A method and device for intelligently locating abnormal line loss in a distribution network

Abstract

This application provides a method and device for intelligently locating line loss anomalies in a distribution network. The method comprises: obtaining real-time operating data and topological parameters for each node in the distribution network, establishing a vertex set, and generating a node-line association matrix; upon detecting a preliminary line loss anomaly at a node, obtaining a set of associated nodes based on the association matrix; calculating the real-time line loss rate of the associated nodes and comparing it with historically normal line loss rates; and determining the root node of the anomaly based on the nodes with deviations exceeding a threshold and the power flow direction. By constructing a topological model and association matrix, this method rapidly locates the root cause of the anomaly, reduces false positives and missed detections, and enhances intelligent management.

Description

Technical Field

[0001] The present application relates to the field of line loss detection, and in particular to a method and device for intelligently locating abnormal line losses in a distribution network. Background Art

[0002] In the operation and management of distribution networks, line loss is a key indicator for measuring the economic operation level of power grids. Its management effect is directly related to the utilization efficiency of power resources and the power supply cost.

[0003] However, traditional line loss anomaly location and correction methods have significant shortcomings.

[0004] First, these methods are limited in their data processing capabilities. When faced with heterogeneous data from multiple sources, such as archives, acquisitions, and parameters, they lack effective integration and analysis tools, making it difficult to accurately filter out abnormal data from massive amounts of data. Second, when locating anomalies, traditional methods often lack precise positioning strategies and are easily affected by single data anomalies, leading to misjudgments and missed detections, making it impossible to quickly and accurately determine the specific location of line loss anomalies.

[0005] In addition, traditional methods for dealing with missing data are usually relatively simple and difficult to ensure the integrity and continuity of the data, which further affects the accuracy and reliability of line loss analysis. Summary of the Invention

[0006] The purpose of this application is to overcome the defects in the above-mentioned prior art and provide a method and device for intelligently locating abnormal line loss in a distribution network.

[0007] This application provides a method for intelligently locating abnormal line loss in a distribution network, comprising:

[0008] Acquire real-time operating data and topological parameters of each node in the distribution network, and establish a vertex set of the distribution network based on the real-time operating data and the topological parameters;

[0009] generating a node-link association matrix based on the vertex set;

[0010] detecting whether a node in the vertex set has a preliminary line loss anomaly; and when it is detected that a node in the vertex set has a preliminary line loss anomaly, acquiring a set of associated nodes directly connected to the abnormal node according to the association matrix;

[0011] Calculating the real-time line loss rate of the associated point of each associated node in the associated node set, and comparing the deviation between the real-time line loss rate of the associated point and the historical normal line loss rate corresponding to each associated node;

[0012] The root node of the line loss anomaly is determined based on the associated nodes whose deviation comparison results exceed a preset threshold and the power flow direction in the real-time operation data.

[0013] Optionally, detecting whether a node in the vertex set has a preliminary line loss anomaly includes:

[0014] Get real-time line loss data series:

[0015] The mean μ and standard deviation σ of the real-time line loss data sequence are calculated. If the real-time line loss rate of any node exceeds the range of μ-3σ to μ+3σ, the corresponding node is a preliminary line loss anomaly.

[0016] Optionally, obtain real-time operating data and topology parameters of each node in the distribution network, including:

[0017] Extract the historical line loss data series of each node in the distribution network;

[0018] Performing a stationary test on the historical line loss data sequence; if the test result indicates an unstationary sequence, performing a differential process on the sequence until a stationary sequence is obtained;

[0019] Determine the optimal time series model parameter combination based on the AIC criterion;

[0020] Establishing a prediction model using the optimal time series model parameter combination, and calculating the filling value of the missing data using the prediction model;

[0021] The filling value is updated to the historical line loss data sequence of the corresponding node.

[0022] Optionally, determining the root node of the line loss anomaly based on the associated nodes whose deviation comparison result exceeds a preset threshold and the power flow direction in the real-time operation data includes:

[0023] Screening out the nodes with preliminary abnormal line loss, and verifying the load similarity between the load curve of the nodes with preliminary abnormal line loss and the load curve of the historical normal load;

[0024] When the line loss deviation rate and the load similarity of the preliminary line loss abnormal node are both beyond the allowable range, it is determined to be the abnormal root node.

[0025] Optionally, generating a node-link association matrix based on the vertex set includes:

[0026] Dynamically updating the vertex set according to the topological parameter change;

[0027] When a change in the physical connection state of a line is detected, a new correlation matrix is ​​regenerated.

