A power distribution district hybrid topology identification method and device

Abstract

The application discloses a power distribution substation hybrid topology identification method and device, and the method comprises the following steps: identifying a first topology relationship by using a signal injection method, wherein the first topology relationship comprises the subordinate relationship of transformers and branch switches at all levels; extracting electrical quantity data of the transformers and the branch switches at all levels, and constructing a training feature data set; constructing a topology relationship model based on the training feature data set and the first topology relationship by using a classification algorithm; constructing a prediction feature data set by using the electrical quantity data of the electric meters and the end switches; bringing the prediction feature data set into the topology relationship model to obtain a second topology relationship, wherein the second topology relationship comprises the subordinate relationship of the electric meters and the end switches; and obtaining the topology relationship of the entire substation based on the first topology relationship and the second topology relationship. The method can improve the accuracy of topology identification and save equipment investment cost.

Description

Technical Field

[0001] This invention belongs to the field of power system and distribution Internet of Things technology, specifically relating to a method and device for identifying hybrid topology of distribution substations. Background Technology

[0002] There are two main methods for low-voltage topology identification: signal injection and data analysis. Signal injection utilizes the principle of current; by injecting a current signal into the line at the end, all low-voltage devices connected to that injection point can receive the signal. Through a time-division control strategy, all branches are eventually identified one by one. Data analysis typically involves analyzing large amounts of data from all devices within a distribution area to obtain the topology relationships of the area.

[0003] Actual research and verification revealed that the signal injection method offers high accuracy, achieving up to 100% recognition accuracy while ensuring 100% communication reliability. However, this method requires adding equipment at each critical node, resulting in high investment costs. Furthermore, some distribution area meter boxes have limited space, lacking redundancy for installing new equipment, making topology identification inaccurate down to the household level. Interference in communication also affects recognition accuracy. Using a time-sharing control strategy increases recognition time exponentially with the addition of equipment. While big data methods reduce equipment investment costs, inconsistent electrical characteristics across different distribution areas, missing key data, and data asynchrony lead to low recognition accuracy, failing to meet requirements.

[0004] Therefore, there is an urgent need to develop a fast and accurate method for identifying the relationship between power distribution equipment and network topology, so as to improve the accuracy of topology identification while reducing equipment investment. Summary of the Invention

[0005] Purpose of the invention: To address the above-mentioned problems, the present invention provides a method and apparatus for identifying hybrid topology of power distribution areas, which can quickly, accurately, and cost-effectively identify the relationship between power distribution equipment and network topology.

[0006] Technical solution: According to a first aspect of the present invention, a method for identifying hybrid topology of distribution radio stations is provided, comprising the following steps:

[0007] The first topological relationship is identified using the signal injection method. The first topological relationship includes the subordinate relationship between the transformer and the branch switches at each level.

[0008] Extract electrical quantity data from transformers and branch switches at all levels to construct a training feature dataset;

[0009] A topological relationship model is constructed based on the training feature dataset and the first topological relationship using a classification algorithm;

[0010] A predictive feature dataset is constructed using electrical quantity data from electricity meters and end switches;

[0011] The predicted feature dataset is fed into the topological relationship model to obtain the second topological relationship, which includes the subordinate relationship between the meter and the terminal switch.

[0012] The topology of the entire substation area is obtained based on the first and second topology relationships.

[0013] Furthermore, constructing the training feature dataset includes: selecting synchronously frozen electrical quantity data of n-day transformers and branch switches at all levels, extracting feature variables, and constructing a training feature dataset C1, C1 = (X, Y), where X is a feature vector constructed based on q feature attributes. Y is a vector of objective variables indicating whether two devices have a parent-child relationship.

[0014] Furthermore, the characteristic attributes consist of the synchronous minute-freezing data of any two devices with a known parent-child relationship, including eight attributes: the three-phase voltage of the parent node, the three-phase current of the parent node, the total active power of the parent node, the three-phase voltage of the child node, the three-phase current of the child node, the total active power of the child node, the three-phase voltage difference between the parent and child nodes, and time.

