classification results and line loss trends, and issue graded early warning information.

[0028] Specifically, based on the generated line loss detection results, combined with the included normal line loss levels and abnormal line loss patterns (such as suspected electricity theft losses, equipment failure losses, data quality anomalies, and anomaly source location information), the overall operating status of the transformer area is classified. The classification must reflect whether the line loss is within a reasonable range, whether abnormal losses exist, and the type of anomaly, clearly distinguishing between normal operating conditions and abnormal operating conditions containing different anomaly patterns. Subsequently, combining historical operating data and time-period characteristic variables from the line loss feature database, and referencing historical line loss change patterns and current line loss detection results, the trend of line loss changes over a subsequent period is predicted to determine whether the line loss is increasing, decreasing, or remaining stable. Finally, based on the transformer area operating status classification results and line loss trend predictions, tiered rules are formulated according to risk level classification standards (such as normal, watch, alarm, and emergency). Corresponding early warning content is generated for different risk levels, clearly informing users of the current line loss status, potential risks (such as abnormal loss types, trend risk points, and preliminary handling suggestions), and finally, tiered early warning information is issued to provide clear guidance for transformer area line loss management and fault handling.

[0029] In one possible implementation, step S200 further includes:

[0030] Step S210: Align the collected timestamps of the main power supply transformer and each user branch in the distribution area, and perform transmission simulation based on the distribution network topology to establish the time alignment relationship of each topology.

[0031] Step S220: Based on the time alignment relationship, synchronize the collected operating data in time, and perform anomaly screening and missing data completion on the collected operating data. Based on the power convergence and branching relationship represented by the distribution network topology, establish the power balance supply relationship between the main power supply transformer of the distribution area and the user branch.

[0032] Step S230: Calculate the energy difference based on the power supply balance relationship to obtain the preliminary line loss result.

[0033] Specifically, for the collected operating data of the main power supply transformer in the distribution area, reflecting the overall power supply status, and the operating data of each user branch, reflecting the end-user power consumption status, the first step is to perform timestamp alignment processing to ensure that the initial time records of the two types of data are consistent and to avoid data time misalignment caused by clock deviations of the acquisition equipment. Then, based on the actual topology of the distribution network, transmission simulation is performed, abstracting the entire distribution area into a tree network composed of nodes and edges, where the root node corresponds to the main power supply transformer in the distribution area, the intermediate nodes correspond to the power branch points at each level, and the leaf nodes correspond to each user branch. Based on this tree network model, the length and speed of power transmission paths between different nodes are analyzed, the transmission time difference between each node is calculated, and the transmission synchronization window of each acquisition node is determined. Finally, according to the transmission synchronization window, the time records of all acquisition nodes, including the main power supply transformer, the branch points at each level, and the user branches, are precisely aligned to establish a time alignment relationship covering the entire distribution area topology, laying a time consistency foundation for subsequent data synchronization and line loss calculation.

[0034] Based on the established time alignment relationships of each topology, the collected operating data of the main power supply transformer and each user branch in the distribution area are processed for time synchronization to ensure that the operating data of different nodes and branches are completely matched in the time dimension, eliminating data misalignment caused by transmission delay or acquisition time difference. Subsequently, data cleaning is carried out. Anomalies are filtered out from the synchronized operating data by using preset reasonable threshold ranges for electrical parameters, such as the normal fluctuation range of voltage and current. Invalid or interference data exceeding the threshold is removed. At the same time, appropriate interpolation algorithms are used, such as linear interpolation based on valid data in adjacent time periods, to fill in missing parts of the data, ensuring data integrity and validity. Finally, combined with the power flow and branching relationship defined by the distribution network topology, that is, the power output of the main power supply transformer is distributed to each user branch through each level of branch nodes, and the power consumption of each user branch is summarized and then correlated with the power output of the main power supply transformer, the transmission and distribution logic of power from the power supply end to the power consumption end is sorted out, and the power balance supply relationship between the total power supply of the main power supply transformer in the distribution area and the total power consumption of each user branch is established, providing a clear logical basis for subsequent energy difference calculation.

[0035] Based on the distribution network topology, multiple line loss monitoring nodes are configured on the critical transmission paths of the main power supply transformer in the distribution area, branch nodes at all levels, and user branches to ensure coverage of the entire power transmission link from the power supply end to the power consumption end. Subsequently, based on the established power balance supply relationship between the main power supply transformer in the distribution area and the user branches—that is, the theoretical power supply energy output by the main power supply transformer should be balanced with the actual metered energy of each user branch and the line loss energy during transmission—energy difference calculations are performed for each monitoring node. Taking the main power supply transformer monitoring node as an example, the theoretical power supply energy output by it is calculated relative to the downstream... The difference is calculated as the sum of the actual metered energy of all branch nodes and user branches. For each level of branch node monitoring node, the difference between the upstream power received and the downstream power allocated and the power consumed by the user is calculated, thus forming the energy difference distribution of the entire link. Finally, based on the difference distribution and combined with parameters such as the transmission path length and line specifications of different monitoring nodes, the line loss results of the multi-node path are standardized to eliminate the calculation deviation caused by the difference in hardware specifications between different nodes, and finally generate a multi-level preliminary line loss result covering the main power supply transformer, branches at all levels, and user branches.

[0036] In one possible implementation, step S210 further includes:

[0037] Step S211: Based on the distribution network topology, the transformer substation is abstracted into a tree network consisting of nodes and edges, where the root node is the main power supply transformer of the substation, the intermediate nodes are the branch points at each level, and the leaf nodes are the user branches.

[0038] Step S212: Perform transmission time difference analysis based on the transmission path between nodes to determine the node transmission synchronization window.

[0039] Step S213: Align the time synchronization relationship of each acquisition node according to the node transmission synchronization window, and establish the time alignment relationship of each topology.

