Distribution network line transfer identification method, system and device based on CatBoost

Abstract

The present invention discloses a method, system and device for identifying power transfer of distribution network lines based on CatBoost, and the method includes: obtaining required data; obtaining a corresponding topological relationship set of a tie switch, a key topological path, a minimum upstream segment of the tie switch, and a pair of opposite-end lines connected to the tie switch through a key topological path; extracting characteristic indicators, applying a distribution transformer power transfer identification model based on CatBoost ensemble learning online, and outputting whether a power transfer occurs in the distribution transformer in the minimum upstream segment of the tie switch at the current moment; calculating the voting probability of power transfer of the distribution transformer in each minimum upstream segment of the tie switch, and determining whether a power transfer occurs in the line to be diagnosed at the current moment; constructing a line power transfer area positioning identification model, and outputting a substitute supply line set and a power transfer area distribution transformer set. The present invention can quickly and accurately identify whether a power transfer occurs in a line and locate a specific power transfer area, providing technical support for line power transfer identification and dynamic topology identification.

Description

Technical Field

[0001] The present invention belongs to the field of power distribution networks, and relates to a line transfer identification technology, and specifically to a distribution network line transfer identification method, system and device based on CatBoost. Background Art

[0002] In recent years, with the continuous growth of my country's economy, all walks of life have become increasingly dependent on power supply. The distribution network system needs to continuously adjust and optimize its operation mode to cope with various possible faults and load changes, so as to ensure the reliability of power supply and improve the economic efficiency of power grid operation. When the distribution network line encounters planned maintenance, operator adjustment, fault repair, engineering transformation, etc., the line load is often transferred to other lines through line transfer to meet the normal power demand of users and minimize the scope of power outages.

[0003] At present, traditional power transfer identification mainly relies on manually recorded dispatching operation tickets, automatic switch uploading signals, etc. for manual verification. Faced with the problems of multiple types of line transfer, large manual verification workload and easy omissions, patent CN202110753039 proposes a medium-voltage distribution network power transfer operation identification method, device and equipment. For the lines to be detected with abnormal power loss, a random forest-based power transfer identification model is used to identify whether the lines to be detected have power transfer. The existing technology only identifies the power transfer line, but does not identify the specific power transfer area, and the algorithm has the problems of low applicability and low recognition accuracy. Summary of the invention

[0004] Purpose of the invention: In order to solve the problems of multiple types of current line transfers, large manual verification workload, easy omissions, and difficulty in accurately locating the transfer area, the present invention proposes a distribution network line transfer identification method, system and device based on CatBoost, which can quickly, accurately and conveniently identify line transfers and locate specific transfer areas.

[0005] Technical solution: The present invention provides a method for identifying power distribution network line transfer based on CatBoost, comprising the following steps:

[0006] Step S1: obtaining the topological structure of the line to be diagnosed, the distribution transformer account, the relationship between the line and the tie switch, the opposite line account connected to the tie switch, and the measurement data of the line to be diagnosed, the opposite line and the distribution transformer under the line, and the distribution transformer is connected to the line;

[0007] Step S2: Obtain a set of topological relationships corresponding to the line to be diagnosed, the tie switch, the key topological path, the smallest upstream segment of the tie switch, and the opposite end line connected to the tie switch through the key topological path of the line to be diagnosed;

[0008] Step S3: traverse the upstream minimum segment set of the line tie switch to be diagnosed, extract the characteristic index of each distribution transformer in the upstream minimum segment of the tie switch with the same parameters as the trained distribution transformer power transfer identification model based on CatBoost ensemble learning in the period before the current moment, apply the distribution transformer power transfer identification model based on CatBoost ensemble learning online according to the extracted characteristic index, and output whether the distribution transformer in the upstream minimum segment of the tie switch has power transfer at the current moment;

[0009] Step S4: according to the output data of the distribution transformer transfer identification model based on CatBoost ensemble learning, the distribution transformer transfer voting probability in each upstream minimum segment of the tie switch is calculated. If the distribution transformer transfer voting probability in the upstream minimum segment of any tie switch exceeds the threshold, the line to be diagnosed has transferred power at the current moment; otherwise, the line to be diagnosed has not transferred power at the current moment.

[0010] Step S5: for the line to be diagnosed that is currently switching power, a line switching area positioning and identification model is constructed, and a set of substitute supply lines and a set of distribution transformers in the switching area are output.

[0011] Furthermore, in step S2, the specific process of obtaining the corresponding topological relationship set of the line to be diagnosed, the tie switch, the key topological path, the smallest upstream segment of the tie switch, and the opposite end line connected to the tie switch through the key topological path of the line to be diagnosed includes:

[0012] The path set formed by the circuit breaker nodes and the interconnection switch nodes of the connecting line is called the main path of the line; the path between any two adjacent switch nodes (including circuit breakers, section switches and interconnection switch nodes) on the main path of the line is called a line segment; the main path of the line is traversed, the bifurcation nodes, load nodes and relay nodes are marked, the load nodes and relay nodes on each line segment are contracted and virtual load nodes are generated, thereby generating the key topological path of the line;

[0013] By traversing the set of contact switches on the line XL to be diagnosed ,in For the tie switches, obtain the key topological path set consisting of the circuit breaker node connecting the line XL to be diagnosed and each tie switch node , where the key topological path Include Minimum segments, each of which contains a set of more than 3 distribution transformers;

[0014] By traversing the key topological path set A and the contact switch set , get the minimum segment set upstream of the tie switch , where the minimum upstream segment of the tie switch Critical topological path Upper connection contact switch The smallest upstream segment;

[0015] Using the relationship between lines and tie switches, traverse the tie switch set KG on the line to be diagnosed, and obtain the opposite line set connected to the tie switch , where the opposite line To connect the contact switch on the line XL to be diagnosed The opposite end line;

[0016] Finally, a set corresponding to the line to be diagnosed XL, the tie switch set KG, the key topological path set A, the tie switch upstream minimum segment set B, and the opposite line set C is formed. ,in .

[0017] Furthermore, the step S3 specifically includes:

[0018] Step A: Using the historical samples marked with whether the power supply is transferred, extract multiple groups of characteristic indicators under different parameters of the distribution transformer and the local line and the opposite line before and after the power supply transfer, perform offline training and model comparison of the distribution transformer power supply transfer identification model based on CatBoost ensemble learning, save the trained model and characteristic indicator parameters and apply them to the subsequent distribution transformer power supply transfer identification;

[0019] Step B: Traverse the set of upstream minimum segments of the line tie switch to be diagnosed, extract the characteristic indicators of each distribution transformer in the upstream minimum segment of the tie switch with the same parameters as the trained model in the period before the current moment, apply the trained distribution transformer power transfer identification model based on CatBoost ensemble learning online, and output whether the distribution transformer in the upstream minimum segment of the tie switch is currently transferring power.

[0020] Furthermore, the specific process of step A includes:

[0021] Step A1: Obtain the local line ledger, the opposite line ledger, the local distribution transformer ledger, the opposite distribution transformer ledger, the local distribution transformer whether to transfer the supply label, and the line and distribution transformer power and voltage measurement data that are marked with historical samples of whether to transfer the supply;

[0022] Step A2: Using sliding windows with different parameters, extract characteristic indicators of the distribution transformer, local line, and opposite line for a period of time before and after the power transfer, including multiple groups of characteristic indicators with different parameters; the specific process includes:

[0023] Step A2-1: Based on the sliding window method with different parameters such as data duration, sliding window width, sliding step length, sliding times, etc., the voltage and power data of the distribution transformer and the local line and the opposite line before and after the power transfer are processed to generate multiple sets of data with different parameters;

[0024] Assume that the measurement data sampling time interval is minutes after the transfer starts The sliding starts at minutes, and the sliding step length is Minutes, the sliding window width is Minutes, slide times, before the transfer starts The sliding window ends at the sampling time point of , forming a sliding window method with different parameters such as data duration, sliding window width, sliding step size, and sliding times;

[0025] For the same set of parameters, the sliding window method is used to obtain the The local distribution transformer, local line, and opposite line of the sliding Minute normalized voltage data, sliding Generated after The group's local distribution transformer, local line, and opposite line The voltage data set after standardization in minutes; at the same time, the end sampling time point of the sliding window is obtained as The power data set of the local distribution transformer, the opposite distribution transformer, the local line, and the opposite line;

[0026] A sliding window method with different parameters such as data duration, sliding window width, sliding step size, and sliding times is used to generate multiple sets of voltage data sets and power data sets with different parameters.

