Power transmission line fault prediction system based on Internet platform

Abstract

The invention discloses a power transmission line fault prediction system based on an Internet platform, and belongs to the technical field of power transmission line fault prediction, and the system comprises an acquisition module which presets a monitoring terminal to capture a traveling wave signal of a power transmission line, and generates a position deviation parameter based on the traveling wave signal; the dynamic compensation module is used for correcting the traveling wave signal through a preset correction unit based on the position deviation parameter to generate a corrected signal; the collaborative learning module comprises a sub-model and a main model, the sub-model generates regional features, and the main model outputs optimized regional features; the positioning prediction module is used for predicting the fault occurrence probability based on the corrected signals and the optimized regional features; generating a health risk level of the monitoring terminal based on a preset historical positioning error library; and the health feedback module generates a position correction instruction and transmits the instruction to the acquisition module. According to the invention, the dynamic compensation module is combined with the position deviation parameter to significantly improve the fault positioning precision.

Description

Technical Field

[0001] The present invention relates to the technical field of power transmission line fault prediction, and in particular to a power transmission line fault prediction system based on an Internet platform. Background Art

[0002] Transmission lines are the core arteries of the power transmission system. They are precisely constructed from components such as conductors, insulator strings, pole tower structures, ground wires, and various connecting hardware. They are erected in a vast natural geographical environment and carry the fundamental mission of efficiently, stably, and reliably transmitting the huge amount of electricity produced by power plants to remote load center areas.

[0003] Existing precise location technologies for transmission line faults generally adopt the dual-end traveling wave location principle as the mainstream solution. Its core working method is to firmly deploy dedicated traveling wave signal monitoring terminals equipped with high-precision time synchronization units at the physical starting and ending points of the transmission line. Once a typical fault event such as a short circuit or ground fault occurs on the line, these terminals will instantly detect and capture the high-frequency transient current or voltage traveling wave physical signal excited by the fault, which propagates extremely rapidly along the line conductor. These signals are converted into electrical signals by highly sensitive sensors and then the absolute timestamp corresponding to the precise arrival of the traveling wave head at the monitoring point is recorded. A remote analysis center uses a communication network to aggregate the absolute timestamp data recorded at both ends and calculate the difference between them. Combined with the known constant propagation velocity of the fault current traveling wave within the conductor of a specific type of transmission line, a preset traveling wave ranging calculation model can be used to infer the theoretical distance of the fault point relative to the physical endpoint of the line, thereby achieving preliminary identification and location of the fault point.

[0004] The existing technology has two defects: on the one hand, during actual installation, the monitoring terminal is often deviated from the theoretical position due to terrain conditions, resulting in attenuation of the traveling wave head signal propagation and time recording deviation, which reduces the positioning accuracy; on the other hand, due to cost or technical limitations, some areas may not have enough monitoring terminals, resulting in data scarcity in these areas and weak ability to identify faults.

[0005] Therefore, there is an urgent need to provide a transmission line fault prediction system based on the Internet platform to solve the above problems. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a power transmission line fault prediction system based on the Internet platform.

[0007] In order to solve the above technical problems, the present invention adopts a technical solution: providing a transmission line fault prediction system based on the Internet platform, including an acquisition module, a dynamic compensation module, a collaborative learning module, a positioning prediction module and a health feedback module;

[0008] An acquisition module, in which a preset monitoring terminal captures the traveling wave signal of the transmission line and generates a position deviation parameter based on the traveling wave signal;

[0009] a dynamic compensation module, which corrects the traveling wave signal based on the position deviation parameter through a preset correction unit to generate a corrected signal, and transmits the corrected signal to the positioning prediction module;

[0010] A collaborative learning module, comprising a sub-model deployed at each of the monitoring terminals and a main model preset on an internet platform, wherein the sub-model generates regional features based on the traveling wave signal and the position deviation parameter, and the main model optimizes the sub-models whose confidence is lower than a preset first threshold and outputs the optimized regional features;

[0011] A positioning prediction module, which predicts the probability of a fault occurrence based on the corrected signal and the optimized regional characteristics. If the probability of a fault occurrence exceeds a preset second threshold, the fault location coordinates are generated and sent to an internet platform. The health risk level of the monitoring terminal is generated based on a preset historical positioning error library.

[0012] The health feedback module updates the position offset record and terrain attenuation coefficient mapping table in the historical positioning error library based on the health risk level, generates a position correction instruction, and transmits it to the acquisition module.

[0013] The present invention is further configured as follows: the monitoring terminal in the acquisition module specifically includes a traveling wave current sensor, an industrial frequency current measurement unit, a power supply module, a communication unit, a data acquisition unit and a GPS clock module.

[0014] The present invention is further configured as follows: the steps for generating the position deviation parameter in the acquisition module are as follows:

[0015] S1. Synchronize the time scale of each monitoring terminal based on the GPS clock module, capture the traveling wave signal detected by the adjacent monitoring terminal, and extract the wave head arrival time and signal amplitude attenuation characteristics in the traveling wave signal;

[0016] S2. Calculate a theoretical distance difference based on the arrival time difference of the wave fronts of the traveling wave signals at adjacent monitoring terminals; synchronously compare the amplitude attenuation parameters of the traveling wave signals monitored by adjacent monitoring terminals, and generate an equivalent propagation path length of the traveling wave signals based on a preset attenuation-distance mapping relationship;

[0017] S3. Input the theoretical distance difference and the equivalent length of the propagation path of the traveling wave signal into a preset spatiotemporal weight function to generate an offset between the actual installation position and the theoretical installation position of the monitoring terminal, and use the offset as a position deviation parameter.