[0028] The present application also provides a distribution network line loss abnormality intelligent positioning device comprising:

[0029] An initialization module, which obtains real-time operation data and topology parameters of each node of the distribution network, and establishes a vertex set of the distribution network based on the real-time operation data and the topology parameters;

[0030] A matrix module, generating a node-line association matrix based on the vertex set;

[0031] a node module, detecting whether a node in the vertex set has a preliminary line loss anomaly, and when detecting that a node in the vertex set has a preliminary line loss anomaly, acquiring a set of associated nodes directly connected to the abnormal node according to the association matrix;

[0032] a comparison module, which calculates the real-time line loss rate of the associated point of each associated node in the associated node set, and compares the deviation between the real-time line loss rate of the associated point and the historical normal line loss rate corresponding to each associated node;

[0033] The judgment module determines the root node of the line loss anomaly based on the associated nodes whose deviation comparison results exceed a preset threshold and the power flow direction in the real-time operation data.

[0034] Optionally, detecting whether a node in the vertex set has a preliminary line loss anomaly includes:

[0035] Get real-time line loss data series:

[0036] The mean μ and standard deviation σ of the real-time line loss data sequence are calculated. If the real-time line loss rate of any node exceeds the range of μ-3σ to μ+3σ, the corresponding node is a preliminary line loss anomaly.

[0037] Optionally, the initial module obtains real-time operating data and topology parameters of each node in the distribution network, including:

[0038] Extract the historical line loss data series of each node in the distribution network;

[0039] Performing a stationary test on the historical line loss data sequence; if the test result indicates an unstationary sequence, performing a differential process on the sequence until a stationary sequence is obtained;

[0040] Determine the optimal time series model parameter combination based on the AIC criterion;

[0041] Establishing a prediction model using the optimal time series model parameter combination, and calculating the filling value of the missing data using the prediction model;

[0042] The filling value is updated to the historical line loss data sequence of the corresponding node.

[0043] Optionally, the judgment module determines the root node of the line loss anomaly based on the associated nodes whose deviation comparison result exceeds a preset threshold and the power flow direction in the real-time operation data, including:

[0044] Screening out the nodes with preliminary abnormal line loss, and verifying the load similarity between the load curve of the nodes with preliminary abnormal line loss and the load curve of the historical normal load;

[0045] When the line loss deviation rate and the load similarity of the preliminary line loss abnormal node are both beyond the allowable range, it is determined to be the abnormal root node.

[0046] Optionally, the matrix module generates a node-link association matrix based on the vertex set, comprising:

[0047] Dynamically updating the vertex set according to the topological parameter change;

[0048] When a change in the physical connection state of a line is detected, a new correlation matrix is ​​regenerated.

[0049] The beneficial effects of this application are:

[0050] The present application provides a method for intelligently locating abnormal line loss in a distribution network, comprising: obtaining real-time operating data and topological parameters of each node in the distribution network, establishing a vertex set of the distribution network based on the real-time operating data and the topological parameters; generating a node-line association matrix based on the vertex set; detecting whether a node in the vertex set has a preliminary line loss abnormality, and when a node in the vertex set has a preliminary line loss abnormality, obtaining a set of associated nodes directly connected to the abnormal node according to the association matrix; calculating the real-time line loss rate of the associated point of each associated node in the associated node set, and performing a deviation comparison between the real-time line loss rate of the associated point and the historical normal line loss rate corresponding to each associated node; determining the root node of the line loss abnormality based on the associated nodes whose deviation comparison results exceed a preset threshold and the power flow direction in the real-time operating data. The present application quickly locates the root node of the line loss abnormality by constructing a distribution network topology model and an association matrix, reduces misjudgments and missed judgments, improves positioning accuracy, ensures safe and stable operation of the power grid, and enhances the intelligent level of line loss management. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a schematic diagram of the dynamic adjustment process of the wide-area energy storage model for peak-shaving needs in this application;

[0052] Figure 2 This is a schematic diagram of the evaluation of the combined coupling coordination degree and principal component analysis method in this application. DETAILED DESCRIPTION

[0053] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it is understood that various forms of implementing the present disclosure are not limited by the embodiments set forth herein. Rather, the embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0054] This application provides a method for intelligently locating abnormal line loss in a distribution network, the implementation principle of which is as follows:

