Power distribution network communication fault diagnosis method based on machine learning

Abstract

The invention relates to the field of power data processing, and discloses a power distribution network communication fault diagnosis method based on machine learning, which comprises the steps of S1, distribution box data acquisition area determination, S2, distribution box communication data acquisition, S3, communication analysis, S4, comprehensive data processing, S5, communication fault judgment and S6, man-machine interaction. The invention provides a power distribution network communication fault diagnosis method based on machine learning. The method comprises the following steps: firstly, dividing distribution box monitoring sub-areas, collecting multiple types of data such as RSSI values at the frequency of 100Hz, calculating communication quality, equipment health and network robustness indexes through a specific formula so as to obtain a communication fault assessment index, and judging a fault risk level and analyzing a fault reason according to the assessment index. Diagnosis efficiency is high, positioning is accurate, targeted decision suggestions can be provided according to fault levels, and power distribution network communication fault diagnosis level and operation and maintenance efficiency are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of power data processing, and more specifically to a distribution network communication fault diagnosis method based on machine learning. Background Art

[0002] With the advancement of smart grid construction, the reliability of distribution network communication systems has become a key factor in ensuring power supply stability. Current mainstream communication fault diagnosis technologies mainly rely on the comprehensive application of multi-dimensional monitoring and analysis methods: Traditional power quality monitoring uses devices such as voltage / current sensors and spectrum analyzers to collect real-time grid static parameters such as effective value and frequency and dynamic parameters such as harmonics and transient fluctuations, and manually interprets power supply anomalies based on preset thresholds; fiber optic sensing technology, based on distributed fiber optic temperature sensors and the principles of optical time domain reflection and Raman scattering, accurately senses cable temperature anomalies or mechanical damage such as breakage and extrusion, and locates physical layer faults; network protocol analysis analyzes communication protocols such as Modbus and IEC61850, and evaluates the transmission quality of the communication network by statistically analyzing link layer packet loss rate, retransmission rate and other indicators.

[0003] However, in practical applications, some problems still exist. For example, existing technologies mainly focus on monitoring limited aspects such as power parameters and cable temperature, and do not fully cover the multi-dimensional information of distribution network communication systems. In communication fault diagnosis, the lack of comprehensive collection and analysis of key data such as communication signal quality, equipment operating status, and network topology leads to incomplete and inaccurate fault diagnosis. Traditional power quality monitoring and fiber optic sensing detection methods rely on manual experience and simple threshold judgments. When faced with large amounts of data, the calculation process is cumbersome and cannot quickly process and analyze data, resulting in low fault diagnosis efficiency. At the same time, traditional methods are insufficient in fault location, making it difficult to accurately determine the location of the fault. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a distribution network communication fault diagnosis method based on machine learning to solve the problems existing in the above-mentioned background technology, such as the lack of comprehensive collection and analysis of key data, low fault diagnosis efficiency, and insufficient ability in fault location, making it difficult to accurately determine the location of the fault.

[0005] The present invention provides the following technical solution: a method for diagnosing distribution network communication faults based on machine learning, comprising:

[0006] S1: Determine the distribution box data acquisition area: Divide the distribution box to be monitored into multiple monitoring sub-areas according to the distribution branch lines, which are denoted as 1, 2, 3, ..., n in sequence, and each monitoring sub-area is determined as the target monitoring area;

[0007] S2: Distribution box communication data collection: After the data collection terminal is deployed, data is collected in real time at a sampling frequency of 100 Hz through the data collection terminal to obtain communication data text;

[0008] The communication data text includes the RSSI value, signal-to-noise ratio, bit error rate, kurtosis, temperature gradient, voltage sag, number of redundant paths, centrality variance and maximum centrality of the target monitoring area;

[0009] S3: Communication analysis: input the communication data text obtained in step S3 into the communication function calculation text to calculate and obtain the analysis result text;

[0010] S4: Comprehensive data processing: input the obtained analysis result text into the communication fault assessment function calculation text to calculate the communication fault assessment index;

[0011] S5: Communication fault judgment: Compare the communication fault evaluation index of each sub-region obtained in the comprehensive data processing of step S4 with the preset communication fault evaluation index judgment interval, output the judgment result of each sub-region, and analyze the cause of the fault;

[0012] S6: Human-computer interaction: output the judgment result to the user information terminal.

