Feeder automation strategy debugging method and system based on neural network

Abstract

The invention discloses a feeder automation strategy debugging method and system based on a neural network. The method comprises the following steps: establishing an FA debugging problem expert database, wherein the FA debugging problem expert database comprises an FA strategy error matrix and a data error matrix which have a corresponding relation; performing neural network construction and training based on the FA debugging problem expert database to obtain a trained neural network; and acquiring FA strategy error data generated in a line debugging process, inputting the FA strategy error data into the trained neural network, outputting the probability that a corresponding type of data error exists in the debugging line database, and generating a data error indication vector based on the probability that the corresponding type of data error exists in the debugging line database and a preset probability threshold. According to the method, the data error type in the database can be positioned through the FA strategy error type occurring in the debugging process, the FA strategy debugging time is greatly shortened, and the FA strategy deployment work quality and efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of feeder automation strategy debugging and deployment, and particularly to a method and system for debugging feeder automation strategies based on a neural network. Background Art

[0002] With the development of smart grids, the feeder automation system (FA) has become the core technology for rapid fault handling in 10kV distribution lines. Based on data such as line topology relationships and addresser action signals, the FA system can automatically generate strategies for fault location, isolation, and power restoration, significantly improving the reliability and operation and maintenance efficiency of the distribution network. However, the generation of FA strategies highly depends on the accuracy of data such as line topologies and equipment parameters in the distribution automation master station database. If the data is incorrect, it may lead to incorrect fault handling strategies generated by the FA, and even risks such as incorrect isolation and incorrect operation.

[0003] Currently, before the deployment of FA strategies, debugging is required by injecting faults through the master station to verify the correctness of the strategies and troubleshoot data errors. However, the following problems exist in actual projects: ‌Low debugging efficiency‌: The debugging personnel of each power supply company are mostly dispatchers, and they lack an in-depth understanding of the internal logic of the FA program and the data association mechanism. They can only locate errors by manually traversing all data related to the line, which is time-consuming and laborious; ‌Difficult error location‌: The existing debugging methods cannot directly infer the specific incorrect data in the database from the errors of the FA strategies, resulting in a lack of pertinence in the data error correction process; ‌Insufficient engineering research‌: The current research mainly focuses on the optimization of FA control logic and algorithm implementation, but lacks a systematic solution to key issues such as engineering debugging methods and data verification technologies in large-scale FA deployment, and it is difficult to meet the requirements for rapid deployment and reliable operation in actual projects.

[0004] The above problems severely restrict the large-scale application of the FA system. There is an urgent need to propose an efficient and accurate FA debugging method and data verification mechanism to reduce the debugging complexity and improve the deployment efficiency. Summary of the Invention

[0005] Aiming at the technical problems pointed out in the above background art, the purpose of the present invention is to provide a method and system for debugging feeder automation strategies based on a neural network.

[0006] To achieve the purpose of the present invention, the technical solutions provided by the present invention are as follows: First Aspect The present invention provides a method for debugging feeder automation strategies based on a neural network, including the following steps: Step S1: Establish an expert library for FA debugging problems. The expert library for FA debugging problems includes an FA policy error matrix and a data error matrix with a corresponding relationship. Step S2: Based on the expert library for FA debugging problems, construct and train a neural network to obtain a trained neural network. Step S3: Collect FA policy error data generated during the line debugging process and input it into the trained neural network to output the probability of data errors of the corresponding type existing in the debugging line database. And based on the probability of data errors of the corresponding type existing in the debugging line database and a preset probability threshold, generate a data error indication vector.

[0007] Further, it further includes Step S4: Based on the data error indication vector, calculate the correlation accuracy rate, and according to the correlation accuracy rate, optimize the parameters of the trained neural network and the preset probability threshold.

[0008] Further, Step S4 specifically includes the following: Step 4.1: Compare the data error indication vector with the actual data error vector, and use their ratio as the correlation accuracy rate. Step 4.2: Adjust the number of neurons in the hidden layer of the neural network and the preset probability threshold, and repeat the calculation of the correlation accuracy rate until the correlation accuracy rate meets the usage requirements, and use the current number of neurons in the hidden layer of the neural network and the preset probability threshold for subsequent calculations.