[0015] Furthermore, the classification algorithm includes any one of the following: random forest algorithm, extreme learning machine algorithm, K-nearest neighbor algorithm, or decision tree algorithm.

[0016] Furthermore, the method also includes: constructing a test feature dataset using electrical quantity data of transformers and branch switches at all levels; verifying the accuracy of the topology relationship model using the test feature dataset; and adjusting the topology relationship model so that the accuracy meets a preset threshold condition, so that the prediction feature dataset can be used for prediction.

[0017] Furthermore, adjusting the topological relationship model includes: adjusting the parameters of the classification algorithm to improve the model accuracy; if adjusting the parameters of the classification algorithm cannot meet the accuracy threshold requirement, then re-identifying the first topological relationship.

[0018] Furthermore, obtaining the second topological relationship includes: by statistically analyzing all the true and false prediction results between any final-stage switch and any meter, the parent-child relationship is established when the number of true prediction results is greater than the number of false prediction results; otherwise, the parent-child relationship is not established.

[0019] According to a second aspect of the present invention, a distribution area hybrid topology identification device is provided, comprising:

[0020] The first topology relationship acquisition module is used to identify the first topology relationship using a signal injection method. The first topology relationship includes the subordinate relationship between the transformer and the branch switches at each level.

[0021] The training feature dataset construction module is used to extract electrical quantity data of transformers and branch switches at all levels to construct the training feature dataset;

[0022] The topology model building module is used to construct a topology model based on the training feature dataset and the first topology using a classification algorithm;

[0023] The Predictive Feature Dataset Construction Module is used to construct a predictive feature dataset using electrical quantity data from electricity meters and end switches.

[0024] The second topology relationship acquisition module is used to input the predicted feature dataset into the topology relationship model to obtain the second topology relationship, which includes the subordinate relationship between the meter and the terminal switch.

[0025] The topology determination module is used to obtain the topology of the entire transformer area based on the first topology relationship and the second topology relationship.

[0026] According to a third aspect of the present invention, a computer device is provided, the device comprising:

[0027] One or more processors;

[0028] Memory; and

[0029] One or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, wherein when the programs are executed by the processors, they implement the steps of the distribution area hybrid topology identification method as described in the first aspect of the invention.

[0030] According to a fourth aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the distribution area hybrid topology identification method as described in the first aspect of the present invention.

[0031] Beneficial Effects: Addressing the issues of high cost and low accuracy in existing topology identification methods, this invention divides the transformer substation topology into two parts. First, a signal injection method is used to identify the first topological relationship, including the hierarchical relationships between transformers and branch switches at various levels. Based on this topological relationship, a classification algorithm is used to construct a topological relationship model based on data features. Then, the topological relationship model is used to identify the second topological relationship, including the hierarchical relationships between final-level switches and meters, ultimately obtaining the overall topological relationship of the transformer substation. This invention employs a hybrid approach of signal injection and classification algorithms for identification. The topology identification module can be deployed only at branch switches, thus saving equipment costs. By obtaining the hierarchical relationships between transformers and branch switches at various levels as the basis for training and modeling the classification prediction algorithm, the low accuracy problem caused by factors such as different electrical characteristics of different transformer substations is avoided. Attached Figure Description

[0032] Figure 1 This is a flowchart of the distribution area hybrid topology identification method according to the present invention;

[0033] Figure 2 This is a topological example diagram according to the present invention. Detailed Implementation

[0034] To address the problems of high cost and low accuracy in distribution radio station topology identification, this invention proposes a hybrid topology identification method for distribution radio stations. To provide a more comprehensive understanding of the characteristics, objectives, and advantages of this method, the technical solution of this invention will be further described below with reference to the accompanying drawings.