[0013] Preferably, the process of acquiring various types of data in the communication data text is:

[0014] A1: Install a communication module with RSSI measurement function on the terminal nodes of the distribution network communication to automatically measure and record the signal strength value. The communication module transmits the measured RSSI value to the data dispatching center via the Ethernet interface. The data dispatching center will store this data and pre-process the RSSI data. By setting upper and lower thresholds, RSSI values ​​that exceed the threshold range are regarded as outliers and eliminated.

[0015] A2: At the receiving end of the communication, a signal processing algorithm is used to analyze the received signal, separate the signal component from the noise component, and calculate the power of the signal and noise respectively. For discrete signals, the power is calculated by summing the squares of the signal samples and taking the average. Suppose the signal sample sequence is x[n] and the length is N, then the signal power is For a noise sample sequence n[n], the noise power According to the formula Calculate the signal-to-noise ratio in dB;

[0016] A3: The transmitting and receiving ends of the communication system record the transmitted and received bit sequences, respectively. The transmitting end encodes the data to be transmitted and then sends it. The receiving end decodes the received data and compares the received bit sequence with the transmitted bit sequence bit by bit. The number of erroneous bits is counted. Let the number of transmitted bits be Ntotal and the number of erroneous bits be Nerror. The bit error rate is calculated according to the formula BER = Nerror / Ntotal.

[0017] A4: Use the scipy.stats library in Python to calculate the kurtosis of the vibration signal data collected by the acquisition terminal. Assuming that the vibration signal samples are x1, x2, …, xn, the kurtosis calculation formula is:

[0018] in is the sample mean;

[0019] A5: Record the temperature value collected by the acquisition terminal and the corresponding acquisition time. Calculate the temperature gradient based on the two adjacent temperature values ​​and the time interval. Assume T1 and T2 are the two adjacent temperature values, and Δt is the time interval. Then, the temperature gradient = (T2 - T1) / Δt.

[0020] A6: Monitor the voltage data collected by the acquisition terminal in real time and record the rated voltage U of the equipment. nominal And the actual measured voltage value U actual According to the formula VoltageSag=(U nominal -U actual ) / U nominal Calculate the voltage sag amplitude;

[0021] A7: Use the network management system to obtain topological information about the distribution network communication network, including the connection relationships between nodes and links. Abstract the network topology into a graph model, where nodes represent network devices and links represent the communication connections between devices. Use the Dijkstra algorithm from graph theory to calculate the shortest path between any two nodes in the network. For each node pair, count the number of shortest paths. The total number of paths with more than one shortest path is considered the number of redundant paths.

[0022] A8: Use betweenness centrality from graph theory to measure the importance of nodes in the network. Betweenness centrality indicates how often a node appears on the shortest path between other pairs of nodes. For each node v in the graph, calculate its betweenness centrality CB(v). Normalize the betweenness centrality values ​​of all nodes to a range of 0-1. Then calculate the variance of these normalized betweenness centrality values ​​to obtain the centrality variance. At the same time, find the maximum value among them, which is the maximum centrality.

[0023] Preferably, the calculation formula of the communication quality index is:

[0024] CQI refers to the communication quality index, RSSI refers to the received signal strength, SNR refers to the signal-to-noise ratio, and BER refers to the bit error rate.

[0025] Preferably, the calculation formula of the device health index is:

[0026] DHI = 0.6 × Kurtosis norm +0.3×Temperature Gradient+0.1×Voltage Sag

[0027]

[0028] DHI stands for Device Health Index, Kurtosis max Refers to the historical maximum kurtosis value, Kurtosis norm Refers to the dynamic normalized value of kurtosis. Kurtosis is the kurtosis value, TemperatureGradient refers to the normalized temperature gradient, and VoltageSag refers to the voltage sag.

[0029] Preferably, the calculation formula of the network robustness index is:

[0030]

[0031] Where NRI represents the network robustness index, Redundancy Ratio represents the normalized number of redundant paths, Redundancy Paths represents the number of redundant paths, Total Nodes represents the total number of nodes, Centrality Variance refers to the centrality variance, and Max Centrality refers to the maximum centrality.

