Fault monitoring method and system for extra-high voltage equipment based on machine learning

Abstract

The invention relates to a fault monitoring method and system for extra-high voltage equipment based on machine learning, and the method comprises the steps: forming a comprehensive diagnosis database of the operation state of the extra-high voltage equipment based on fault related data in the historical operation process of the extra-high voltage equipment; based on the fault feature data in the ultra-high voltage equipment operation state comprehensive diagnosis database, a quantifiable oil gas data distribution rule description index is established; training an artificial neural network based on the fault related data and quantifiable oil and gas data distribution rule description indexes, and establishing an extra-high voltage equipment abnormal working condition analysis model; and collecting concentration signals and proportion signals of a plurality of gas components of the oil chromatography of the extra-high voltage equipment, inputting the concentration signals and the proportion signals into the abnormal working condition analysis model of the extra-high voltage equipment, and judging whether the concentration signals and the proportion signals are abnormal or not.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical equipment operation and maintenance of ultra-high voltage power grids, and in particular to a method and system for fault monitoring of ultra-high voltage equipment based on machine learning. Background Art

[0002] Major accidents during the operation of UHV power grids will affect the normal operation of the power system and cause damage to industrial production facilities. UHV transformers and reactors are important equipment for maintaining the stable operation of UHV power grids. Transformer handover and preventive tests are all carried out when the equipment is offline. The detection cycle is long and cannot reflect the real-time status of the equipment. Therefore, transformer and reactor diagnosis and early warning based on online monitoring devices have become an important means of production management, realizing early warning of internal defects of equipment and avoiding the occurrence of failures. Oil chromatography online monitoring technology is currently the most widely used and effective transformer monitoring technology.

[0003] However, the monitoring data of UHV transformers and reactors are characterized by multiple sources, large total volume, and complex attributes. The online monitoring data sent back is often incomplete, noisy, and inconsistent, mainly manifested as missing, noise, outliers, over-limit, inconsistency, etc. The data quality is difficult to meet the requirements of subsequent equipment and data source evaluation, resulting in a large number of false alarms and multiple missed reports. In addition, the current oil chromatography fault diagnosis and analysis method often uses empirical methods such as three ratios and David's triangle. This method has a certain degree of universality, but in the actual specific operating environment, there is a large deviation in the judgment of the operating status of the transformer. Summary of the invention

[0004] The present invention provides a fault monitoring method and system for UHV equipment based on machine learning, which can ensure the high integration and automation of the monitoring system, improve the operating efficiency and reliability of the system, and also facilitate maintenance and upgrading, thereby enhancing the practicality and market competitiveness of the system.

[0005] The present invention solves the above technical problems in the following way: A method for fault monitoring of UHV equipment based on machine learning, comprising:

[0006] Based on the fault-related data of the UHV equipment during its historical operation, a comprehensive diagnosis database for the operation status of UHV equipment is formed;

[0007] Based on the fault feature data in the comprehensive diagnosis database of UHV equipment operation status, a quantifiable description index of oil and gas data distribution law is established;

[0008] Based on the fault-related data and quantifiable oil and gas data distribution law description indicators, an artificial neural network is trained to establish an abnormal operating condition analysis model for UHV equipment;

[0009] The concentration signals and proportion signals of multiple gas components of the oil chromatogram of the UHV equipment are collected, and the concentration signals and proportion signals are input into the abnormal operating condition analysis model of the UHV equipment to determine whether the concentration signals and proportion signals are abnormal.

[0010] It should be noted that the ultra-high voltage range referred to in the present invention is the voltage level of ±800 kV and above for direct current and 1000 kV and above for alternating current, and ultra-high frequency refers to the radio wave frequency band with a frequency range between 300 MHz and 3 GHz.

