Low-voltage transformer area electric energy meter operation error detection method and system based on graph calculation

Abstract

This invention relates to the field of power data analysis, and more particularly to a method and system for detecting operational errors of low-voltage distribution transformer energy meters based on graph computation. The method first acquires the energy nodes connected to the power grid, then establishes an energy meter graph model based on the node attributes of the energy nodes and the edge relationships between multiple energy nodes. Next, based on a preset distribution transformer determination engine, detectable distribution transformers are selected from the energy meter graph model. Then, based on the law of conservation of energy, a set of energy conservation equations for the detectable distribution transformers is constructed. Finally, the energy conservation equations are substituted into a preset application graph optimization model and graph neural network algorithm to solve for the operational errors of each energy meter in the detectable distribution transformers. Compared to existing technologies, this invention mainly utilizes graph neural networks to mine complex relationships between nodes and a graph optimization model, combined with a distribution transformer determination engine to filter out abnormal data. This approach significantly improves the accuracy and efficiency of error detection.

Description

Technical Field

[0001] This invention relates to the field of power data analysis. More specifically, this invention relates to a method and system for detecting operational errors of low-voltage distribution area energy meters based on graph calculations. Background Technology

[0002] In low-voltage power monitoring networks, error detection of electricity meters is a core component in ensuring grid operation quality and fair billing for users. Traditional electricity meter error detection typically employs a sampling method, specifically involving the removal and verification of expired residential smart meters. This method has the following drawbacks:

[0003] 1) It is impossible to achieve real-time perception of operational errors, which often lags behind the actual occurrence of anomalies.

[0004] 2) The operation and maintenance costs are too high, especially in the context of large-scale deployment of smart meters, the error detection efficiency is difficult to meet the requirements.

[0005] To address the two major technical challenges mentioned above, Chinese patent document CN116794593B discloses a data-driven online estimation method for DC energy meter metering errors. This method mainly estimates meter errors by constructing a system error function and finding its global maximum value. At the same time, it optimizes the error extrema using genetic algorithms and pattern search algorithms, and uses nonlinear programming methods to cross-validate the evaluation results to enhance reliability.

[0006] On the one hand, such methods rely on prior estimates of line resistance parameters, and the estimation error of these parameters accumulates gradually during the calculation process, causing interference and significant deviations in the final error assessment results. On the other hand, such methods require processing a large amount of data and involve complex data processing models, resulting in a long time consumption in generating the final results.

[0007] Therefore, the existing technology mainly suffers from low accuracy and efficiency in error inspection. Summary of the Invention

[0008] To address the aforementioned technical problems of low accuracy and efficiency in error detection, this invention discloses a method and system for detecting operational errors of low-voltage distribution area energy meters based on graph calculation.

[0009] In a first aspect, the present invention discloses a method for detecting the operating error of a low-voltage distribution area energy meter based on graph calculation, comprising:

[0010] Obtain the power nodes connected to the power grid;

[0011] A graph model of the energy meter is established based on the node attributes of the energy nodes and the edge relationships between multiple energy nodes.

[0012] Based on the preset transformer area determination engine, detectable transformer areas are selected from the electricity meter diagram model;

[0013] Based on the law of conservation of energy, a set of energy conservation equations for the detectable transformer area is constructed.

[0014] By substituting the energy conservation equations into the preset application graph optimization model and graph neural network algorithm, the operating errors of each energy meter in the detectable distribution area can be obtained.

[0015] Beneficial Effects: The method of this invention first obtains the power nodes connected to the power grid, and then establishes a power meter graph model based on the node attributes of the power nodes and the edge relationships between multiple power nodes, realizing an accurate expression of the topological description of each power node in the power grid and the energy transmission relationship between nodes. Next, using a transformer area determination engine, normally operating transformer areas in the power meter graph model are screened out to obtain detectable transformer areas. This reduces the amount of data processing and eliminates the contamination of the data processing source by abnormal operating transformer areas. Then, the energy conservation equation system of the detectable transformer areas is constructed using the law of energy conservation, and the energy conservation equation system is substituted into a preset application graph optimization model and graph neural network algorithm to solve for the operating error of each power meter in the detectable transformer area. Compared with the prior art, the method of this invention mainly utilizes graph neural networks to mine the complex relationships between nodes and graph optimization models, and combines the transformer area determination engine to screen out abnormal data. This approach significantly improves the accuracy and efficiency of error detection.