[0032] Preferably, the communication fault assessment function calculates the text as CFI stands for Communication Fault Index, CQI stands for Communication Quality Index, DHI stands for Device Health Index, and NRI stands for Network Robustness Index.

[0033] Preferably, the communication fault assessment index judgment interval is specifically:

[0034] Low risk (0.0≤CFI<0.3). The CFI value is low, indicating that communication quality, device health, and network robustness are all in good condition. This is based on the fact that the combined value of the risk factors in the three dimensions is small. In this case, only routine monitoring is required, and relevant data is collected according to the established period to continuously monitor the system status.

[0035] Medium risk (0.3 ≤ CFI < 0.6): CFI values ​​in this range indicate a system problem, but not yet a significant impact on overall operations. This could be a slight decrease in communication quality, minor equipment anomalies, or a slight reduction in network robustness. In this case, inspections should be strengthened, data collection frequency increased, and communication equipment, lines, and network topology inspected more thoroughly to identify potential faults.

[0036] High risk (0.6 ≤ CFI < 0.8): A high CFI value indicates a significant risk of system failure. This could be due to significant interference in the communication link, equipment aging or damage, or potential vulnerabilities in the network structure. In this case, proactive maintenance measures should be initiated, such as performance testing and calibration of communication equipment, evaluation and replacement planning of aging equipment, and optimization of the network structure, to reduce the likelihood of failure.

[0037] Extremely high risk (0.8 ≤ CFI < 1.0): A CFI value close to 1 indicates that the system is in a very dangerous state and a serious failure has occurred or is about to occur. This could be due to a communication link interruption, severe equipment damage, or network paralysis. Immediate troubleshooting is required. Professional maintenance personnel should conduct a comprehensive inspection of the system to quickly locate and repair the fault point to restore normal system operation.

[0038] Technical effects and advantages of the present invention:

[0039] 1. This invention deploys data collection equipment at multiple locations, including terminal nodes, communication receivers, device terminals, and network management systems, to comprehensively collect multi-dimensional data such as RSSI values, signal-to-noise ratios, bit error rates, kurtosis, temperature gradients, voltage sags, number of redundant paths, centrality variance, and maximum centrality. This data is systematically analyzed using comprehensive indicators such as the communication quality index, device health index, network robustness index, and communication fault assessment index to comprehensively and accurately assess the operating status of the distribution network communication system, thereby improving the accuracy of fault diagnosis.

[0040] 2. This invention collects data in real time at a sampling frequency of 100Hz, and combined with a real-time calculated fault assessment index, can promptly detect and respond to faults. When a fault occurs, the system can quickly output the fault diagnosis result and notify the operation and maintenance personnel to handle it, shortening the time for fault detection and repair, and ensuring the stability and reliability of power supply;

[0041] 3. By analyzing the contribution of each component of the communication fault assessment index, the cause and location of the fault can be accurately located. For example, if the contribution of 1-CQI is high, it indicates a communication quality issue, and targeted inspections of communication equipment, antennas, and lines can be conducted. If the contribution of DHI is high, the focus can be on equipment hardware. If the contribution of 1-NRI is high, the network topology needs to be considered. This precise location method facilitates rapid fault repair and reduces troubleshooting costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic flow diagram of the present invention. DETAILED DESCRIPTION

[0043] The technical solutions of the present invention will be clearly and completely described below in conjunction with the drawings in the present invention. In addition, the forms of the various structures described in the following embodiments are merely examples. The distribution network communication fault diagnosis method based on machine learning involved in the present invention is not limited to the various structures described in the following embodiments. All other implementations obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0044] As attached Figure 1 A distribution network communication fault diagnosis method based on machine learning is shown, comprising:

[0045] S1: Determine the distribution box data acquisition area: Divide the distribution box to be monitored into multiple monitoring sub-areas according to the distribution branch lines, which are denoted as 1, 2, 3, ..., n in sequence, and each monitoring sub-area is determined as the target monitoring area;

[0046] In this embodiment, it should be specifically explained that the division of the monitoring sub-areas is conducive to detecting the communication status of the distribution boxes in each sub-area, facilitating timely investigation of possible communication anomalies, and providing convenience for the maintenance of the distribution boxes.