[0011] Preferably, the formation of a comprehensive diagnosis database for the operation status of UHV equipment based on the fault-related data during the historical operation of the UHV equipment includes:

[0012] The fault-related data of UHV equipment during historical operation are collected and normalized. The formula is as follows:

[0013]

[0014] Where X is the actual value in the fault-related data, x max and x min are the maximum and minimum values ​​in the fault-related data, respectively, max and min is the normalized range, y is the normalized value of x;

[0015] The back propagation artificial neural network is trained based on the normalized fault-related data to form a comprehensive diagnosis database for the operation status of UHV equipment. The formula is as follows:

[0016]

[0017]

[0018] In the formula, x i is the input value of the normalized fault-related data, ω ijis the connection weight between neurons i and j, which can represent the influence of a specific input feature (such as a certain gas concentration) on the output of the hidden layer neurons. b is a bias term in the form of a threshold, which represents the minimum signal strength required to trigger the neuron to identify a specific fault mode. s is the weighted signal received by each neuron in the hidden layer from all input nodes, which represents the fault probability calculated based on all input features (such as different gas concentrations). N represents the number of input signal sources received by each neuron, that is, the total number of input layer nodes, which represents the total number of oil and gas data features used for analysis. y is the output value of the sigmoid activation function, which represents the probability of a fault. In fault monitoring, this can represent the possibility of a specific fault mode. e represents the base of the natural logarithm, also known as Euler's number, and its approximate value is 2.71828.

[0019] By collecting fault-related data from the historical operation of UHV equipment and normalizing it, the consistency and comparability of the data are ensured, so that data of different dimensions and magnitudes can be processed fairly in the neural network. Normalization can prevent certain features from excessively affecting the model results during training due to their large numerical range, while improving the convergence speed of the algorithm and the generalization ability of the model.

[0020] Preferably, the fault-related data during the historical operation of the UHV equipment include basic parameters of the UHV equipment, maintenance statistics, insulating oil test data, main body offline oil chromatography, bushing, gas oil chromatography test data, insulation resistance measurement data, capacitance and dielectric loss measurement data, bushing test data, core clamp grounding current data, DC leakage current data, infrared testing, vibration, sound level, pulse current method partial discharge, UHF, high frequency and ultrasonic data.

[0021] By comprehensively considering the multi-dimensional fault-related data generated by UHV equipment during historical operation, fault modes and risk factors can be more accurately identified, thereby improving the accuracy and timeliness of fault prediction, optimizing maintenance strategies, reducing unexpected downtime, extending equipment life, and enhancing the stability and security of the entire power grid. In addition, this comprehensive data-driven approach can also help experts discover new fault associations and promote the advancement of fault diagnosis technology.

[0022] Preferably, the establishment of quantifiable oil and gas data distribution law description indicators based on the fault feature data in the comprehensive diagnosis database of the UHV equipment operation status specifically includes:

[0023] Extract fault feature data from the comprehensive diagnostic database of UHV equipment operation status, including basic parameters, maintenance statistics, and experimental data, and extract key features related to oil and gas distribution patterns, including dissolved gas concentration and changes in the physical and chemical properties of oil;

[0024] Based on the extracted features, quantifiable indicators describing the distribution patterns of oil and gas data are constructed.

[0025] The quantifiable oil and gas data distribution law description indicators, such as gas concentration level (such as H 2 ,CO,CO 2 , CH 4 , C 2 H 6 , C 2 H 4 , C 2 H 2 etc.); gas growth rate; gas concentration ratio; temperature change; pressure fluctuation. It can objectively reflect the operating status and potential failure risks of the equipment, provide a scientific basis for equipment failure monitoring and preventive maintenance, thereby improving the accuracy of fault prediction and equipment reliability, reducing unexpected downtime events, reducing maintenance costs, and ultimately enhancing the stability and security of the power grid. It can also be updated according to changes in real-time signals and indicators to reflect the latest status of the equipment.

[0026] Preferably, the artificial neural network is trained based on the fault-related data and the quantifiable oil and gas data distribution law description index to establish an abnormal operating condition analysis model for UHV equipment, specifically including:

[0027] Divide the normalized fault-related data into a training set and a test set;

[0028] Based on the training set, quantifiable indicators describing the distribution of oil and gas data are used as characteristic variables of the input layer of the back-propagation artificial neural network. The weights and biases of the back-propagation artificial neural network are optimized by genetic algorithm. The determination coefficient, mean absolute error, mean absolute percentage error, mean square error and root mean square error of the back-propagation artificial neural network are evaluated based on the test set, and the abnormal operating condition analysis model of UHV equipment is obtained.