[0040] Specifically, based on the actual topology of the distribution network, the power transmission and distribution system of the transformer substation is simplified and abstracted, and a tree network composed of nodes and edges is constructed. Nodes correspond to key power points with different functions within the transformer substation, while edges represent the power transmission paths between nodes. In this tree network, the roles of various nodes are clearly defined: the main power supply transformer, which is the source of the total power output of the transformer substation, is set as the root node of the tree network, responsible for the initial transmission of power from the power supply end to the entire transformer substation; the branch switches or junction boxes at all levels within the transformer substation that perform power transfer and distribution functions are set as intermediate nodes, connecting the root node and the end power consumption nodes to realize hierarchical distribution of power; and the user branches that directly connect to users and are responsible for terminal power consumption metering are set as leaf nodes, serving as the ends of the tree network and receiving power distributed by the intermediate nodes. This abstract approach clearly presents the transmission logic of electrical energy flowing from the main power supply transformer (root node) through various branch points (intermediate nodes) to the leaf nodes of each user branch, providing an intuitive network model to support subsequent analysis of transmission time differences between nodes and the establishment of time alignment relationships.

[0041] Based on the constructed tree network, the actual transmission paths between each node are first determined, namely, the specific route from the root node's main power supply transformer to each intermediate node, i.e., the branch points at each level, and from each intermediate node to each leaf node's user branch. Simultaneously, key parameters such as the line length and conductor material of each transmission path are obtained. This is combined with the transmission speed of electrical energy in conductors of different materials, such as approximately 3 × 10⁻⁶ for copper core conductors. 8 The time difference of power transmission between the root node and each intermediate node, and between each intermediate node and each leaf node is calculated using m / s. The time difference is calculated as the transmission path length divided by the power transmission speed. Based on the calculated transmission time difference between each node, the data acquisition time deviation range of each acquisition node is determined, and then the node transmission synchronization window is defined. This window needs to cover the maximum transmission time difference between all nodes to ensure that the running data collected by the root node, intermediate nodes, and leaf nodes can be accurately matched in the time dimension, avoiding data time misalignment due to differences in transmission time, and providing a clear time range basis for the subsequent alignment of the data time synchronization relationship between each node.

[0042] Using a defined node transmission synchronization window as the basis for time calibration, the root node in the tree network, i.e., the main power supply transformer of the distribution area, is first selected as the time reference node, and the timestamp of its collected data is set as a unified reference standard. Then, for each intermediate node, i.e., each level of branch point and leaf node user branch, the timestamp of its own collected data is offset and corrected according to the time difference corresponding to its transmission path with the reference node, i.e., the root node. For example, if the transmission time difference between an intermediate node and the root node is 0.02 seconds, the timestamp of the data collected by the intermediate node is uniformly shifted backward by 0.02 seconds to ensure that its collected data at the corresponding time of the root node is accurately matched in the time dimension. After completing the timestamp correction of all nodes, the time synchronization consistency between each node is further verified to ensure that any two nodes, such as adjacent intermediate nodes, or intermediate nodes and leaf nodes, collect data in the same time period, are all within the node transmission synchronization window. Finally, through the above calibration and verification, a time alignment relationship covering all topological structures of the root node, intermediate nodes, and leaf nodes in the tree network is established, realizing the time reference unification of the entire distribution area's operational data collection, and eliminating time dimension deviations for subsequent data synchronization processing and power balance calculation.

[0043] In one possible implementation, step S230 further includes:

[0044] Step S231: Based on the distribution network topology, configure multiple line loss monitoring nodes, which cover the main power supply transformers of the distribution area and the key transmission path nodes of each user branch.

[0045] Step S232: Based on the power supply balance relationship, calculate the energy difference for each of the multiple line loss monitoring nodes in sequence to obtain the difference distribution.

[0046] Step S233: Standardize the multi-node path loss results according to the difference distribution to obtain the preliminary line loss results at multiple levels.

[0047] Specifically, based on the tree-structure diagram of the distribution network and the results of on-site surveys, the locations of the main power supply transformer at the root node, the branch switch boxes / cabinets at intermediate nodes, and the user branch access distribution boxes at leaf nodes are mapped one by one to the physical equipment locations, clarifying the power transmission role and path association of each node. Secondly, key monitoring points are selected: three-phase smart meters with real-time power and energy metering functions are installed at the low-voltage side of the main power supply transformer, serving as core monitoring nodes; current transformers and voltage acquisition modules are installed in the branch switch boxes at each level, synchronously connected to edge computing terminals, supporting local data storage and preliminary processing, covering the branch transmission paths. At the distribution box where the user branch line enters the transformer area, replace the smart energy meter with one equipped with remote communication function to ensure that there is a monitoring device at the entrance of each user branch line. Then, using wireless public network or power line carrier communication technology, connect the acquisition devices of all monitoring nodes to the transformer area power monitoring master station system to achieve real-time data upload. Finally, through the master station system, issue calibration instructions to adjust the metering accuracy of each monitoring node, such as current and voltage acquisition errors and time synchronization, to ensure that the monitoring data can accurately reflect the power flow status of the corresponding node, and finally complete the configuration of multi-line loss monitoring nodes covering the main transformer, branch paths, and key points of user branches.