[0047] S2: Distribution box communication data collection: After the data collection terminal is deployed, data is collected in real time at a sampling frequency of 100 Hz through the data collection terminal to obtain communication data text;

[0048] It should be specifically noted that the communication data text includes the RSSI value, signal-to-noise ratio, bit error rate, kurtosis, temperature gradient, voltage drop, number of redundant paths, centrality variance, and maximum centrality of the target monitoring area;

[0049] The acquisition terminal is used to perform data acquisition tasks in the distribution network communication fault diagnosis system, and includes a vibration sensor, a temperature sensor and a voltage sensor;

[0050] The vibration sensor is used to collect vibration signal data to calculate kurtosis. A piezoelectric vibration sensor, such as the PCB352C65, can be used. This type of sensor has high sensitivity and a wide frequency response range. It can accurately capture the vibration signal of the equipment during operation and meet the requirements of calculating kurtosis to assess the risk of mechanical failure of the equipment.

[0051] The temperature sensor is responsible for collecting temperature values ​​and then calculating the temperature gradient. A DS18B20 digital temperature sensor can be used. This type of sensor has the characteristics of small size, low power consumption, high precision, and can directly output digital signals, which is convenient for data acquisition terminal processing and can accurately monitor the temperature changes of the equipment in real time.

[0052] The voltage sensor is used to monitor the voltage data in real time and calculate the voltage drop amplitude. A Hall voltage sensor such as the LV25-P type can be used. It can achieve electrical isolation, accurately measure voltage values, and has a fast response speed, and can promptly reflect voltage changes.

[0053] In this embodiment, it should be specifically explained that the process of obtaining various types of data in the communication data text is as follows:

[0054] A1: Communication modules with RSSI measurement capabilities are installed on the terminal nodes of the distribution network communication. These modules automatically measure and record the signal strength upon receiving a signal. For example, in wireless communication devices, the built-in RF chip detects the power of the received signal and converts the detection result into an RSSI value. The communication module transmits the measured RSSI value via the Ethernet interface to the data dispatching center. The data dispatching center stores this data and pre-processes the RSSI data. By setting upper and lower thresholds, RSSI values ​​that exceed the threshold range are considered outliers and are eliminated.

[0055] A2: At the receiving end of a communication, a signal processing algorithm is used to analyze the received signal and separate the signal component from the noise component. For example, filtering techniques such as low-pass filtering, high-pass filtering, or band-pass filtering are used to filter out the noise from the signal based on the frequency characteristics of the signal and noise, and the power of the signal and noise are calculated separately. For discrete signals, the power can be calculated by summing the squares of the signal samples and taking the average. Suppose the signal sample sequence is x[n] and its length is N. Then the signal power is For a noise sample sequence n[n], the noise power According to the formula Calculate the signal-to-noise ratio in dB;

[0056] A3: The transmitting and receiving ends of the communication system record the transmitted and received bit sequences, respectively. The transmitting end encodes the data to be transmitted and then sends it. The receiving end decodes the received data and compares the received bit sequence with the transmitted bit sequence bit by bit. The number of erroneous bits is counted. Let the number of transmitted bits be Ntotal and the number of erroneous bits be Nerror. The bit error rate is calculated according to the formula BER = Nerror / Ntotal.

[0057] A4: Use the scipy.stats library in Python to calculate the kurtosis of the vibration signal data collected by the acquisition terminal. Assuming that the vibration signal samples are x1, x2, …, xn, the kurtosis calculation formula is:

[0058] in is the sample mean;

[0059] A5: Record the temperature value collected by the acquisition terminal and the corresponding acquisition time. Calculate the temperature gradient based on the two adjacent temperature values ​​and the time interval. Let T1 and T2 be the two adjacent temperature values, and Δt be the time interval. Then the temperature gradient = (T2-T1) / Δt.

[0060] A6: Monitor the voltage data collected by the acquisition terminal in real time and record the rated voltage U of the equipment. nominal And the actual measured voltage value U actual According to the formula VoltageSag=(U nominal -U actual ) / U nominal Calculate the voltage sag amplitude;

[0061] A7: Use the network management system to obtain topological information about the distribution network communication network, including the connection relationships between nodes and links. Abstract the network topology into a graph model, where nodes represent network devices and links represent the communication connections between devices. Use the Dijkstra algorithm from graph theory to calculate the shortest path between any two nodes in the network. For each node pair, count the number of shortest paths. The total number of paths with more than one shortest path is considered the number of redundant paths.

