Electric power communication network fault detection method based on transfer learning

Abstract

The invention discloses an electric power communication network fault detection method based on transfer learning, and relates to the technical field of electric power communication. According to the power communication network fault detection method based on transfer learning, feature processing is performed on source power communication data of a source power communication network to obtain source power communication feature data, each pre-selected fault detection model is evaluated to obtain an optimal pre-selected fault detection model, and an optimal adjusted fault detection model is determined. The pre-selected model is evaluated and adjusted to be adaptive to the target power communication network by using the data features of the source power communication network, so that the accuracy and efficiency of fault detection are improved, the false alarm and leak detection risks caused by data insufficiency or feature difference are reduced, the stable operation and safety of the power communication network are ensured, and the fault detection efficiency is improved. The problem of poor applicability of a fault detection model between different power communication networks can be effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of power communication, and in particular to a power communication network fault detection method based on transfer learning. Background Art

[0002] The power communication network is responsible for the transmission of various information in the power system. During operation, the power communication network may experience various faults, such as link failures, equipment failures, etc. These faults will not only affect the normal operation of the power system, but may also cause serious economic losses and safety hazards. Therefore, it is very important to detect and diagnose faults quickly and accurately.

[0003] However, different power communication networks have different parameters, and usually use different models for prediction, which leads to the problem of poor applicability of fault detection models between different power communication networks. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a power communication network fault detection method based on transfer learning, which solves the problem of poor applicability of fault detection models between different power communication networks.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a method for fault detection of an electric power communication network based on transfer learning, comprising the following steps: obtaining source electric power communication data of a source electric power communication network, performing feature processing on the source electric power communication data of the source electric power communication network, and obtaining source electric power communication feature data, wherein the source electric power communication feature data includes a source communication network frequency offset, a source communication network phase difference, and a source communication network transmission delay; obtaining each preselected fault detection model stored in a database, inputting the source electric power communication feature data into each preselected fault detection model, evaluating each preselected fault detection model, and obtaining an optimal preselected fault detection model; obtaining each adjustment scheme stored in the database and adjusting the optimal preselected fault detection model. , obtain each adjusted fault detection model, apply each adjusted fault detection model to the target power communication network; obtain the labeled target power communication data of the target power communication network, perform feature processing on the labeled target power communication data of the target power communication network, and obtain the labeled target power communication feature data, wherein the target power communication feature data includes the target communication network frequency offset, the target communication network phase difference and the target communication network network transmission delay; input the labeled target power communication feature data and the fault type corresponding to the labeled target power communication feature data into each adjusted fault detection model, obtain the processing result of each adjusted fault detection model, evaluate each adjusted fault detection model, and determine the optimal adjusted fault detection model.

[0006] Furthermore, the evaluation of each preselected fault detection model to obtain the optimal preselected fault detection model includes the following steps: obtaining effect data of each preselected fault detection model on processing source power communication characteristic data, the effect data including the average time consumed to obtain fault detection results, the number of correct fault detection results, the number of times the fault detection model does not respond, the CPU temperature when the fault detection model is running, and the CPU usage rate when the fault detection model is running; analyzing and processing the effect data of each preselected fault detection model on processing source power communication characteristic data with reference effect data stored in a database to obtain a processing effect evaluation value, the processing effect evaluation value is used as an analysis basis for obtaining the optimal preselected fault detection model, the reference effect data stored in the database including the reference average time consumed for fault detection results, the reference number of correct fault detection results, the reference number of times the fault detection model does not respond, the CPU reference temperature when the fault detection model is running, and the CPU reference usage rate when the fault detection model is running; arranging the processing effect evaluation values ​​in ascending order, and the preselected fault detection model corresponding to the smallest processing effect evaluation value is the optimal preselected fault detection model.

[0007] Furthermore, the effect data of each pre-selected fault detection model processing source power communication characteristic data is analyzed and processed with the reference effect data stored in the database to obtain a processing effect evaluation value, including the following steps: comparing and analyzing the average time consumed by the fault detection results, the number of correct fault detection results, and the number of times the fault detection model does not respond with the reference average time consumed, the reference number of correct fault detection results, and the reference number of times the fault detection model does not respond to obtain a detection result evaluation coefficient; comparing and analyzing the CPU temperature when the fault detection model is running and the CPU usage rate when the fault detection model is running with the CPU reference temperature when the fault detection model is running and the CPU reference usage rate when the fault detection model is running to obtain a CPU state evaluation coefficient; fusing the detection result evaluation coefficient with the CPU state evaluation coefficient to obtain a processing effect evaluation value.