[0027] Step A2-2: traverse multiple sets of data with different parameters, extract characteristic indicators of the distribution transformer, the local line, and the opposite line for each set of data with different parameters, and generate multiple sets of characteristic indicators for different parameters;

[0028] Each set of characteristic indicators includes: voltage correlation coefficient difference sequence, voltage difference standard deviation ratio sequence, voltage Euclidean distance ratio sequence, local line loss rate sequence, opposite line loss rate sequence, and package line loss rate sequence. Each sequence indicator has Feature indicators, a total of 6 characteristic index; the index calculation is as follows:

[0029] Voltage correlation coefficient difference sequence: Based on the The local distribution transformer, local line, and opposite line of the sliding The voltage data after standardization for 1 minute is used to calculate the voltage correlation coefficient between the local distribution transformer and the local line, and the voltage correlation coefficient between the local distribution transformer and the opposite line respectively using the Pearson correlation coefficient formula, and then the difference between the voltage correlation coefficients between the local distribution transformer and the local line and the opposite line is calculated; the sliding The voltage correlation coefficient difference sequence formed after the first The local distribution transformer is sliding The voltage correlation coefficient between the local distribution transformer and the local line corresponding to this time The calculation formula is as follows:

[0030] ;

[0031] in, , Respectively The local distribution transformer and the local line are sliding The corresponding The average voltage after normalization in minutes, , For the The local distribution transformer and the local line are in the Sliding window Minutes Standardized voltage data of sampling points;

[0032] Similarly, calculate The local distribution transformer is sliding The voltage correlation coefficient between the local distribution transformer and the opposite line corresponding to this time , and then calculate the difference in voltage correlation coefficient between the local distribution transformer and the local line and the opposite line, then The local distribution transformer is sliding The corresponding voltage correlation coefficient difference index The calculation formula is as follows:

[0033] ;

[0034] No. The local distribution transformer is sliding The voltage correlation coefficient difference sequence index formed after for:

[0035] ;

[0036] Ratio sequence of voltage difference standard deviation: based on The local distribution transformer, local line, and opposite line of the sliding After the voltage data is standardized for 1 minute, the difference between the voltage of the local distribution transformer and the local line is calculated, and then the standard deviation is calculated as the standard deviation of the voltage difference between the local distribution transformer and the local line; similarly, the standard deviation of the voltage difference between the local distribution transformer and the opposite line is calculated; finally, the ratio of the standard deviation of the voltage difference between the local distribution transformer and the local line and the opposite line is calculated; sliding The ratio sequence of the voltage difference standard deviation formed after the first The local distribution transformer is sliding The corresponding standard deviation of the voltage difference between the local distribution transformer and the local line The calculation formula is as follows:

[0037] ;

[0038] in, , Respectively The local distribution transformer and the local line are sliding The corresponding The average voltage after normalization in minutes, , For the The local distribution transformer and the local line are in the Sliding window Minutes Standardized voltage data of sampling points;

[0039] Similarly, calculate The local distribution transformer is sliding The corresponding standard deviation of the voltage difference between the local distribution transformer and the opposite line , and then calculate the ratio of the standard deviation of the voltage difference between the local distribution transformer and the local line and the opposite line, then The local distribution transformer is sliding The ratio index of the standard deviation of the corresponding voltage difference The calculation formula is as follows:

[0040] ;

[0041] No. The local distribution transformer is sliding The ratio series index of the voltage difference standard deviation formed after times for:

[0042] ;

[0043] Voltage Euclidean distance ratio sequence: Based on the The local distribution transformer, local line, and opposite line of the sliding After the voltage data is standardized for 1 minute, the Euclidean distance between the distribution transformer at the local end and the line voltage at the local end, and the Euclidean distance between the distribution transformer at the local end and the line voltage at the opposite end are calculated respectively. Then, the ratio of the Euclidean distance between the distribution transformer at the local end and the line at the local end and the line at the opposite end is calculated. The voltage Euclidean distance ratio sequence formed after the first The local distribution transformer is sliding The corresponding European distance between the local distribution transformer and the local line voltage The calculation formula is as follows:

[0044] ;

[0045] in, , For the The local distribution transformer and the local line are in the Sliding window Minutes Standardized voltage data of sampling points;

[0046] Similarly, calculate The local distribution transformer is sliding The corresponding European distance between the local distribution transformer and the opposite line voltage , and then calculate the ratio of the voltage Euclidean distance between the local distribution transformer and the local line and the opposite line, then The local distribution transformer is sliding The ratio index of the Euclidean distance of the corresponding voltage The calculation formula is as follows:

[0047] ;

[0048] No. The local distribution transformer is sliding The ratio sequence index of the voltage Euclidean distance formed after for:

[0049] ;

[0050] Line loss rate sequence of the local line: set the power greater than or equal to 0 as forward, and the power less than 0 as reverse. Calculate the sum of the forward power and reverse power of all distribution transformers under the local line, and calculate the line forward power and line reverse power of the local line. Use the rectangular area method to calculate the line at the sampling time point of the local line. The forward power of the line minus the reverse power of the distribution transformer and multiplied by the sampling time interval is taken as the input power. The output power is the sum of the forward power of the distribution transformer minus the reverse power of the line multiplied by the sampling time interval. The input power minus the output power is defined as the loss power, and the loss power divided by the input power is defined as the line at this end at the sampling time point. The time-sharing line loss rate; the local line at the sampling time point is The time-sharing line loss rate sequence is used as the line loss rate sequence of the local line; The local line of the distribution transformer Distribution transformer calculation sampling time point The forward power and , reverse power and :

[0051] ;

[0052] ;

[0053] in, For the The local distribution transformer belongs to the local line The local distribution transformer at the sampling time Power;

[0054] For The sampling time point of the local line calculation of the distribution transformer Forward power at the moment , Reverse power :

[0055] ;

[0056] ;

[0057] in, For the At the sampling time point, the local line to which the local distribution transformer belongs Power;

[0058] Calculate the area of ​​the rectangle using the The local line of the distribution transformer is at the sampling time point arrive Input power in the time period , Output power :

[0059] ;

[0060] ;

[0061] The input power minus the output power is defined as the loss power. The power loss of the local line of a distribution transformer for:

[0062] ;

[0063] The line loss rate is defined as the power loss divided by the input power. The local line of the distribution transformer is at the sampling time point Time-sharing line loss rate for:

[0064] ;

[0065] No. The local distribution transformer at the sampling time point is The line loss rate sequence index of the local line for:

[0066] ;

[0067] The line loss rate sequence of the opposite end line: The calculation logic of the line loss rate of the opposite end line is the same as that of the local line line. The opposite end line is The time-sharing line loss rate sequence is used as the opposite line line loss rate sequence; The opposite end line of the distribution transformer is at the sampling time point Time-sharing line loss rate :

[0068] ;

[0069] in , The sampling time point of the opposite line arrive The input power and loss power in the time period;

[0070] No. The local distribution transformer at the sampling time point is The peer line loss rate sequence index for:

[0071] ;

[0072] Package line loss rate sequence: The local line and the opposite line are packaged together as a packaged line, and the local line and the opposite line are packaged at the sampling time point. The input power of the packaged line is added as the total input power of the packaged line, and the local line and the opposite line are added at the sampling time point. The total power loss of the packaging line is added as the total power loss of the packaging line, and the total power loss of the packaging line divided by the total input power is defined as the power loss of the packaging line at the sampling time point The time-sharing line loss rate; the packaging line is The time-sharing line loss rate sequence is used as the packaging line loss rate sequence; The local line and the opposite line belonging to the local distribution transformer are packaged and combined together at the sampling time point. Baling line loss rate at the moment for:

[0073] ;

[0074] in , For the The local line to which the local distribution transformer belongs is at time point arrive The input power and loss power in the time period; , For the The opposite end line to which the local distribution transformer belongs is at time point arrive The input power and loss power in the time period;

[0075] No. The local distribution transformer at the sampling time point is The line loss rate index of the opposite line for:

[0076] ;

[0077] Step A3: A random partitioning method is used to select a part of the samples in the historical sample data set as the training set, and the remaining samples are used as the test set. The model training and comparison are performed for multiple groups of feature indicators with different parameters: the feature indicators in the training set are used as the model input, and whether the distribution transformer is switching power is used as the model output. The model is input into the CatBoost ensemble learning algorithm to train the distribution transformer switching recognition model based on CatBoost ensemble learning;

[0078] Step A4: Apply the trained model to the test set and use the harmonic mean of precision and recall The scores are used to evaluate the model and compare and select The score is the largest and exceeds the threshold Model and characteristic index parameters; The scores are defined as follows:

[0079] ;

[0080] In the formula, The model predicts the number of distribution transformers that are actually transferring power as the number of distribution transformers that are actually transferring power; The model predicts the number of distribution transformers that are not actually transferring power as the number that are transferring power; The model predicts the number of distribution transformers that are actually transferring power as not transferring power;

[0081] Step A5: Save the distribution transformer power transfer identification model and characteristic index parameters based on CatBoost ensemble learning trained in step A4, and apply them to the subsequent distribution transformer power transfer identification.