[0016] Preferably, the energy nodes include user energy meters and transformers; the node attributes include basic energy meter attributes, dynamic energy meter attribute attributes, basic transformer attributes, and dynamic transformer attribute data. Based on the node attributes of the energy nodes and the edge relationships between multiple energy nodes, an energy meter graph model is established, including:

[0017] Define the first location feature of the user's electricity meter based on the basic attributes of the electricity meter;

[0018] The second position characteristic of a transformer is defined based on its basic properties.

[0019] Establish the network topology between the first and second positional features to obtain the edge relationship model;

[0020] Based on the dynamic characteristic attributes of electricity meters and the dynamic characteristic data of transformers, the characteristics of electricity transmission in the edge relationship model are determined, and the electricity meter graph model is obtained.

[0021] Beneficial effects: The method of the present invention can accurately depict the location association and energy interaction relationship between the electricity meter and the transformer in the power grid, effectively integrate the static topology and dynamic operation data, and provide structured and dynamic graph data support for power grid status monitoring, fault tracing and energy efficiency analysis.

[0022] Preferably, the execution steps built into the area determination engine are as follows:

[0023] The testing area is divided from the electricity meter diagram model;

[0024] Obtain the power supply to the area to be tested. If the number of consecutive days with the same power supply is greater than the threshold for the first day, the area is determined to be undetectable.

[0025] The line loss power of the area to be tested is obtained. If the number of consecutive days with the same line loss power is greater than the threshold of the second day, the area is determined to be an undetectable area.

[0026] Obtain the average load rate of the area to be tested. If the average load rate is less than the load threshold, the area is determined to be undetectable.

[0027] Beneficial effects: Through the above technical solution, the method of the present invention can accurately divide the area to be tested from the electricity meter diagram model, effectively avoid the investment of ineffective monitoring resources, improve the detection efficiency, and ensure that the subsequent analysis process relies on high-quality data, providing reliable evidence for power grid operation status assessment and anomaly diagnosis.

[0028] Preferably, after selecting detectable transformer areas from the electricity meter diagram model, the method of the present invention further includes:

[0029] Extract power data from the detectable transformer area;

[0030] Redundant / abnormal data is removed and missing data is filled in for power data.

[0031] Beneficial effects: In the preparatory stage of power data processing, by removing redundant / abnormal data and filling in missing data, we can reduce data pollution caused by duplicate or abnormal data and make up for the interruption / error problem caused by missing data.

[0032] Preferably, the power data includes the power loss of the distribution area and the line loss rate.

[0033] Preferably, redundant / abnormal data is removed from the power data, including:

[0034] If both the line loss and line loss rate of the transformer area exceed the corresponding preset range, the power data of the detectable transformer area for that day will be removed and the number of days for recording power data will be reset.

[0035] Preferably, missing data imputation for power data includes:

[0036] If the power data of any power node in the detectable distribution area is missing;

[0037] Search for similar nodes of power nodes with missing data and extract the filling data corresponding to the similar nodes;

[0038] Interpolation or K-nearest neighbor methods are used to fill in the missing data in the power nodes.

[0039] Preferably, the process of removing redundant / abnormal data from power data also includes:

[0040] Data on line loss in the distribution area, average user power consumption, and number of missing user power consumption were statistically analyzed, and the statistical results were summarized.

[0041] Based on the statistical results, users with zero or low power consumption were identified, and their power data were removed.