[0009] Further, it further includes Step S5: Expand the FA policy error vector and the actual data error vector generated during the line debugging process into the expert library for FA debugging problems, and further train the trained neural network.

[0010] Further, Step S1 specifically includes the following: Step S1.1: Obtain the FA policy errors in the FA historical debugging records of the debugging line and the data errors in the database and classify them. Step S1.2: Establish an FA policy error vector and a data error vector for each debugging line. Step S1.3: Compose all the established FA policy error vectors into an FA policy error matrix, compose the corresponding data error vectors into a data error matrix, and write the two matrices into Matlab. The two matrices and their corresponding relationships constitute the expert library for FA debugging problems.

[0011] Further, in step S1.1, the FA policy errors are divided into 11 categories, specifically including: failure to read the set fault, reading an error fault, expansion of the fault isolation range, incomplete transfer supply policy, inability to identify false dual users, incorrect load rate reading, non - line - specific switch in the operation steps, inability to identify automated stations, incorrect optimal transfer supply plan, inability to pop up the policy, and remote control policy given by non - automated stations; The data errors in the database are divided into the following 14 categories, specifically including: unassociated device diagram models, different node numbers at the connection points of adjacent devices, repeated node numbers in this line, the same node numbers existing in different lines in the system, the corresponding switch column data in the protection node table being empty, two pieces of data for the same signal in the pre - placed tele - signal definition table, incorrect data in the signal type column of the protection node table, incorrect data in the coefficient column of the pre - placed tele - measurement definition table, incorrect data in the switch column of the four - area related table, incorrect data in the feeder column of the distribution network load table, incorrect data in the feeder column of the distribution network switch table, incorrect data in the feeder column of the distribution network feeder section table, incorrect data in the feeder column of the distribution network bus table, and non - database problems.

[0012] Further, step S2 specifically includes the following: Step S2.1: Determine the neural network structure; 11 neurons are set in the input layer of the neural network to input the FA policy error vector; 25 neurons are set in the hidden layer, and the Sigmoid activation function is used; 14 neurons are set in the output layer, and the ReLU activation function is used, and finally the probability of 14 - bit data errors is output; Step S2.2: Use the FA policy error matrix and the data error matrix in the FA debugging problem expert database as the sample input and sample output respectively to train the neural network, and obtain the trained neural network.

[0013] Further, step S3 specifically includes the following: Step S3.1: Collect the FA policy error data generated during the line debugging process; Step S3.2: Input the collected FA policy error data into the trained neural network, and output the probability that there are corresponding type data errors in the database of this debugging line; Step S3.3: Compare the probability that there are corresponding type data errors in the database of this debugging line with the preset probability threshold. If the row data is greater than or equal to the preset probability threshold, set it to 1, and if it is less than the probability threshold, set it to 0, to obtain a data error indication vector indicating the data error types existing in the database of this debugging line.

[0014] Second aspect The present invention provides a feeder automation strategy debugging system based on a neural network, including: an FA debugging problem expert database establishment module, a neural network construction and training module, and a data error indication vector generation module; The FA debugging problem expert database establishment module is used to establish an FA debugging problem expert database, and the FA debugging problem expert database includes an FA policy error matrix and a data error matrix with a corresponding relationship; The neural network construction and training module is used to construct and train a neural network based on the FA debugging problem expert database to obtain a trained neural network; The data error indication vector generation module is used to collect FA policy error data generated during the line debugging process, input it into the trained neural network, output the probability that there are data errors of the corresponding type in the debugging line database, and generate a data error indication vector based on the probability that there are data errors of the corresponding type in the debugging line database and a preset probability threshold.

[0015] Furthermore, it further includes an optimization module, which is used to calculate the correlation accuracy rate based on the data error indication vector, and optimize the parameters of the trained neural network and the preset probability threshold according to the correlation accuracy rate.