[0082] Furthermore, the step B specifically includes:

[0083] Traverse the minimum segment set B upstream of the XL tie switch of the line to be diagnosed, and extract the characteristic index of each distribution transformer in the minimum segment upstream of the tie switch with the same parameters as the trained model in the period before the current moment: for each distribution transformer in the minimum segment b upstream of a tie switch, start sliding from the current moment, and the sliding step length is Minutes, the sliding window width is Minutes, slide times, until the current time The sliding window ends at the sampling time point of Each slide obtains the local distribution transformer, local line, and opposite line The voltage data is normalized after 1 minute; at the same time, the end sampling time point of the sliding window is obtained. The power data of the local distribution transformer, the opposite distribution transformer, the local line, and the opposite line; the calculation logic of the characteristic indicators of model training is the same, and the same parameters as the trained model can be calculated characteristic indicators, including: voltage correlation coefficient difference sequence, voltage difference standard deviation ratio sequence, voltage Euclidean distance ratio sequence, local line line loss rate sequence, opposite line line loss rate sequence, and package line loss rate sequence;

[0084] The trained distribution transformer transfer identification model based on CatBoost ensemble learning is applied online to output whether the distribution transformer in the smallest section upstream of the tie switch is transferring power at the current moment.

[0085] Furthermore, the step S4 specifically includes:

[0086] Traverse the minimum segment set B upstream of the XL tie switch of the line to be diagnosed, and calculate the ratio of the number of distribution transformers in each upstream minimum segment of the tie switch to the total number of distribution transformers in the upstream minimum segment of the tie switch according to the output data of the distribution transformer transfer identification model based on CatBoost ensemble learning, and use it as the distribution transformer transfer voting probability in the upstream minimum segment of the tie switch;

[0087] If the probability of the distribution transformer switching in any of the upstream minimum sections of the tie switch exceeds the threshold, the line to be diagnosed XL switches power at the current moment, and outputs the set of upstream minimum sections of the tie switch where the switching occurs. Otherwise, the line XL to be diagnosed has not transferred power at the current moment.

[0088] Furthermore, the step S5 specifically includes:

[0089] Step S5-1: For the line to be diagnosed XL that is currently transferring power, obtain the minimum segment set upstream of the tie switch that is transferring power Corresponding contact switch set , key topological path set , peer line set ;

[0090] Step S5-2: Traverse the set of tie switches of the line XL to be diagnosed , get the Tie switch Corresponding key topological path , the key topological path Include Minimum segments, filter out the upstream minimum segment of the tie switch where the transfer occurs The other smallest upstream segment ;

[0091] Step S5-3: divide the upstream other smallest segments Each distribution transformer applies the trained CatBoost ensemble learning-based distribution transformer transfer identification model online to output the other minimum segments upstream. Whether the internal distribution transformer is transferring power at the current moment;

[0092] Step S5-4: Calculate other upstream minimum segments The internal distribution transformer transfer voting probability, if the upstream other minimum segment If the probability of voting for power transfer exceeds the threshold, the other upstream minimum segments The transfer occurs at the current moment;

[0093] Step S5-5: Continue to filter out the smallest segment According to steps S5-3 to S5-4, it is determined whether the other upstream smallest segments are currently being transferred, until a certain upstream smallest segment is No transfer or tracing back to the smallest upstream segment Then stop tracing the upstream segment, and finally form the Tie switch The corresponding transfer areas are:

[0094] ;

[0095] Step S5-6: Traverse the set of tie switches of the line XL to be diagnosed , the output transfer area distribution transformer set is , the corresponding supply line set is .

[0096] The present invention also provides a distribution network line transfer identification system based on CatBoost, comprising:

[0097] Data acquisition module: obtain the topological structure of the line to be diagnosed, the distribution transformer account, the relationship between the line and the tie switch, the opposite line account connected to the tie switch, and the measurement data of the line to be diagnosed, the opposite line and the distribution transformer under the line; obtain the corresponding topological relationship set of the line to be diagnosed, the tie switch, the key topological path, the upstream minimum segment of the tie switch, and the opposite line connected to the tie switch through the key topological path of the line to be diagnosed;

[0098] Segmented power transfer identification module: traverses the upstream minimum segment set of the line tie switch to be diagnosed, extracts the characteristic index of each distribution transformer in the upstream minimum segment of the tie switch with the same parameters as the trained distribution transformer power transfer identification model based on CatBoost ensemble learning in the period before the current moment, applies the distribution transformer power transfer identification model based on CatBoost ensemble learning online according to the extracted characteristic index, and outputs whether the distribution transformer in the upstream minimum segment of the tie switch has power transfer at the current moment;

[0099] Line transfer identification module: According to the output data of the distribution transformer transfer identification model based on CatBoost ensemble learning, the distribution transformer transfer voting probability in each upstream minimum segment of the tie switch is calculated. If the distribution transformer transfer voting probability in the upstream minimum segment of any tie switch exceeds the threshold, the line to be diagnosed has transferred power at the current moment, otherwise, the line to be diagnosed has not transferred power at the current moment;

[0100] Transfer area identification module: For the lines to be diagnosed that are currently transferring supply, a line transfer area positioning and identification model is constructed to output the set of substitute supply lines and the set of distribution transformers in the transfer area.

[0101] The present invention also provides a distribution network line transfer identification device based on CatBoost, comprising a network interface, a memory and a processor;

[0102] The network interface is used to receive and send signals during the process of sending and receiving information with other external network elements;

[0103] The memory is used to store computer program instructions that can be executed on the processor;

[0104] The processor is used to execute the steps of the CatBoost-based distribution network line transfer identification method when running the computer program instructions.

[0105] The present invention also provides a computer storage medium, which stores a program of a distribution network line transfer identification method based on CatBoost. When the program of the distribution network line transfer identification method based on CatBoost is executed by at least one processor, the steps of the distribution network line transfer identification method based on CatBoost are implemented.

[0106] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0107] (1) This invention proposes a distribution network line transfer identification method based on CatBoost, which can identify whether the line is transferred and locate the specific transfer area, providing technical support for line transfer identification and dynamic topology identification;

[0108] (2) The present invention proposes a line transfer identification method based on the upstream segmentation of the tie switch and the CatBoost ensemble learning algorithm, which greatly improves the accuracy of line transfer identification. Compared with the existing identification methods, the present invention can quickly, accurately and conveniently identify whether the line is transferred, and assist the operation and maintenance personnel to quickly check the line transfer behavior;

[0109] (3) The present invention proposes a line transfer area identification method based on key topological paths and CatBoost ensemble learning, which can locate the specific transfer area, reduce the error rate of power outage information release caused by unclear power outage scope, and greatly improve user satisfaction;

[0110] (4) The present invention has simple calculation and clear principle, and can realize line transfer identification and specific transfer area positioning, helping operation and maintenance personnel to quickly verify transfer behavior and locate specific transfer areas, providing a powerful tool for line transfer management and having good application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0111] Figure 1 is a flow chart of the method of the present invention;

[0112] Figure 2 is a schematic diagram of a key topological path of the line XL to be diagnosed in an embodiment of the present invention;

[0113] Figure 3 In the embodiment of the present invention, the line XL to be diagnosed and the opposite line Schematic diagram of the time-sharing line loss rate curve. DETAILED DESCRIPTION

[0114] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0115] Embodiment 1:

[0116] This embodiment takes a line transfer sample of a certain province as an example to further illustrate a distribution network line transfer identification method based on CatBoost proposed by the present invention.

[0117] Figure 1The invention describes the application of the CatBoost-based distribution network line transfer identification method to the line transfer in the medium voltage distribution network, including but not limited to the medium voltage distribution network line transfer identification and the specific transfer area location identification. Figure 1 As shown, this embodiment adopts a distribution network line transfer identification method based on CatBoost, Figure 1 The left side shows the training process of the distribution transformer transfer identification model based on CatBoost ensemble learning. Figure 1 The right side shows the application process of the distribution network line transfer identification method based on CatBoost. In the following description, the training of the distribution transformer transfer identification model based on CatBoost ensemble learning and the application of the distribution network line transfer identification method based on CatBoost are described in detail. In this embodiment, 6493 transfer line samples in a certain province are used to obtain more than 220,000 distribution transformer transfer samples, and the distribution transformer transfer identification model based on CatBoost ensemble learning is trained; the distribution network line transfer identification method based on CatBoost is applied using 1 line in a certain province.