[0018] The present invention is further configured such that: the offset includes a terrain relief parameter and a high-frequency attenuation gradient parameter, the terrain relief parameter being calculated based on terrain elevation difference data at the location where the monitoring terminal is installed; the high-frequency attenuation gradient parameter being obtained by comparing amplitude attenuation parameters of the traveling wave signals of adjacent monitoring terminals;

[0019] The method for generating the corrected signal in the dynamic compensation module is as follows:

[0020] Q1. Based on the position deviation parameter, extract the terrain relief parameter and the high-frequency attenuation gradient parameter, and train a preset machine learning model based on the actual installation location of the monitoring terminal;

[0021] Q2. Input the traveling wave signal into the machine learning model to generate a simulated propagation path of the traveling wave signal in the transmission line, and collect the traveling wave signal attenuation gradient parameter and the traveling wave head arrival time during the propagation of the traveling wave signal along the simulated propagation path;

[0022] Q3. Compare the simulated attenuation gradient parameter of the traveling wave signal with the actual attenuation gradient parameter of the traveling wave signal monitored by the monitoring terminal to generate a first deviation; and simultaneously compare the simulated arrival time of the traveling wave head with the actual arrival time of the traveling wave head of the traveling wave signal monitored by the monitoring terminal to generate a second deviation;

[0023] Q4. Compensate the amplitude attenuation parameter of the traveling wave signal based on the first deviation to generate a compensation parameter, and calibrate the wave head arrival time of the traveling wave signal based on the second deviation, adjust the compensation parameter according to the calibration result, and feed the adjusted compensation parameter back to the traveling wave signal to generate a corrected signal.

[0024] The present invention is further configured as follows: the specific content of step Q4 is as follows:

[0025] Q41. Calculate the difference between the first deviation and a preset amplitude attenuation threshold of the traveling wave signal. If the difference exceeds a set range, adjust the amplitude attenuation parameter of the traveling wave signal according to a preset ratio to generate a preliminary compensation parameter. Based on the difference between the second deviation and a preset wave front time threshold, if the difference is positive and exceeds the set range, advance the wave front arrival time of the traveling wave signal in the machine learning model; if the difference is negative and exceeds the set range, delay the wave front arrival time of the traveling wave signal in the machine learning model to obtain a calibration result.

[0026] Q42. Further adjust the preliminary compensation parameter based on the calibration result: if the calibration result is to advance the arrival time of the wave crest of the traveling wave signal, increase the preliminary compensation parameter by a preset first proportional coefficient; if the preliminary calibration result is to delay the arrival time of the wave crest of the traveling wave signal, decrease the preliminary compensation parameter by a preset second proportional coefficient to generate an adjusted compensation parameter;

[0027] Q43. Apply the adjusted compensation parameter to the traveling wave signal, correct the amplitude attenuation parameter of the traveling wave signal and the wave crest arrival time of the traveling wave signal, and generate a corrected signal.

[0028] The present invention is further configured as follows: a first connection channel is provided between the main model and each sub-model in the collaborative learning module, and a second connection channel is provided between two adjacent sub-models;

[0029] The regional features in the collaborative learning module include local fault area features between adjacent sub-models, propagation path features of the traveling wave signal, frequency distribution features of the traveling wave signal, and load features of the transmission line.

[0030] The present invention is further configured as follows: the steps for calculating the confidence of the sub-model in the collaborative learning module are:

[0031] W1. Transferring the regional features of each sub-model to the main model, and adding up the data amounts of the regional features of all the sub-models as the main weight;

[0032] W2. The main model selects part of the main weight as sub-weight according to the data amount of the regional features in each sub-model in a dynamically set ratio, and distributes it to each sub-model. The scoring unit preset in the main model scores each sub-model according to the regional features, and the confidence of the sub-model is obtained by multiplying the sub-weight multiplied by the scoring result multiplied by the preset adjustment factor.

[0033] The present invention is further configured as follows: the step of generating the optimized regional features in the collaborative learning module is:

[0034] H1. Compare the confidence of the sub-model with the first threshold. If the confidence of the sub-model is less than the first threshold, select multiple sets of regional features transmitted by the sub-model with higher sub-weight as training data, and transmit the training data to the sub-model with confidence lower than the first threshold. If the confidence of the sub-model is greater than or equal to the first threshold, maintain the operating state of the sub-model.

[0035] H2. The sub-model with a confidence level lower than the first threshold uses its own parameters as the main training block, divides several parameters in the sub-model according to the set area as training nodes in the main training block, inputs the training data into the training node located at the center of the main training block to generate first regional feature data, inputs the first regional feature data into other adjacent training nodes in turn to generate multiple second regional feature data, and then follows the above steps until all training data are input to generate the Nth regional feature data, inputs the sub-model that combines the first regional feature data, the second regional feature data and the Nth regional feature data into the main model, trains to obtain an optimized sub-model, inputs the regional features into the optimized sub-model to generate optimized regional features.

[0036] The present invention is further configured as follows: the prediction of the fault occurrence probability in the positioning prediction module specifically comprises: synchronizing the corrected signal with the optimized regional features in time and space to generate synchronized input data; inputting the synchronized input data into a preset fault probability model; combining preset feature allocation weights to output the fault occurrence probability of each segment of the transmission line;

[0037] The method for generating the fault location coordinates is as follows: when the probability of the fault occurrence exceeds the second threshold, the propagation path characteristics of the corrected signal, the local fault area characteristics in the optimized regional characteristics, the frequency distribution characteristics of the traveling wave signal and the load characteristics of the transmission line are combined, and the precise location of the fault is calculated through a preset fault location algorithm to generate the fault location coordinates.