[0048] Relying on the power monitoring master station system of the distribution area, the system retrieves the associated data corresponding to the established power balance supply relationship, including the data collected by each line loss monitoring node within a preset time period. For the root node, i.e., the monitoring node of the main power supply transformer in the distribution area, the system needs to retrieve the theoretical power supply energy data calculated from voltage, current, and power factor. For the intermediate nodes, i.e., the branch monitoring nodes, the system needs to retrieve the upstream input power energy data and the downstream output power energy data. For the leaf nodes, i.e., the monitoring nodes of user branches, the system needs to retrieve the access distribution power energy data and the actual metered consumption power energy data. Then, the energy difference calculation formula is applied sequentially according to the root node, intermediate node, and leaf node: for the root node, the formula is calculated as theoretical power supply energy minus the sum of the actual metered energy of all downstream intermediate nodes and leaf nodes; for the intermediate nodes, the formula is calculated as upstream input energy minus the sum of the output energy of downstream branches and leaf nodes; for the leaf nodes, the formula is calculated as access distribution energy minus the actual metered consumption energy. During the calculation process, the data calculation is completed by the built-in automated calculation module of the master station system to avoid errors from manual calculation. Next, the calculation results of each node are verified hierarchically to confirm whether the difference at the root node is approximately equal to the sum of the differences at all intermediate nodes, and whether the sum of the differences at all intermediate nodes is approximately equal to the sum of the differences at all leaf nodes. This ensures compliance with the power balance logic. If any deviation is found, a data review mechanism is triggered to investigate problems in the collected data or the calculation process. Finally, through the data visualization module of the main station system, the energy differences calculated by each node are displayed according to the topological hierarchy of the main transformer layer, branch layer, and tributary layer. This forms a difference distribution chart that includes the specific values ​​of the differences at each node, the percentage of the differences, and the hierarchical summary results, thus completing the acquisition of the difference distribution.

[0049] First, based on the distribution network operation and maintenance archives and on-site measurements, basic parameters such as line material, conductor cross-sectional area, line length, average load current, and power factor within a preset period are collected for each line loss monitoring node. These parameters are then entered into the line loss analysis module of the distribution area power monitoring master station system. Next, within the master station system, a standardized correction model is constructed based on industry standards such as the Distribution Network Line Loss Calculation Technical Guidelines, combined with the collected path parameters. The model includes a built-in table of line loss correction coefficients for different line parameters and load conditions, automatically matching the corresponding correction coefficients based on the input parameters. Then, the energy difference values ​​of each node in the obtained difference distribution are input into the standardized correction model along with the corresponding node's path parameters. The model eliminates deviations in line loss results caused by differences in line specifications (e.g., copper vs. aluminum conductors, 10mm² vs. 25mm² cross-sectional areas) and load levels (e.g., high-load vs. low-load nodes) by calculating the energy difference multiplied by the correction coefficient, resulting in a unified standard for line loss values. Finally, using the hierarchical classification function of the main station system, the standardized line loss values ​​are classified into the main power supply transformer layer, the root node line loss, the intermediate branch layer, the intermediate node line loss, the user branch layer, and the leaf node line loss according to the distribution network topology. A summary table of line loss at each level and a line loss distribution map of the entire distribution area are generated, and finally, preliminary line loss results at multiple levels are obtained.

[0050] In one possible implementation, step S300 further includes:

[0051] Step S310: Fit the line loss relationship to the historical operation data, meteorological data, and time period characteristic variables respectively, and establish the line loss influence relationship of each dimension.

[0052] Step S320: Based on the influence relationship of line loss in each dimension and the distribution network topology, extract the features of the influencing variables and establish the mapping relationship between the features of the influencing variables and the line loss quantification coefficient.

[0053] Step S330: Based on the mapping relationship, perform association indexing according to the corresponding data time and topology nodes to establish the line loss feature library.

[0054] Specifically, line loss relationship fitting is performed on historical operating data. Longer-term operating records for the transformer area, such as those over a year, are collected, including current, voltage, and power factor data for each time period, along with corresponding line loss values. Algorithms such as linear regression and time series analysis are used to analyze the intrinsic relationship between changes in operating parameters and line loss fluctuations. For example, the proportional relationship between changes in current values ​​and increases in line loss, and the impact of power factor adjustments on line loss reduction, are calculated to establish a quantitative impact relationship between historical operating data and line loss. Next, line loss relationship fitting is performed on meteorological data. Meteorological records such as temperature, humidity, and precipitation from the same period are summarized, and line loss data under different meteorological conditions are categorized and compared. For example, the difference in line loss rates between high-temperature and normal-temperature weather is statistically analyzed, and the impact of increased humidity on conductor insulation performance on line loss is analyzed. Through data comparison and trend analysis, the direction and extent of meteorological factors' influence on line loss are clarified, establishing a corresponding impact relationship between meteorological data and line loss. Finally, a fitting of the line loss relationship was carried out for the time-period characteristic variables. The time dimension was divided into different characteristic periods, including weekdays and weekends, peak and off-peak electricity consumption periods, statutory holidays and ordinary days. The average line loss rate and line loss fluctuation range of each characteristic period were statistically analyzed, and the impact of changes in user electricity load characteristics on line loss in different periods was analyzed. For example, the law that the concentrated load during peak periods leads to an increase in the line loss rate was established, thereby establishing the correlation and influence relationship between time-period characteristic variables and line loss.

[0055] Focusing on the core of electrical operation, the voltage deviation rate of the distribution area outlet and key nodes at each level is extracted, namely the ratio of the difference between the actual voltage and the rated voltage, the three-phase imbalance, and the total harmonic distortion rate, to form electrical state characteristics. Then, combined with the user's electricity consumption patterns, the daily load rate and peak-valley difference rate are calculated based on the user branch data, and the similarity of the actual electricity consumption curve and the curve shape of the typical pattern is calculated through the dynamic time warping algorithm to obtain load behavior characteristics. Subsequently, based on the tree topology of the distribution network, the power supply radius from each node to the main transformer is measured, the equivalent impedance of the line is calculated, and the node level depth is marked to extract network topology characteristics. Finally, environmental and time period information is integrated, and temperature and humidity data are converted into temperature and humidity influence coefficients. Date type factors are generated according to date type, and holiday pattern labels are generated according to holiday marking to form spatiotemporal environmental characteristics. Finally, historical data on the four types of features and their corresponding line loss impact were collected, i.e., the difference between actual line loss and baseline line loss. A multiple linear regression algorithm was used, with the four types of features as independent variables and the line loss impact as dependent variable. The regression coefficients of each feature were determined through iterative calculation, i.e., the change in line loss impact corresponding to each unit change in the feature. Finally, a mapping relationship between all influencing variable features and line loss quantification coefficients was constructed.