[0016] Compared with the prior art, the present invention can display the correlation between the FA policy error type and the data error type with a correlation accuracy rate exceeding 95%, which can help FA policy debuggers locate the data error type in the database through the FA policy error type that appears during the debugging process, overcome the difficulty that debuggers can only use the traversal method to eliminate data errors without understanding the internal principle of the FA program, greatly shorten the FA policy debugging time, improve the quality and efficiency of the FA policy deployment work, and the process of the present invention is simple. After the neural network training is completed, it has high portability, can assist the 10kV distribution line FA policy deployment of each power supply enterprise, and thus promote the construction work of the new power system. Description of the Drawings

[0017] Figure 1 It is a schematic diagram of the method flow provided by the embodiment of the present invention; Figure 2 It is a graph showing the relationship between the correlation accuracy rate and parameters in the embodiment of the present invention. Detailed Embodiment

[0018] Next, in combination with the drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0019] The present invention discloses a method and system for debugging feeder automation strategies based on a neural network. The method analyzes and summarizes historical debugging records, establishes an expert database for FA debugging problems, trains a neural network using the expert database for FA debugging problems, analyzes the correlation between FA strategy errors and database data errors in the expert database, and uses the trained neural network to guide FA strategy debugging, locates database data errors through FA strategy errors, improves the efficiency of FA strategy debugging, speeds up the FA deployment speed, and improves the power supply reliability of the distribution network. Embodiment 1

[0020] As Figure 1 shown, this embodiment provides a method for debugging feeder automation strategies based on a neural network, including the following steps: Step S1: Establish an expert database for FA debugging problems, where the expert database for FA debugging problems includes an FA strategy error matrix and a data error matrix with a corresponding relationship; Preferably, step S1 specifically includes the following: Step S1.1: Obtain FA strategy errors in the FA historical debugging records of the debugging line and data errors in the database and classify them; Specifically, the FA historical debugging records of 282 debugging lines debugged by the power grid enterprise were analyzed, and the FA strategy errors and data errors in the database during the debugging process were classified according to their characteristics.

[0021] The FA strategy errors are divided into 11 categories, specifically including: (1) Failure to read the set fault: The phenomenon is that after a fault indicator at a certain location on the line acts, FA determines that the fault point is at the outlet cable.

[0022] (2) Reading an incorrect fault: The phenomenon is that after a fault indicator at a certain location on the line acts, FA determines that the fault point is not within the range indicated by the fault indicator.

[0023] (3) Expansion of the fault isolation range: The phenomenon is that in the "Fault Area Determination" column on the "Fault Summary" page of FA, an expanded fault isolation range is displayed, and the isolation operation given is to open the previous three-remote switch of the last-acting fault indicator.

[0024] (4) Incomplete transfer supply strategy: The phenomenon is that not all operation strategies for three-remote tie switches are given on the "Load Transfer Supply" page of FA.

[0025] (5) Unable to identify false dual users: The phenomenon is that there is a transfer supply scheme on the line that causes false dual power supply for users, but the preferred scheme in the "Downstream Restoration Scheme" column on the "Fault Summary" page of FA will cause false dual power supply for users.

[0026] (6) Load rate reading error: The phenomenon is that the load rate of the transfer line on the FA load transfer page does not match the actual situation.

[0027] (7) The switch of a non-target line appears in the operation steps: The phenomenon is that the three-remote switch given in the operation strategy on the FA "fault isolation" or "load transfer" page is not on the debugging line.

[0028] (8) Unable to identify an automated site: The phenomenon is that FA identifies a three-remote switch as a non-three-remote switch, and no operation strategy for this switch is given on the FA "fault summary", "fault isolation" or "load transfer" page in any fault situation.

[0029] (9) Incorrect optimal transfer plan: The phenomenon is that the preferred plan in the "downstream recovery plan" column on the FA "fault summary" page will select a line with important users for transfer.

[0030] (10) Unable to pop up the strategy: The phenomenon is that no corresponding strategy is generated on the FA interface after simulating a fault.

[0031] (11) Remote control strategy given for a non-automated site: The phenomenon is the operation strategy of a non-three-remote switch on the FA "fault isolation" or "load transfer" page.

[0032] Data errors in the database are divided into the following 14 categories, specifically including: (1) Device graphic and model not associated: The graphic and model of the device are not associated.

[0033] (2) Different node numbers for connection points of adjacent devices: The adjacent device graphics are not connected or the node numbers of the common connection points of adjacent devices are inconsistent.