[0118] Part 1: Offline training of distribution transformer transfer identification model based on CatBoost ensemble learning

[0119] Reference Figure 1 In the left part of FIG. 1 , the model training process in this embodiment specifically includes the following steps:

[0120] Step 1: Obtain the local line ledger, the opposite line ledger, the local distribution transformer ledger, the opposite distribution transformer ledger, the local distribution transformer whether to transfer the supply label, and the line and distribution transformer power and voltage measurement data with the historical samples marked whether to transfer the supply;

[0121] In the model training of this embodiment, 6493 samples of transfer lines and more than 220,000 distribution transformer transfer samples were obtained in a province. The ratio of transfer lines to non-transfer lines was 1:2, and the ratio of transfer distribution transformers to non-transfer distribution transformers was about 1:3. The distribution transformer transfer samples obtained from the PMS (equipment asset lean management) system and the marketing system involved the local line ledger, the opposite line ledger, the local distribution transformer ledger, and the opposite distribution transformer ledger; the distribution transformer transfer samples obtained from the D5000 system involved 96 points of power and voltage data of the line, where the measurement data sampling time interval was t=15 minutes; the distribution transformer transfer samples obtained from the power consumption collection system involved 96 points of power and voltage data of the distribution transformer, where the measurement data sampling time interval was t=15 minutes; and the historical distribution transformer transfer samples were obtained, which marked whether the local distribution transformer was transferred.

[0122] Step 2: Use sliding windows with different parameters to extract characteristic indicators of the distribution transformer, local line, and opposite line before and after the power transfer, including multiple groups of characteristic indicators with different parameters. The specific process includes:

[0123] Step 2-1: Based on the sliding window method with different parameters such as data duration, sliding window width, sliding step, sliding times, etc., the voltage and power data of the distribution transformer and the local line and the opposite line before and after the power transfer are processed to generate multiple sets of data with different parameters;

[0124] Assume that the measurement data sampling time interval is minutes after the transfer starts The sliding starts at minutes, and the sliding step length is Minutes, the sliding window width is Minutes, slide times, before the transfer starts The sliding window ends at the sampling time point of , forming a sliding window method with different parameters such as data duration, sliding window width, sliding step size, and sliding times;

[0125] For the same set of parameters, the sliding window method is used to obtain the The local distribution transformer, local line, and opposite line of the sliding Minute normalized voltage data, sliding Generated after The group's local distribution transformer, local line, and opposite line The voltage data set after standardization in minutes; at the same time, the end sampling time point of the sliding window is obtained as The power data set of the local distribution transformer, the opposite distribution transformer, the local line, and the opposite line;

[0126] A sliding window method with different parameters such as data duration, sliding window width, sliding step size, and sliding times is used to generate multiple sets of voltage data sets and power data sets with different parameters.

[0127] Step 2-2: traverse multiple sets of data with different parameters, extract characteristic indicators of the distribution transformer, the local line, and the opposite line for each set of data with different parameters, and generate multiple sets of characteristic indicators for different parameters;

[0128] Each set of characteristic indicators includes: voltage correlation coefficient difference sequence, voltage difference standard deviation ratio sequence, voltage Euclidean distance ratio sequence, local line loss rate sequence, opposite line loss rate sequence, and package line loss rate sequence. Each sequence indicator has characteristic indicators, a total of characteristic index; the index calculation is as follows:

[0129] Voltage correlation coefficient difference sequence: Based on the The local distribution transformer, local line, and opposite line of the sliding The voltage data after standardization for 1 minute is used to calculate the voltage correlation coefficient between the local distribution transformer and the local line, and the voltage correlation coefficient between the local distribution transformer and the opposite line respectively using the Pearson correlation coefficient formula, and then the difference between the voltage correlation coefficients between the local distribution transformer and the local line and the opposite line is calculated; the sliding The voltage correlation coefficient difference sequence formed after the first The local distribution transformer is sliding The voltage correlation coefficient between the local distribution transformer and the local line corresponding to this time The calculation formula is as follows:

[0130] ;

[0131] in, , Respectively The local distribution transformer and the local line are sliding The corresponding The average voltage after normalization in minutes, , For the The local distribution transformer and the local line are in the Sliding window Minutes Standardized voltage data of sampling points;

[0132] Similarly, calculate The local distribution transformer is sliding The voltage correlation coefficient between the local distribution transformer and the opposite line corresponding to this time , and then calculate the difference in voltage correlation coefficient between the local distribution transformer and the local line and the opposite line, then The local distribution transformer is sliding The corresponding voltage correlation coefficient difference index The calculation formula is as follows:

[0133] ;

[0134] No. The local distribution transformer is sliding The voltage correlation coefficient difference sequence index formed after for:

[0135] ;

[0136] Ratio series of voltage difference standard deviation: based on the The local distribution transformer, local line, and opposite line of the sliding After the voltage data is standardized for 1 minute, the difference between the voltage of the local distribution transformer and the local line is calculated, and then the standard deviation is calculated as the standard deviation of the voltage difference between the local distribution transformer and the local line; similarly, the standard deviation of the voltage difference between the local distribution transformer and the opposite line is calculated; finally, the ratio of the standard deviation of the voltage difference between the local distribution transformer and the local line and the opposite line is calculated; sliding The ratio sequence of the voltage difference standard deviation formed after the first The local distribution transformer is sliding The corresponding standard deviation of the voltage difference between the local distribution transformer and the local line The calculation formula is as follows:

[0137] ;

[0138] in, , Respectively The local distribution transformer and the local line are sliding The corresponding The average voltage after normalization in minutes, , For the The local distribution transformer and the local line are in the Sliding window Minutes Standardized voltage data of sampling points;

[0139] Similarly, calculate The local distribution transformer is sliding The corresponding standard deviation of the voltage difference between the local distribution transformer and the opposite line , and then calculate the ratio of the standard deviation of the voltage difference between the local distribution transformer and the local line and the opposite line, then The local distribution transformer is sliding The ratio index of the standard deviation of the corresponding voltage difference The calculation formula is as follows:

[0140] ;

[0141] No. The local distribution transformer is sliding The ratio series index of the voltage difference standard deviation formed after times for:

[0142] ;

[0143] Voltage Euclidean distance ratio sequence: Based on the The local distribution transformer, local line, and opposite line of the sliding After the voltage data is standardized for 1 minute, the Euclidean distance between the distribution transformer at the local end and the line voltage at the local end, and the Euclidean distance between the distribution transformer at the local end and the line voltage at the opposite end are calculated respectively. Then, the ratio of the Euclidean distance between the distribution transformer at the local end and the line at the local end and the line at the opposite end is calculated. The voltage Euclidean distance ratio sequence formed after the first The local distribution transformer is sliding The corresponding European distance between the local distribution transformer and the local line voltage The calculation formula is as follows:

[0144] ;

[0145] in, , For the The local distribution transformer and the local line are in the Sliding window Minutes Standardized voltage data of sampling points;

[0146] Similarly, calculate The local distribution transformer is sliding The corresponding European distance between the local distribution transformer and the opposite line voltage , and then calculate the ratio of the voltage Euclidean distance between the local distribution transformer and the local line and the opposite line, then The local distribution transformer is sliding The ratio index of the Euclidean distance of the corresponding voltage The calculation formula is as follows:

[0147] ;

[0148] No. The local distribution transformer is sliding The ratio sequence index of the voltage Euclidean distance formed after for:

[0149] ;

[0150] Line loss rate sequence of the local line: set the power greater than or equal to 0 as forward, and the power less than 0 as reverse. Calculate the sum of the forward power and reverse power of all distribution transformers under the local line, and calculate the line forward power and line reverse power of the local line. Use the rectangular area method to calculate the line at the sampling time point of the local line. The forward power of the line minus the reverse power of the distribution transformer and multiplied by the sampling time interval is taken as the input power. The output power is the sum of the forward power of the distribution transformer minus the reverse power of the line multiplied by the sampling time interval. The input power minus the output power is defined as the loss power, and the loss power divided by the input power is defined as the line at this end at the sampling time point. The time-sharing line loss rate; the local line at the sampling time point is The time-sharing line loss rate sequence is used as the line loss rate sequence of the local line; The local line of the distribution transformer Distribution transformer calculation sampling time point The forward power and , reverse power and :

[0151] ;

[0152] ;

[0153] in, For the The local distribution transformer belongs to the local line The local distribution transformer at the sampling time Power;

[0154] For The sampling time point of the local line calculation of the distribution transformer Forward power at the moment , Reverse power :

[0155] ;

[0156] ;

[0157] in, For the At the sampling time point, the local line to which the local distribution transformer belongs Power;

[0158] Calculate the area of ​​the rectangle using the The local line of the distribution transformer is at the sampling time point arrive Input power in the time period , output power :

[0159] ;

[0160] ;

[0161] The input power minus the output power is defined as the loss power. The power loss of the local line of a distribution transformer for:

[0162] ;

[0163] The line loss rate is defined as the power loss divided by the input power. The local line of the distribution transformer is at the sampling time point Time-sharing line loss rate for:

[0164] ;

[0165] No. The local distribution transformer at the sampling time point is The line loss rate sequence index of the local line for:

[0166] ;

[0167] The line loss rate sequence of the opposite end line: The calculation logic of the line loss rate of the opposite end line is the same as that of the local line line. The opposite end line is The time-sharing line loss rate sequence is used as the opposite line line loss rate sequence; The opposite end line of the distribution transformer is at the sampling time point Time-sharing line loss rate :

[0168] ;

[0169] in , The sampling time point of the opposite line arrive The input power and loss power in the time period;

[0170] No. The local distribution transformer at the sampling time point is The line loss rate sequence index of the opposite line for:

[0171] ;

[0172] Package line loss rate sequence: The local line and the opposite line are packaged together as a packaged line, and the local line and the opposite line are packaged at the sampling time point. The input power of the packaged line is added as the total input power of the packaged line, and the local line and the opposite line are added at the sampling time point. The total power loss of the packaging line is added as the total power loss of the packaging line, and the total power loss of the packaging line divided by the total input power is defined as the power loss of the packaging line at the sampling time point The time-sharing line loss rate; the packaging line is The time-sharing line loss rate sequence is used as the packaging line loss rate sequence; The local line and the opposite line belonging to the local distribution transformer are packaged and combined together at the sampling time point. Baling line loss rate at the moment for:

[0173] ;

[0174] in , For the The local line to which the local distribution transformer belongs is at time point arrive The input power and loss power in the time period; , For the The opposite end line to which the local distribution transformer belongs is at time point arrive The input power and loss power in the time period;

[0175] No. The local distribution transformer at the sampling time point is The line loss rate index of the opposite line for:

[0176] .