[0035] Observations revealed that the power distribution topology follows a pyramid structure, with more equipment at lower levels, and the largest number of meters at the very end, accounting for an average of 80%. If a signal injection method is used, additional topology identification equipment is required to inject and identify characteristic signals. Since switches and transformers belong to the power distribution system, while meters belong to the marketing system, they are under different management departments, making the installation of such equipment administratively difficult. Furthermore, according to research, the meter boxes on-site have limited space and no redundant space for installing new equipment. Deploying the topology identification module only up to the last-level switch can avoid these problems and save 80% of equipment costs. However, identifying equipment below the switch level remains a challenge. This invention proposes a hybrid topology identification method for power distribution areas, combining signal injection and classification algorithms for identification, improving accuracy while reducing equipment costs. Here, the topology identification module refers to the equipment controlling signal injection and identification. Additionally, "branch switch" is a general term for all switches in the topology, as the distribution network topology is divided into branches, and "end switch" refers to the last-level switch in the topology.

[0036] In this invention, topology identification includes identifying the topological relationships of transformers, branch switches at all levels, and meters in a low-voltage distribution area. The hybrid topology identification method includes using a signal injection method to identify the topological relationships of transformers and branch switches at all levels, and using a classification prediction algorithm to identify the topological relationships of the final-level switch and meters.

[0037] like Figure 1 As shown, a specific embodiment of the present invention is: a method for identifying hybrid topology of distribution radio stations, comprising the following steps:

[0038] Step (1): Use the signal injection method to identify the topological relationship between the transformer and each level of branch switches, thereby obtaining the parent-child dependency relationship between the devices;

[0039] Here, the topology identification module or device of the signal injection method is deployed to the final-stage switch. The parent-child relationship between the devices obtained at this time, that is, the parent-child relationship between the transformer and the branch switches at each stage, is also referred to as the first topology relationship in this invention.

[0040] Step (2): Collect synchronous frozen electrical quantity data of transformers, branch switches at all levels and meters for n days; select synchronous frozen electrical quantity data of transformers and branch switches at all levels; extract feature variables to construct a dataset; use bootstrap sampling to divide the dataset into training dataset C1 and validation dataset C2; select the last-level switch and meter to construct prediction dataset C3.

[0041] The training feature dataset C1 is: C1=(X,Y), where X is a feature vector constructed based on q feature attributes. Y is a vector of target variables indicating whether two devices have a parent-child relationship. Specifically, the feature attributes consist of the synchronous minute-freezing data of any two devices with a known parent-child relationship in step (1), including the three-phase voltage (A, B, C) of the parent node, the three-phase current (A, B, C) of the parent node, the total active power of the parent node, the three-phase voltage (A, B, C) of the child node, the three-phase current (A, B, C) of the child node, the total active power of the child node, and the three-phase voltage difference between the parent and child nodes, and time, totaling eight attributes. Y is represented by 0 and 1, where 1 indicates that the parent-child relationship is established and 0 indicates that the parent-child relationship is not established. The definition of the verification dataset C2 is the same as that of C1. The prediction feature dataset C3 is: C3 = (X, Y′), where Y′ is the target result vector to be predicted, represented by 0 and 1, where 1 indicates that the parent-child relationship is established and 0 indicates that the parent-child relationship is not established. The final stage switch is also called the end switch.

[0042] The following is an example of the C1 dataset:

[0043] Table 1 C1 dataset

[0044]

[0045] In the table, U1A-W1 represents the three-phase voltage, current, and energy of the parent node. U2A-W2 represents the three-phase voltage, current, and energy of the child node. UA-UC represents the voltage difference between the parent and child nodes, and T is the discrete value of the time period (0-24) in which the data is located.

[0046] The C2 and C1 datasets have the same format, only the content is different.

[0047] The following is an example of the C3 dataset:

[0048] Table 2 C3 Dataset

[0049]

[0050] In the table, U1A-W1 represents the three-phase voltage, current, and quantity of a certain final-stage switch. U2A-W2 represents the three-phase voltage, current, and quantity of a certain meter. UA-UC represents the voltage difference between the switch and the meter. T is the discrete quantity (0-24) of the time period in which the data is located.