[0034] (3) Repeated node numbers on the same line: Two devices that are not adjacent and belong to the same line have the same node number.

[0035] (4) The same node numbers exist in different lines of the system: Two devices that are not adjacent and belong to different lines have the same node number.

[0036] (5) The data in the corresponding switch column of the protection node table is empty: The corresponding switch column in the protection node table is not associated with the device model.

[0037] (6) Two pieces of data for the same signal in the front-end telemetry definition table: There are two rows of information with the same point number for a certain device in the front-end telemetry definition table.

[0038] (7) Incorrect data in the signal type column of the protection node table: The type column data of the fault merging or other indication fault signal rows of the relevant switches in the protection node table is "other", which should be "fault signal".

[0039] (8) There is an error in the coefficient column data in the pre - telemetry definition table: The telemetry coefficient of the relevant switch in the pre - telemetry definition table does not match the actual situation.

[0040] (9) There is an error in the switch column data in the four - zone related class table: The switch column in the four - zone related class table is not associated with the device model.

[0041] (10) There is an error in the feeder column data of the distribution network load table: The feeder column data of the distribution network load table of the users on the debugging line is not the name of the debugging line.

[0042] (11) There is an error in the feeder column data of the distribution network switch table: The feeder column data of the distribution network switch table of the switches on the debugging line is not the name of the debugging line.

[0043] (12) There is an error in the feeder column data of the distribution network feeder section table: The feeder column data of the distribution network feeder section table of the feeder sections on the debugging line is not the name of the debugging line.

[0044] (13) There is an error in the feeder column data of the distribution network bus table: The feeder column data of the distribution network bus section table of the switching station bus on the debugging line is not the name of the debugging line.

[0045] (14) Non - database problem: The server crashes or the memory occupancy rate is too high during debugging.

[0046] It should be noted here that the above FA page names, database structures, and names are applicable to the distribution automation system to which they are applied.

[0047] Step S1.2: Establish an FA policy error vector and a data error vector for each debugging line; Specifically, establish an FA policy error vector and a data error vector for each debugging line in the form of Table 1. For example, if a line has the first - type problem of "failure to read set faults" and the fifth - type problem of "unable to identify false dual users", then the FA policy error vector is [1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0] T , if the database problem types corresponding to these two FA policy errors are the second - type of "different node numbers at the connection points of adjacent devices", the seventh - type of "error in the signal type column data of the protection node table", and the eighth - type of "error in the signal type column data of the protection node table", then the data error vector is [0, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0] T .

[0048] Table 1

[0049] Step S1.3: Combine all the FA policy error vectors established to form an FA policy error matrix, combine the corresponding data error vectors to form a data error matrix, and write the two matrices into Matlab. The two matrices and their corresponding relationships constitute an expert library for FA debugging problems.

[0050] Specifically, combine all the FA policy error vectors of 282 debugged lines to form an FA policy error matrix with 11 rows and 282 columns. Similarly, combine the corresponding data error vectors in sequence to form a data error matrix with 14 rows and 282 columns, and write the two matrices into Matlab. The above two matrices and their corresponding relationships are the expert library for FA debugging problems.

[0051] Step S2: Based on the expert library for FA debugging problems, construct and train a neural network to obtain a trained neural network. The above step S2 specifically includes the following: Step S2.1: Determine the neural network structure; set 11 neurons in the input layer of the neural network to input the FA policy error vectors; set 25 neurons in the hidden layer and use the Sigmoid activation function; set 14 neurons in the output layer and use the ReLU activation function, and finally output the probability of 14-bit data errors; this neural network is implemented through a program example in Matlab.

[0052] Among them, the number of neurons in the input layer and the output layer is the same as the number of rows of the FA policy error vector and the data error vector respectively, and the number of neurons in the hidden layer is an empirical value. In addition, the output of the neural network should show the probability of various types of data errors occurring, and should be a percentage value between 0 and 1. Therefore, the Sigmoid function and the ReLU function are selected as the activation functions for the hidden layer and the output layer respectively, and their expressions are shown in formulas (1) and (2).