[0177] In the model training of this embodiment, for more than 220,000 distribution transformer transfer samples, the sliding window method is used to extract the characteristic indicators of the distribution transformer and the local line and the opposite line for a period of time before and after the transfer, including multiple groups of characteristic indicators of parameters such as data duration, sliding window width, sliding step length, and sliding times. The values ​​of parameters x, y, z, w, and k involved in variables such as data duration, sliding window width, sliding step length, and sliding times are shown in Table 1 below. A total of 17 groups of parameters are obtained, and 17 groups of characteristic indicators are calculated using the above-mentioned characteristic indicator calculation logic. Each group of characteristic indicators includes the following: voltage correlation coefficient difference sequence, voltage difference standard deviation ratio sequence, voltage Euclidean distance ratio sequence, local line line loss rate sequence, opposite line line loss rate sequence, and package line loss rate sequence, a total of A characteristic indicator.

[0178] Table 1 Parameter value table

[0179]

[0180] Step 3: Using the historical samples of whether the power supply was transferred or not, and targeting multiple groups of characteristic indicators under different parameter conditions, offline training and model comparison of the distribution transformer power transfer identification model based on CatBoost ensemble learning are carried out, and the trained model and characteristic indicator parameters are saved and applied to the subsequent distribution transformer power transfer identification;

[0181] In the model training of this embodiment, based on more than 220,000 distribution transformer transfer samples in a province, the ratio of transferred distribution transformers to non-transferred distribution transformers is about 1:3. 17 groups of feature indicators are used as model inputs, and whether the distribution transformer transfers is used as model output, and the model training is performed using the CatBoost ensemble learning algorithm.

[0182] When training the CatBoost model, a random partitioning method is used to select 70% of the samples in the historical sample data set as the training set, and the remaining 30% of the samples are used as the test set. The characteristic index is used as the model input, and whether the distribution transformer transfers power is used as the model output, which is input into the CatBoost ensemble learning algorithm to train the distribution transformer transfer identification model based on CatBoost ensemble learning.

[0183] Apply the trained model to the test set and use the harmonic mean of precision and recall The scores are used to evaluate the model and compare and select The score is the largest and exceeds the threshold The model and characteristic index parameters of the distribution transformer transfer identification model based on CatBoost ensemble learning are saved; The scores are defined as follows:

[0184]

[0185] In the formula, The model predicts the number of distribution transformers that are actually transferring power as the number of distribution transformers that are actually transferring power; The model predicts the number of distribution transformers that are not actually transferring power as the number that are transferring power; The model predicts the number of distribution transformers that are actually transferring power as not transferring power;

[0186] In order to evaluate the classification performance of the distribution transformer transfer identification model based on CatBoost ensemble learning, this embodiment uses five classification algorithms, namely, CatBoost, Lightweight Gradient Boosting Machine (LightGBM), Extreme Gradient Boosting Machine (XGBoost), Gradient Boosting Decision Tree (GBDT), and Random Forest (RF), to compare the distribution transformer transfer identification model. For 17 sets of feature indicators with different parameters, these five classification algorithms are trained on the training set, and the trained models are applied to the test set to obtain the performance evaluation indicators of the five classification algorithms, as shown in Table 2. As can be seen from Table 2, the performance of the CatBoost distribution transformer transfer identification model under 17 sets of parameters is The scores of the CatBoost distribution transformer transfer identification model under the 16th group of parameters are higher than those of the other four algorithms. The highest score, The score is 95.4%, exceeding the threshold It is 90%. From Table 1, we can see that the values ​​of the 16th group of parameters x, y, z, w, k are 12, 8, 1, 8, 13. The trained CatBoost distribution transformer transfer identification model is saved, and the characteristic parameters x, y, z, w, k are saved with the values ​​of 12, 8, 1, 8, 13, and then applied to the actual distribution transformer transfer identification.

[0187] Table 2 Comparison results of five models under different parameters

[0188]

[0189] Part II: Application and implementation of the distribution network line transfer identification method based on CatBoost

[0190] Reference Figure 1 Based on the distribution transformer power transfer identification model based on CatBoost ensemble learning trained in the first part, the distribution network line power transfer identification method based on CatBoost provided by the present invention is applied in this embodiment, which specifically includes the following steps:

[0191] Step 1): Obtain the topological structure of the line to be diagnosed, the distribution transformer account, the relationship between the line and the tie switch, and the opposite line account connected to the tie switch; and obtain the measurement data of the line to be diagnosed, the opposite line and the distribution transformer under the line, and the distribution transformer is connected to the line;

[0192] In the model application of this embodiment, the line topology, distribution transformer ledger, relationship between line and interconnection switch, and opposite line ledger connected to interconnection switch of a line in a certain province are obtained, where there are 69 distribution transformers, 2 interconnection switches, and 2 opposite lines; and the measurement data of the line to be diagnosed, opposite line, and distribution transformer under the line are obtained. The line topology, distribution transformer ledger, relationship between line and interconnection switch, and opposite line ledger connected to interconnection switch are obtained from the PMS (equipment asset lean management) system and marketing system; 96 points of power and voltage data related to the line are obtained from the D5000 system, where the measurement data sampling time interval is t=15 minutes; 96 points of power and voltage data of the distribution transformer are obtained from the power consumption collection system, where the measurement data sampling time interval is t=15 minutes.

[0193] Step 2): Obtain the corresponding topological relationship set of the line to be diagnosed, the tie switch, the key topological path, the upstream minimum segment of the tie switch, and the opposite end line connected to the tie switch through the key topological path of the line to be diagnosed;

[0194] By traversing the set of contact switches on the line XL to be diagnosed ,in For the tie switches, obtain the key topological path set consisting of the circuit breaker node connecting the line XL to be diagnosed and each tie switch node , where the key topological path Include Minimum segments, each of which contains a set of more than 3 distribution transformers;

[0195] By traversing the key topological path set A and the contact switch set , get the minimum segment set upstream of the tie switch , where the minimum upstream segment of the tie switch Critical topological path Upper connection contact switch The smallest upstream segment;

[0196] Using the relationship between lines and tie switches, traverse the tie switch set KG on the line to be diagnosed, and obtain the opposite line set connected to the tie switch , where the opposite line To connect the contact switch on the line XL to be diagnosed The opposite end line;

[0197] Finally, a set corresponding to the line to be diagnosed XL, the tie switch set KG, the key topological path set A, the tie switch upstream minimum segment set B, and the opposite line set C is formed. ,in .

[0198] In the model application of this embodiment, the key topological path of a line XL to be diagnosed in a certain province is as follows: Figure 2 As shown in the figure, the diagnosis line XL has a total of 69 distribution transformers. Through the above steps, the topological relationship set corresponding to the line to be diagnosed, the tie switch, the key topological path, the smallest upstream segment of the tie switch, and the opposite end line connected to the tie switch is obtained, as shown in Table 3 below. Key topological path It is the circuit breaker BRK of the line XL to be diagnosed to the tie switch The key topological path of the 4 smallest segments, the tie switch The upstream minimum segment The smallest segment is fd3-kg1, the smallest upstream segment Contains 17 distribution transformers, connected to tie switches The opposite line is line ; Critical topological path It is the circuit breaker BRK of the line XL to be diagnosed to the tie switch The key topological path of the , including 3 minimum segments, tie switch The upstream minimum segment The smallest segment is fd3-kg1, the smallest upstream segment Contains 16 distribution transformers, connected to tie switches The opposite line is line .