[0056] Based on the established mapping relationship between influencing variable characteristics and line loss quantification coefficients, an association indexing mechanism is constructed in the database system. This mechanism timestamps various influencing variable characteristics, such as electrical state characteristics and load behavior characteristics, according to the time dimension of data collection (e.g., hourly, daily) and spatially categorizes them according to distribution network topology nodes (e.g., root nodes, intermediate nodes, leaf nodes). Through a dual index of timestamps and topology node identifiers, the influencing variable characteristics, line loss quantification coefficients, and historical line loss data corresponding to the same time and node are associated and stored, ensuring that subsequent queries can quickly match complete line loss impact data for a specific time and node. Finally, through this association indexing mechanism, all dimensional data and mapping relationships are integrated to form a clearly structured and highly efficient line loss feature library.

[0057] In one possible implementation, step S320 further includes:

[0058] Step S321: Extract electrical condition characteristics, including voltage deviation rate, three-phase unbalance, and total harmonic distortion rate at the transformer substation outlet and key nodes at each level.

[0059] Step S322: Extract load behavior characteristics, including daily load factor, peak-valley difference rate, and curve shape similarity to typical electricity consumption patterns calculated based on user branch data.

[0060] Step S323: Extract network topology features, including the power supply radius, line equivalent impedance, and node hierarchy depth in the tree network calculated based on the topology.

[0061] Step S324: Extract spatiotemporal environmental features, including temperature and humidity influence coefficients generated from meteorological data, and date type factors and holiday pattern labels generated from time period feature variables.

[0062] Step S325: Quantify and map the electrical state characteristics, load behavior characteristics, network topology characteristics, and spatiotemporal environmental characteristics with the corresponding line loss impact to construct the mapping relationship.

[0063] Specifically, the focus is on extracting electrical state characteristics from core electrical operation parameters: Real-time voltage data from the transformer substation's power monitoring master station system is retrieved from the substation outlet and key nodes at each level, such as the low-voltage side of the main transformer and branch line junctions. The voltage deviation rate is calculated, which is the percentage of the difference between the actual voltage and the rated voltage relative to the rated voltage. Three-phase current data from each node is collected, and the three-phase current imbalance is calculated using the three-phase current imbalance formula, which is the percentage of the difference between the maximum and minimum current relative to the average current. Simultaneously, voltage waveforms are monitored, and harmonic components are analyzed using Fourier transform to obtain the total harmonic distortion rate and the percentage of the square root of the sum of the squares of the effective values ​​of each harmonic voltage relative to the effective value of the fundamental voltage, thus completing the extraction of electrical state characteristics.

[0064] Based on historical electricity consumption data of each user branch, the daily maximum load and average load are statistically analyzed to calculate the daily load factor, which is the ratio of the daily average load to the daily maximum load. The peak-valley difference rate is obtained by comparing the daily peak load with the minimum load during off-peak hours, which is the ratio of the peak-valley load difference to the peak load. At the same time, the actual electricity consumption curves of each branch are matched with preset typical electricity consumption pattern curves, such as the early morning low and evening high pattern for residential users and the daytime concentrated pattern for commercial users. The similarity of the curve shape is calculated by using a dynamic time warping algorithm to complete the extraction of load behavior features.

[0065] Network topology features are extracted by combining the distribution network topology structure: Based on the tree network topology diagram, the shortest line distance from each node to the main power supply transformer is measured to determine the power supply radius; based on parameters such as line material, cross-sectional area, and length, the equivalent impedance of the line where each node is located is calculated using the resistance calculation formula: resistance = resistivity × length / cross-sectional area, combined with the series and parallel relationship of the lines; according to the tree network hierarchy, the root node is level 0, the direct subordinate branches are level 1, and so on, the level depth of each node is marked to complete the extraction of network topology features.

[0066] Integrating environmental and time-period information to extract spatiotemporal environmental features: Meteorological data, i.e., temperature and humidity, are substituted into a preset formula, such as temperature and humidity influence coefficient = 0.002 × temperature + 0.001 × humidity. The coefficient value increases with the increase of temperature and humidity, and is converted into a temperature and humidity influence coefficient; Based on the time-period feature variables, the date is divided into weekdays and weekends, and a date type factor is generated, with weekdays being 1 and weekends being 0; Statutory holidays are marked to generate holiday mode labels, with holidays being 1 and non-holidays being 0, thus completing the extraction of spatiotemporal environmental features.

[0067] Establish a quantitative mapping relationship between features and line loss impact: Collect historical data of various features and corresponding actual line loss values ​​for the same period. Use a multiple linear regression algorithm with electrical state features, load behavior features, network topology features, and spatiotemporal environment features as independent variables, and the amount of line loss impact and the difference between the actual line loss value and the baseline line loss value as dependent variables. Determine the regression coefficient of each feature through iterative calculation, that is, the change in the amount of line loss impact corresponding to each unit change in the feature. For example, each 1% increase in voltage deviation rate corresponds to an increase of 0.8 kWh in the amount of line loss impact, and each 100-meter increase in power supply radius corresponds to an increase of 0.5 kWh in the amount of line loss impact. Finally, construct a quantitative mapping relationship between all features and the amount of line loss impact.

[0068] In one possible implementation, step S400 further includes:

[0069] Step S410: Based on the line loss feature library, learn the theoretical line loss changes through machine learning algorithms to construct a baseline line loss model.

[0070] Step S420: Judge the preliminary line loss result using the baseline line loss model to determine whether it is a normal theoretical line loss, identify and mark abnormal theoretical line losses, and generate the line loss detection result.