[0062] A8: Use betweenness centrality from graph theory to measure the importance of nodes in a network. Betweenness centrality indicates how often a node appears on the shortest path between other pairs of nodes. For each node v in the graph, calculate its betweenness centrality CB(v). Normalize the betweenness centrality values ​​of all nodes to a range between 0 and 1. Then calculate the variance of these normalized betweenness centrality values ​​to obtain the centrality variance. Also, find the maximum betweenness centrality, which is the maximum centrality. You can use the networkx library in Python to implement graph construction and centrality calculation.

[0063] S3: Communication analysis: input the communication data text obtained in step S3 into the communication function calculation text to calculate and obtain the analysis result text;

[0064] It should be noted that the communication function calculation text includes the communication quality calculation text, the device health calculation text and the network robustness calculation text;

[0065] The analysis result text includes a communication quality index, a device health index, and a network robustness index;

[0066] Specifically, the calculation formula of the communication quality index is:

[0067] Where CQI stands for Communication Quality Index, RSSI refers to Received Signal Strength, SNR refers to Signal-to-Noise Ratio, and BER refers to Bit Error Rate;

[0068] In this embodiment, it should be specifically noted that the RSSI range is -100dBm to 0dBm. This value reflects the signal strength. The larger the value, the better the link quality. The SNR range is: -10dB noise dominant to 20dB clear signal. This value is used to measure the signal anti-interference ability. The larger the value, the smaller the noise impact. The BER range is 0 to 10 -4 Industry standard threshold, which directly reflects the data transmission error rate;

[0069] The calculation formula of the equipment health index is:

[0070] DHI = 0.6 × Kurtosis norm +0.3×Temperature Gradient+0.1×Voltage Sag

[0071]

[0072] DHI represents the device health index, Kurtosismax refers to the historical maximum kurtosis value, Kurtosisnorm refers to the dynamic normalized value of kurtosis, Kurtosis is the kurtosis value, TemperatureGradient refers to the standardized temperature gradient, and VoltageSag refers to the voltage sag.

[0073] It should be noted that the kurtosis ranges from 3 Gaussian distribution to positive infinity. A larger value indicates a stronger signal impact and a higher risk of mechanical failure of the device. The TemperatureGradient refers to the standardized temperature gradient, which reflects the heat dissipation capability of the device. A larger gradient indicates a higher risk of thermal runaway. The voltage sag ranges from 0 to 1. A larger voltage sag indicates worse power supply stability of the device.

[0074] The calculation formula of the network robustness index is:

[0075]

[0076]

[0077] Where NRI represents the network robustness index, Redundancy Ratio represents the normalized number of redundant paths, Redundancy Paths represents the number of redundant paths, Total Nodes represents the total number of nodes, Centrality Variance refers to the centrality variance, and Max Centrality refers to the maximum centrality.

[0078] It should be specifically noted that the more redundant paths there are, the stronger the network's ability to resist failures, the larger the variance, and the more concentrated the network's key nodes are, the higher the vulnerability;

[0079] S4: Comprehensive data processing: input the obtained analysis result text into the communication fault assessment function calculation text to calculate the communication fault assessment index;

[0080] In this embodiment, it should be specifically explained that the communication fault assessment function calculation text is CFI stands for Communication Fault Index, CQI stands for Communication Quality Index, DHI stands for Device Health Index, and NRI stands for Network Robustness Index.

[0081] S5: Communication fault judgment: Compare the communication fault evaluation index of each sub-region obtained in the comprehensive data processing of step S4 with the preset communication fault evaluation index judgment interval, output the judgment result of each sub-region, and analyze the cause of the fault;

[0082] It should be noted that the communication fault assessment index judgment interval is specifically as follows:

[0083] Low risk (0.0≤CFI<0.3). The CFI value is low, indicating that communication quality, device health, and network robustness are all in good condition. This is based on the fact that the combined value of the risk factors in the three dimensions is small. In this case, only routine monitoring is required, and relevant data is collected according to the established period to continuously monitor the system status.