[0008] Furthermore, the calculation formula of the treatment effect evaluation value is as follows: ; In the formula, is the treatment effect evaluation value, is the test result evaluation coefficient, is the CPU status evaluation coefficient, is the weight factor of the test result evaluation coefficient stored in the database, The weight factor of the CPU status evaluation coefficient stored in the database.

[0009] Furthermore, the calculation formula of the test result evaluation coefficient is as follows: ; In the formula, is the average time consumed, is the number of correct fault detection results, is the number of times the fault detection model does not respond, For reference, the average time consumption is Refer to the correct number of fault detection results. The number of no responses is the reference of the fault detection model. is a natural constant.

[0010] Furthermore, the calculation formula of the test result evaluation coefficient is as follows: ; In the formula, CPU temperature when the fault detection model is running, The CPU usage when the fault detection model is running. The CPU reference temperature when the fault detection model is running. The reference CPU usage when the fault detection model is running.

[0011] Furthermore, the correct number of fault detection results is obtained as follows: the source power communication characteristic data is merged with the source power communication parameter characteristic data stored in the database to obtain a fault detection evaluation value, wherein the source power communication parameter characteristic data stored in the database includes the source communication network frequency parameter offset, the source communication network parameter phase difference and the source communication network network parameter transmission delay; the fault detection evaluation value is compared with the predicted fault type corresponding to each fault detection evaluation value stored in the database to obtain the predicted fault type corresponding to the fault detection evaluation value; the actual fault type of the fault detection evaluation value is compared with the predicted fault type; if the actual fault type is the same as the predicted fault type, the correct number of fault detection results is increased by one; if the actual fault type is the same as but different from the predicted fault type, the correct number of fault detection results is not increased by one.

[0012] Furthermore, the calculation formula of the fault detection evaluation value is as follows: ; In the formula, is the fault detection evaluation value, is the target communication network frequency offset, is the target communication network phase difference, is the target communication network transmission delay, The frequency parameter offset of the source communication network is set. Determine the phase difference for the source communication network, Determine the transmission delay for the source communication network parameters.

[0013] Furthermore, the evaluation of each adjusted fault detection model to determine the optimal adjusted fault detection model includes the following steps: inputting the labeled target power communication characteristic data into each adjusted fault detection model to obtain the processing results of each adjusted fault detection model, wherein the processing results include the number of correct fault diagnoses and the average time consumed to obtain the correct diagnosis results; comparing and analyzing the processing results with the reference processing results stored in the database to obtain a comparison evaluation value, and determining the adjusted fault detection model corresponding to the largest comparison evaluation value as the optimal adjusted fault detection model, wherein the reference processing results stored in the database include the reference number of correct fault diagnoses and the reference average time consumed to obtain the correct diagnosis results.

[0014] Furthermore, the calculation formula of the comparison evaluation value is as follows: ; In the formula, To compare the evaluation values, is the number of correct fault diagnoses, The average time to obtain a correct diagnosis result is For fault diagnosis, refer to the correct number. The average time required to obtain a correct diagnosis.

[0015] The present invention has the following beneficial effects: This power communication network fault detection method based on transfer learning utilizes the data characteristics of the source power communication network to evaluate and adjust the pre-selected model to adapt to the target power communication network, thereby improving the accuracy and efficiency of fault detection, reducing the risks of false alarms and missed detections caused by insufficient data or feature differences, and thus ensuring the stable operation and security of the power communication network. It can effectively solve the problem of poor applicability of fault detection models between different power communication networks.

[0016] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of the power communication network fault detection method based on transfer learning of the present invention. DETAILED DESCRIPTION

[0018] The embodiment of the present application can effectively solve the problem of poor applicability of fault detection models between different power communication networks through a power communication network fault detection method based on transfer learning.