[0055] First, through intelligent devices and management systems, we collect and integrate archival, collection, and parameter data to build a data foundation. We use the Laida criterion to calculate the mean μ and standard deviation σ of the line loss data series, screen out preliminary abnormal data, and narrow the scope of investigation. We build a distribution network topology model based on graph theory, and use the correlation matrix to analyze the operating data and line loss rate of abnormal nodes and their connected nodes to accurately locate problem nodes. Using the ARIMA time series model, after stationarity testing and parameter determination, we predict and fill in missing data to ensure data integrity. We use a dual-threshold verification mechanism, combining the line loss rate threshold and load curve similarity judgment to accurately identify abnormal line loss nodes, achieve intelligent positioning and correction of distribution network line loss anomalies, and improve the level of grid operation and management.

[0056] Please refer to Figure 1 As shown, a method for intelligently locating abnormal line loss in a distribution network includes:

[0057] S101, obtaining real-time operating data and topological parameters of each node in the distribution network, and establishing a vertex set of the distribution network based on the real-time operating data and the topological parameters;

[0058] Deploy smart meters, current transformers, voltage transformers and other data acquisition devices at each node of the distribution network to periodically collect real-time data from each node at a set sampling frequency (e.g., 15 minutes / time), including voltage U, current I, active power P, reactive power Q and other collection data.

[0059] Retrieve line parameters from the power grid equipment management system, such as the conductor model's resistivity ρ, line length L, conductor cross-sectional area S, and other archival data. Obtain transformer parameters such as the rated capacity SN, transformation ratio k, no-load loss P0, and short-circuit loss Pk.

[0060] Collect user profile information, such as user type (residential, industrial, commercial, etc.), power consumption capacity C, etc. Obtain system operating parameters from the power grid dispatching automation system, such as current operating mode, load distribution ratio in each area, and other parameter data.

[0061] Integrate all types of collected data, establish a unified data storage structure, and store them in tables according to time series.

[0062] Then use the following formula to calculate the line loss power:

[0063]

[0064] in, is the line loss power, Input active power for the node, where Output active power for the node, and According to the collected voltage U, current I and other data, combined with the power calculation formula ( is the power factor).

[0065] After obtaining the line loss power, combined with other related calculation requirements (such as calculating the line loss rate, etc.), we can obtain a node line loss data sequence for screening based on the Laida criterion. , ,…, .

[0066] The line loss rate calculation formula is:

[0067]

[0068] in, is the line loss rate of the i-th node, Input active power for the i-th node, Output active power for the i-th node.

[0069] Preliminary screening of abnormal line loss data:

[0070] The Laida criterion is used to quickly screen massive amounts of line loss data, calculating the mean μ and standard deviation σ of each node's line loss data series. Each data point is then individually determined to see if it falls outside the abnormality judgment interval. Abnormal data is then marked and a preliminary abnormal data list is generated.

[0071] The collected line loss data series of each node , ,…, , calculate the mean μ of the sequence, the calculation formula is:

[0072]

[0073] in, is the mean of the line loss data series, n is the number of line loss data points, is the i-th line loss data point.

[0074] Calculate the standard deviation σ of the data series using the formula:

[0075]

[0076] in, is the standard deviation of the line loss data series, reflecting the degree of data dispersion; n is the number of line loss data points; is the i-th line loss data point; is the mean of the line loss data series.

[0077] According to the Laida criterion, the judgment interval of abnormal data is determined as ( -3 , +3 ), the line loss data will exceed this range.

[0078] It is initially judged to be abnormal data.

[0079] S102, generating a node-line association matrix based on the vertex set;

[0080] The distribution network topology model and node-line association matrix are pre-built and maintained. When preliminary abnormal data appears, the topology association analysis program is started to obtain the associated nodes and their data based on the abnormal nodes.

[0081] Calculate the line loss rate of associated nodes, compare and analyze it with historical normal data, and identify the root node of the anomaly. This can accurately determine the root node of the line loss anomaly at the network level, avoiding misjudgments caused by single data anomalies and improving the accuracy of anomaly location.

[0082] The distribution network topology model is constructed using graph theory, and the nodes in the distribution network are abstracted as vertices of the graph. .

[0083] The lines between nodes are abstracted as edges of the graph , and establish the node-line association matrix A.

[0084] in Representation node With node Is there a line connection between them? If yes, ,otherwise .