[0038] The present invention is further configured as follows: the specific content of the health feedback module is: based on the positioning error records in the historical positioning error library, combined with the health risk level, dynamically adjust the position offset records in the historical positioning error library, and synchronously update the attenuation compensation parameters in the terrain attenuation coefficient mapping table according to the terrain undulation parameter and the high-frequency attenuation gradient parameter to generate a dynamic attenuation compensation value; based on the adjusted position offset record and the updated terrain attenuation coefficient mapping table, generate a position correction instruction, the position correction instruction includes the position calibration amount of the monitoring terminal and the attenuation compensation parameter of the traveling wave signal, and transmit the position correction instruction to the acquisition module.

[0039] The beneficial effects of the present invention are as follows:

[0040] 1. This invention utilizes a dynamic compensation module combined with position deviation parameters and a machine learning model to simulate the propagation path of traveling wave signals, correcting signal amplitude attenuation and wave front arrival time in real time. It also calculates position offsets based on a spatiotemporal weight function, generates compensation parameters, and feeds them back to the signal acquisition terminal, effectively offsetting installation position deviations caused by terrain and significantly improving fault location accuracy.

[0041] 2. The present invention generates local regional features through the sub-models of the collaborative learning module. The main model optimizes the low-confidence sub-model and outputs the globally optimized regional features. The data sharing and weight distribution strategy of adjacent sub-models is used to make up for the lack of signal features in sparse areas of the monitoring terminal, so that the system can still accurately identify fault modes in areas with insufficient data. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a system flow chart of the present invention;

[0043] Figure 2 A flow chart of the steps for generating optimized regional features of the present invention. DETAILED DESCRIPTION

[0044] The preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more precise definition of the protection scope of the present invention.

[0045] See also Figure 1-Figure 2 , a transmission line fault prediction system based on the Internet platform, including an acquisition module, a dynamic compensation module, a collaborative learning module, a positioning prediction module and a health feedback module;

[0046] The acquisition module uses a preset monitoring terminal to capture the traveling wave signal of the transmission line and generate position deviation parameters based on the traveling wave signal;

[0047] The dynamic compensation module corrects the traveling wave signal based on the position deviation parameter through a preset correction unit to generate a corrected signal, and transmits the corrected signal to the positioning prediction module;

[0048] The collaborative learning module includes a sub-model deployed at each monitoring terminal and a main model preset on the Internet platform. The sub-model generates regional features based on the traveling wave signal and position deviation parameters. The main model optimizes the sub-models whose confidence is lower than a preset first threshold and outputs the optimized regional features.

[0049] The positioning prediction module predicts the probability of fault occurrence based on the corrected signal and optimized regional characteristics. If the probability of fault occurrence exceeds a preset second threshold, the fault location coordinates are generated and sent to the Internet platform. If the probability of fault occurrence does not exceed the preset threshold, the fault probability for the next preset period is continuously predicted. The health risk level of the monitoring terminal is generated based on a preset historical positioning error library.

[0050] The health feedback module, based on the historical positioning error library, updates the position offset records and terrain attenuation coefficient mapping table in the historical positioning error library according to the health risk level, generates a position correction instruction and transmits it to the acquisition module.

[0051] The acquisition module captures the traveling wave signal and generates position deviation parameters, which are corrected by the dynamic compensation module to generate a corrected signal, thereby improving signal accuracy. The collaborative learning module uses the regional features generated by the sub-model, and the main model optimizes the low-confidence sub-model and outputs the optimized regional features to enhance feature reliability. The positioning prediction module combines the corrected signal with the optimized regional features to predict the probability of fault occurrence. When the probability of fault occurrence exceeds the second threshold, the fault location coordinates are generated, and the health risk level is generated based on the historical positioning error library. The health feedback module updates the position offset record and terrain attenuation coefficient mapping table according to the health risk level, generates a position correction instruction closed-loop feedback to the acquisition module, and realizes dynamic optimization of signal acquisition-prediction-calibration, which significantly improves the accuracy of fault location, reduces the false alarm rate, and ensures the long-term stable operation of the transmission line.

[0052] Sub-model: This is a neural network learning model in existing technology, deployed in each monitoring terminal, and generates a localized analysis module of regional characteristics based on traveling wave signals and position deviation parameters;

[0053] Main model: A neural network learning model in the prior art, preset on an internet platform, which receives regional features of sub-models through a first connection channel, optimizes sub-models whose confidence is lower than a preset first threshold, and outputs a global collaborative module of optimized regional features;

[0054] One of the embodiments of the present invention is: the monitoring terminal in the acquisition module specifically includes a traveling wave current sensor, an industrial frequency current measurement unit, a power supply module (such as solar power supply), a communication unit (4G / 5G wireless transmission), a data acquisition unit and a GPS clock module.

[0055] Specifically, the steps for generating the position deviation parameters in the acquisition module are as follows:

[0056] S1. Synchronize the time scale of each monitoring terminal based on the GPS clock module, capture the traveling wave signal detected by the adjacent monitoring terminal, and extract the wave head arrival time and signal amplitude attenuation characteristics in the traveling wave signal;

[0057] S2. Calculate the theoretical distance difference based on the arrival time difference of the traveling wave signals of adjacent monitoring terminals; synchronously compare the amplitude attenuation parameters of the traveling wave signals monitored by adjacent monitoring terminals, and generate the equivalent propagation path length of the traveling wave signal based on a preset attenuation-distance mapping relationship;

[0058] The theoretical distance difference between adjacent monitoring terminals is calculated as follows: the velocity of the traveling wave signal multiplied by the arrival time difference of the traveling wave signal heads at adjacent monitoring terminals. In actual calculations, the velocity of the traveling wave signal can be calculated based on the material, structure and other parameters of the transmission line using relevant electromagnetic theory formulas to obtain a relatively accurate theoretical value.