[0071] Specifically, a baseline line loss model is constructed using a line loss feature library as data support. This library retrieves historical electrical state characteristics from the same period, such as voltage deviation rate, three-phase imbalance, load behavior characteristics (e.g., daily load rate, peak-valley difference rate), network topology characteristics (e.g., power supply radius, line equivalent impedance), and spatiotemporal environmental characteristics (e.g., temperature and humidity influence coefficient, date type factor). Simultaneously, normal line loss records for the corresponding time period are extracted, excluding high-loss data caused by abnormal faults as labels. Gradient Boosting Tree (XGBoost) is selected as the machine learning algorithm. The aforementioned feature data is input into the algorithm for model training. The algorithm iteratively learns the variation law of theoretical line loss under different feature combinations, such as the reasonable range of theoretical line loss under high load and high temperature environments, continuously optimizing model parameters to reduce the error between predicted values ​​and actual normal line loss values. After training, the model performance is verified using test set data to ensure a prediction accuracy of over 95%, ultimately forming a baseline line loss model that can accurately output the reasonable line loss range under different operating conditions.

[0072] The system retrieves operating condition data corresponding to the preliminary line loss results from the real-time monitoring system of the distribution network. This includes electrical state characteristics during the period in which the line loss occurred, such as the voltage deviation rate at the transformer substation outlet, the three-phase imbalance at branch nodes, load behavior characteristics such as the daily load rate and peak-valley difference rate of user branches, spatiotemporal environmental characteristics such as the real-time temperature and humidity influence coefficient and date type factor, as well as the network topology characteristics of the corresponding nodes, such as the power supply radius and the equivalent impedance of the line. These characteristic data are then organized into an input dataset according to the format required by the baseline line loss model. Subsequently, the organized characteristic dataset is input into the baseline line loss model. Based on its built-in line loss prediction logic, the model outputs the theoretical normal range of line loss under the current operating conditions. For example, under the current load and temperature and humidity conditions, the theoretical line loss of a certain user branch should be in the range of 4.2-5.8 kWh. The model also outputs the confidence level of this range, such as the 98% confidence interval. Next, the preliminary line loss results at the corresponding level are compared with the theoretical normal range output by the model. If the preliminary line loss value falls within the normal range and meets the confidence level requirements, it is determined to be a normal theoretical line loss, marked as normal in the detection record, and the corresponding feature data and theoretical range are associated and stored. If the preliminary line loss value exceeds the normal range or is lower than the reasonable lower limit, it is determined to be an abnormal theoretical line loss. The abnormal node number, abnormal line loss value, and the difference exceeding the range are automatically recorded. Combined with the feature data, the abnormal causes are preliminarily analyzed. For example, a three-phase imbalance of 9% may lead to high loss. The abnormality and suspected causes are marked in the detection record. Finally, the judgment results of all nodes are integrated and summarized according to the distribution network topology level: main transformer layer, branch layer, and branch line layer. A line loss detection report is generated, which includes normal line loss statistics, an abnormal line loss list, node location, abnormal value, suspected causes, and abnormal percentage. It also supports exporting spatiotemporal distribution charts of abnormal nodes, providing accurate basis for subsequent maintenance personnel to investigate line loss anomalies, and finally forming a complete line loss detection result.

[0073] In one possible implementation, step S420 further includes:

[0074] Step S421: Calculate the deviation rate by comparing the preliminary line loss results of each level with the corresponding benchmark value.

[0075] Step S422: Based on the deviation rate, compare it with the pre-stored correlation patterns in the abnormal feature library, identify abnormal line loss patterns and classify them, and generate the line loss detection result containing at least one abnormal line loss pattern.

[0076] Specifically, the theoretical line loss baseline values ​​for each level—the main power supply transformer level, intermediate branch level, and user branch level—are extracted from the baseline line loss model. These baseline values ​​are reasonable line loss values ​​generated by the model based on characteristic data under normal operating conditions, such as voltage deviation rate, daily load rate, and temperature and humidity influence coefficient. For example, the baseline value for a node in the main transformer level is 18 kW·h / day, and the baseline value for a user branch node is 4.5 kW·h / day. Then, the preliminary line loss results for each level and node are retrieved. For example, the preliminary line loss for the node in the main transformer level is 22.5 kW·h / day, and the preliminary line loss for the user branch node is 6.3 kW·h / day. The deviation rate is calculated node by node using the unified formula: Deviation Rate = (Preliminary Line Loss Result - Theoretical Line Loss Baseline Value) ÷ Theoretical Line Loss Baseline Value × 100%. For example, the deviation rate for the main transformer level node is (22.5 - 18) ÷ 18 × 100% = 25%, and the deviation rate for the user branch node is (6.3 - 4.5) ÷ 4.5 × 100% = 40%. Ultimately, a deviation rate dataset covering all levels of nodes is formed, which can intuitively reflect the degree of deviation of line loss, providing a quantitative basis for subsequent anomaly identification.

[0077] The calculation of deviation rates for each node is retrieved, and simultaneously, real-time operational data for the corresponding nodes is extracted, such as user branch power consumption curves, line voltage values, impedance monitoring data, and raw data quality records, to construct a comprehensive analysis dataset of deviation rate, operational data, and data quality. This dataset is then compared one by one with three pre-stored correlation patterns in the anomaly feature database: For suspected electricity theft patterns, it is determined whether the deviation rate exceeds a preset threshold (e.g., >30%), and it is checked whether there are any sudden changes in the power consumption behavior of one or more user branches, such as a sudden drop in power consumption of more than 50% and a high correlation between the branch's electrical quantities (e.g., current, power) and the remaining current of the main line (e.g., correlation coefficient >0.8). If all conditions are met, it is matched as a suspected electricity theft pattern; For equipment failure and loss patterns, it is first confirmed that the deviation rate exceeds the threshold, and then the abnormal section is located through line segment monitoring data, and the abnormal section is checked. If the voltage drop is accompanied by an abnormal decrease, such as being more than 10% lower than the rated voltage, or if the impedance characteristics change, such as the impedance value increasing by more than 20% compared to the historical average, then it is determined to be an equipment failure loss mode. For data quality abnormal mode, under the premise that the deviation rate exceeds the threshold, check whether there are logical contradictions in the original data, such as mismatch between current and power calculations, missing key data, such as no voltage record for 1 consecutive hour or data freezing, such as the metering value remaining unchanged for a long time, and if the abnormal data can be reconstructed through the correlation of the distribution network topology, such as by extrapolating data from adjacent nodes, then it is classified as a data quality abnormal mode.