[0051] Step (3): Use the classification prediction algorithm to train the feature dataset C1 and obtain the topological relationship model;

[0052] Based on the fact that electrical characteristics within the same transformer substation are identical, it can be determined that some basic circuit principles are also the same. For example, the voltage, current, and electrical quantity of upstream equipment must be greater than or equal to those of downstream equipment. Therefore, a model of feature data and equipment hierarchy can be established using known topological relationships. The classification and prediction algorithms include any one of the following: Random Forest (RF), Extreme Learning Machine (ELM), k-Nearest Neighbor (KNN), or Decision Tree (DT).

[0053] The modeling process is as follows:

[0054] 1) Data preprocessing: First, missing values ​​are filled by filling missing data with zeros based on timestamps. Time enumerations are converted into numerical values. Finally, the numerical values ​​are normalized to map all data to the range of 0-1.

[0055] 2) Algorithm implementation: Based on the algorithm principle, micro-applications are implemented by programming on the edge computing platform.

[0056] 3) Model Acquisition: The data processed in step 1) is used as the input of the micro-application in step 2). By setting reasonable algorithm parameters, the output of the micro-application is obtained: the logical relationship between feature attributes and prediction targets, i.e., the topological relationship model.

[0057] Step (4) involves inputting the verification feature dataset C2 into the topological relationship model identified in step (3) to obtain the prediction results and verify whether the accuracy of the model reaches the threshold. If the threshold is not reached, the process returns to step (1). Verifying the results of the signal injection method based on the classification prediction algorithm helps to identify single topological identification errors caused by factors such as communication.

[0058] Step (5): If the prediction feature dataset C3 is reached, then the prediction result set Y′ is obtained by inputting the prediction feature dataset C3 into the topological relationship model.

[0059] Step (6) Statistically determine the parent-child relationship based on the prediction result set Y′;

[0060] Since it's unknown which meters are connected to a particular switch, a dataset needs to be built for this switch and all meters at the same timestamp. This means pairwise matching of switches and meters is required. Statistical determination of parent-child relationships involves counting all true and false predictions between any final-level switch and any meter. If the number of true predictions is greater than the number of false predictions, the parent-child relationship is established; otherwise, it is not.

[0061] For the prediction result set Y′ obtained in step (6), since it is a continuous n days, there are multiple prediction results at different times for a certain switch and a certain meter. These prediction results may exist due to data loss, interference, etc., and some results may be 1 (parent-child relationship is established) and some results may be 0 (parent-child relationship is not established). In this invention, the result with the most results is taken as the final result. At this time, the parent-child relationship between the final-level switch and the meter is obtained, which is also called the second topological relationship.

[0062] Step (7) merges the topology relationships of the devices generated in steps (1) and (6) to obtain the topology relationship of the entire distribution area, namely the topology relationship of transformers, branch switches at all levels and meters.

[0063] In one instance, such as Figure 2 As shown, the distribution area has 1 transformer, 21 switches, and 103 meters. 22 topology identification modules are installed, saving 103 modules and reducing equipment investment by 82.4%. A smart distribution transformer terminal (TTU) is installed on the transformer to collect frozen data from the transformer, switches, and meters, and to implement the software of this method. The TTU selected is the NARI PDZ833 smart distribution transformer terminal. NARI PDZ976 topology identification modules are installed at the switches for signal injection topology identification and data collection from the switches. The communication method between the smart distribution terminal and branch and terminal devices is power line carrier HPLC, using the protocols "Q / GDW 1376.2-2013 Power User Electricity Information Collection System Communication Protocol Part 2: Concentrator Local Communication Module Interface Protocol" and "DL-T645 Multifunctional Energy Meter Communication Protocol". The specific steps are as follows:

[0064] (1) Obtain the transformer and switch topology:

[0065] The TTU and PDZ976 topology identification module communicate via carrier wave. The TTU employs time-division control testing, identifying each branch individually. The identification principle is as follows: the device sending the characteristic signal and the device receiving the signal are on the same branch, and the branch with more accumulated received signals is placed at the upper level, and the branch with fewer signals is placed at the lower level, thus obtaining the final hierarchical relationship. Through the association of primary and secondary devices, the transformer and switch topology relationship is ultimately obtained.