[0059] The preset attenuation-distance mapping relationship specifically refers to the functional correspondence between signal amplitude attenuation and propagation distance established under a specific transmission line environment, taking into account factors such as energy diffusion, conductor resistance loss, dielectric absorption, and impact corona effect during the propagation of traveling wave signals.

[0060] S3. Input the theoretical distance difference and the equivalent length of the propagation path of the traveling wave signal into a preset spatiotemporal weight function to generate an offset between the actual installation position and the theoretical installation position of the monitoring terminal, and use the offset as a position deviation parameter.

[0061] The calculation formula of the spatiotemporal weight function is:

[0062] ΔP=k(ΔDt-ΔDs)+βΔDs;

[0063] Among them, ΔP is the spatiotemporal weight function; k is the time dimension weight factor, which defaults to 0.7 and increases dynamically with the terrain undulation; ΔDt is the theoretical distance difference; ΔDs is the equivalent length of the propagation path of the traveling wave signal; β is the spatial dimension compensation coefficient, which defaults to 0.3 and increases dynamically with the high-frequency attenuation gradient.

[0064] Specifically, the offset includes the terrain undulation parameter and the high-frequency attenuation gradient parameter. The terrain undulation parameter is calculated based on the terrain elevation difference data at the installation location of the monitoring terminal (used to characterize the impact of terrain changes on signal propagation); the high-frequency attenuation gradient parameter is obtained by comparing the amplitude attenuation parameters of the traveling wave signals of adjacent monitoring terminals (used to quantify the degree of signal loss in the propagation path).

[0065] The beneficial effects of this embodiment are as follows: high-precision time synchronization of each monitoring terminal is achieved through the GPS clock module, ensuring that the calculation error of the arrival time difference of the traveling wave signal head is reduced; the traveling wave signal velocity is calculated based on the electromagnetic theory formula, so that the accuracy of the theoretical distance difference is improved; the equivalent length of the propagation path is generated by using the preset attenuation and distance mapping relationship (comprehensive energy diffusion, conductor resistance loss, dielectric absorption and impact corona effect and other factors), which effectively quantifies the actual propagation loss of the signal in complex terrain; the difference between the theoretical distance difference and the equivalent length of the propagation path is dynamically integrated through the time-space weight function (where the time dimension weight factor 0.7 increases with the terrain undulation, and the space dimension compensation coefficient 0.3 increases with the high-frequency attenuation gradient), so that the calculation error of the position offset is reduced; the synergistic effect of the terrain undulation parameter and the high-frequency attenuation gradient parameter accurately characterizes the impact of terrain changes and high-frequency losses on signal propagation, and ultimately improves the fault location accuracy.

[0066] In one embodiment of the present invention, the method for generating the corrected signal in the dynamic compensation module is as follows:

[0067] Q1. Based on the position deviation parameters, the terrain relief parameters and high-frequency attenuation gradient parameters are extracted, and the preset machine learning model is trained based on the actual installation location of the monitoring terminal.

[0068] The actual installation location of the monitoring terminal can be obtained by a preset GPS positioning module. The machine learning model is a learning model in the existing technology. It is trained by inputting data such as terrain relief parameters, high-frequency attenuation gradient parameters, and the actual installation location of the monitoring terminal to learn the propagation law of traveling wave signals under different terrain and attenuation conditions.

[0069] Q2. Input the traveling wave signal into the machine learning model to generate a simulated propagation path of the traveling wave signal in the transmission line, and collect the traveling wave signal attenuation gradient parameters and the traveling wave head arrival time during the propagation process along the simulated propagation path;

[0070] Q3. Compare the attenuation gradient parameters of the simulated traveling wave signal with the attenuation gradient parameters of the actual traveling wave signal monitored by the monitoring terminal to generate a first deviation; and simultaneously compare the arrival time of the simulated traveling wave head with the arrival time of the actual traveling wave head monitored by the monitoring terminal to generate a second deviation;

[0071] The attenuation gradient parameter of the actual traveling wave signal and the wave head arrival time of the actual traveling wave signal can be monitored and obtained by the monitoring terminal;

[0072] Q4. Compensate the amplitude attenuation parameter of the traveling wave signal based on the first deviation to generate a compensation parameter, and calibrate the wave head arrival time of the traveling wave signal based on the second deviation. Adjust the compensation parameter according to the calibration result, and feed the adjusted compensation parameter back to the traveling wave signal to generate a corrected signal.

[0073] Specifically, the specific content of step Q4 is as follows:

[0074] Q41. Calculate the difference between the first deviation and the preset amplitude attenuation threshold of the traveling wave signal. If the difference exceeds the set range, adjust the amplitude attenuation parameter of the traveling wave signal according to the preset ratio to generate a preliminary compensation parameter. According to the difference between the second deviation and the preset wave head time threshold, if the difference is positive and exceeds the set range (indicating that the simulated wave head arrival time is later than the actual wave head arrival time), advance the wave head arrival time of the traveling wave signal in the machine learning model. If the difference is negative and exceeds the set range (indicating that the simulated wave head arrival time is earlier than the actual wave head arrival time), postpone the wave head arrival time of the traveling wave signal in the machine learning model to obtain a calibration result. If the difference does not exceed the set range, take the wave head arrival time of the current traveling wave signal as the calibration result.