[0078] Finally, all successfully matched abnormal line loss patterns are summarized, and the information is organized according to the structure of abnormal pattern type, involved nodes, key feature evidence, and deviation rate value to generate line loss detection results. The results must clearly indicate at least one abnormal pattern. If a branch is matched with both the suspected electricity theft pattern and the data quality abnormal pattern, the corresponding judgment basis must be explained separately, and preliminary handling suggestions must be given. For example, the suspected electricity theft pattern suggests on-site verification, and the equipment failure pattern suggests line repair, so as to provide clear guidance for subsequent operation and maintenance.

[0079] In one possible implementation, step S422 further includes:

[0080] The abnormal line loss patterns include suspected electricity theft losses, equipment failure losses, and data quality anomalies. For the identified abnormal line loss patterns, electrical correlation analysis and distribution network topology are used to trace and locate the source, determine the level with the highest anomaly contribution in the preliminary line loss results, calculate the correlation coefficient between the current of each user branch and the total residual current of the system within the abnormal level, or calculate the equivalent loss power of each line section; the object with the highest correlation coefficient or the largest equivalent loss power is identified as a high-probability anomaly source and its location information is output.

[0081] Specifically, suspected electricity theft loss refers to abnormal line loss caused by users illegally altering electrical wiring or illegally connecting lines. It typically manifests as a line loss deviation rate exceeding the normal threshold, accompanied by sudden changes in the electricity consumption behavior of specific user branches, such as a sharp drop in electricity consumption or abnormal electricity usage times. Simultaneously, the current, power, and other electrical quantities of this branch show a high correlation with the residual current of the main line, reflecting abnormal current diversion caused by electricity theft. Equipment failure loss refers to additional line loss caused by aging, malfunction, or abnormal operation of distribution network equipment. Its core characteristic is a line loss deviation rate exceeding the threshold, and the anomaly can be located to a specific section through segmented line monitoring. This section is often accompanied by abnormal voltage drops, such as more than 10% below the rated voltage, and an increase in the line's equivalent impedance. Large fluctuations, exceeding 20% ​​above historical normal values, are essentially due to equipment malfunctions, such as damaged wires, poor switch contact, or transformer abnormalities leading to extra energy consumption. Data quality anomalies, however, are not true line loss anomalies but rather false deviations caused by data acquisition, transmission, or storage issues. These manifest as a line loss deviation rate exceeding the threshold, but the original line loss calculation data, such as voltage, current, and electricity consumption records, contain logical contradictions, such as mismatches between current and power calculations, missing key data, no collected values ​​for consecutive periods, or frozen data. Abnormal data can be reconstructed and corrected through network topology correlation, such as extrapolation from adjacent node data or comparison with historical data from the same period. After correction, the line loss usually returns to the normal range.

[0082] The line loss deviation rate and line loss value of each level are extracted from the preliminary line loss results. Through electrical correlation analysis, such as comparing the line loss transmission relationship between levels and verifying the current and voltage balance law, it is determined which level contributes the most to the overall abnormal line loss. For example, if the line loss deviation rate of the main transformer level is 12%, the deviation rate of a certain branch level reaches 38%, and the total line loss of the branch's subordinate branches is abnormally higher than the input line loss of the branch level, then the branch level can be determined as the level with the highest abnormal contribution, and subsequent positioning will focus on this level. Within the locked anomaly level, the corresponding calculation method is selected according to the anomaly line loss mode type: If it is suspected electricity theft loss, the real-time current data of all user branches at this level and the total residual current data of the system are retrieved. The total residual current = the total input current of the upstream line - the sum of the normal power consumption current of each branch. The correlation coefficient between the current and the total residual current of each branch is calculated by the Pearson correlation coefficient algorithm. The closer the coefficient is to 1, the stronger the correlation between the change in the current of the branch and the change in the total residual current, and the more likely there is an abnormal current diversion caused by electricity theft; if it is equipment failure loss, the anomaly level is divided into several line sections, such as a branch level divided into A, B, ... C consists of three sections. Based on the real-time current data and line parameters of each section, the equivalent resistance is calculated. Then, using the formula Equivalent Loss Power = Current² × Equivalent Resistance, the equivalent loss power of each section is obtained. The larger the power value, the more severe the additional loss caused by equipment failure, such as aging wires or poor contact at joints, in that section. If the data quality is abnormal, the interference of data problems on positioning needs to be eliminated first through topology correlation verification. Then, the above two calculation logics are combined. For example, when the data abnormality is accompanied by the current abnormality, the correlation coefficient is calculated. When it is accompanied by the power abnormality, the equivalent loss power is calculated to assist in positioning the potential source of data problems, such as a meter failure in a branch.

[0083] Identify high-probability anomaly sources and output location information. The user branch with the highest correlation coefficient, corresponding to suspected electricity theft or the line segment with the highest equivalent power loss, is identified as a high-probability anomaly source. For data quality anomalies, the system also outputs the data anomaly node and its associated high-probability potential source. The final generated location information must include: the topology level of the anomaly source, its specific location (e.g., XX branch layer - XX line segment), the corresponding anomaly mode type, and key calculated data such as correlation coefficient and equivalent power loss value, providing precise location guidance for on-site verification and fault handling.