[0066] (2) Obtain the relationship model between the final-stage switch and the meter:

[0067] Develop micro-applications based on decision tree algorithms on TTU. These micro-applications have functions such as data preprocessing, parameter tuning, model training, model validation, and topology prediction. The parameter tuning function allows for model optimization by adjusting the parameter max_depth (maximum depth of the decision tree).

[0068] 1) First, start the data preprocessing module: collect feature data of all transformers, switches and meters for 7 days at a sampling frequency of 1440 samples per day at a time interval of 1 minute. According to the timestamp, select the feature data of any two transformers and switches, and construct the feature dataset based on the parent-child relationship of the two devices determined in step (1). Then, use bootstrap sampling to divide it into feature dataset C1 and test dataset C2. According to the timestamp, select the data of any two final switches and all meters to construct dataset C3. Finally, the micro-application performs data normalization processing on C1, C2 and C3.

[0069] 2) Start the micro-application model training module, input the dataset C1, and output the topology model.

[0070] 3) Start the micro-application model verification module, input the feature vector X and topology model of dataset C2, output the predicted target vector Y′ of C2, and finally count the number N of identical target vector Y and predicted target vector Y′ of C2 to obtain the model accuracy as (N / total number of Y × 100%). If the accuracy does not meet the requirements, start the parameter tuning module first to improve the model accuracy. If starting the parameter tuning module still cannot meet the accuracy threshold requirements, it means that the topology relationship generated in step (1) is inaccurate and it is necessary to return to step (1) to regenerate the transformer and switch topology relationship (this is equivalent to verifying the signal input method). Based on multiple experiments, and taking into account data loss and synchronization issues, this paper sets the accuracy threshold to 90%.

[0071] 4) Start the micro-application topology prediction module. The input is dataset C3 and the topology model, and the output is the prediction target vector Y′. By counting all true and false prediction results between any final-stage switch and any meter, if the number of true predictions is greater than the number of false predictions, the parent-child relationship is valid; otherwise, the parent-child relationship is invalid. For example, in all time-point prediction results for the final switch 4 and meter 7 in Y′, the number of true predictions is 10045, and the number of false predictions is 35. Therefore, final switch 4 is the parent node of meter 7.

[0072] (3) Obtain the topology of the entire transformer area:

[0073] Based on the topology obtained in steps (1) and (2), the parent and child nodes of each device are sorted out from bottom to top to obtain the entire topology.

[0074] The present invention also provides a distribution area hybrid topology identification device, comprising:

[0075] The first topology relationship acquisition module is used to identify the first topology relationship using a signal injection method. The first topology relationship includes the subordinate relationship between the transformer and the branch switches at each level.

[0076] The training feature dataset construction module is used to extract electrical quantity data of transformers and branch switches at all levels to construct the training feature dataset;

[0077] The topology model building module is used to construct a topology model based on the training feature dataset and the first topology using a classification algorithm;

[0078] The Predictive Feature Dataset Construction Module is used to construct a predictive feature dataset using electrical quantity data from electricity meters and end switches.

[0079] The second topology relationship acquisition module is used to input the predicted feature dataset into the topology relationship model to obtain the second topology relationship, which includes the subordinate relationship between the meter and the terminal switch.

[0080] The topology determination module is used to obtain the topology of the entire transformer area based on the first topology relationship and the second topology relationship.

[0081] Furthermore, the training feature dataset construction module constructs the training feature dataset and the test feature dataset according to the following method: Selecting the synchronously frozen electrical quantity data of the transformer and branch switches at all levels for n days, extracting feature variables, and constructing the training feature dataset C1, C1 = (X, Y), where X is a feature vector constructed based on q feature attributes. Y is a vector of objective variables indicating whether two devices have a parent-child relationship.