[0019] See also Figure 1, an embodiment of the present invention provides a technical solution: a method for fault detection of a power communication network based on transfer learning, comprising the following steps: obtaining source power communication data of a source power communication network, performing feature processing on the source power communication data of the source power communication network, and obtaining source power communication feature data, wherein the source power communication feature data has multiple groups, each group corresponding to a type of fault, and the source power communication feature data includes source communication network frequency offset, source communication network phase difference, and source communication network network transmission delay. The source power communication data of the source power communication network is obtained through historical data or a database, including current and voltage signals, network transmission logs, equipment status information, etc. These data can be collected through sensors, measuring instruments, or network monitoring tools. The process of feature extraction to obtain source power communication feature data is as follows: the collected raw data needs to be preprocessed to remove noise and outliers, and the data is normalized to the same scale. This process may include steps such as data cleaning, missing value processing, and filtering. Based on the preprocessed data, specific features are extracted through analysis and calculation, including: performing spectrum analysis on the electrical signal through frequency domain analysis methods such as fast Fourier transform (FFT), and calculating the frequency offset. By comparing the phase information of multiple signals at the same time, the phase loop technology is used to calculate the phase difference between the signals. The timestamp technology is used to record the time when the data packet is transmitted between the networks, so as to calculate the network transmission delay. It can be estimated by the time difference between sending the data packet and receiving the confirmation packet.

[0020] Obtain each preselected fault detection model stored in the database, input the source power communication characteristic data into each preselected fault detection model, evaluate each preselected fault detection model, and obtain the optimal preselected fault detection model; obtain each adjustment scheme stored in the database to adjust the optimal preselected fault detection model, obtain each adjusted fault detection model, and apply each adjusted fault detection model to the target power communication network; obtain the labeled target power communication data of the target power communication network, perform feature processing on the labeled target power communication data of the target power communication network, and obtain the labeled target power communication characteristic data, wherein the target power communication characteristic data includes the target communication network frequency offset, the target communication network phase difference, and the target communication network network transmission delay. The method for obtaining the labeled target power communication characteristic data is the same as the method for obtaining the source power communication characteristic data.

[0021] The labeled target power communication characteristic data and the fault type corresponding to the labeled target power communication characteristic data are input into each adjusted fault detection model, the processing result of each adjusted fault detection model is obtained, each adjusted fault detection model is evaluated, and the optimal adjusted fault detection model is determined.

[0022] By processing the features of the data of the source power communication network, multiple sets of power communication feature data are extracted. These feature data can effectively reflect the network status and provide support for fault detection. The source feature data is input into a variety of pre-established fault detection models to evaluate their performance and select the best model to ensure accurate fault identification in the target power communication network. The selected optimal model is adapted using the adjustment scheme stored in the database to ensure that the model can better adapt to the characteristics of the target power communication network and further improve the detection effect. The adjusted model is trained and verified through the labeled target power communication data, and the optimal model is finally determined to ensure the reliability of the detection results.

[0023] Through transfer learning, the model is adjusted to better adapt to the target environment, effectively improve the accuracy of fault detection, and reduce false alarms and missed detections. Since the rich data of the source power communication network can be used for training and adjustment, the reliance on a large amount of labeled data of the target power grid is reduced, and the cost of data collection and labeling is reduced. With the help of the existing pre-selected fault detection model, it can quickly respond to and adapt to the fault detection needs of different power communication networks, which helps to improve the overall operation and maintenance efficiency. Through the logical steps of the system and the effective adaptation of the model, it can not only solve the adaptability problem of fault detection between different power communication networks, but also improve the detection efficiency and system security, which has significant practical value.

[0024] Specifically, the evaluation of each pre-selected fault detection model to obtain the optimal pre-selected fault detection model includes the following steps: obtaining effect data of each pre-selected fault detection model processing source power communication characteristic data, the effect data including the average time consumed to obtain the fault detection result, the number of correct fault detection results, the number of times the fault detection model does not respond, the CPU temperature when the fault detection model is running, and the CPU usage rate when the fault detection model is running. The average time consumed for the fault detection result can be obtained by recording the start time and the end time when the detection model is running, and then calculating the average time consumed for each detection. The number of correct fault detection results is calculated by comparing the detection results of the model with the actual fault situation to count the number of correct detections. The number of times the fault detection model does not respond is calculated by monitoring the situation where the model fails to return any output within a certain period of time. The CPU temperature when the fault detection model is running is obtained by using a system monitoring tool or an API. The CPU usage rate when the fault detection model is running is obtained by using a system monitoring tool to obtain the CPU usage rate.