[0075] Advancing the arrival time of the wave head of the traveling wave signal in the machine learning model and delaying the arrival time of the wave head of the traveling wave signal in the machine learning model, wherein the advancement and delay are as follows: on the time axis of the machine learning model, moving the arrival time mark point of the wave head of the traveling wave signal forward or backward according to a certain time step. When the difference is positive and exceeds the set range, the operation of advancing the arrival time of the wave head is to make the simulated traveling wave signal more consistent with the actual monitoring situation and to more accurately reflect the actual propagation process of the traveling wave signal in the transmission line. For example, if the difference is positive and exceeds the set range of 0.1 seconds, the arrival time of the wave head of the traveling wave signal can be advanced by 0.1 seconds in the machine learning model, so as to correct the deviation of the simulated signal in the time dimension.

[0076] When the difference is negative and exceeds the set range, the arrival time of the wave head of the traveling wave signal is delayed to calibrate the time difference between the simulated signal and the actual signal. For example, if the difference is negative and exceeds the set range of 0.08 seconds, the arrival time of the wave head of the traveling wave signal is delayed by 0.08 seconds in the machine learning model.

[0077] Amplitude attenuation threshold of traveling wave signal: set to 1.5dB / km (reference value for lines with voltage levels of 110kV and above) based on typical operating conditions of transmission lines;

[0078] Preset ratio: The preset ratio of the amplitude attenuation parameter adjustment is fixed at 0.6 (i.e. 60% of the compensation deviation);

[0079] Setting range: The setting range of the wave head time difference is ±0.5μs (corresponding to the traveling wave propagation distance error of ±150 meters);

[0080] Q42. Further adjust the preliminary compensation parameters based on the calibration results: if the calibration result is to advance the arrival time of the wave front of the traveling wave signal, increase the preliminary compensation parameters by a preset first proportional coefficient; if the preliminary calibration result is to delay the arrival time of the wave front of the traveling wave signal, decrease the preliminary compensation parameters by a preset second proportional coefficient to generate adjusted compensation parameters;

[0081] Preset the first proportional coefficient: When the calibration result is to advance the wave front arrival time, the compensation parameter increase ratio is fixed to 1.25 (i.e. 25% increase);

[0082] Preset the second proportional coefficient: when the calibration result is to delay the wave head arrival time, the compensation parameter reduction ratio is fixed to 0.85 (i.e. 15% attenuation);

[0083] Q43. Apply the adjusted compensation parameters to the traveling wave signal, correct the amplitude attenuation parameter of the traveling wave signal and the wave head arrival time of the traveling wave signal, and generate a corrected signal.

[0084] The beneficial effects of this embodiment are as follows: the influence of terrain undulation parameters and high-frequency attenuation gradient parameters on the propagation law of traveling wave signals is learned through a machine learning model, a simulated propagation path is generated, and the attenuation gradient parameters and wave head arrival time are extracted; the amplitude attenuation parameters and wave head arrival time are dynamically adjusted by comparing the first deviation and the second deviation generated by the simulation value and the actual monitoring value; the wave head time difference calibration mechanism (the wave head mark is advanced or delayed when the difference exceeds ±0.5μs) and the compensation parameter linkage adjustment (the compensation parameter is adjusted by 1.25 times the amplification when advancing the wave head and by 0.85 times the attenuation when delaying the wave head) are utilized to achieve coordinated optimization of the propagation path characteristics and the signal time domain characteristics, so that the corrected signal accurately reflects the real propagation process, and the fault monitoring accuracy is significantly improved.

[0085] One embodiment of the present invention is as follows: a first connection channel is provided between the main model and each sub-model in the collaborative learning module, and a second connection channel is provided between two adjacent sub-models;

[0086] The regional features in the collaborative learning module include the local fault area characteristics between adjacent sub-models, the propagation path characteristics of the traveling wave signal, the frequency distribution characteristics of the traveling wave signal, and the load characteristics of the transmission line. These propagation path characteristics, the frequency distribution characteristics of the traveling wave signal, and the load characteristics of the transmission line comprehensively reflect the operating status and potential risks of the transmission line in different regions.

[0087] Specifically, the confidence calculation steps of the sub-model in the collaborative learning module are:

[0088] W1, transfer the regional features of each sub-model to the main model, and add up the data volume of the regional features of all sub-models as the main weight;

[0089] W2. The main model selects part of the main weight as the sub-weight according to the amount of data of the regional characteristics in each sub-model in a dynamically set ratio and distributes it to each sub-model. The scoring unit preset in the main model scores each sub-model according to the regional characteristics, and the confidence of the sub-model is obtained by multiplying the sub-weight multiplied by the scoring result multiplied by the preset adjustment factor.

[0090] Dynamic setting ratio: The dynamic setting ratio can be adjusted according to the actual operation conditions and data characteristics of the transmission line. During the period when the operation of the transmission line is relatively stable and the data fluctuations are small, the dynamic setting ratio can be appropriately lowered, preferably set to 0.3, so that each sub-model is assigned relatively fewer sub-weights, the main model has a stronger control over the overall regional characteristics, and pays more attention to global feature optimization. When the transmission line is in a special operating state, such as when the load changes greatly or it is subject to external interference, the data fluctuates greatly. At this time, the dynamic setting ratio can be appropriately increased, preferably set to 0.7, so that each sub-model can be assigned more sub-weights, and the sub-model's ability to analyze local regional characteristics is more prominent, so as to better adapt to the complex and changing operating environment.