[0084] Medium risk (0.3 ≤ CFI < 0.6): CFI values ​​in this range indicate a system problem, but not yet a significant impact on overall operations. This could be a slight decrease in communication quality, minor equipment anomalies, or a slight reduction in network robustness. In this case, inspections should be strengthened, data collection frequency increased, and communication equipment, lines, and network topology inspected more thoroughly to identify potential faults.

[0085] High risk (0.6 ≤ CFI < 0.8): A high CFI value indicates a significant risk of system failure. This could be due to significant interference in the communication link, equipment aging or damage, or potential vulnerabilities in the network structure. In this case, proactive maintenance measures should be initiated, such as performance testing and calibration of communication equipment, evaluation and replacement planning of aging equipment, and optimization of the network structure, to reduce the likelihood of failure.

[0086] Extremely high risk (0.8 ≤ CFI < 1.0): A CFI value close to 1 indicates that the system is in a very dangerous state and a serious failure has occurred or is about to occur. This could be due to a communication link interruption, severe equipment damage, or network paralysis. Immediate troubleshooting is required. Professional maintenance personnel should conduct a comprehensive inspection of the system to quickly locate and repair the fault point to restore normal system operation.

[0087] In this embodiment, it should be specifically noted that by analyzing the contribution of each component of CFI (1-CQI, DHI, and 1-NRI), the cause of the fault can be located. If the contribution of 1-CQI is high, it indicates that there is a major problem with communication quality, which may be caused by interference in the communication link, insufficient signal strength, or an excessively high bit error rate. In this case, the communication equipment, antennas, and lines should be inspected, and the communication parameters should be adjusted and optimized. If the contribution of DHI is high, it indicates that the equipment is in poor health, which may be caused by hardware failure, overheating, aging, etc., and the equipment needs to be comprehensively inspected and tested, damaged components should be replaced in a timely manner, and aging equipment should be maintained or upgraded. If the contribution of 1-NRI is high, it indicates that the network robustness is poor, which may be caused by an unreasonable network topology, low network redundancy, or a failure in a key node. In this case, the network structure should be re-evaluated, network redundancy should be increased, and key nodes should be backed up and protected.

[0088] S6: Human-computer interaction: output the judgment result to the user information terminal.

[0089] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0090] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for diagnosing distribution network communication faults based on machine learning, characterized by: include: S1: Determine the distribution box data acquisition area: Divide the distribution box to be monitored into multiple monitoring sub-areas according to the distribution branch lines, which are denoted as 1, 2, 3, ..., n in sequence, and each monitoring sub-area is determined as the target monitoring area; S2: Distribution box communication data collection: After the data collection terminal is deployed, data is collected in real time at a sampling frequency of 100 Hz through the data collection terminal to obtain communication data text; The communication data text includes the RSSI value, signal-to-noise ratio, bit error rate, kurtosis, temperature gradient, voltage sag, number of redundant paths, centrality variance and maximum centrality of the target monitoring area; S3: Communication analysis: input the communication data text obtained in step S3 into the communication function calculation text to calculate and obtain the analysis result text; S4: Comprehensive data processing: input the obtained analysis result text into the communication fault assessment function calculation text to calculate the communication fault assessment index; S5: Communication fault judgment: Compare the communication fault evaluation index of each sub-region obtained in the comprehensive data processing of step S4 with the preset communication fault evaluation index judgment interval, output the judgment result of each sub-region, and analyze the cause of the fault; S6: Human-computer interaction: output the judgment result to the user information terminal.