[0082] The characteristic attributes consist of the synchronous minute-freezing data of any two devices with a known parent-child relationship, including eight attributes: the three-phase voltage of the parent node, the three-phase current of the parent node, the total active power of the parent node, the three-phase voltage of the child node, the three-phase current of the child node, the total active power of the child node, the three-phase voltage difference between the parent and child nodes, and time.

[0083] In this embodiment of the invention, the topology relationship model building module utilizes any one of the following classification algorithms: random forest algorithm, extreme learning machine algorithm, K-nearest neighbor algorithm, or decision tree algorithm.

[0084] To further improve recognition accuracy, the hybrid topology recognition device for distribution substations also includes a topology model verification module. This module verifies the accuracy of the topology model using a verification feature dataset and adjusts the model to meet a preset threshold condition for use in predictive feature datasets. The verification feature dataset is constructed based on electrical quantity data of transformers and branch switches at various levels. Furthermore, the topology model verification module includes: a parameter tuning module for adjusting the parameters of the classification algorithm to improve model accuracy; and a model retuning module for reconstructing the feature dataset and topology model when adjusting the classification algorithm parameters fails to meet the accuracy threshold requirement. This module calls the first topology acquisition module to re-identify the first topology relationship, and the first feature dataset construction module and the topology model verification module reconstruct the feature dataset and the topology model.

[0085] The second topology relationship acquisition module obtains the second topology relationship according to the following method: count all the prediction results that are true and false between any final-stage switch and any meter. When the number of true entries is greater than the number of false entries, the parent-child relationship is established; otherwise, the parent-child relationship is not established.

[0086] It should be understood that the distribution area hybrid topology identification device in the embodiments of the present invention can realize all the technical solutions in the above method embodiments. The functions of each functional module can be specifically implemented according to the methods in the above method embodiments. The specific implementation process can be referred to the relevant descriptions in the above embodiments, which will not be repeated here.

[0087] Based on the same technical concept as the method embodiments, according to another embodiment of the present invention, a computer device is provided, the device comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the programs, when executed by the processors, implement the steps in the method embodiments.

[0088] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0089] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0090] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0091] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for identifying hybrid topology of distribution radio stations, characterized in that, The method includes the following steps: The first topological relationship is identified using the signal injection method. The first topological relationship includes the subordinate relationship between the transformer and the branch switches at each level. Extract electrical quantity data from transformers and branch switches at all levels to construct a training feature dataset; A topological relationship model is constructed based on a training feature dataset and a first topological relationship using a classification algorithm. The classification algorithm includes any one of the following: random forest algorithm, extreme learning machine algorithm, K-nearest neighbor algorithm, or decision tree algorithm. A predictive feature dataset is constructed using electrical quantity data from electricity meters and end switches; The predicted feature dataset is fed into the topological relationship model to obtain the second topological relationship, which includes the subordinate relationship between the meter and the terminal switch. Obtaining the second topological relationship includes: counting all the prediction results that are true and false between any terminal switch and any meter. When the number of true entries is greater than the number of false entries, the parent-child relationship is established; otherwise, the parent-child relationship is not established. The topology of the entire substation area is obtained based on the first and second topology relationships.

2. The method for identifying hybrid topology of distribution radio areas according to claim 1, characterized in that, Constructing the training feature dataset involves: selecting synchronously frozen electrical quantity data of transformers and branch switches at all levels over n days, extracting feature variables, and constructing a training feature dataset C1, where C1 = (X, Y), and X is a feature vector constructed based on q feature attributes. Y is a vector of objective variables indicating whether two devices have a parent-child relationship.

3. The method for identifying hybrid topology of distribution radio areas according to claim 2, characterized in that, The characteristic attributes consist of the synchronous minute-freezing data of any two devices with a known parent-child relationship, including eight attributes: the three-phase voltage of the parent node, the three-phase current of the parent node, the total active power of the parent node, the three-phase voltage of the child node, the three-phase current of the child node, the total active power of the child node, the three-phase voltage difference between the parent and child nodes, and time.