[0025] The effect data of each pre-selected fault detection model processing source power communication characteristic data are analyzed and processed with the reference effect data stored in the database to obtain a processing effect evaluation value, and the processing effect evaluation value is used as an analysis basis for obtaining the optimal pre-selected fault detection model. The reference effect data stored in the database includes a reference average time consumption of fault detection results, a reference number of correct fault detection results, a reference number of no responses of the fault detection model, a reference temperature of the CPU when the fault detection model is running, and a reference CPU usage rate when the fault detection model is running; the processing effect evaluation values ​​are arranged in order from small to large, and the pre-selected fault detection model corresponding to the smallest processing effect evaluation value is the optimal pre-selected fault detection model.

[0026] In this implementation scheme, by collecting a variety of effect data (such as average time consumption, correct number, number of unresponsive times, CPU temperature and usage rate, etc.), we can fully understand the performance of each model in processing source power communication characteristic data. Comparing the effect data of the pre-selected model with the reference effect data stored in the database can provide an objective standard for the evaluation of the model, and can eliminate the accidental influence of individual models under specific operating environments or data conditions, so as to draw more general conclusions. By converting the collected effect data into processing effect evaluation values, this process can integrate different evaluation indicators to form a unified evaluation standard for easy sorting and comparison. By sorting the processing effect evaluation values ​​from small to large, the best performing model can be quickly identified to ensure that the final selected model performs well in all indicators and meets the actual application needs.

[0027] Selecting the best performing fault detection model can help improve the overall performance of fault detection and enhance the stability and reliability of the system. Considering CPU temperature and usage during the evaluation process can help identify models with excessive resource consumption and ensure that overload does not occur in actual applications, thereby optimizing resource usage and extending the service life of the equipment. By selecting a model with a shorter response time, the timeliness of fault detection can be improved, ensuring that the system can respond quickly when a fault occurs, reducing potential losses.

[0028] Specifically, the effect data of each pre-selected fault detection model processing source power communication characteristic data is analyzed and processed with the reference effect data stored in the database to obtain a processing effect evaluation value, including the following steps: comparing and analyzing the average time consumption of the fault detection results, the number of correct fault detection results, and the number of no responses of the fault detection model with the reference average time consumption, the reference correct number of fault detection results, and the reference number of no responses of the fault detection model to obtain a detection result evaluation coefficient; comparing and analyzing the CPU temperature when the fault detection model is running and the CPU utilization rate when the fault detection model is running with the CPU reference temperature when the fault detection model is running and the CPU reference utilization rate when the fault detection model is running to obtain a CPU state evaluation coefficient; fusing the detection result evaluation coefficient with the CPU state evaluation coefficient to obtain a processing effect evaluation value.

[0029] The calculation formula of the treatment effect evaluation value is as follows: ; In the formula, is the treatment effect evaluation value, is the test result evaluation coefficient, is the CPU status evaluation coefficient, is the weight factor of the test result evaluation coefficient stored in the database, The weight factor of the CPU status evaluation coefficient stored in the database.

[0030] The calculation formula of the test result evaluation coefficient is as follows: ; In the formula, is the average time consumed, is the number of correct fault detection results, is the number of times the fault detection model does not respond, For reference, the average time consumption is Refer to the correct number of fault detection results. The number of no responses is the reference of the fault detection model. is a natural constant.

[0031] The calculation formula of the test result evaluation coefficient is as follows: ; In the formula, CPU temperature when the fault detection model is running, The CPU usage when the fault detection model is running. The CPU reference temperature when the fault detection model is running. The reference CPU usage when the fault detection model is running.