[0199] Table 3 Topological relationship table of lines to be diagnosed

[0200]

[0201] Step 3): Traverse the set of upstream minimum segments of the line tie switch to be diagnosed, extract the characteristic indicators of each distribution transformer in the upstream minimum segment of the tie switch for a period of time before the current moment, apply the distribution transformer power transfer identification model based on CatBoost ensemble learning trained in the first part of this embodiment online, and output whether the distribution transformer in the upstream minimum segment of the tie switch has power transfer at the current moment.

[0202] In the model application of this embodiment, the parameters The value is , the current time is 2024-11-03 08:30:00, and the suspected power transfer start time is 2024-11-03 06:30:00. Traverse the minimum segment set B upstream of the XL tie switch of the line to be diagnosed, and extract the feature index of each distribution transformer in the minimum segment upstream of the tie switch with the same parameters as the trained model in the period before the current time: For each distribution transformer in any minimum segment b upstream of the tie switch, start sliding from the current time, and the sliding step length is Minutes, the sliding window width is Minutes, slide times, until the current time The sliding window ends at the sampling time point of Each slide obtains the local distribution transformer, local line, and opposite line The voltage data is normalized after 1 minute; at the same time, the end sampling time point of the sliding window is obtained. The power data of the local distribution transformer, the opposite distribution transformer, the local line, and the opposite line; the calculation logic of the characteristic indicators of the model training in the first part is the same, and the same parameters as the trained model in the first part can be calculated characteristic indicators, including: voltage correlation coefficient difference sequence, voltage difference standard deviation ratio sequence, voltage Euclidean distance ratio sequence, local line line loss rate sequence, opposite line line loss rate sequence, and package line loss rate sequence;

[0203] The trained distribution transformer transfer identification model based on CatBoost ensemble learning is applied online to output whether the distribution transformer in the smallest section upstream of the tie switch is transferring power at the current moment.

[0204] In the model application of this embodiment, the upstream minimum segment of the tie switch under a line XL to be diagnosed in a certain province is respectively , , using the trained distribution transformer transfer identification model based on CatBoost ensemble learning, output the upstream minimum segment of the tie switch , Whether the internal distribution transformer is transferring power at the current moment, the statistical identification results are shown in Table 4. From Table 4, it can be seen that the minimum upstream segment of the tie switch There are 17 distribution transformers, and the smallest upstream section of the tie switch is There are 0 power supply and distribution transformers.

[0205] Table 4 Application results of distribution transformer transfer identification model based on CatBoost ensemble learning

[0206]

[0207] Step 4): According to the output data of the distribution transformer transfer identification model based on CatBoost ensemble learning, the distribution transformer transfer voting probability in each upstream minimum segment of the tie switch is calculated. If the distribution transformer transfer voting probability in the upstream minimum segment of any tie switch exceeds the threshold, the line to be diagnosed has transferred power at the current moment, otherwise, the line to be diagnosed has not transferred power at the current moment. The specific steps are as follows:

[0208] Traverse the minimum segment set B upstream of the XL tie switch of the line to be diagnosed, and calculate the ratio of the number of distribution transformers in each upstream minimum segment of the tie switch to the total number of distribution transformers in the upstream minimum segment of the tie switch according to the output data of the distribution transformer transfer identification model based on CatBoost ensemble learning, and use it as the distribution transformer transfer voting probability in the upstream minimum segment of the tie switch;

[0209] If the probability of the distribution transformer switching in any of the upstream minimum sections of the tie switch exceeds the threshold, the line to be diagnosed XL switches power at the current moment, and outputs the set of upstream minimum sections of the tie switch where the switching occurs. Otherwise, the line XL to be diagnosed has not transferred power at the current moment.

[0210] In this embodiment, the above steps are used to calculate the minimum upstream segment of the tie switch under a line XL to be diagnosed in a certain province. , The probability of voting for power transfer of distribution transformers is shown in Table 5. Assuming the probability threshold of power transfer voting is 80%, the minimum upstream segment of the tie switch is If the probability of voting for power transfer of the distribution transformer exceeds the threshold, the line to be diagnosed XL will transfer power at the current moment, and the upstream minimum segment of the tie switch where the transfer occurs will be output. . Figure 3 The following table shows the relationship between the line XL to be diagnosed and the opposite line The time-sharing line loss rate curve shows that the line XL to be diagnosed is There is indeed a transfer between them. At 2024-11-03 06:30:00, the line loss rate of the line XL to be diagnosed begins to drop suddenly, and the line at the opposite end The line loss rate of the line XL to be diagnosed begins to increase suddenly. At 2024-11-03 09:30:00, the line loss rate of the line XL to be diagnosed begins to return to normal. The line loss rate also began to return to normal.

[0211] Table 5 Calculation results of distribution transformer transfer voting probability

[0212]

[0213] Step 5): For the line to be diagnosed that is currently being transferred, a line transfer area positioning and identification model is constructed to output the set of substitute supply lines and the set of distribution transformers in the transfer area. The specific steps include:

[0214] Step 5-1): For the line to be diagnosed XL that is currently transferring power, obtain the minimum segment set upstream of the tie switch that is transferring power Corresponding contact switch set , key topological path set , peer line set ;

[0215] Step 5-2): Traverse the set of tie switches of the line XL to be diagnosed , get the Tie switch Corresponding key topological path , the key topological path Include Minimum segments, filter out the upstream minimum segment of the tie switch where the transfer occurs The other smallest upstream segment ;

[0216] Step 5-3): Divide the upstream into other smallest segments Each distribution transformer applies the trained CatBoost ensemble learning-based distribution transformer transfer identification model online to output the other minimum segments upstream. Whether the internal distribution transformer is transferring power at the current moment;

[0217] Step 5-4): Calculate other minimum upstream segments The internal distribution transformer transfer voting probability, if the upstream other minimum segment If the probability of voting for power transfer exceeds the threshold, the other upstream minimum segments The transfer occurs at the current moment;

[0218] Step 5-5): Continue to filter out the smallest segment The other upstream smallest segments are determined according to steps 5-3) to 5-4) to determine whether the other upstream smallest segments are currently being transferred, until a certain upstream smallest segment is No transfer or tracing back to the smallest upstream segment Then stop tracing the upstream segment, and finally form the Tie switch The corresponding transfer area is ;

[0219] Step 5-6): Traverse the set of tie switches of the line XL to be diagnosed , the output transfer area distribution transformer set is , the corresponding supply line set is .

[0220] In this embodiment, for the line to be diagnosed XL that is currently switching power, the minimum segment set upstream of the tie switch that is switching power is obtained. Corresponding contact switch set , key topological path set , peer line set . Contact switch Corresponding key topological path , minimum segment

[0221] For segment BRK-fd1, the minimum segment The set of distribution transformers includes , and All distribution transformers in these three sections, the smallest section For segment fd1-fd2, the minimum segment For segment fd2-fd3, the smallest segment For segmentation Assuming the transfer voting probability threshold is 80%, the above line transfer area positioning identification model is used to output the line transfer area positioning identification model results as shown in Table 6. As can be seen from Table 6, the transfer area distribution transformer set of the line XL to be diagnosed is , that is, the transfer area is , the corresponding supply line set is .

[0222] Table 6 Results of the line transfer area location identification model

[0223]

[0224] Embodiment 2:

[0225] This embodiment provides a distribution network line transfer identification system based on CatBoost, including:

[0226] Data acquisition module: obtain the topological structure of the line to be diagnosed, the distribution transformer account, the relationship between the line and the tie switch, the opposite line account connected to the tie switch, and the measurement data of the line to be diagnosed, the opposite line and the distribution transformer under the line; obtain the corresponding topological relationship set of the line to be diagnosed, the tie switch, the key topological path, the upstream minimum segment of the tie switch, and the opposite line connected to the tie switch through the key topological path of the line to be diagnosed;

[0227] Segmented power transfer identification module: traverses the upstream minimum segment set of the line tie switch to be diagnosed, extracts the characteristic index of each distribution transformer in the upstream minimum segment of the tie switch with the same parameters as the trained distribution transformer power transfer identification model based on CatBoost ensemble learning in the period before the current moment, applies the distribution transformer power transfer identification model based on CatBoost ensemble learning online according to the extracted characteristic index, and outputs whether the distribution transformer in the upstream minimum segment of the tie switch has power transfer at the current moment;

[0228] Line transfer identification module: According to the output data of the distribution transformer transfer identification model based on CatBoost ensemble learning, the distribution transformer transfer voting probability in each upstream minimum segment of the tie switch is calculated. If the distribution transformer transfer voting probability in the upstream minimum segment of any tie switch exceeds the threshold, the line to be diagnosed has transferred power at the current moment, otherwise, the line to be diagnosed has not transferred power at the current moment;

[0229] Transfer area identification module: For the lines to be diagnosed that are currently transferring supply, a line transfer area positioning and identification model is constructed to output the set of substitute supply lines and the set of distribution transformers in the transfer area.