[0053] (1) (2) In the formula, x is the input value of the activation function of the hidden layer and the output layer, S ( x ) is the output value of the activation function of the hidden layer, F ( x ) is the output value of the activation function of the output layer.

[0054] Step S2.2: Use the FA policy error matrix and the data error matrix in the expert library for FA debugging problems as the sample input and sample output respectively to train the neural network to obtain a trained neural network.

[0055] Specifically, the FA policy error matrix and data error matrix in the FA debugging problem expert library are used as sample inputs and sample outputs respectively to train the neural network. The gradient descent algorithm with momentum and a changing learning rate is adopted. The learning rate is set to 0.01, the maximum number of iterations is 1000, and the target value is 0.001. The accuracy of the model reaches the target value after 176 iterations, and the training is completed.

[0056] Step S3: Collect the FA policy error data generated during the line debugging process and input it into the trained neural network to output the probability of data errors of the corresponding type existing in the debugging line database. Based on the probability of data errors of the corresponding type existing in the debugging line database and a preset probability threshold, a data error indication vector is generated.

[0057] Furthermore, step S3 specifically includes the following: Step S3.1: Collect the FA policy error data generated during the line debugging process; Step S3.2: Input the collected FA policy error data into the trained neural network to output the probability of data errors of the corresponding type existing in the debugging line database; In the above steps, the FA policy errors generated during the debugging of a certain 10 kV line are classified, and the FA policy error vector is established according to steps S1.1 - S1.2 and written into the Data data table. For example, during the debugging of a certain line, there are two types of problems: "expansion of the fault isolation range" and "failure to identify the automation site", and its FA policy error vector is recorded as [0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0], as shown in column B of Table 2. The FA policy error vector in the Data data table is read through the Matlab program, input into the neural network for calculation, and then the output data is written into the Data data table, as shown in column E of Table 2. The data shows that the probabilities of 14 types of data errors existing in the database related to the debugging line are 7.31%, 2.73%, 3.88%, 79.11%, 85.79%, 3.94%, 6.26%, 13.62%, 11.61%, 7.85%, 14.77%, 8.40%, 80.89%, and 4.90% respectively. T As shown in column B of Table 2. The FA policy error vector in the Data data table is read through the Matlab program, input into the neural network for calculation, and then the output data is written into the Data data table, as shown in column E of Table 2. The data shows that the probabilities of 14 types of data errors existing in the database related to the debugging line are 7.31%, 2.73%, 3.88%, 79.11%, 85.79%, 3.94%, 6.26%, 13.62%, 11.61%, 7.85%, 14.77%, 8.40%, 80.89%, and 4.90% respectively.

[0058] Table 2

[0059] Step S3.3: Compare the probability of data errors of the corresponding type existing in the debugging line database with the preset probability threshold. If the row data is greater than or equal to the preset probability threshold, it is set to 1; if it is less than the probability threshold, it is set to 0, to obtain a data error indication vector for indicating the types of data errors existing in the debugging line database.

[0060] Specifically, the output of the neural network is a 14-row vector, and each row of the vector is a decimal number with a value from 0 to 1, representing the probability of data errors of the corresponding type existing in the debugging line database. By using the set preset probability threshold, the row data with a probability greater than or equal to the probability threshold is set to 1, and the row data with a probability less than the probability threshold is set to 0, thus obtaining a vector indicating the types of data errors existing in the debugging line database, denoted as the data error indication vector, as shown in Table 2, where the preset probability threshold takes the empirical value of 50%.

[0061] Furthermore, it further includes step S4: Based on the data error indication vector, calculate the correlation accuracy rate, and optimize the trained neural network parameters and the preset probability threshold according to the correlation accuracy rate.

[0062] Furthermore, step S4 specifically includes the following:[[]]END]] Step S4.1: Compare the data error indication vector with the actual data error vector, and take their ratio as the correlation accuracy rate. Specifically, conduct FA debugging on 30 10kV distribution lines with FA strategy errors, summarize and analyze the debugging reports to obtain 30 groups of FA strategy vulnerability vectors and data error vectors, and calculate 30 groups of data error indication vectors. Compare the data error indication vectors with the actually excluded data error vectors. The results show that the data error indication vectors of 23 lines are consistent with the data error vectors, and the correlation accuracy rate is 23 / 30 = 76.67%.