[0084] Example 2, based on the same inventive concept as the distribution network transformer area line loss detection method in the previous examples, such as... Figure 2 As shown, this application provides a distribution network transformer area line loss detection system. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0085] The data acquisition module 10 is used to collect the operating data of the main power supply transformer in the distribution area and each user branch, including at least voltage, current and power factor.

[0086] The preliminary line loss result acquisition module 20 is used to synchronize the collected operating data with time and clean the data, calculate the difference between the theoretical power supply energy and the actual metered energy based on the principle of power balance, and obtain the preliminary line loss result.

[0087] The line loss feature database establishment module 30 is used to establish a line loss feature database by combining historical operating data, meteorological data, and time period characteristic variables.

[0088] The line loss detection result generation module 40 is used to identify and correct the preliminary line loss results based on the line loss feature library and using machine learning algorithms to generate line loss detection results.

[0089] The early warning information release module 50 is used to classify the operating status of the transformer area based on the line loss detection results, and to release graded early warning information based on the classification results and line loss trend prediction.

[0090] Furthermore, the system is also used to implement the following functions:

[0091] The collected operating data of the main power supply transformer and each user branch in the distribution area are timestamped and the transmission is simulated based on the distribution network topology to establish the time alignment relationship of each topology. Based on the time alignment relationship, the collected operating data is synchronized in time, and the collected operating data is filtered for anomalies and missing data is filled in. Based on the power convergence and branching relationship represented by the distribution network topology, the power balance supply relationship between the main power supply transformer and the user branch is established. The energy difference is calculated according to the power balance supply relationship to obtain the preliminary line loss result.

[0092] Furthermore, the system is also used to implement the following functions:

[0093] Based on the distribution network topology, the transformer substation is abstracted as a tree network composed of nodes and edges, where the root node is the main power supply transformer of the substation, the intermediate nodes are the branch points of each level, and the leaf nodes are the user branches. Based on the transmission path between nodes, the transmission time difference analysis is performed to determine the node transmission synchronization window. According to the node transmission synchronization window, the time synchronization relationship of each acquisition node is aligned to establish the time alignment relationship of each topology.

[0094] Furthermore, the system is also used to implement the following functions:

[0095] Based on the aforementioned distribution network topology, multiple line loss monitoring nodes are configured, covering the main power supply transformers of the distribution area and the key transmission path nodes of each user branch. According to the aforementioned power balance supply relationship, energy difference calculations are performed on the multiple line loss monitoring nodes sequentially to obtain the difference distribution. Based on the difference distribution, the line loss results of multiple nodes are standardized to obtain the preliminary line loss results at multiple levels.

[0096] Furthermore, the system is also used to implement the following functions:

[0097] Line loss relationships are fitted to the historical operational data, meteorological data, and time-period characteristic variables respectively to establish the line loss impact relationships in each dimension; based on the line loss impact relationships in each dimension and the distribution network topology, the features of the influencing variables are extracted, and a mapping relationship between the features of the influencing variables and the line loss quantification coefficient is established; based on the mapping relationship, the line loss feature library is established by associating and indexing the corresponding data time and topology nodes.

[0098] Furthermore, the system is also used to implement the following functions:

[0099] Electrical condition features are extracted, including voltage deviation rate, three-phase imbalance, and total harmonic distortion rate at transformer substation outlets and key nodes at each level; load behavior features are extracted, including daily load factor, peak-valley difference rate, and curve shape similarity to typical power consumption patterns calculated based on user branch data; network topology features are extracted, including power supply radius, line equivalent impedance, and node hierarchy depth in the tree network calculated based on the topology; spatiotemporal environmental features are extracted, including temperature and humidity influence coefficients generated from meteorological data, date type factors, and holiday mode labels generated from time period characteristic variables; the electrical condition features, load behavior features, network topology features, and spatiotemporal environmental features are quantified and mapped with the corresponding line loss impact, and the mapping relationship is constructed.

[0100] Furthermore, the system is also used to implement the following functions:

[0101] Based on the line loss feature library, a baseline line loss model is constructed by learning theoretical line loss changes through machine learning algorithms. The preliminary line loss results are then judged by the baseline line loss model to determine whether they are normal theoretical line losses. Abnormal theoretical line losses are identified and marked to generate the line loss detection results.

[0102] Furthermore, the system is also used to implement the following functions:

[0103] The deviation rate is obtained by calculating the deviation between the preliminary line loss results at each level and the corresponding benchmark value. The deviation rate is then compared with the pre-stored correlation patterns in the abnormal feature library to identify and classify abnormal line loss patterns, and to generate the line loss detection result containing at least one abnormal line loss pattern.

[0104] Furthermore, the system is also used to implement the following functions:

[0105] For the identified abnormal line loss patterns, the source is located by combining electrical correlation analysis and distribution network topology. The level with the highest abnormal contribution in the preliminary line loss results is determined. Within the abnormal level, the correlation coefficient between the current of each user branch and the total residual current of the system is calculated, or the equivalent loss power of each line section is calculated. The object with the highest correlation coefficient or the largest equivalent loss power is identified as a high-probability abnormal source and the location information is output.

[0106] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0107] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0108] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A method for detecting line loss in a distribution network area, characterized in that, include: Collect operating data of the main power supply transformer and each user branch in the distribution area, including at least voltage, current and power factor; The collected operational data is synchronized with time and cleaned. Based on the principle of power balance, the difference between the theoretical power supply energy and the actual metered energy is calculated to obtain preliminary line loss results. By combining historical operational data, meteorological data, and time-specific variables, a line loss characteristic database is established; Based on the aforementioned line loss feature library, machine learning algorithms are used to identify and correct the preliminary line loss results, generating line loss detection results. Based on the line loss detection results, the operating status of the transformer area is classified, and the classification results and line loss trend predictions are used to issue graded early warning information.