[0230] Embodiment 3:

[0231] The present embodiment provides a distribution network line transfer identification device based on CatBoost, comprising a network interface, a memory and a processor; wherein the network interface is used to realize signal reception and transmission in the process of sending and receiving information between other external network elements; the memory is used to store computer program instructions that can be run on the processor; the processor is used to execute the steps of a distribution network line transfer identification method based on CatBoost provided in Example 1 when running the computer program instructions.

[0232] Embodiment 4:

[0233] The present embodiment also provides a computer storage medium, which stores a computer program, and the method described above can be implemented when the processor executes the computer program. The computer readable medium can be considered to be tangible and non-temporary. Non-limiting examples of non-temporary tangible computer-readable media include non-volatile memory circuits (such as flash memory circuits, erasable programmable read-only memory circuits or mask read-only memory circuits), volatile memory circuits (such as static random access memory circuits or dynamic random access memory circuits), magnetic storage media (such as analog or digital tapes or hard drives) and optical storage media (such as CDs, DVDs or Blu-ray discs), etc. The computer program includes processor executable instructions stored on at least one non-temporary tangible computer-readable medium. The computer program may also include or rely on stored data. The computer program may include a basic input / output system (BIOS) that interacts with the hardware of a special-purpose computer, a device driver that interacts with a specific device of a special-purpose computer, one or more operating systems, user applications, background services, background applications, etc.

[0234] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0235] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0236] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0237] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

Claims

1. A method for identifying power transfer in a distribution network based on CatBoost, characterized in that: The following steps are involved: Step S1: obtaining the topological structure of the line to be diagnosed, the distribution transformer account, the relationship between the line and the tie switch, the opposite line account connected to the tie switch, and the measurement data of the line to be diagnosed, the opposite line and the distribution transformer under the line, and the distribution transformer is connected to the line; Step S2: Obtain a set of topological relationships corresponding to the line to be diagnosed, the tie switch, the key topological path, the smallest upstream segment of the tie switch, and the opposite end line connected to the tie switch through the key topological path of the line to be diagnosed; Step S3: traverse the upstream minimum segment set of the line tie switch to be diagnosed, extract the characteristic index of each distribution transformer in the upstream minimum segment of the tie switch with the same parameters as the trained distribution transformer power transfer identification model based on CatBoost ensemble learning in the period before the current moment, apply the distribution transformer power transfer identification model based on CatBoost ensemble learning online according to the extracted characteristic index, and output whether the distribution transformer in the upstream minimum segment of the tie switch has power transfer at the current moment; Step S4: according to the output data of the distribution transformer transfer identification model based on CatBoost ensemble learning, the distribution transformer transfer voting probability in each upstream minimum segment of the tie switch is calculated. If the distribution transformer transfer voting probability in the upstream minimum segment of any tie switch exceeds the threshold, the line to be diagnosed has transferred power at the current moment; otherwise, the line to be diagnosed has not transferred power at the current moment. Step S5: for the line to be diagnosed that is currently transferring power, a line transfer area positioning and identification model is constructed, and a set of substitute supply lines and a set of distribution transformers in the transfer area are output; In step S2, the specific process of obtaining the topological relationship set corresponding to the line to be diagnosed, the tie switch, the key topological path, the smallest upstream segment of the tie switch, and the opposite end line connected to the tie switch through the key topological path of the line to be diagnosed includes: By traversing the tie switch set KG on the line to be diagnosed XL, a key topological path set A consisting of the circuit breaker nodes connecting the line to be diagnosed and each tie switch node is obtained, wherein each set in the set A contains multiple minimum segments, and each minimum segment contains a set of more than 3 distribution transformers; By traversing the key topological path set A and the tie switch set KG, the minimum segment set B upstream of the tie switch is obtained; Using the relationship between lines and tie switches, traverse the tie switch set KG on the line to be diagnosed, and obtain the opposite line set C connected to the tie switch; Finally, a topological relationship set D corresponding to the line to be diagnosed XL, the tie switch set KG, the key topological path set A, the tie switch upstream minimum segment set B, and the opposite line set C is formed; Step S4 specifically includes: Traverse the minimum segment set B upstream of the XL tie switch of the line to be diagnosed, and calculate the ratio of the number of distribution transformers in each upstream minimum segment of the tie switch to the total number of distribution transformers in the upstream minimum segment of the tie switch according to the output data of the distribution transformer transfer identification model based on CatBoost ensemble learning, and use it as the distribution transformer transfer voting probability in the upstream minimum segment of the tie switch; If the probability of the distribution transformer switching in any of the upstream minimum sections of the tie switch exceeds the threshold, the line to be diagnosed XL switches power at the current moment, and outputs the set of upstream minimum sections of the tie switch where the switching occurs. Otherwise, the line XL to be diagnosed has not been transferred at the current moment; Step S5 specifically includes: Step S5-1: For the line to be diagnosed XL that is currently transferring power, obtain the minimum segment set upstream of the tie switch that is transferring power Corresponding contact switch set Critical topological path collection Peer Line Collection Step S5-2: Traverse the set of tie switches of the line XL to be diagnosed Get the mth tie switch kg m Corresponding key topological path Critical topological patha m Contains m The smallest segment b upstream of the tie switch where the transfer occurs is selected m The other smallest upstream segment Step S5-3: divide the upstream other smallest segments Each distribution transformer applies the trained CatBoost ensemble learning-based distribution transformer transfer identification model online to output the other minimum segments upstream. Whether the internal distribution transformer is transferring power at the current moment; Step S5-4: Calculate other upstream minimum segments The internal distribution transformer transfer voting probability, if the upstream other minimum segment If the probability of voting for power transfer exceeds the threshold, the other upstream minimum segments The transfer occurs at the current moment; Step S5-5: Continue to filter out the smallest segment According to steps S5-3 to S5-4, it is determined whether the other upstream smallest segments are currently being transferred, until a certain upstream smallest segment is If there is no transfer or the upstream smallest segment e1 is traced back, the upstream segment tracing will stop, and the mth contact switch kg will be formed. m The corresponding transfer area is Step S5-6: Traverse the set of tie switches of the line XL to be diagnosed The set of distribution transformers in the output transfer area is F = {f1,f2,...,f m }, the corresponding supply line set is 2. According to a CatBoost-based distribution network line transfer identification method according to claim 1, it is characterized in that: The step S3 specifically includes: Step A: Using the historical samples marked with whether the power supply is transferred, extract multiple groups of characteristic indicators under different parameters of the distribution transformer and the local line and the opposite line before and after the power supply transfer, perform offline training and model comparison of the distribution transformer power supply transfer identification model based on CatBoost ensemble learning, save the trained model and characteristic indicator parameters and apply them to the subsequent distribution transformer power supply transfer identification; Step B: Traverse the set of upstream minimum segments of the line tie switch to be diagnosed, extract the characteristic indicators of each distribution transformer in the upstream minimum segment of the tie switch with the same parameters as the trained model in the period before the current moment, apply the trained distribution transformer power transfer identification model based on CatBoost ensemble learning online, and output whether the distribution transformer in the upstream minimum segment of the tie switch is currently transferring power.

3. According to a CatBoost-based distribution network line transfer identification method according to claim 2, it is characterized in that: The specific process of step A includes: Step A1: Obtain the local line ledger, the opposite line ledger, the local distribution transformer ledger, the opposite distribution transformer ledger, the local distribution transformer whether to transfer the supply label, and the line and distribution transformer power and voltage measurement data that are marked with historical samples of whether to transfer the supply; Step A2: extracting characteristic indicators of the distribution transformer, the local line and the opposite line for a period of time before and after the power transfer by using a sliding window method with different parameters, including multiple groups of characteristic indicators with different parameters; Step A3: A random partitioning method is used to select a part of the samples in the historical sample data set as the training set, and the remaining samples are used as the test set. The model training and comparison are performed for multiple groups of feature indicators with different parameters: the feature indicators in the training set are used as the model input, and whether the distribution transformer is switching power is used as the model output. The model is input into the CatBoost ensemble learning algorithm to train the distribution transformer switching recognition model based on CatBoost ensemble learning; Step A4: Apply the trained model to the test set, use the F1 score of the harmonic mean of precision and recall to evaluate the model, and select the one with the largest F1 score that exceeds the threshold ε. F The model and feature index parameters of the model; the definition of the F1 score is as follows: Where, TP is the number of distribution transformers that are actually transferring power and predicted by the model to be transferring power; FP is the number of distribution transformers that are actually not transferring power and predicted by the model to be transferring power; FN is the number of distribution transformers that are actually transferring power and predicted by the model to be not transferring power; Step A5: Save the distribution transformer power transfer identification model and characteristic index parameters based on CatBoost ensemble learning trained in step A4, and apply them to the subsequent distribution transformer power transfer identification.