[0042] Preferably, the energy conservation equations are substituted into a pre-defined application graph optimization model and graph neural network algorithm to solve for the operating errors of each energy meter in the detectable distribution area, including:

[0043] Substitute the power grid topology diagram and energy conservation equations of the detectable transformer area into the preset application graph optimization model, and define the objective function and constraints of the graph neural network algorithm.

[0044] The training set is extracted from the energy conservation equations and input into a preset graph neural network algorithm to solve for the operating error of each electricity meter in the detectable distribution area.

[0045] Preferably, the constraints include:

[0046]

[0047] In the formula, The first one in the energy meter operating error estimation vector One element, The first one in the energy meter operating error estimation vector One element; Indicates from the first From the first power node to the second The set of paths for each power node; Representing a path Length, Representing a path Weight parameters; Indicates the first The first power node and the first The correlation threshold between power nodes.

[0048] Beneficial effects: The above constraints can guide the graph neural network algorithm to estimate the operating error of the electricity meter more accurately, thereby improving the accuracy and reliability of the estimation.

[0049] Secondly, the present invention also discloses a low-voltage distribution area energy meter operation error detection system based on graph calculation, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the low-voltage distribution area energy meter operation error detection method based on graph calculation described in the first aspect is implemented.

[0050] The beneficial effects of this invention are as follows:

[0051] (1) Compared with the prior art, the method of the present invention mainly utilizes graph neural networks to mine the complex relationships between nodes and graph optimization models, and combines the area judgment engine to filter out abnormal data, which has higher error detection accuracy and error detection efficiency.

[0052] (2) Compared with the prior art, the method of the present invention has a built-in transformer area determination engine, which can automatically determine undetectable transformer areas as the basis for filtering out abnormal data, so as to reduce the pollution of the data source and improve the accuracy of subsequent algorithm calculations.

[0053] (3) Compared with the prior art, the method of the present invention is applicable to various power grid structures and data environments with different sampling accuracies, and has good generalization ability. Attached Figure Description

[0054] Figure 1 This is a flowchart of the low-voltage distribution area energy meter operation error detection method based on graph calculation in Embodiment 1 of the present invention;

[0055] Figure 2 This is a network topology diagram of the low-voltage distribution area in Embodiment 1 of the present invention;

[0056] Figure 3 This is an information display board diagram showing the operating error of the low-voltage distribution area energy meter in Embodiment 1 of the present invention;

[0057] Figure 4 This is a schematic diagram of the low-voltage distribution area energy meter operation error detection system based on graph calculation in Embodiment 2 of the present invention. Detailed Implementation

[0058] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0059] This embodiment discloses a method and system for detecting operational errors of low-voltage distribution area energy meters based on graph calculation, which is used to solve the technical problems of low accuracy and efficiency of error inspection in the prior art. The specific implementation of the present invention will be described in detail below with reference to the accompanying drawings.

[0060] Example 1

[0061] like Figure 1 As shown, this embodiment discloses a method for detecting the operating error of a low-voltage distribution area energy meter based on graph calculation, including:

[0062] S10: Obtain the power nodes connected to the power grid.

[0063] In this embodiment, the aforementioned power grid mainly refers to power networks below 10kV, which typically use transformers for voltage transformation when connecting to users. These users include residential and commercial users. The transformer can be a 10kV / 0.4kV three-phase isolation transformer. Residential or commercial users can choose one or more phases from the three-phase isolation transformer as their power source. All of these transformers are power nodes connected to the power grid, and their attributes can be collected through current transformers (CTs), voltage transformers (PTs), and positioning and communication devices.

[0064] In contrast, such as Figure 2 As shown, on the secondary side of the aforementioned transformer, a main energy meter and multiple branch energy meters are typically connected to monitor the user's energy consumption. The method for collecting the node attributes of the energy meters is similar to that of the transformer and will not be repeated here. It should be noted that the aforementioned energy meters also serve as energy nodes connected to the power grid.

[0065] S20: Based on the node attributes of power nodes and the edge relationships between multiple power nodes, establish a power meter graph model.