[0032] In this implementation scheme, the actual detection results of the model (such as average time consumption, number of correct answers, and number of no responses) are compared with the preset reference values, which can effectively identify the gap between the performance of the model in actual operation and the expected results. The detection result evaluation coefficient JcG is calculated by comparison, which reflects the performance of the model in identifying faults in an exponential form. The larger the value, the better the performance. At the same time, the CPU temperature and usage rate of the fault detection model are monitored during operation, and compared with the reference value to obtain the CPU status evaluation coefficient. This part of the analysis aims to ensure that the model runs smoothly at the technical level without significant overload or temperature anomalies, which is crucial to the long-term stability and accuracy of the model. Finally, the detection result evaluation coefficient JcG and the CPU status evaluation coefficient GP are weighted and fused to form a comprehensive processing effect evaluation value. This evaluation logic involves multiple important performance indicators, not only focusing on the recognition accuracy of the model, but also considering the operating system conditions, thereby providing a more comprehensive performance analysis.

[0033] It provides a comprehensive, dynamic and adjustable performance inspection and optimization method for the fault detection model, which can effectively support the stable operation, performance improvement and resource optimization configuration of the system, and has significant practical value. Through quantitative analysis, the performance of the model can be scientifically evaluated and improved, thereby improving the overall performance of the fault detection system.

[0034] Specifically, the method for obtaining the correct number of fault detection results is as follows: the source power communication characteristic data is merged with the source power communication parameter characteristic data stored in the database to obtain a fault detection evaluation value, wherein the source power communication parameter characteristic data stored in the database includes the source communication network frequency parameter offset, the source communication network parameter phase difference and the source communication network network parameter transmission delay; the fault detection evaluation value is compared with the predicted fault type corresponding to each fault detection evaluation value stored in the database to obtain the predicted fault type corresponding to the fault detection evaluation value; the actual fault type of the fault detection evaluation value is compared with the predicted fault type; if the actual fault type is the same as the predicted fault type, the correct number of fault detection results is increased by one; if the actual fault type is the same as or different from the predicted fault type, the correct number of fault detection results is not increased by one.

[0035] The calculation formula of the fault detection evaluation value is as follows: ; In the formula, is the fault detection evaluation value, is the target communication network frequency offset, is the target communication network phase difference, is the target communication network transmission delay, The frequency parameter offset of the source communication network is set. Determine the phase difference for the source communication network, Determine the transmission delay for the source communication network parameters.

[0036] In this implementation, by fusing the source power communication characteristic data (i.e., frequency offset, phase difference, and transmission delay during transmission) with the parameter characteristic data stored in the database, a fault detection evaluation value can be extracted and calculated. This process can help the system more accurately reflect the current network status. The acquired fault detection evaluation value is compared with the corresponding evaluation value in the database to identify potential fault types. This link helps the system quickly locate the source of the fault by establishing a mapping relationship between the fault evaluation value and the fault type.

[0037] Compare the actual fault type with the predicted fault type. If the two are consistent, the number of correct fault detection results is increased by one; if they are inconsistent, no count is performed. This step realizes the evaluation of detection accuracy and helps with subsequent algorithm optimization and model improvement. The calculation formula of the fault detection evaluation value quantifies the differences in frequency offset, phase difference, and transmission delay, and uses logarithmic functions to reduce the impact of outliers, thereby introducing a certain degree of stability and robustness in the calculation.

[0038] By fusing and processing feature data and calculating fault detection evaluation values, the system can accurately identify which parameter deviations lead to the occurrence of faults, which helps to locate problems in a timely manner. This fault detection result acquisition logic not only has good data-driven characteristics, but also shows significant advantages in accuracy, dynamic response, robustness, traceability and economic benefits, providing strong technical support for the stable operation of power communication systems.

[0039] Specifically, the evaluation of each adjusted fault detection model to determine the optimal adjusted fault detection model includes the following steps: inputting the labeled target power communication characteristic data into each adjusted fault detection model to obtain the processing results of each adjusted fault detection model, wherein the processing results include the number of correct fault diagnoses and the average time consumed to obtain the correct diagnosis results; comparing and analyzing the processing results with the reference processing results stored in the database to obtain a comparison evaluation value, and determining the adjusted fault detection model corresponding to the largest comparison evaluation value as the optimal adjusted fault detection model, wherein the reference processing results stored in the database include the reference number of correct fault diagnoses and the reference average time consumed to obtain the correct diagnosis results.