[0085] S103, when it is detected that a node in the vertex set has a preliminary line loss anomaly, obtaining a set of associated nodes directly connected to the abnormal node according to the association matrix;

[0086] When the node where the abnormal data initially screened out is located is According to the correlation matrix A, find Directly connected nodes , ,…, .

[0087] S104: Calculate the real-time line loss rate of the associated point of each associated node in the associated node set, and compare the deviation between the real-time line loss rate of the associated point and the historical normal line loss rate corresponding to each associated node;

[0088] Obtain real-time operating data (such as voltage, current, power, and line loss rate) for these nodes and their upstream and downstream nodes. Compare the line loss rates and correlations of each node under normal operating conditions.

[0089] S105: Determine the root node of the line loss anomaly based on the associated nodes whose deviation comparison results exceed a preset threshold and the power flow direction in the real-time operation data;

[0090] If the line loss rate of a certain associated node deviates greatly from the historical normal data, and its operating status changes If there is a logical correlation between the abnormality (such as abnormal power flow direction), the node is determined to be the root node causing the line loss abnormality.

[0091] Furthermore, when missing data exists in the line loss data series for each node, dynamic historical data supplementation is required. This is done by testing the historical line loss data for stationarity. If it is not stationary, differential processing is performed until it becomes stationary. By traversing different p and q values, the AIC is calculated to determine the optimal ARIMA model parameters. Based on the determined model, missing data is predicted and filled, and the data series is updated.

[0092] Historical line loss data series , ,…, Perform a stationary test. If the data is not stationary, perform a d-order difference operation to obtain a stationary sequence:

[0093]

[0094] Where Δ is the difference operator, , .

[0095] The calculation formula of the AIC criterion is:

[0096] AIC=2k-2ln(L)

[0097] Here, k = p + q + 1, where p is the autoregressive order, which determines the degree of influence of past data on current data in the model; q is the moving average order, which reflects the influence of past error terms on current data; and the addition of 1 accounts for the constant term in the model. L is the likelihood function value.

[0098] By trying different p and q values, calculating the corresponding AIC values, and selecting the p and q combination that minimizes the AIC value as the model parameters, a balance can be achieved between model complexity and goodness of fit.

[0099] The missing data are predicted using the determined ARIMA(p,q) model.

[0100] The ARIMA(p,q) model is an autoregressive integrated moving average model used to forecast missing data. p represents the number of lagged observations included in the model, and q represents the number of lagged forecast error terms included in the model.

[0101] For missing data points y, the missing data is predicted according to the prediction formula of the model and filled into the data sequence to update the data sequence.

[0102] This application adopts a dual-threshold verification mechanism, combining the line loss rate threshold and load curve similarity judgment to accurately identify abnormal line loss nodes, including:

[0103] 1. Line loss rate threshold verification:

[0104] Reasonable thresholds are set based on the voltage level of the distribution network, historical operation data and industry standards, such as: high-voltage distribution network Set to ±8%, low voltage distribution network Set to ±15%.

[0105] like > , then the line loss rate of the node is determined to be abnormal and marked as a "suspected abnormal node".

[0106] in, Line loss rate Absolute value, not affected by positive or negative direction (loss or reverse power), i is the node number, and th represents the threshold.

[0107] 2. Load curve similarity verification:

[0108] Calculate the distance matrix: Construct an m×n distance matrix D, where m and n are the number of data points of the real-time load curve and the historical load curve, respectively, and D(i, j) represents the real-time load curve. and the normal load curve during the same period in history The Euclidean distance between them is:

[0109]

[0110] Path planning: Find a path P from the upper left corner to the lower right corner of the matrix so that the sum of the elements on the path is minimized. The path P consists of a series of matrix elements. ,in , and the boundary conditions must be met =(1,1), =(m,n), and the continuity condition | ∣≤1.

[0111] Calculate the cumulative distance matrix: Define the cumulative distance matrix S, S(i, j) represents the minimum cumulative distance from the upper left corner of the matrix to the (i, j) position, and the calculation formula is:

[0112]

[0113] Calculate similarity: The final DTW distance DTW=S(m,n). DTW=S(m,n) is the core calculation formula for evaluating the similarity of load curves, where m is the number of data points of the real-time load curve and n is the number of data points of the historical normal load curve.

[0114] Load curve similarity:

[0115]

[0116] Similarity determination: Set the load curve similarity threshold ,like < , then the load curve similarity is judged to be abnormal.