[0066] It should be explained that node attributes include basic attributes of the electricity meter, dynamic characteristic attributes of the electricity meter, basic attributes of the transformer, and dynamic characteristic data of the transformer. Basic attributes of the electricity meter at least include the user account number, address coordinates, and electricity usage type (residential / commercial). Dynamic characteristic attributes of the electricity meter at least include real-time current, voltage, power, active energy, and historical electricity consumption. Basic attributes of the transformer at least include the transformer substation number and address coordinates. Dynamic characteristic data of the transformer at least includes real-time current, voltage, power, active energy, and historical power supply.

[0067] Step S20 above includes:

[0068] S21: Define the first position feature of the user's electricity meter based on the basic attributes of the electricity meter.

[0069] Specifically, the user account number, address coordinates, and electricity usage type are digitized. The user account number is generally a unique numerical identifier; the address coordinates can be normalized using digital encoding; and the electricity usage type can be defined as 0 for residential electricity and 1 for commercial electricity. After the above data is encoded and normalized, a feature vector can be formed as the first location feature mentioned above.

[0070] S22: Define the second position characteristics of the transformer based on the basic properties of the transformer.

[0071] It should be noted that the principle is the same as that for generating feature vectors in step S21 above, and will not be repeated here. The difference is that the basic attributes of a transformer do not include the nature of electricity consumption, and the secondary side of a single transformer is generally connected to multiple electricity meters, forming a relatively independent distribution area.

[0072] In this embodiment, the number of transformers is used to determine the specific number of distribution zones in the low-voltage power grid.

[0073] S23: Establish the network topology between the first positional feature and the second positional feature to obtain the edge relationship model.

[0074] Specifically, by establishing the network topology between the transformer and multiple electricity meters, a boundary relationship model is obtained, which is used to describe the power grid topology relationship of each distribution area.

[0075] S24: Based on the dynamic characteristic attributes of the electricity meter and the dynamic characteristic data of the transformer, determine the power transmission characteristics in the edge relationship model to obtain the electricity meter graph model.

[0076] It's important to clarify that simply establishing an edge relationship model only reveals the topological connections between energy nodes, not their energy transmission relationships, such as current transmission direction, power supply time, and / or node power consumption. Therefore, by assigning the dynamic characteristic attributes of the energy meter and the dynamic characteristic data of the transformer to the corresponding energy nodes in the edge relationship model, a graph model of the energy meter can be obtained. At the data processing level, this can be achieved by extending the length of the energy node feature vectors.

[0077] Combining steps S21-S24 above, the expression for the above electricity meter diagram model can be obtained as follows:

[0078]

[0079] in, Represents a diagram model of an electricity meter; Represents the set of electrical energy nodes; This represents the set of edges, which is calculated based on the edge relationship model. This represents the node adjacency matrix, used to describe the connection relationships between power nodes; This represents the dynamic characteristic data of the power nodes.

[0080] Through the above steps S21-S24, a graph model of the electricity meter is constructed to accurately depict the location association and energy interaction relationship between the electricity meter and the transformer in the power grid, providing dynamic and reliable graph data support for the subsequent data analysis process.

[0081] S30: Based on the preset transformer area determination engine, detectable transformer areas are selected from the electricity meter diagram model.

[0082] The execution steps built into the aforementioned district determination engine are as follows:

[0083] S31: Divide the testing area from the electricity meter diagram model.

[0084] In this embodiment, the division of the testing area can be based on the area number.

[0085] S32: Obtain the power supply to the area to be tested. If the number of consecutive days with the same power supply is greater than the threshold of the first day, it is determined to be an undetectable area.

[0086] Specifically, first, obtain the power supply data for any testing area for M days. If the power supply data for the testing area remains unchanged from day one to day M, then the testing area is designated as an untestable area, thereby filtering out worthless data. Otherwise, proceed to step S33.

[0087] S33: Obtain the line loss power of the area to be detected. If the number of consecutive days with the same line loss power is greater than the threshold of the second day, the area is determined to be an undetectable area.