2. The method for detecting line losses in a distribution network area according to claim 1, characterized in that, The collected operational data is synchronized with time and cleaned. Based on the principle of power balance, the difference between the theoretical power supply energy and the actual metered energy is calculated to obtain preliminary line loss results, including: The collected operating data of the main power supply transformers and user branches in the distribution area are time-stamped and the transmission is simulated based on the distribution network topology to establish the time alignment relationship of each topology. Based on the time alignment relationship, the collected operating data is synchronized in time, and the collected operating data is filtered for anomalies and missing data is filled in. Based on the power convergence and branching relationship represented by the distribution network topology, the power balance supply relationship between the main power supply transformer of the distribution area and the user branch is established. The energy difference is calculated based on the power supply balance relationship to obtain the preliminary line loss result.

3. The method for detecting line losses in a distribution network area according to claim 2, characterized in that, Transmission simulation is performed based on the distribution network topology to establish the time alignment relationship between various topologies, including: Based on the distribution network topology, the transformer substation is abstracted into a tree network consisting of nodes and edges, where the root node is the main power supply transformer of the substation, the intermediate nodes are the branch points at each level, and the leaf nodes are the user branches. Based on the transmission path between nodes, the transmission time difference analysis is performed to determine the node transmission synchronization window; The time synchronization relationship of each acquisition node is aligned according to the node transmission synchronization window, and the time alignment relationship of each topology is established.

4. The method for detecting line losses in a distribution network area according to claim 3, characterized in that, Based on the aforementioned power supply balance, the energy difference is calculated to obtain the preliminary line loss results, including: Based on the aforementioned distribution network topology, multiple line loss monitoring nodes are configured, covering the main power supply transformers of the distribution area and the critical transmission path nodes of each user branch. Based on the power supply balance relationship, the energy difference is calculated sequentially for the multiple line loss monitoring nodes to obtain the difference distribution; Based on the difference distribution, the multi-node path loss results are standardized to obtain the preliminary line loss results at multiple levels.

5. The method for detecting line losses in a distribution network area according to claim 2, characterized in that, By combining historical operational data, meteorological data, and time-specific variables, a line loss characteristic database is established, including: Line loss relationships were fitted to the historical operational data, meteorological data, and time-period characteristic variables respectively to establish the influence relationships of line loss in each dimension. Based on the influence relationship of line loss in each dimension and the distribution network topology, the characteristics of the influencing variables are extracted, and the mapping relationship between the characteristics of the influencing variables and the line loss quantification coefficient is established. Based on the mapping relationship, the line loss feature library is established by associating and indexing the corresponding data time and topology nodes.

6. The method for detecting line losses in a distribution network area according to claim 5, characterized in that, Establishing the mapping relationship between the characteristics of the influencing variables and the line loss quantification coefficient includes: Extract electrical condition characteristics, including voltage deviation rate, three-phase unbalance, and total harmonic distortion rate at transformer substation outlets and key nodes at each level; Extract load behavior characteristics, including daily load factor, peak-valley difference rate, and curve shape similarity with typical electricity consumption patterns calculated based on user branch data; Extract network topology features, including power supply radius, line equivalent impedance, and node hierarchy depth in the tree network calculated based on the topology. Extract spatiotemporal environmental features, including temperature and humidity influence coefficients generated from meteorological data, and date type factors and holiday pattern labels generated from time period feature variables; The electrical state characteristics, load behavior characteristics, network topology characteristics, and spatiotemporal environmental characteristics are quantified and mapped with the corresponding line loss impact to construct the mapping relationship.

7. The method for detecting line losses in a distribution network area according to claim 5, characterized in that, Based on the aforementioned line loss feature library, machine learning algorithms are used to identify and correct preliminary line loss results, generating line loss detection results, including: Based on the aforementioned line loss feature library, theoretical line loss changes are learned through machine learning algorithms to construct a baseline line loss model; The baseline line loss model is used to judge the preliminary line loss results to determine whether they are normal theoretical line losses. Abnormal theoretical line losses are identified and marked to generate the line loss detection results.

8. The method for detecting line losses in a distribution network area according to claim 7, characterized in that, Also includes: The deviation rate is obtained by calculating the deviation between the preliminary line loss results of each level and the corresponding benchmark value; Based on the deviation rate, the abnormal line loss patterns are compared with the pre-stored correlation patterns in the abnormal feature library to identify and classify abnormal line loss patterns, and to generate the line loss detection result containing at least one abnormal line loss pattern.

9. The method for detecting line losses in a distribution network area according to claim 8, characterized in that, The abnormal line loss modes include suspected electricity theft loss, equipment failure loss, and abnormal data quality. The method also includes: For the identified abnormal line loss patterns, the source is located by combining electrical correlation analysis and distribution network topology to determine the level with the highest abnormal contribution in the preliminary line loss results. Within the abnormal level, the correlation coefficient between the current of each user branch and the total residual current of the system is calculated, or the equivalent loss power of each line section is calculated. The object with the highest correlation coefficient or the largest equivalent loss power is identified as a high-probability anomaly source and its location information is output.

10. A distribution network transformer area line loss detection system, characterized in that, The system is used to implement the distribution network transformer area line loss detection method according to any one of claims 1-9, the system comprising: The data acquisition module is used to collect operating data of the main power supply transformer in the distribution area and each user branch, including at least voltage, current and power factor; The preliminary line loss result acquisition module is used to synchronize and clean the collected operational data, calculate the difference between the theoretical power supply energy and the actual metered energy based on the power balance principle, and obtain the preliminary line loss result. The line loss feature database creation module is used to create a line loss feature database by combining historical operational data, meteorological data, and time-period characteristic variables. The line loss detection result generation module is used to identify and correct the preliminary line loss results based on the line loss feature library and using machine learning algorithms to generate line loss detection results; The early warning information release module is used to classify the operating status of the transformer area based on the line loss detection results, and to release graded early warning information based on the classification results and line loss trend prediction.

Images