[0091] The preset adjustment factor is designed to further balance the differences between different sub-models. It is set based on factors such as the sub-model's historical performance and the importance of the region in which it is located. For sub-models with good historical performance and a region that has a greater impact on the overall operation of the transmission line, the preset adjustment factor can be appropriately increased, preferably set to 1.2, to improve its confidence. For sub-models with average historical performance and a smaller regional impact, the preset adjustment factor can be appropriately reduced, preferably set to 0.8, to reduce its confidence. In this way, the main model can more reasonably optimize low-confidence sub-models, improving the overall collaborative learning effect.

[0092] Specifically, the steps for generating optimized regional features in the collaborative learning module are:

[0093] H1. Compare the confidence of the sub-model with a first threshold. If the confidence of the sub-model is less than the first threshold, select multiple sets of regional features transmitted by the sub-model with higher sub-weight as training data, and transmit the training data to the sub-model with confidence lower than the first threshold. If the confidence of the sub-model is greater than or equal to the first threshold, maintain the operation state of the sub-model.

[0094] H2. The sub-model with a confidence level lower than the first threshold uses its own parameters as the main training block, divides several parameters in the sub-model according to the set area as training nodes in the main training block, inputs the training data into the training node located at the center of the main training block to generate first regional feature data, inputs the first regional feature data into other adjacent training nodes in turn to generate multiple second regional feature data, and then follows the above steps until all training data are input to generate the Nth regional feature data. The sub-model that combines the first regional feature data, the second regional feature data, and the Nth regional feature data is input into the main model, and the optimized sub-model is trained. The regional features are input into the optimized sub-model to generate optimized regional features.

[0095] The beneficial effect of this embodiment is that by setting the first connection channel and the second connection channel, the information interaction between the main model and the sub-model and between the sub-models is strengthened. Taking into account the local fault area characteristics, propagation path characteristics, frequency distribution characteristics and transmission line load characteristics, the operation status and potential risks of the transmission line can be more comprehensively reflected. The calculation method of the sub-model confidence dynamically allocates sub-weights according to the amount of regional feature data and combines the scores to make the confidence assessment more reasonable. For sub-models with confidence lower than the first threshold, the regional characteristics of the sub-models with higher sub-weights are used for training and optimization. Multiple sets of regional feature data are generated by multiple rounds of input training data, which can effectively improve the performance of the sub-model, so that the optimized regional characteristics can more accurately reflect the actual situation of the transmission line, further improving the accuracy and reliability of transmission line fault prediction. At the same time, the operation status of the sub-model with a confidence greater than or equal to the first threshold is maintained to ensure the stability and efficiency of the system.

[0096] In one embodiment of the present invention, the prediction of the fault probability in the positioning prediction module specifically comprises: synchronizing the corrected signal with the optimized regional features in time and space to generate synchronized input data; inputting the synchronized input data into a preset fault probability model; assigning weights based on preset features; and outputting the fault probability for each segment of the transmission line.

[0097] Fault probability model: This is a mathematical model built based on historical fault data, transmission line parameters, and real-time monitoring data. It uses machine learning or deep learning algorithms to analyze and learn from synchronized input data to capture the potential relationship between transmission line operating status and fault occurrence.

[0098] The preset feature allocation weights are set according to the degree of influence of different characteristics of the transmission line on the occurrence of faults. These characteristics include but are not limited to the amplitude, frequency, wave head arrival time, propagation path characteristics, regional characteristics, etc. of the traveling wave signal.

[0099] The method for generating the fault location coordinates is as follows: when the probability of a fault occurring exceeds a second threshold, the precise location of the fault is calculated through a preset fault location algorithm based on the propagation path characteristics of the corrected signal, the local fault area characteristics in the optimized regional characteristics, the frequency distribution characteristics of the traveling wave signal, and the load characteristics of the transmission line, to generate the fault location coordinates.

[0100] Fault location algorithm: This algorithm is based on the topological structure of the transmission line, the principle of traveling wave propagation, and the above-mentioned characteristics. It uses the propagation path characteristics of the corrected signal to determine the propagation trajectory of the traveling wave on the transmission line. The local fault area characteristics in the optimized regional characteristics are combined to narrow the scope of possible fault occurrence. The frequency distribution characteristics of the traveling wave signal are used to analyze the impact of the fault on the signal frequency. The load characteristics of the transmission line are then considered to determine the possibility of fault occurrence under different load conditions.

[0101] When calculating the fault location coordinates, the fault location algorithm first estimates the approximate range where the fault may occur based on the propagation speed and time information of the traveling wave signal in the transmission line, combined with the propagation path characteristics of the corrected signal. This approximate range is then further screened and refined using the local fault area characteristics within the optimized regional features, eliminating areas where faults are unlikely to occur. The fault location algorithm then analyzes the frequency distribution characteristics of the traveling wave signal. Different types of faults can cause varying degrees of variation in the frequency of the traveling wave signal. By analyzing the frequency distribution characteristics, the type and severity of the fault can be more accurately determined, further narrowing the fault location. Furthermore, the load characteristics of the transmission line also affect fault location. The propagation characteristics of the traveling wave signal vary under different load conditions. The algorithm takes these load characteristics into account and corrects and adjusts the previously estimated fault location.