[0063] Step S4.2: Adjust the number of neurons in the hidden layer of the neural network and the preset probability threshold, repeat the calculation of the correlation accuracy rate until the correlation accuracy rate meets the usage requirements, and use the current number of neurons in the hidden layer of the neural network and the preset probability threshold for subsequent calculations.

[0064] In order to further improve the correlation accuracy rate, adjust the neural network parameters and the probability threshold, repeat the operation of step S4.1 for the above 30 10kV distribution line numbers, and the relationship between the correlation accuracy rate and the parameters is as Figure 2 shown. The results show that after setting the number of neurons in the hidden layer of the neural network to 23 and setting the preset probability threshold to 60%, the correlation accuracy rate reaches the maximum value of 96.67%, meeting the relevant work requirements, and determining the parameters.

[0065] Furthermore, it further includes step S5: Expand the FA strategy error vector and the actual data error vector generated during the line debugging process into the FA debugging problem expert database, further train the trained neural network, further improve the correlation accuracy rate, guide the FA strategy debugging work, and accelerate the promotion of FA deployment.

[0066] Example 2 Corresponding to the above method, this embodiment provides a feeder automation strategy debugging system based on a neural network, including: an FA debugging problem expert database establishment module, a neural network construction and training module, and a data error indication vector generation module; The FA debugging problem expert database establishment module is used to establish an FA debugging problem expert database, and the FA debugging problem expert database includes an FA policy error matrix and a data error matrix with a corresponding relationship; The neural network construction and training module is used to construct and train a neural network based on the FA debugging problem expert database to obtain a trained neural network; The data error indication vector generation module is used to collect FA policy error data generated during the line debugging process, input it into the trained neural network, output the probability that there are data errors of the corresponding type in the debugging line database, and generate a data error indication vector based on the probability that there are data errors of the corresponding type in the debugging line database and a preset probability threshold.

[0067] Further, it further includes an optimization module, and the optimization module is used to calculate the association accuracy rate based on the data error indication vector, and optimize the parameters of the trained neural network and the preset probability threshold according to the association accuracy rate.

[0068] Finally, it should be noted that: the above embodiments are only used for exemplifying and illustrating the present invention, and are not intended to limit the present invention to the scope of the described embodiments. In addition, those skilled in the art can understand that the present invention is not limited to the above embodiments, and more variations and modifications can be made according to the teachings of the present invention, and these variations and modifications all fall within the scope of protection required by the present invention.