[0029]

[0030]

[0031] Where N is the number of real training data points. is the predicted value, i.e. the failure probability or state prediction output by the neural network, is the average value, representing the average value of all true values; i is the true value, that is, the actually observed fault data or equipment status.

[0032] MAE and MSE measure the absolute size of the deviation between the true value and the predicted value, while MAPE measures the relative size of the deviation. Relatively speaking, MAE and MAPE are not easily affected by extreme values, while MSE and RMSE calculate the square of the error, are more sensitive to outliers, and can highlight the error values ​​with greater impact. Generally speaking, a satisfactory model should have lower MAE, MAPE, MSE values, and higher RMSE and R values. 2 Should be close to 1.

[0033] Preferably, the quantifiable oil and gas data distribution law description index based on the training set is used as the characteristic variable of the input layer of the back-propagation artificial neural network, the weight and bias of the back-propagation artificial neural network are optimized by a genetic algorithm, and the determination coefficient, mean absolute error, mean absolute percentage error, mean square error, and root mean square error of the back-propagation artificial neural network are evaluated based on the test set to obtain an abnormal operating condition analysis model for UHV equipment, which specifically includes:

[0034] Set the structure of the artificial neural network, including the number of layers and neurons in each layer, network weights and biases, and initially encode the weights and biases as individuals of the genetic algorithm;

[0035] The back propagation artificial neural network is trained through the training set, and the quantifiable oil and gas data distribution law description index is used as the characteristic variable of the input layer of the back propagation artificial neural network, and the error of the training result is used as the fitness evaluation standard of the individual in the genetic algorithm;

[0036] According to the error, individuals with better performance are selected for reproduction, the selected individuals are paired and recombined to produce new offspring, and the weights and biases of the offspring are mutated to introduce new genetic diversity;

[0037] Calculate the error between the output of the back-propagation artificial neural network and the expected output. If the error meets the set termination condition (such as reaching a certain accuracy or the number of training iterations reaches the upper limit), stop the optimization process, output the final weights and biases, and evaluate the coefficient of determination, mean absolute error, mean absolute percentage error, mean square error, and root mean square error of the back-propagation artificial neural network through the test set;

[0038] If the set termination condition is not met, the selection operation, crossover operation, and mutation operation will continue to be performed under the set optimal weight threshold until the set termination condition is met;

[0039] The final weights and biases are substituted into the structure of the back-propagation artificial neural network to obtain the abnormal operating condition analysis model of UHV equipment.

[0040] The use of genetic algorithms to optimize the weights and biases of artificial neural networks and to comprehensively evaluate multiple error indicators on the test set is to achieve global optimization, avoid local minima, enhance the generalization ability of the model, comprehensively evaluate model performance, improve prediction accuracy, and enhance the adaptability and robustness of the system, thereby ensuring the efficient, accurate and stable operation of the UHV equipment fault monitoring system. The advantage of this is that it can improve the prediction accuracy and robustness of the fault monitoring system, while enhancing the generalization ability of the model to ensure stable operation under different conditions.

[0041] Preferably, the collecting of concentration signals and ratio signals of multiple gas components of the oil chromatogram of the UHV equipment, inputting the concentration signals and ratio signals into an abnormal operating condition analysis model of the UHV equipment, determining whether the concentration signals and ratio signals are abnormal and reporting the abnormal signals specifically includes:

[0042] Collect the oil chromatogram H of the UHV equipment 2 ,CO,CO 2 , CH 4 , C 2 H 6 , C 2 H 4 , C 2 H 2 The concentration signal and ratio signal of the total hydrocarbon are evaluated. If they meet the set integrity, effectiveness and stability ranges, they are valid concentration signals and ratio signals;

[0043] The effective concentration signal and proportion signal are input into the abnormal operating condition analysis model of the UHV equipment to determine whether the concentration signal and proportion signal are abnormal and report the abnormal signal.