[0102] The beneficial effects of this embodiment are as follows: by synchronizing the corrected signal with the optimized regional features in time and space, generating synchronized input data and inputting it into the fault probability model, and combining the feature allocation weights to output the probability of fault occurrence in each segmented interval, the fault risk of each part of the transmission line can be more accurately assessed. When the probability of fault occurrence exceeds the second threshold, the fault location coordinates are calculated using a preset fault location algorithm, taking into account the propagation path characteristics of the corrected signal, the local fault area characteristics in the optimized regional features, the frequency distribution characteristics of the traveling wave signal, and the load characteristics of the transmission line, thereby realizing a complete process from fault risk assessment to accurate fault location. This method of comprehensively predicting and locating faults based on multiple features can more comprehensively and deeply analyze the operating status of the transmission line, effectively improving the accuracy and reliability of fault monitoring and location, providing strong technical support for the maintenance and management of the transmission line, helping to timely discover and handle potential faults, ensuring the safe and stable operation of the transmission line, and reducing losses such as power outages caused by faults, with significant economic and social benefits.

[0103] One of the embodiments of the present invention is: the specific content of the health feedback module is: based on the positioning error records in the historical positioning error library, combined with the health risk level, dynamically adjust the position offset records in the historical positioning error library, and synchronously update the attenuation compensation parameters in the terrain attenuation coefficient mapping table according to the terrain undulation parameter and the high-frequency attenuation gradient parameter to generate a dynamic attenuation compensation value; based on the adjusted position offset record and the updated terrain attenuation coefficient mapping table, generate a position correction instruction, the position correction instruction includes the position calibration amount of the monitoring terminal and the attenuation compensation parameter of the traveling wave signal, and transmit the position correction instruction to the acquisition module.

[0104] The beneficial effects of this embodiment are: by dynamically adjusting the position offset records in the historical positioning error library and updating the terrain attenuation coefficient mapping table, dynamic attenuation compensation values ​​and position correction instructions that are more consistent with actual conditions are generated. The position correction instructions can accurately calibrate the position of the monitoring terminal, enabling the monitoring terminal to more accurately collect traveling wave signals and reduce monitoring errors caused by inaccurate position;

[0105] At the same time, updating the attenuation compensation parameters of the traveling wave signal can more effectively compensate for the attenuation of the traveling wave signal during propagation, improving the quality and accuracy of the signal. This series of operations further enhances the reliability and stability of the entire transmission line fault prediction system, enabling the system to more accurately reflect the actual operating conditions of the transmission line, providing a solid guarantee for the safe operation of the transmission line.

[0106] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention's description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A power transmission line fault prediction system based on an Internet platform, characterized by: It includes acquisition module, dynamic compensation module, collaborative learning module, positioning prediction module and health feedback module; An acquisition module, in which a preset monitoring terminal captures the traveling wave signal of the transmission line and generates a position deviation parameter based on the traveling wave signal; a dynamic compensation module, which corrects the traveling wave signal based on the position deviation parameter through a preset correction unit to generate a corrected signal, and transmits the corrected signal to the positioning prediction module; A collaborative learning module, comprising a sub-model deployed at each of the monitoring terminals and a main model preset on an internet platform, wherein the sub-model generates regional features based on the traveling wave signal and the position deviation parameter, and the main model optimizes the sub-models whose confidence is lower than a preset first threshold and outputs the optimized regional features; A positioning prediction module, which predicts the probability of a fault occurrence based on the corrected signal and the optimized regional characteristics, and generates fault location coordinates and sends them to an Internet platform if the probability of a fault occurrence exceeds a preset second threshold; Generate health risk level of monitoring terminal based on preset historical positioning error library; The health feedback module updates the position offset record and terrain attenuation coefficient mapping table in the historical positioning error library based on the health risk level, generates a position correction instruction, and transmits it to the acquisition module.

2. The Internet-based power transmission line fault prediction system according to claim 1, characterized in that: The monitoring terminal in the acquisition module specifically includes a traveling wave current sensor, an industrial frequency current measurement unit, a power supply module, a communication unit, a data acquisition unit and a GPS clock module.

3. The Internet-based power transmission line fault prediction system according to claim 2, characterized in that: The steps for generating the position deviation parameters in the acquisition module are as follows: S1. Synchronize the time scale of each monitoring terminal based on the GPS clock module, capture the traveling wave signal detected by the adjacent monitoring terminal, and extract the wave head arrival time and signal amplitude attenuation characteristics in the traveling wave signal; S2. Calculate a theoretical distance difference based on the arrival time difference of the wave fronts of the traveling wave signals at adjacent monitoring terminals; synchronously compare the amplitude attenuation parameters of the traveling wave signals monitored by adjacent monitoring terminals, and generate an equivalent propagation path length of the traveling wave signals based on a preset attenuation-distance mapping relationship; S3. Input the theoretical distance difference and the equivalent length of the propagation path of the traveling wave signal into a preset spatiotemporal weight function to generate an offset between the actual installation position and the theoretical installation position of the monitoring terminal, and use the offset as a position deviation parameter.

4. The Internet-based power transmission line fault prediction system according to claim 3, characterized in that: The offset includes a terrain relief parameter and a high-frequency attenuation gradient parameter. The terrain relief parameter is calculated based on the terrain elevation difference data at the location where the monitoring terminal is installed. The high-frequency attenuation gradient parameter is obtained by comparing the amplitude attenuation parameters of the traveling wave signals of adjacent monitoring terminals. The method for generating the corrected signal in the dynamic compensation module is as follows: Q1. Based on the position deviation parameter, extract the terrain relief parameter and the high-frequency attenuation gradient parameter, and train a preset machine learning model based on the actual installation location of the monitoring terminal; Q2. Input the traveling wave signal into the machine learning model to generate a simulated propagation path of the traveling wave signal in the transmission line, and collect the traveling wave signal attenuation gradient parameter and the traveling wave head arrival time during the propagation of the traveling wave signal along the simulated propagation path; Q3. Compare the simulated attenuation gradient parameter of the traveling wave signal with the actual attenuation gradient parameter of the traveling wave signal monitored by the monitoring terminal to generate a first deviation; and simultaneously compare the simulated arrival time of the traveling wave head with the actual arrival time of the traveling wave head of the traveling wave signal monitored by the monitoring terminal to generate a second deviation; Q4. Compensate the amplitude attenuation parameter of the traveling wave signal based on the first deviation to generate a compensation parameter, and calibrate the wave head arrival time of the traveling wave signal based on the second deviation, adjust the compensation parameter according to the calibration result, and feed the adjusted compensation parameter back to the traveling wave signal to generate a corrected signal.