[0117] When the line loss deviation rate of the preliminary line loss abnormal node and the load similarity of the node are both beyond the allowable range and are determined to be abnormal, the node is determined to be the abnormal root node.

[0118] Please refer to Figure 2 As shown, the present application also provides an intelligent positioning device for abnormal line loss in a distribution network, comprising:

[0119] Initial module 201, obtaining real-time operating data and topological parameters of each node in the distribution network, and establishing a vertex set of the distribution network based on the real-time operating data and the topological parameters;

[0120] A matrix module 202 generates a node-link association matrix based on the vertex set;

[0121] The node module 203 detects whether a node in the vertex set has a preliminary line loss anomaly. When it is detected that a node in the vertex set has a preliminary line loss anomaly, a set of associated nodes directly connected to the abnormal node is obtained according to the association matrix.

[0122] The comparison module 204 calculates the real-time line loss rate of the associated point of each associated node in the associated node set, and compares the deviation between the real-time line loss rate of the associated point and the historical normal line loss rate corresponding to each associated node;

[0123] The judgment module 205 determines the root node of the line loss anomaly based on the associated nodes whose deviation comparison results exceed a preset threshold and the power flow in the real-time operation data.

[0124] Furthermore, detecting whether a node in the vertex set has a preliminary line loss anomaly includes:

[0125] Get real-time line loss data series:

[0126] The mean μ and standard deviation σ of the real-time line loss data sequence are calculated. If the real-time line loss rate of any node exceeds the range of μ-3σ to μ+3σ, the corresponding node is a preliminary line loss anomaly.

[0127] Furthermore, the initial module obtains real-time operating data and topology parameters of each node in the distribution network, including:

[0128] Extract the historical line loss data series of each node in the distribution network;

[0129] Performing a stationary test on the historical line loss data sequence; if the test result indicates an unstationary sequence, performing a differential process on the sequence until a stationary sequence is obtained;

[0130] Determine the optimal time series model parameter combination based on the AIC criterion;

[0131] Establishing a prediction model using the optimal time series model parameter combination, and calculating the filling value of the missing data using the prediction model;

[0132] The filling value is updated to the historical line loss data sequence of the corresponding node.

[0133] Furthermore, the judgment module determines the root node of the line loss anomaly based on the associated nodes whose deviation comparison result exceeds a preset threshold and the power flow direction in the real-time operation data, including:

[0134] Screening out the nodes with preliminary abnormal line loss, and verifying the load similarity between the load curve of the nodes with preliminary abnormal line loss and the load curve of the historical normal load;

[0135] When the line loss deviation rate and the load similarity of the preliminary line loss abnormal node are both beyond the allowable range, it is determined to be the abnormal root node.

[0136] Furthermore, the matrix module generates a node-line association matrix based on the vertex set, including:

[0137] Dynamically updating the vertex set according to the topological parameter change;

[0138] When a change in the physical connection state of a line is detected, a new correlation matrix is ​​regenerated.

[0139] The above description of the embodiments is intended to facilitate understanding and application of this application by those skilled in the art. It will be readily apparent to those skilled in the art that various modifications to the above embodiments can be made, and the general principles described herein can be applied to other embodiments without requiring creative effort. Therefore, this application is not limited to the above embodiments. Any improvements or modifications made to this application by those skilled in the art based on the disclosure of this application should fall within the scope of protection of this application.

Claims

1. A method for intelligently locating abnormal line loss in a distribution network, characterized in that: include: Acquire real-time operating data and topological parameters of each node in the distribution network, and establish a vertex set of the distribution network based on the real-time operating data and the topological parameters; generating a node-link association matrix based on the vertex set; detecting whether a node in the vertex set has a preliminary line loss anomaly; and when it is detected that a node in the vertex set has a preliminary line loss anomaly, acquiring a set of associated nodes directly connected to the abnormal node according to the association matrix; Calculating the real-time line loss rate of the associated point of each associated node in the associated node set, and comparing the deviation between the real-time line loss rate of the associated point and the historical normal line loss rate corresponding to each associated node; The root node of the line loss anomaly is determined based on the associated nodes whose deviation comparison results exceed a preset threshold and the power flow direction of the real-time operation data.

2. The method for intelligently locating abnormal line loss in a distribution network according to claim 1, characterized in that: Detecting whether a node in the vertex set has a preliminary line loss anomaly includes: Get real-time line loss data series: The mean μ and standard deviation σ of the real-time line loss data sequence are calculated. If the real-time line loss rate of any node exceeds the range of μ-3σ to μ+3σ, the corresponding node is a preliminary line loss anomaly.