[0040] The calculation formula of the comparison evaluation value is as follows: ; In the formula, To compare the evaluation values, is the number of correct fault diagnoses, The average time to obtain a correct diagnosis result is For fault diagnosis, refer to the correct number. The average time required to obtain a correct diagnosis.

[0041] In this implementation scheme, the marked target power communication characteristic data is input into each adjusted fault detection model for evaluation. Each model processes the input data and outputs the corresponding results, including the correct number of fault diagnoses and the average time required to obtain the correct diagnosis results. The results of the model processing are compared with the reference processing results stored in the database. The reference results include the reference correct number of fault diagnoses and the reference average time required to obtain the correct diagnosis results. The comparison evaluation value is calculated using a formula. This formula comprehensively considers the number of correct diagnoses and the time consumption, and makes a quantitative comparison of the performance. Based on the calculated comparison evaluation value, the adjusted fault detection model with the highest evaluation value is selected as the optimal model.

[0042] The comparison evaluation value takes into account not only accuracy (number of correct answers) but also efficiency (average time consumed), thereby ensuring that the selected model has an advantage in performance, not just accuracy or time. By introducing dynamic data input and comparison with reference data, the evaluation criteria of the model can be updated in real time to adapt to different data and changing environments, thereby enhancing the adaptability of the system.

[0043] An electronic device comprises: a processor and a memory, wherein computer program instructions are stored in the memory, and when the computer program instructions are executed by the processor, the processor executes the power communication network fault detection method based on transfer learning as described above.

[0044] A computer-readable storage medium is used to store a program, and when the program is executed by a processor, the power communication network fault detection method based on transfer learning as described above is implemented.

[0045] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0046] The present invention is described with reference to flowcharts and / or block diagrams of systems, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks 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 produce 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.

[0047] 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.

[0048] 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.

[0049] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0050] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A power communication network fault detection method based on transfer learning, characterized in that: The following steps are involved: Acquire source power communication data of a source power communication network, perform feature processing on the source power communication data of the source power communication network, and obtain source power communication feature data, wherein the source power communication feature data includes a source communication network frequency offset, a source communication network phase difference, and a source communication network network transmission delay; Obtaining each preselected fault detection model stored in a database, inputting source power communication characteristic data into each preselected fault detection model, evaluating each preselected fault detection model, and obtaining an optimal preselected fault detection model; Acquire each adjustment scheme stored in the database to adjust the optimal pre-selected fault detection model, obtain each adjusted fault detection model, and apply each adjusted fault detection model to the target power communication network; Acquire the marked target power communication data of the target power communication network, perform feature processing on the marked target power communication data of the target power communication network, and obtain the marked target power communication feature data, wherein the target power communication feature data includes the target communication network frequency offset, the target communication network phase difference and the target communication network network transmission delay; The labeled target power communication characteristic data and the fault type corresponding to the labeled target power communication characteristic data are input into each adjusted fault detection model, the processing result of each adjusted fault detection model is obtained, each adjusted fault detection model is evaluated, and the optimal adjusted fault detection model is determined.

2. The power communication network fault detection method based on transfer learning according to claim 1 is characterized in that: The step of evaluating each pre-selected fault detection model to obtain the optimal pre-selected fault detection model comprises the following steps: Obtaining effect data of each preselected fault detection model processing source power communication characteristic data, the effect data including average time consumption to obtain fault detection results, number of correct fault detection results, number of times the fault detection model does not respond, CPU temperature when the fault detection model is running, and CPU usage when the fault detection model is running; Analyze and process the effect data of each pre-selected fault detection model processing source power communication characteristic data with the reference effect data stored in the database to obtain a processing effect evaluation value, which is used as an analysis basis for obtaining the optimal pre-selected fault detection model. The reference effect data stored in the database includes a reference average time consumption of the fault detection result, a reference number of correct fault detection results, a reference number of no responses of the fault detection model, a reference CPU temperature when the fault detection model is running, and a reference CPU usage rate when the fault detection model is running; The processing effect evaluation values ​​are arranged in order from small to large, and the pre-selected fault detection model corresponding to the smallest processing effect evaluation value is the optimal pre-selected fault detection model.