5. The Internet-based power transmission line fault prediction system according to claim 4, characterized in that: The specific content of step Q4 is as follows: Q41. Calculate the difference between the first deviation and a preset amplitude attenuation threshold of the traveling wave signal. If the difference exceeds a set range, adjust the amplitude attenuation parameter of the traveling wave signal according to a preset ratio to generate a preliminary compensation parameter. Based on the difference between the second deviation and a preset wave front time threshold, if the difference is positive and exceeds the set range, advance the wave front arrival time of the traveling wave signal in the machine learning model; if the difference is negative and exceeds the set range, delay the wave front arrival time of the traveling wave signal in the machine learning model to obtain a calibration result. Q42. Further adjust the preliminary compensation parameter based on the calibration result: if the calibration result is to advance the arrival time of the wave crest of the traveling wave signal, increase the preliminary compensation parameter by a preset first proportional coefficient; if the preliminary calibration result is to delay the arrival time of the wave crest of the traveling wave signal, decrease the preliminary compensation parameter by a preset second proportional coefficient to generate an adjusted compensation parameter; Q43. Apply the adjusted compensation parameter to the traveling wave signal, correct the amplitude attenuation parameter of the traveling wave signal and the wave crest arrival time of the traveling wave signal, and generate a corrected signal.

6. The Internet-based power transmission line fault prediction system according to claim 5, characterized in that: A first connection channel is provided between the main model and each sub-model in the collaborative learning module, and a second connection channel is provided between two adjacent sub-models; The regional features in the collaborative learning module include local fault area features between adjacent sub-models, propagation path features of the traveling wave signal, frequency distribution features of the traveling wave signal, and load features of the transmission line.

7. The Internet-based power transmission line fault prediction system according to claim 6, characterized in that: The calculation steps of the confidence of the sub-model in the collaborative learning module are: W1. Transferring the regional features of each sub-model to the main model, and adding up the data amounts of the regional features of all the sub-models as the main weight; W2. The main model selects part of the main weight as sub-weight according to the data amount of the regional features in each sub-model in a dynamically set ratio, and distributes it to each sub-model. The scoring unit preset in the main model scores each sub-model according to the regional features, and the confidence of the sub-model is obtained by multiplying the sub-weight multiplied by the scoring result multiplied by the preset adjustment factor.

8. The Internet-based power transmission line fault prediction system according to claim 7, characterized in that: The steps for generating the optimized regional features in the collaborative learning module are: H1. Compare the confidence of the sub-model with the first threshold. If the confidence of the sub-model is less than the first threshold, select multiple sets of regional features transmitted by the sub-model with higher sub-weight as training data, and transmit the training data to the sub-model with confidence lower than the first threshold. If the confidence of the sub-model is greater than or equal to the first threshold, maintain the operating state of the sub-model. H2. The sub-model with a confidence level lower than the first threshold uses its own parameters as the main training block, divides several parameters in the sub-model according to the set area as training nodes in the main training block, inputs the training data into the training node located at the center of the main training block to generate first regional feature data, inputs the first regional feature data into other adjacent training nodes in turn to generate multiple second regional feature data, and then follows the above steps until all training data are input to generate the Nth regional feature data, inputs the sub-model that combines the first regional feature data, the second regional feature data and the Nth regional feature data into the main model, trains to obtain an optimized sub-model, inputs the regional features into the optimized sub-model to generate optimized regional features.

9. The Internet-based power transmission line fault prediction system according to claim 8, characterized in that: The prediction of the fault probability in the positioning prediction module specifically comprises: synchronizing the corrected signal with the optimized regional features in time and space to generate synchronized input data, inputting the synchronized input data into a preset fault probability model, assigning weights based on preset features, and outputting the fault probability of each segment of the transmission line; The method for generating the fault location coordinates is as follows: when the probability of the fault occurrence exceeds the second threshold, the propagation path characteristics of the corrected signal, the local fault area characteristics in the optimized regional characteristics, the frequency distribution characteristics of the traveling wave signal and the load characteristics of the transmission line are combined, and the precise location of the fault is calculated through a preset fault location algorithm to generate the fault location coordinates.

10. The Internet-based power transmission line fault prediction system according to claim 9, characterized in that: The specific content of the health feedback module is: based on the positioning error records in the historical positioning error library and combined with the health risk level, the position offset records in the historical positioning error library are dynamically adjusted, and the attenuation compensation parameters in the terrain attenuation coefficient mapping table are synchronously updated according to the terrain undulation parameter and the high-frequency attenuation gradient parameter to generate a dynamic attenuation compensation value; based on the adjusted position offset record and the updated terrain attenuation coefficient mapping table, a position correction instruction is generated, the position correction instruction includes the position calibration amount of the monitoring terminal and the attenuation compensation parameter of the traveling wave signal, and the position correction instruction is transmitted to the acquisition module.

Images