[0088] It should be explained that the judgment logic of step S32 is similar to that of step S33, the difference being that its judgment criteria are line loss power and the second day's number of days threshold. If the judgment result of step S33 is negative, then step S34 is executed.

[0089] S34: Obtain the average load rate of the area to be tested. If the average load rate is less than the load threshold, it is determined to be an undetectable area.

[0090] Specifically, the load rates of the monitoring area over multiple hours or days are summed and averaged to obtain the average load rate. If the average load rate is less than the load threshold, the monitoring area is deemed undetectable and is removed. Otherwise, it indicates that the monitoring data within the monitoring area are all valid data.

[0091] Through the above steps S31-S34, the method of this embodiment accurately removes undetectable transformer areas from the electricity meter diagram model to avoid the input of invalid monitoring data. This method can reduce the actual amount of data processing and avoid the problem of inaccurate monitoring results due to data pollution.

[0092] Furthermore, after step S30 above, the method of this embodiment may further include:

[0093] S100: Extracts power data from the detectable transformer area.

[0094] The aforementioned power data includes the line loss amount and line loss rate of the transformer area, and may also include some dynamic characteristic data of the detectable transformer area.

[0095] S200: Removes redundant / abnormal data and fills in missing data in power data.

[0096] Through the above steps S100-S200, the method of this embodiment overcomes the problem of data pollution caused by data duplication or data abnormality by removing redundant / abnormal data and filling in missing data in the power data, and can also overcome the problem of calculation interruption or calculation error caused by missing data.

[0097] More specifically, step S200, which involves removing redundant / abnormal data from the power data, may include:

[0098] S201: If the line loss power and line loss rate of the transformer area both exceed the corresponding preset range, the power data of the detectable transformer area for that day will be removed and the number of days for recording power data will be reset.

[0099] By taking the above step S201, the interference of abnormal data on subsequent analysis and judgment is effectively avoided, and the quality and reliability of the data are improved.

[0100] S202: Collect statistical data on line loss power in the distribution area, average power consumption per user, and number of missing power consumption per user, and summarize the statistical results.

[0101] S203: Based on the statistical results, identify users with zero or low power consumption and remove the power data of users with zero or low power consumption.

[0102] It should be explained that the above statistical results include the quartiles, maximum and minimum values ​​of the power loss in the distribution area. The identification logic of step S203 is as follows: if any one of the statistical results of the distribution area power loss, the average power consumption of users, and the power consumption of users is 0, then the user is determined to be a user with zero power consumption; if any one of the statistical results of the distribution area power loss, the average power consumption of users, and the power consumption of users is lower than the statistical threshold, then the user is determined to be a user with low power consumption.

[0103] Regarding the aforementioned step S200, which involves filling in missing data in the power data, the method in this embodiment includes:

[0104] S204: If power data for any power node in the detectable distribution area is missing.

[0105] It should be noted that, unlike power data with a value of 0, missing power data refers to the status of the data being "null".

[0106] S205: Search for similar nodes of power nodes with missing data and extract the filling data corresponding to the similar nodes.

[0107] For example, if the power node is a 1kV / 0.4kV three-phase isolation transformer, the graph is searched to see if there is a 1kV / 0.4kV three-phase isolation transformer or a transformer with similar performance in other detectable areas. If so, the power data of the similar node is extracted as the aforementioned filler data.

[0108] S206: Use interpolation or K-nearest neighbor method to fill in the missing data in the power nodes.

[0109] It should be explained that the above data missing includes point missing and line missing. For point missing, the filling data from step S205 can be directly filled into the "null" position. For line missing, that is, multiple consecutive locations are missing, interpolation or K-nearest neighbor method is required to fill in the data.

[0110] Through the above steps S204-S206, the method of this embodiment can automatically complete the missing data of the power node, ensure the integrity of the data, and avoid abnormal interruption of the data operation process.

[0111] S40: Based on the law of conservation of energy, construct a set of energy conservation equations for the detectable transformer area.