[0044] By evaluating the integrity, validity and stability of the key gas component signals in the oil chromatogram of UHV equipment and inputting the concentration signal and the proportion signal into the abnormal operating condition analysis model, this method can achieve early warning of faults, improve the accuracy of fault diagnosis, reduce maintenance costs, enhance the reliability and safety of equipment, and realize the automation and intelligence of fault monitoring.

[0045] The present invention also provides a fault monitoring system for UHV equipment based on machine learning, comprising:

[0046] A database establishment module is used to form a comprehensive diagnosis database of the operation status of UHV equipment based on the fault-related data during the historical operation of UHV equipment;

[0047] An index generation module is used to establish quantifiable oil and gas data distribution law description indicators based on the fault feature data in the comprehensive diagnosis database of the UHV equipment operation status;

[0048] A model building module, used to train an artificial neural network based on the fault-related data and quantifiable oil and gas data distribution law description indicators to establish an abnormal operating condition analysis model for UHV equipment;

[0049] The fault detection module is used to collect the concentration signals and proportion signals of multiple gas components of the oil chromatogram of the ultra-high voltage equipment, input the concentration signals and proportion signals into the abnormal operating condition analysis model of the ultra-high voltage equipment, and determine whether the concentration signals and proportion signals are abnormal.

[0050] The present invention also provides a computer storage medium, wherein the computer-readable storage medium stores a computer program, and is characterized in that when the computer program is executed by a processor, the fault monitoring method for ultra-high voltage equipment based on machine learning as described above is implemented.

[0051] The present invention also provides an electronic device, comprising a memory and a processor: the memory is used to store computer-executable instructions, the processor is used to execute the computer-executable instructions, and when the computer-executable instructions are executed by the processor, the fault monitoring method for ultra-high voltage equipment based on machine learning as described above is implemented.

[0052] The beneficial effects of the present invention are:

[0053] Improve monitoring efficiency: The automated fault monitoring system reduces manual intervention and improves monitoring efficiency and response speed;

[0054] Enhanced prediction accuracy: Machine learning-based methods can learn patterns from large amounts of data, improving the accuracy of fault prediction;

[0055] Reduce maintenance costs: By detecting and preventing failures early, repair costs and potential equipment damage risks are reduced;

[0056] Improve equipment reliability: Real-time monitoring and early warning mechanisms help maintain stable operation of equipment and improve overall reliability;

[0057] Data-driven decision support: Provides data-based decision support to make maintenance and operation decisions more scientific and reasonable.

[0058] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention and implement it according to the contents of the specification, the following is a detailed description of the preferred embodiments of the present invention in conjunction with the accompanying drawings. The specific implementation of the present invention is given in detail by the following embodiments and their accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0060] Figure 1 A flow chart of a method for fault monitoring of UHV equipment based on machine learning provided in Example 1;

[0061] Figure 2 This is a flow chart of optimizing the back propagation artificial neural network by genetic algorithm in Example 1;

[0062] Figure 3 A module diagram of a fault monitoring system for UHV equipment based on machine learning provided in Example 2. DETAILED DESCRIPTION

[0063] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0064] like Figure 1 As shown, a fault monitoring method for UHV equipment based on machine learning includes the following steps:

[0065] 1. Based on the fault-related data of the UHV equipment during its historical operation, a comprehensive diagnosis database of the UHV equipment operation status is formed, as follows:

[0066] The fault-related data of UHV equipment during historical operation are collected and normalized. The formula is as follows:

[0067]

[0068] Where X is the actual value in the fault-related data, x max and x min are the maximum and minimum values ​​in the fault-related data, respectively, max and min is the normalized range, y is the normalized value of x;

[0069] The back propagation artificial neural network is trained based on the normalized fault-related data to form a comprehensive diagnosis database for the operation status of UHV equipment. The formula is as follows:

[0070]

[0071] In the formula, x i is the input value of the normalized fault-related data, ω ijis the connection weight between neurons i and j, b is a bias term in the form of a threshold, s is the weighted signal received by each neuron in the hidden layer from all input nodes, N represents the number of input signal sources received by each neuron, that is, the total number of input layer nodes, y is the output value of the sigmoid activation function, and e represents the base of the natural logarithm, also known as Euler's number, whose approximate value is 2.71828.