3. The method for intelligently locating abnormal line loss in a distribution network according to claim 1, characterized in that: Obtain real-time operating data and topology parameters of each node in the distribution network, including: Extract the historical line loss data series of each node in the distribution network; Performing a stationarity test on the historical line loss data sequence; if the test result indicates an unstationary sequence, performing a differential process on the historical line loss data sequence until a stationary sequence is obtained; Determine the optimal time series model parameter combination based on the AIC criterion; Establishing a prediction model using the optimal time series model parameter combination, and calculating the filling value of the missing data using the prediction model; The filling value is updated to the historical line loss data sequence of the corresponding node.

4. The method for intelligently locating abnormal line loss in a distribution network according to claim 1, characterized in that: Determining the root node of the line loss anomaly based on the associated nodes whose deviation comparison results exceed a preset threshold and the power flow direction in the real-time operation data includes: Screening out the nodes with preliminary abnormal line loss, and verifying the load similarity between the load curve of the nodes with preliminary abnormal line loss and the load curve of the historical normal load; When the line loss deviation rate and the load similarity of the preliminary line loss abnormal node are both beyond the allowable range, it is determined to be the abnormal root node.

5. The method for intelligently locating abnormal line loss in a distribution network according to claim 1, characterized in that: Generating a node-link association matrix based on the vertex set includes: Dynamically updating the vertex set according to the topological parameter change; When a change in the physical connection state of a line is detected, a new correlation matrix is ​​regenerated.

6. An intelligent positioning device for abnormal line loss in a distribution network, characterized in that: include: An initialization module, which obtains real-time operation data and topology parameters of each node of the distribution network, and establishes a vertex set of the distribution network based on the real-time operation data and the topology parameters; A matrix module, generating a node-line association matrix based on the vertex set; a node module, detecting whether a node in the vertex set has a preliminary line loss anomaly, and when detecting that a node in the vertex set has a preliminary line loss anomaly, acquiring a set of associated nodes directly connected to the abnormal node according to the association matrix; a comparison module, which calculates the real-time line loss rate of the associated point of each associated node in the associated node set, and compares the deviation between the real-time line loss rate of the associated point and the historical normal line loss rate corresponding to each associated node; The judgment module determines the root node of the line loss anomaly based on the associated nodes whose deviation comparison results exceed a preset threshold and the power flow direction in the real-time operation data.

7. The intelligent positioning device for abnormal line loss in a distribution network according to claim 6, characterized in that: Detecting whether a node in the vertex set has a preliminary line loss anomaly includes: Get real-time line loss data series: The mean μ and standard deviation σ of the real-time line loss data sequence are calculated. If the real-time line loss rate of any node exceeds the range of μ-3σ to μ+3σ, the corresponding node is a preliminary line loss anomaly.

8. The intelligent positioning device for abnormal line loss in a distribution network according to claim 6, characterized in that: The initial module obtains real-time operating data and topology parameters of each node in the distribution network, including: Extract the historical line loss data series of each node in the distribution network; Performing a stationary test on the historical line loss data sequence; if the test result indicates an unstationary sequence, performing a differential process on the sequence until a stationary sequence is obtained; Determine the optimal time series model parameter combination based on the AIC criterion; Establishing a prediction model using the optimal time series model parameter combination, and calculating the filling value of the missing data using the prediction model; The filling value is updated to the historical line loss data sequence of the corresponding node.

9. The intelligent positioning device for abnormal line loss in a distribution network according to claim 6, characterized in that: The judgment module determines the root node of the line loss anomaly based on the associated nodes whose deviation comparison result exceeds a preset threshold and the power flow direction in the real-time operation data, including: Screening out the nodes with preliminary abnormal line loss, and verifying the load similarity between the load curve of the nodes with preliminary abnormal line loss and the load curve of the historical normal load; When the line loss deviation rate and the load similarity of the preliminary line loss abnormal node are both beyond the allowable range, it is determined to be the abnormal root node.

10. The intelligent positioning device for abnormal line loss in a distribution network according to claim 6, characterized in that: The matrix module generates a node-line association matrix based on the vertex set, including: Dynamically updating the vertex set according to the topological parameter change; When a change in the physical connection state of a line is detected, a new correlation matrix is ​​regenerated.

Images