[0112] In this embodiment, the basic definition of the above-mentioned law of conservation of energy is that the electricity supplied to the transformer area is equal to the sum of the electricity consumption of all users under the transformer area and the line loss.

[0113] S50: Substitute the energy conservation equations into the preset application graph optimization model and graph neural network algorithm to solve for the operating error of each energy meter in the detectable distribution area.

[0114] Specifically, step S50 includes:

[0115] S51: Substitute the power grid topology diagram and energy conservation equations of the detectable transformer area into the preset application graph optimization model, and define the objective function and constraints of the graph neural network algorithm.

[0116] More specifically, the objective function described above is:

[0117]

[0118] In the formula, Represent the objective function; This represents the estimated vector of the operating error of the electricity meter; Represents the graph Laplace matrix; This represents the transpose symbol.

[0119] It should be noted that the above power grid topology diagram is mathematically represented using the electricity meter diagram model G, and the above graph Laplace matrix can be obtained through data conversion.

[0120] By designing the above objective function, on the one hand, it can improve the accuracy of energy meter operation error estimation by integrating multiple factors and promptly detect and correct measurement deviations; on the other hand, the above objective function enhances the adaptability of the graph neural network algorithm to power grids with different topologies and operating states.

[0121] More specifically, the constraints include the energy meter error range constraint, energy conservation constraint, and topology consistency constraint. The energy meter error range constraint is set based on the allowable error of the energy meter itself. The topology consistency constraint is expressed as follows:

[0122]

[0123] In the formula, The first one in the energy meter operating error estimation vector One element, The first one in the energy meter operating error estimation vector One element; Indicates from the first From the first power node to the second The set of paths for each power node; Representing a path Length; Representing a path The weight parameters can be based on the path. Measured by reliability; Indicates the first The first power node and the first The correlation threshold between individual power nodes is mainly determined by historical data analysis.

[0124] Among them, for path Regarding the confirmation of path length, it should be noted that in network topology, a path consists of a series of interconnected edges, and the path length can be obtained by aggregating some metric of all edges on the path. Specifically, in this embodiment, the path... The formula for calculating the length is:

[0125]

[0126] In the formula, Includes edges , Indicates the first Power transmission time of the strip edge, Representing a path The number of edges.

[0127] The above topology consistency constraint design fully considers the case of error propagation along multiple paths. With the help of path set, length and weight parameters, it accurately characterizes the error propagation characteristics, making the error estimation more in line with the actual power grid.

[0128] S52: Extract the training set from the energy conservation equations, input the training set into the preset graph neural network algorithm, and solve for the operating error of each electricity meter in the detectable distribution area.

[0129] It should be noted that graph neural network algorithms are a type of neural network model specifically designed for processing graph-structured data.

[0130] Through steps S10-S50 described above, the method of this embodiment eliminates manual inspection, achieving remote, online, and batch intelligent processing. Furthermore, this method combines a transformer area determination engine to filter out abnormal data and relies on a graph neural network algorithm for error calculation, resulting in higher error detection accuracy and efficiency. In addition, this method is applicable to various power grid structures and data environments with different sampling accuracies, demonstrating good versatility.

[0131] Finally, the operating errors of each of the above-mentioned electricity meters and related data for each distribution area will be summarized into a file such as... Figure 3 The data analysis system dashboard shown is designed to support visual analysis for power industry professionals. It should be noted that some data involves confidential information. Figure 3 The data displayed in the document has been blurred accordingly.

[0132] Example 2

[0133] like Figure 4 As shown, this embodiment discloses a low-voltage distribution area energy meter operation error detection system based on graph computing, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the low-voltage distribution area energy meter operation error detection method based on graph computing described in Embodiment 1 is implemented.