3. The power communication network fault detection method based on transfer learning according to claim 2 is characterized in that: The method of analyzing and processing the effect data of each pre-selected fault detection model processing source power communication characteristic data and the reference effect data stored in the database to obtain the processing effect evaluation value comprises the following steps: Compare and analyze the average time consumption of fault detection results, the number of correct fault detection results, and the number of no responses of the fault detection model with the reference average time consumption, the reference number of correct fault detection results, and the reference number of no responses of the fault detection model to obtain the detection result evaluation coefficient; Compare and analyze the CPU temperature and CPU usage rate when the fault detection model is running, and the CPU reference temperature and CPU reference usage rate when the fault detection model is running, to obtain a CPU state evaluation coefficient; The detection result evaluation coefficient and the CPU status evaluation coefficient are fused to obtain the processing effect evaluation value.

4. The method for fault detection of electric power communication network based on transfer learning according to claim 3, characterized in that: The calculation formula of the treatment effect evaluation value is as follows: ； In the formula, is the treatment effect evaluation value, is the test result evaluation coefficient, is the CPU status evaluation coefficient, is the weight factor of the test result evaluation coefficient stored in the database, The weight factor of the CPU status evaluation coefficient stored in the database.

5. The method for fault detection in electric power communication network based on transfer learning according to claim 4, characterized in that: The calculation formula of the test result evaluation coefficient is as follows: ； In the formula, is the average time consumed, is the number of correct fault detection results, is the number of times the fault detection model does not respond, For reference, the average time consumption is Refer to the correct number of fault detection results. The number of no responses is the reference of the fault detection model. is a natural constant.

6. The method for fault detection in electric power communication network based on transfer learning according to claim 4, characterized in that: The calculation formula of the test result evaluation coefficient is as follows: ； In the formula, CPU temperature when the fault detection model is running, The CPU usage when the fault detection model is running. The CPU reference temperature when the fault detection model is running. The reference CPU usage when the fault detection model is running.

7. The method for fault detection in a power communication network based on transfer learning according to claim 2, characterized in that: The method for obtaining the correct number of fault detection results is as follows: The source power communication characteristic data is merged with the source power communication parameter characteristic data stored in the database to obtain a fault detection evaluation value, wherein the source power communication parameter characteristic data stored in the database includes a source communication network frequency parameter offset, a source communication network parameter phase difference, and a source communication network network parameter transmission delay; Compare the fault detection evaluation value with the predicted fault type corresponding to each fault detection evaluation value stored in the database to obtain the predicted fault type corresponding to the fault detection evaluation value; The actual fault type of the fault detection evaluation value is compared with the predicted fault type. If the actual fault type is the same as the predicted fault type, the number of correct fault detection results is increased by one. If the actual fault type is the same as the predicted fault type but different, the number of correct fault detection results is not increased by one.

8. The method for fault detection in electric power communication network based on transfer learning according to claim 7, characterized in that: The calculation formula of the fault detection evaluation value is as follows: ； In the formula, is the fault detection evaluation value, is the target communication network frequency offset, is the target communication network phase difference, is the target communication network transmission delay, The frequency parameter offset of the source communication network is set. Determine the phase difference for the source communication network, Determine the transmission delay for the source communication network parameters.

9. The method for fault detection in electric power communication network based on transfer learning according to claim 1, characterized in that: The step of evaluating each adjusted fault detection model to determine the optimal adjusted fault detection model includes the following steps: Inputting the marked target power communication characteristic data into each adjusted fault detection model to obtain processing results of each adjusted fault detection model, wherein the processing results include the number of correct fault diagnoses and the average time consumed to obtain the correct diagnosis results; The processing results are compared and analyzed with the reference processing results stored in the database to obtain a comparison evaluation value, and the adjusted fault detection model corresponding to the largest comparison evaluation value is determined as the optimal adjusted fault detection model. The reference processing results stored in the database include the reference correct number of fault diagnoses and the reference average time for obtaining correct diagnosis results.

10. The method for fault detection in electric power communication network based on transfer learning according to claim 9, characterized in that: The calculation formula of the comparison evaluation value is as follows: ； In the formula, To compare the evaluation values, is the number of correct fault diagnoses, The average time to obtain a correct diagnosis result is For fault diagnosis, refer to the correct number. The average time required to obtain a correct diagnosis.

Images