4. According to claim 3, a method for identifying power distribution network line transfer based on CatBoost, characterized in that: The specific process of step A2 includes: Step A2-1: Using a sliding window method with different parameters, the voltage and power data of the distribution transformer and the local and opposite lines for a period of time before and after the power transfer are processed to generate multiple sets of data with different parameters; Step A2-2: traverse multiple sets of data with different parameters, extract characteristic indicators of the distribution transformer and the local line and the opposite line for each set of data with different parameters, and generate multiple sets of characteristic indicators for different parameters.

5. The method for identifying power distribution network line transfer based on CatBoost according to claim 4, characterized in that: The specific process of step A2-1 includes: Assume that the measurement data sampling interval is t minutes, and the sliding starts from yt minutes after the transfer start time, the sliding step is zt minutes, the sliding window width is wt minutes, and the sliding is k times, and the sliding ends xt minutes before the transfer start time. The end sampling time of the sliding window is t1, t2, ..., t j ,...,t k , forming a sliding window method with different parameters; For the sliding window method of the same set of parameters, the wt-minute standardized voltage data of the local distribution transformer, local line, and opposite line of the j-th sliding are obtained. After sliding k times, k sets of wt-minute standardized voltage data sets of the local distribution transformer, local line, and opposite line are generated; at the same time, the end sampling time points of the sliding window are obtained as t1, t2, ..., t j ,...,t k The power data set of the local distribution transformer, the opposite distribution transformer, the local line, and the opposite line; By traversing the sliding window method of different parameters, multiple sets of voltage data sets and power data sets with different parameters are generated.

6. A method for identifying power distribution network line transfer based on CatBoost according to claim 4, characterized in that: In step A2-2, each group of characteristic indicators includes: voltage correlation coefficient difference sequence, voltage difference standard deviation ratio sequence, voltage Euclidean distance ratio sequence, local line line loss rate sequence, opposite line line loss rate sequence, and package line loss rate sequence. Each sequence indicator has k characteristic indicators, and a total of 6k characteristic indicators. The indicator calculation is specifically as follows: Voltage correlation coefficient difference sequence: Based on the wt-minute standardized voltage data of the local distribution transformer, local line, and opposite line of the j-th sliding, the voltage correlation coefficient between the local distribution transformer and the local line, and the voltage correlation coefficient between the local distribution transformer and the opposite line are calculated using the Pearson correlation coefficient formula, and then the difference between the voltage correlation coefficients between the local distribution transformer and the local line and the opposite line is calculated; the voltage correlation coefficient difference sequence formed after sliding k times is used as the voltage correlation coefficient difference sequence; Ratio sequence of voltage difference standard deviation: Based on the wt-minute standardized voltage data of the local distribution transformer, local line, and opposite line of the j-th sliding, calculate the difference between the voltage of the local distribution transformer and the local line, and then calculate its standard deviation as the standard deviation of the voltage difference between the local distribution transformer and the local line; similarly calculate the standard deviation of the voltage difference between the local distribution transformer and the opposite line; finally calculate the ratio of the standard deviation of the voltage difference between the local distribution transformer and the local line and the opposite line; the ratio sequence of the voltage difference standard deviation formed after sliding k times is used as the ratio sequence of the voltage difference standard deviation; Voltage Euclidean distance ratio sequence: Based on the wt-minute standardized voltage data of the local distribution transformer, local line, and opposite line of the j-th sliding, the Euclidean distance between the local distribution transformer and the local line voltage, and the Euclidean distance between the local distribution transformer and the opposite line voltage are calculated respectively; then the ratio of the Euclidean distance between the local distribution transformer and the local line and the opposite line is calculated; the voltage Euclidean distance ratio sequence formed after sliding k times is used as the voltage Euclidean distance ratio sequence; Line loss rate sequence of the local line: set the power greater than or equal to 0 as forward, and the power less than 0 as reverse. Calculate the sum of the forward power and reverse power of all distribution transformers under the local line, and calculate the line forward power and line reverse power of the local line. Use the rectangular area method to calculate the line at the sampling time point t j The forward power of the line minus the reverse power of the distribution transformer multiplied by the sampling time interval is taken as the input power. j The output power is the sum of the forward power of the distribution transformer minus the reverse power of the line multiplied by the sampling time interval. The input power minus the output power is defined as the loss power. The loss power divided by the input power is defined as the line at the sampling time point t j The time-sharing line loss rate of the local line is t1, t2, ..., t j ,...,t k The time-sharing line loss rate sequence is used as the line loss rate sequence of the local line; The line loss rate sequence of the other end line: The calculation logic of the line loss rate of the other end line is the same as that of the local line line. The sampling time points of the other end line are t1, t2, ..., t j ,...,t k The time-sharing line loss rate sequence is used as the opposite line line loss rate sequence; Package line loss rate sequence: The local line and the opposite line are packaged together as a packaged line, and the local line and the opposite line are sampled at the sampling time point t j The input power of the packaged line is added as the total input power of the packaged line, and the local line and the opposite line are sampled at the sampling time point t j The total power loss of the packaging line is added as the total power loss of the packaging line, and the total power loss of the packaging line divided by the total input power is defined as the power loss of the packaging line at the sampling time point t j The time-sharing line loss rate; the packaging line is sampled at time points t1, t2, ..., t j ,...,t k The time-sharing line loss rate sequence is used as the packaging line loss rate sequence.

7. A method for identifying power distribution network line transfer based on CatBoost according to claim 2, characterized in that: The step B specifically comprises: Traverse the minimum segment set B upstream of the XL tie switch of the line to be diagnosed, and extract the characteristic index of each distribution transformer in the upstream minimum segment of the tie switch with the same parameters as the trained model in the period before the current moment: for each distribution transformer in the upstream minimum segment b of a tie switch, start sliding from the current moment, the sliding step is zt minutes, the sliding window width is wt minutes, slide k times, and end the sliding at (x+y)t minutes before the current moment. The end sampling time point of the sliding window is t1, t2, ..., t j ,...,t k ; Each time the sliding is performed, the standardized voltage data of the local distribution transformer, the local line, and the opposite line after wt minutes are obtained; at the same time, the end sampling time point of the sliding window is obtained as t1, t2, ..., t j ,...,t k The power data of the local distribution transformer, the opposite distribution transformer, the local line, and the opposite line; the calculation logic of the characteristic indicators of model training is the same, and 6k characteristic indicators with the same parameters as the trained model can be calculated. The characteristic indicators include: voltage correlation coefficient difference sequence, voltage difference standard deviation ratio sequence, voltage Euclidean distance ratio sequence, local line line loss rate sequence, opposite line line loss rate sequence, and package line loss rate sequence; The trained distribution transformer transfer identification model based on CatBoost ensemble learning is applied online to output whether the distribution transformer in the smallest section upstream of the tie switch is transferring power at the current moment.

8. The identification system of the CatBoost-based distribution network line transfer identification method according to claim 1, characterized in that: include: Data acquisition module: obtain the topological structure of the line to be diagnosed, the distribution transformer account, the relationship between the line and the tie switch, the opposite line account connected to the tie switch, and the measurement data of the line to be diagnosed, the opposite line and the distribution transformer under the line; obtain the corresponding topological relationship set of the line to be diagnosed, the tie switch, the key topological path, the upstream minimum segment of the tie switch, and the opposite line connected to the tie switch through the key topological path of the line to be diagnosed; Segmented power transfer identification module: traverses the upstream minimum segment set of the line tie switch to be diagnosed, extracts the characteristic index of each distribution transformer in the upstream minimum segment of the tie switch with the same parameters as the trained distribution transformer power transfer identification model based on CatBoost ensemble learning in the period before the current moment, applies the distribution transformer power transfer identification model based on CatBoost ensemble learning online according to the extracted characteristic index, and outputs whether the distribution transformer in the upstream minimum segment of the tie switch has power transfer at the current moment; Line transfer identification module: According to the output data of the distribution transformer transfer identification model based on CatBoost ensemble learning, the distribution transformer transfer voting probability in each upstream minimum segment of the tie switch is calculated. If the distribution transformer transfer voting probability in the upstream minimum segment of any tie switch exceeds the threshold, the line to be diagnosed has transferred power at the current moment, otherwise, the line to be diagnosed has not transferred power at the current moment; Transfer area identification module: For the lines to be diagnosed that are currently transferring supply, a line transfer area positioning and identification model is constructed to output the set of substitute supply lines and the set of distribution transformers in the transfer area.

9. A distribution network line transfer identification device based on CatBoost, characterized in that: Includes network interface, memory and processor; The network interface is used to receive and send signals during the process of sending and receiving information with other external network elements; The memory is used to store computer program instructions that can be executed on the processor; The processor is used to execute the steps of a distribution network line transfer identification method based on CatBoost according to any one of claims 1 to 7 when running the computer program instructions.

10. A computer storage medium, characterized in that: The computer storage medium stores a program of a distribution network line transfer identification method based on CatBoost, and when the program of a distribution network line transfer identification method based on CatBoost is executed by at least one processor, the steps of a distribution network line transfer identification method based on CatBoost described in any one of claims 1 to 7 are implemented.

Images