2. A method for diagnosing distribution network communication faults based on machine learning according to claim 1, characterized in that: The process of obtaining various types of data in the communication data text is as follows: A1: Install a communication module with RSSI measurement function on the terminal nodes of the distribution network communication to automatically measure and record the signal strength value. The communication module transmits the measured RSSI value to the data dispatching center via the Ethernet interface. The data dispatching center will store this data and pre-process the RSSI data. By setting upper and lower thresholds, RSSI values ​​that exceed the threshold range are regarded as outliers and eliminated. A2: At the receiving end of the communication, a signal processing algorithm is used to analyze the received signal, separate the signal component from the noise component, and calculate the power of the signal and noise respectively. For discrete signals, the power is calculated by summing the squares of the signal samples and taking the average. Suppose the signal sample sequence is x[n] and the length is N, then the signal power is For a noise sample sequence n[n], the noise power According to the formula Calculate the signal-to-noise ratio in dB; A3: The transmitting and receiving ends of the communication system record the transmitted and received bit sequences, respectively. The transmitting end encodes the data to be transmitted and then sends it. The receiving end decodes the received data and compares the received bit sequence with the transmitted bit sequence bit by bit. The number of erroneous bits is counted. Let the number of transmitted bits be Ntotal and the number of erroneous bits be Nerror. The bit error rate is calculated according to the formula BER = Nerror / Ntotal. A4: Use the scipy.stats library in Python to calculate the kurtosis of the vibration signal data collected by the acquisition terminal. Assuming that the vibration signal samples are x1, x2, …, xn, the kurtosis calculation formula is: in is the sample mean; A5: Record the temperature value collected by the acquisition terminal and the corresponding acquisition time. Calculate the temperature gradient based on the two adjacent temperature values ​​and the time interval. Assume T1 and T2 are the two adjacent temperature values, and Δt is the time interval. Then, the temperature gradient = (T2 - T1) / Δt. A6: Monitor the voltage data collected by the acquisition terminal in real time and record the rated voltage U of the equipment. nominal And the actual measured voltage value U actual According to the formula VoltageSag=(U nominal -U actual ) / U nominal Calculate the voltage sag amplitude; A7: Use the network management system to obtain topological information about the distribution network communication network, including the connection relationships between nodes and links. Abstract the network topology into a graph model, where nodes represent network devices and links represent the communication connections between devices. Use the Dijkstra algorithm from graph theory to calculate the shortest path between any two nodes in the network. For each node pair, count the number of shortest paths. The total number of paths with more than one shortest path is considered the number of redundant paths. A8: Use betweenness centrality from graph theory to measure the importance of nodes in the network. Betweenness centrality indicates how often a node appears on the shortest path between other pairs of nodes. For each node v in the graph, calculate its betweenness centrality CB(v). Normalize the betweenness centrality values ​​of all nodes to a range of 0-1. Then calculate the variance of these normalized betweenness centrality values ​​to obtain the centrality variance. At the same time, find the maximum value among them, which is the maximum centrality.

3. The method for diagnosing distribution network communication faults based on machine learning according to claim 1, characterized in that: The calculation formula of the communication quality index is: CQI refers to the communication quality index, RSSI refers to the received signal strength, SNR refers to the signal-to-noise ratio, and BER refers to the bit error rate.

4. The method for diagnosing distribution network communication faults based on machine learning according to claim 1, wherein: The calculation formula of the equipment health index is: DHI=0.6×Kurtosis norm +0.3×Temperature Gradient+0.1×Voltage Sag DHI represents the device health index, Kurtosismax refers to the historical maximum kurtosis value, Kurtosisnorm refers to the dynamic normalized value of kurtosis, Kurtosis is the kurtosis value, TemperatureGradient refers to the standardized temperature gradient, and VoltageSag refers to the voltage sag.

5. The method for diagnosing distribution network communication faults based on machine learning according to claim 1, characterized in that: The calculation formula of the network robustness index is: Where NRI represents the network robustness index, Redundancy Ratio represents the normalized number of redundant paths, Redundancy Paths represents the number of redundant paths, Total Nodes represents the total number of nodes, Centrality Variance refers to the centrality variance, and Max Centrality refers to the maximum centrality.

6. The method for diagnosing distribution network communication faults based on machine learning according to claim 1, characterized in that: The communication fault evaluation function calculates the text as follows: CFI stands for Communication Fault Index, CQI stands for Communication Quality Index, DHI stands for Device Health Index, and NRI stands for Network Robustness Index.

7. The method for diagnosing distribution network communication faults based on machine learning according to claim 1, characterized in that: The communication fault assessment index judgment interval is as follows: Low risk: 0.0≤CFI<0.3, medium risk: 0.3≤CFI<0.6, high risk: 0.6≤CFI<0.8, and extremely high risk: 0.8≤CFI<1.0.

Images