[0072] Among them, x i is the input value of the normalized fault-related data, ω ij is the connection weight between neurons i and j, which can represent the influence of a specific input feature (such as a certain gas concentration) on the output of the hidden layer neurons. b is a bias term in the form of a threshold, which represents the minimum signal strength required to trigger the neuron to identify a specific fault mode. s is the weighted signal received by each neuron in the hidden layer from all input nodes, which represents the fault probability calculated based on all input features (such as different gas concentrations). N represents the number of input signal sources received by each neuron, that is, the total number of input layer nodes, which represents the total number of oil and gas data features used for analysis. y is the output value of the sigmoid activation function, which represents the probability of a fault. In fault monitoring, this can represent the possibility of a specific fault mode. e represents the base of the natural logarithm, also known as Euler's number, and its approximate value is 2.71828.

[0073] 2. Based on the fault feature data in the comprehensive diagnosis database of UHV equipment operation status, establish quantifiable oil and gas data distribution law description indicators, including:

[0074] Extract fault feature data from the comprehensive diagnostic database of UHV equipment operation status, including basic parameters, maintenance statistics, and experimental data, and extract key features related to oil and gas distribution patterns, including dissolved gas concentration and changes in the physical and chemical properties of oil;

[0075] Based on the extracted features, quantifiable indicators describing the distribution patterns of oil and gas data are constructed.

[0076] 3. Based on the fault-related data and the quantifiable oil and gas data distribution law description indicators, the artificial neural network is trained by genetic algorithm to establish an abnormal operating condition analysis model for UHV equipment, which specifically includes:

[0077] Divide the normalized fault-related data into a training set and a test set;

[0078] Set the structure of the artificial neural network, including the number of layers and neurons in each layer, network weights and biases, and initially encode the weights and biases as individuals of the genetic algorithm;

[0079] The back propagation artificial neural network is trained through the training set. The quantifiable description index of the oil and gas data distribution law is used as the characteristic variable of the input layer of the back propagation artificial neural network, and the error of the training result is used as the fitness evaluation standard of the individual in the genetic algorithm.

[0080] According to the error, individuals with better performance are selected for reproduction, the selected individuals are paired and recombined to produce new offspring, and the weights and biases of the offspring are mutated to introduce new genetic diversity;

[0081] Calculate the error between the output of the back-propagation artificial neural network and the expected output. If the error meets the set termination condition, stop the optimization process, output the final weights and biases, and evaluate the coefficient of determination, mean absolute error, mean absolute percentage error, mean square error, and root mean square error of the back-propagation artificial neural network through the test set;

[0082] If the set termination condition is not met, the selection operation, crossover operation, and mutation operation will continue to be performed under the set optimal weight threshold until the set termination condition is met;

[0083] The final weights and biases are substituted into the structure of the back-propagation artificial neural network to obtain the abnormal operating condition analysis model of UHV equipment.

[0084]

[0085] In the formula, N is the number of real training data points, is the predicted value, i.e. the failure probability or state prediction output by the neural network, is the average value, representing the average value of all true values; i is the true value, that is, the actually observed fault data or equipment status.

[0086] MAE and MSE measure the absolute size of the deviation between the true value and the predicted value, while MAPE measures the relative size of the deviation. Relatively speaking, MAE and MAPE are not easily affected by extreme values, while MSE and RMSE calculate the square of the error, are more sensitive to outliers, and can highlight the error values ​​with greater impact. Generally speaking, a satisfactory model should have lower MAE, MAPE, MSE values, and higher RMSE and R values. 2 Should be close to 1.