4. The method for identifying hybrid topology of distribution radio areas according to claim 1, characterized in that, The method further includes: constructing a test feature dataset using electrical quantity data of transformers and branch switches at all levels; verifying the accuracy of the topology relationship model using the test feature dataset; and adjusting the topology relationship model so that the accuracy meets a preset threshold condition, so that the prediction feature dataset can be used for prediction.

5. The method for identifying hybrid topology of distribution radio areas according to claim 4, characterized in that, Adjusting the topological relationship model includes: adjusting the parameters of the classification algorithm to improve the model accuracy; if adjusting the parameters of the classification algorithm cannot meet the accuracy threshold requirements, then the first topological relationship is re-identified.

6. A distribution area hybrid topology identification device, characterized in that, include: The first topology relationship acquisition module is used to identify the first topology relationship using a signal injection method. The first topology relationship includes the subordinate relationship between the transformer and the branch switches at each level. The training feature dataset construction module is used to extract electrical quantity data of transformers and branch switches at all levels to construct the training feature dataset; The topology model building module is used to construct a topology model based on the training feature dataset and the first topology using a classification algorithm; the classification algorithm includes any one of the following: random forest algorithm, extreme learning machine algorithm, K-nearest neighbor algorithm, or decision tree algorithm. The Predictive Feature Dataset Construction Module is used to construct a predictive feature dataset using electrical quantity data from electricity meters and end switches. The second topology relationship acquisition module is used to input the predicted feature dataset into the topology relationship model to acquire the second topology relationship, which includes the subordinate relationship between the meter and the terminal switch. The acquisition of the second topology relationship includes: counting all prediction results that are true and false between any terminal switch and any meter. When the number of true entries is greater than the number of false entries, the parent-child relationship is true; otherwise, the parent-child relationship is false. The topology determination module is used to obtain the topology of the entire transformer area based on the first topology relationship and the second topology relationship.

7. The distribution area hybrid topology identification device according to claim 6, characterized in that, The training feature dataset construction module constructs the training feature dataset and the test feature dataset according to the following method: Selecting the synchronously frozen electrical quantity data of the transformer and branch switches at all levels over n days, extracting feature variables, and constructing the training feature dataset C1, C1 = (X, Y), where X is a feature vector constructed based on q feature attributes. Y is a vector of objective variables indicating whether two devices have a parent-child relationship.

8. The distribution area hybrid topology identification device according to claim 7, characterized in that, The characteristic attributes consist of the synchronous minute-freezing data of any two devices with a known parent-child relationship, including eight attributes: the three-phase voltage of the parent node, the three-phase current of the parent node, the total active power of the parent node, the three-phase voltage of the child node, the three-phase current of the child node, the total active power of the child node, the three-phase voltage difference between the parent and child nodes, and time.

9. The distribution area hybrid topology identification device according to claim 6, characterized in that, The device also includes a topology model verification module, which is used to verify the accuracy of the topology model using a verification feature dataset and adjust the topology model so that the accuracy meets a preset threshold condition, so that the prediction feature dataset can be used for prediction. The verification feature dataset is constructed based on the electrical quantity data of the transformer and branch switches at all levels.

10. The power distribution area hybrid topology identification device according to claim 9, characterized in that, The topology relationship model verification module includes: a parameter tuning module, used to adjust the parameters of the classification algorithm to improve the model accuracy; a model remodeling module, used to call the first topology relationship acquisition module to re-identify the first topology relationship when the adjusted parameters of the classification algorithm cannot meet the accuracy threshold requirements; and a first feature dataset construction module and a topology relationship model verification module to reconstruct the feature dataset and reconstruct the topology relationship model.

11. A computer device, characterized in that, The device includes: One or more processors; Memory; and One or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, wherein when the programs are executed by the processors, they implement the steps of the distribution area hybrid topology identification method as described in any one of claims 1-5.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the distribution radio area hybrid topology identification method as described in any one of claims 1-5.

Images