Claims

1. A method for detecting operational errors of low-voltage distribution area energy meters based on graph calculation, characterized in that, include: Obtain the power nodes connected to the power grid; Based on the node attributes of the power nodes and the edge relationships between multiple power nodes, an energy meter graph model is established. Based on the preset transformer area determination engine, detectable transformer areas are selected from the electricity meter diagram model; Based on the law of conservation of energy, a set of energy conservation equations is constructed for the detectable area; Substituting the energy conservation equations into a preset application graph optimization model and graph neural network algorithm, the operating error of each energy meter in the detectable transformer area is obtained; Substituting the energy conservation equations into a preset application graph optimization model and graph neural network algorithm, the operating errors of each energy meter in the detectable distribution area are obtained, including: Substitute the power grid topology diagram of the detectable transformer area and the energy conservation equations into the preset application graph optimization model, and define the objective function and constraints of the graph neural network algorithm. A training set is extracted from the energy conservation equations, and the training set is input into a preset graph neural network algorithm to solve for the operating error of each electricity meter in the detectable area. The constraints include: In the formula, The first one in the energy meter operating error estimation vector One element, The first one in the energy meter operating error estimation vector One element; Indicates from the first From the first power node to the second The set of paths for each power node; Representing a path Length, Representing a path Weight parameters; Indicates the first The first power node and the first The correlation threshold between power nodes.

2. The method for detecting operating errors of low-voltage distribution area energy meters based on graph calculation according to claim 1, characterized in that, The energy nodes include user energy meters and transformers; the node attributes include basic energy meter attributes, dynamic energy meter attribute attributes, basic transformer attributes, and dynamic transformer attribute data. Based on the node attributes of the energy nodes and the edge relationships between multiple energy nodes, an energy meter graph model is established, including: The first location feature of the user's electricity meter is defined based on the basic attributes of the electricity meter; the second location feature of the transformer is defined based on the basic attributes of the transformer. A network topology is established between the first location feature and the second location feature to obtain an edge relationship model; based on the dynamic feature attributes of the electricity meter and the dynamic feature data of the transformer, the power transmission features in the edge relationship model are determined to obtain the electricity meter graph model.

3. The method for detecting operating errors of low-voltage distribution area energy meters based on graph calculation according to claim 1, characterized in that, The execution steps built into the transformer area determination engine are as follows: The areas to be tested are divided from the energy meter diagram model; The power supply to the area to be tested is obtained. If the number of consecutive days with the same power supply is greater than the threshold of the first day, it is determined to be an undetectable area. The line loss power of the area to be tested is obtained. If the number of consecutive days with the same line loss power is greater than the threshold of the second day, the area is determined to be an undetectable area. The average load rate of the area to be tested is obtained. If the average load rate is less than the load threshold, it is determined to be an undetectable area.

4. The method for detecting operating errors of low-voltage distribution area energy meters based on graph calculation according to claim 1, characterized in that, After selecting detectable transformer areas from the electricity meter diagram model, the method further includes: Extract power data from the detectable transformer area; Redundant / abnormal data is removed and missing data is filled in the power data.

5. The method for detecting operating errors of low-voltage distribution area energy meters based on graph calculation according to claim 4, characterized in that, The power data includes the power loss of the transformer area and the line loss rate.

6. The method for detecting operating errors of low-voltage distribution area energy meters based on graph calculation according to claim 5, characterized in that, Redundant / abnormal data removal from the power data includes: If both the line loss power and the line loss rate of the transformer area exceed the corresponding preset range, the power data of the detectable transformer area for that day will be removed and the number of days for recording the power data will be reset.

7. The method for detecting operating errors of low-voltage distribution area energy meters based on graph calculation according to claim 5, characterized in that, The missing data imputation for the power data includes: If the power data of any power node in the detectable distribution area is missing; Search for similar nodes of power nodes with missing data, and extract the filling data corresponding to the similar nodes; The missing data is filled into the power nodes using interpolation or K-nearest neighbor methods.

8. A low-voltage distribution area energy meter operation error detection system based on graph calculation, characterized in that, It includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the method for detecting the operating error of a low-voltage distribution area energy meter based on graph calculation as described in any one of claims 1-7 is implemented.

Images