[0087] 4. Collect the concentration signals and proportion signals of multiple gas components of the oil chromatogram of the UHV equipment, input the concentration signals and proportion signals into the abnormal operating condition analysis model of the UHV equipment, determine whether the concentration signals and proportion signals are abnormal and report the abnormal signals, specifically including:

[0088] Collect the oil chromatogram H of the UHV equipment2 ,CO,CO 2 , CH 4 , C 2 H 6 , C 2 H 4 , C 2 H 2 The concentration signal and ratio signal of the total hydrocarbon are evaluated. If they meet the set integrity, effectiveness and stability ranges, they are valid concentration signals and ratio signals;

[0089] The concentration signals and proportion signals of various gas components in the oil chromatogram of UHV equipment represent the concentration of the gas in the oil chromatogram and the proportion of the gas in the total hydrocarbons. The following information can be obtained from the concentration signals and proportion signals of the gas components: fault type (the presence and concentration of a specific gas can indicate a specific fault type, such as arcing, partial discharge or overheating); fault severity (the level of gas concentration can reflect the severity of the fault); fault location (the distribution of certain gases may help locate the specific location of the fault); fault development speed (the rate of change of gas concentration over time can indicate the speed of fault development); equipment health (the overall pattern of gas components can reflect the overall health of the equipment).

[0090] In this embodiment, the rules for evaluating the integrity, validity and stability of the signal are as follows:

[0091] Device data integrity analysis: For a single device, the unit is day. According to the UHV station operation and maintenance management requirements, the standard measurement cycle of the oil and gas online monitoring device is 4 hours, that is, the daily measurement data baseline value is N = 6, and each measurement component H 2 ,CO,CO 2 , CH 4 , C 2 H 6 , C 2 H 4 , C 2 H 2 And 8 other components including total hydrocarbons, the benchmark value measured on that day was 48.

[0092] When N=6, the number of data that violates the constraint rules on that day (i.e., the number of null values) is counted as n (n≤48), and the data integrity calculation result is S=1-(n / 48)×100%;

[0093] When the device measurement data N is less than 6, the benchmark value of the measurement value on that day is still 48, and the number of data that violates the constraint rules on that day (i.e., the number of null values) is counted as n (n≤48), then the data integrity calculation result is S=1-(n / 48)×100%;

[0094] When the device measurement data N>6, the benchmark value of the measured values ​​on that day is 8*N, and the number of data that violates the constraint rules on that day (i.e. the number of null values) is n, then the data integrity calculation result is S=1-(n / 8N)×100%.

[0095] Validity analysis: For abnormal data such as duplication, null values, garbled characters, and over-limit data in the UHV substation platform, mark whether each piece of data is valid.

[0096] Stability analysis: A single device is analyzed on a daily basis. Based on the real-time data of the operating monitoring device, the difference and ratio between each piece of data and its median are calculated based on the median comparison method, with duplicate data and non-numeric values ​​eliminated, to determine its stability.

[0097] The concentration signal and the ratio signal are input into an abnormal operating condition analysis model of the UHV equipment to determine whether the concentration signal and the ratio signal are abnormal and report the abnormal signal.

[0098] Example 2

[0099] The present invention also provides a fault monitoring system for UHV equipment based on machine learning, comprising:

[0100] A database establishment module is used to form a comprehensive diagnosis database of the operation status of UHV equipment based on the fault-related data during the historical operation of UHV equipment;

[0101] An index generation module is used to establish quantifiable oil and gas data distribution law description indicators based on the fault feature data in the comprehensive diagnosis database of the UHV equipment operation status;

[0102] A model building module, used to train an artificial neural network based on the fault-related data and quantifiable oil and gas data distribution law description indicators to establish an abnormal operating condition analysis model for UHV equipment;

[0103] The fault detection module is used to collect the concentration signals and proportion signals of multiple gas components of the oil chromatogram of the ultra-high voltage equipment, input the concentration signals and proportion signals into the abnormal operating condition analysis model of the ultra-high voltage equipment, and determine whether the concentration signals and proportion signals are abnormal.

[0104] Example 3

[0105] The present invention also provides a computer storage medium, wherein the computer-readable storage medium stores a computer program, and is characterized in that when the computer program is executed by a processor, it implements the fault monitoring method for ultra-high voltage equipment based on machine learning as described in Example 1.

[0106] Example 4

[0107] The present invention also provides an electronic device, comprising a memory and a processor: the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the fault monitoring method for ultra-high voltage equipment based on machine learning as described in Example 1 is implemented.