Claims

1. A method for debugging feeder automation strategies based on neural networks, characterized in that, It includes the following steps: Step S1: Establish an FA debugging problem expert database, where the FA debugging problem expert database includes an FA policy error matrix and a data error matrix with a corresponding relationship; Step S2: Based on the FA debugging problem expert database, construct and train a neural network to obtain a trained neural network; Step S3: Collect the FA policy error data generated during the line debugging process and input it into the trained neural network to output the probability of data errors of the corresponding type existing in the debugging line database, and generate a data error indication vector based on the probability of data errors of the corresponding type existing in the debugging line database and a preset probability threshold; 2. The method for debugging a feeder automation strategy based on a neural network according to claim 1, characterized in that, It further includes Step S4: Calculate the correlation accuracy based on the data error indication vector, and optimize the parameters of the trained neural network and the preset probability threshold according to the correlation accuracy; 3. The method for debugging a feeder automation strategy based on a neural network according to claim 2, wherein The specific content of Step S4 is as follows: Step 4.1: Compare the data error indication vector with the actual data error vector, and use their ratio as the correlation accuracy; Step 4.2: Adjust the number of neurons in the hidden layer of the neural network and the preset probability threshold, and repeat the calculation of the correlation accuracy until the correlation accuracy meets the usage requirements, and use the current number of neurons in the hidden layer of the neural network and the preset probability threshold for subsequent calculations; 4. A method for debugging a feeder automation strategy based on a neural network according to claim 3, characterized in that, It further includes Step S5: Expand the FA policy error vector and the actual data error vector generated during the line debugging process into the FA debugging problem expert database, and further train the trained neural network; 5. A method for debugging a feeder automation strategy based on a neural network according to any one of claims 1-4, characterized in that, The specific content of Step S1 is as follows: Step S1.1: Obtain the FA policy errors in the FA historical debugging records of the debugging line and the data errors in the database and classify them; Step S1.2: Establish an FA policy error vector and a data error vector for each debugging line; Step S1.3: Compose all the established FA policy error vectors into an FA policy error matrix, compose the corresponding data error vectors into a data error matrix, and write the two matrices into Matlab. The two matrices and their corresponding relationships constitute the FA debugging problem expert database; 6. The feeder automation strategy debugging method based on neural network according to claim 5, characterized in that In Step S1.1, the FA policy errors are divided into 11 categories, specifically including: failure to read the set fault, reading the wrong fault, expansion of the fault isolation range, incomplete transfer supply strategy, inability to identify false dual users, incorrect load rate reading, operation steps involving switches not on this line, inability to identify automated stations, incorrect optimal transfer supply plan, inability to pop up the strategy, and non-automated stations giving remote control strategies; Data errors in the database are divided into the following 14 categories, specifically including: device diagram models not associated, adjacent device connection point node numbers different, node numbers on this line repeating, the same node numbers existing in different lines in the system, the corresponding switch column data in the protection node table being empty, there being two pieces of data for the same signal in the pre - front - end tele - signal definition table, the data in the signal type column of the protection node table being incorrect, the data in the coefficient column of the pre - front - end tele - measurement definition table being incorrect, the data in the switch column of the four - zone related table being incorrect, the data in the feeder column of the distribution network load table being incorrect, the data in the feeder column of the distribution network switch table being incorrect, the data in the feeder column of the distribution network feeder segment table being incorrect, the data in the feeder column of the distribution network bus table being incorrect, and non - database problems.

7. A method for debugging a feeder automation strategy based on a neural network according to claim 5, characterized in that, The specific steps of step S2 are as follows: Step S2.1: Determine the neural network structure; 11 neurons are set in the input layer of the neural network to input the FA policy error vector; 25 neurons are set in the hidden layer, and the Sigmoid activation function is used; 14 neurons are set in the output layer, and the ReLU activation function is used, and finally the probability of 14 - bit data errors is output. Step S2.2: Use the FA policy error matrix and the data error matrix in the FA debugging problem expert library as the sample input and sample output respectively to train the neural network, and obtain the trained neural network.

8. A method for debugging a feeder automation strategy based on a neural network according to claim 7, characterized in that The specific steps of step S3 are as follows: Step S3.1: Collect the FA policy error data generated during the line debugging process. Step S3.2: Input the collected FA policy error data into the trained neural network, and output the probability that there are data errors of the corresponding type in the database of this debugging line. Step S3.3: Compare the probability that there are data errors of the corresponding type in the database of this debugging line with the preset probability threshold. If the row data greater than or equal to the preset probability threshold is set to 1, and the row data less than the probability threshold is set to 0, a data error indication vector is obtained, which is used to indicate the data error types existing in the database of this debugging line.

9. A feeder automation strategy debugging system based on a neural network, characterized in that, Including: FA debugging problem expert library establishment module, neural network construction and training module, and data error indication vector generation module; The FA debugging problem expert library establishment module is used to establish the FA debugging problem expert library, and the FA debugging problem expert library includes an FA policy error matrix and a data error matrix with a corresponding relationship. The neural network construction and training module is used to construct and train the neural network based on the FA debugging problem expert library, and obtain the trained neural network. The data error indication vector generation module is used to collect the FA policy error data generated during the line debugging process, input it into the trained neural network, output the probability that there are data errors of the corresponding type in the database of this debugging line, and generate a data error indication vector based on the probability that there are data errors of the corresponding type in the database of this debugging line and the preset probability threshold.

10. A feeder automation strategy debugging system based on a neural network according to claim 9, characterized in that, It further includes an optimization module, which is used to calculate the association accuracy rate based on the data error indication vector, and optimize the parameters of the trained neural network and the preset probability threshold according to the association accuracy rate.

Images