[0108] Those skilled in the art will appreciate that the embodiments of the present disclosure may be provided as methods, systems or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0109] The present disclosure is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram and the combination of processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0110] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0111] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0112] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any ordinary technician in the industry can smoothly implement the present invention as shown in the drawings and described above. However, any equivalent changes, modifications and evolutions made by technicians familiar with the profession without departing from the scope of the technical solution of the present invention using the technical content disclosed above are all equivalent embodiments of the present invention. At the same time, any equivalent changes, modifications and evolutions made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the technical solution of the present invention.

Claims

1. A fault monitoring method for UHV equipment based on machine learning, characterized in that: include: Based on the fault-related data of the UHV equipment during its historical operation, a comprehensive diagnosis database for the operation status of UHV equipment is formed; Based on the fault feature data in the comprehensive diagnosis database of UHV equipment operation status, a quantifiable description index of oil and gas data distribution law is established; Based on the fault-related data and quantifiable oil and gas data distribution law description indicators, an artificial neural network is trained to establish an abnormal operating condition analysis model for UHV equipment; The concentration signals and proportion signals of multiple gas components of the oil chromatogram of the UHV equipment are collected, and the gas concentration signals and proportion signals are input into the abnormal operating condition analysis model of the UHV equipment to determine whether the concentration signals and proportion signals are abnormal.

2. According to claim 1, a method for fault monitoring of UHV equipment based on machine learning is characterized in that: The comprehensive diagnosis database of the operation status of the UHV equipment is formed based on the fault-related data in the historical operation process of the UHV equipment, including: The fault-related data of UHV equipment during historical operation are collected and normalized. The formula is as follows: Where X is the actual value in the fault-related data, x max and x min are the maximum and minimum values ​​in the fault-related data, respectively, max and min is the normalized range, y is the normalized value of x; The back propagation artificial neural network is trained based on the normalized fault-related data to form a comprehensive diagnosis database for the operation status of UHV equipment. The formula is as follows: In the formula, x i is the input value of the normalized fault-related data, ω ij is the set input feature, b is the minimum signal strength required to trigger the neuron to recognize a specific fault mode, s is the fault probability calculated by all input features, N represents the total number of oil and gas data features used for analysis, that is, the total number of input layer nodes, y is the output value of the sigmoid activation function, indicating the probability of fault occurrence, and e represents the base of the natural logarithm.

3. The method for fault monitoring of UHV equipment based on machine learning according to claim 1, characterized in that: The fault-related data during the historical operation of the UHV equipment include basic parameters of the UHV equipment, maintenance statistics, insulating oil test data, main body offline oil chromatography, bushing, gas oil chromatography test data, insulation resistance measurement data, capacitance and dielectric loss measurement data, bushing test data, core clamp grounding current data, DC leakage current data, infrared test, vibration, sound level, pulse current method partial discharge, UHF, high frequency and ultrasonic data.

4. The method for fault monitoring of UHV equipment based on machine learning according to claim 1, characterized in that: The fault feature data in the comprehensive diagnosis database of the UHV equipment operation status are used to establish a quantifiable description index of the oil and gas data distribution law, which specifically includes: Extract fault feature data from the comprehensive diagnostic database of UHV equipment operation status, including basic parameters, maintenance statistics, and experimental data, and extract key features related to oil and gas distribution patterns, including dissolved gas concentration and changes in the physical and chemical properties of oil; Based on the extracted fault features, quantifiable indicators describing the distribution patterns of oil and gas data are constructed.

5. The method for fault monitoring of UHV equipment based on machine learning according to claim 2, characterized in that: The artificial neural network is trained based on the fault-related data and the quantifiable oil and gas data distribution law description index to establish an abnormal operating condition analysis model for UHV equipment, specifically including: Divide the normalized fault-related data into a training set and a test set; Based on the training set, quantifiable indicators describing the distribution of oil and gas data are used as characteristic variables of the input layer of the back-propagation artificial neural network. The weights and biases of the back-propagation artificial neural network are optimized by genetic algorithm. The determination coefficient, mean absolute error, mean absolute percentage error, mean square error and root mean square error of the back-propagation artificial neural network are evaluated based on the test set, and the abnormal operating condition analysis model of UHV equipment is obtained.

6. The method for fault monitoring of UHV equipment based on machine learning according to claim 4, characterized in that: The quantifiable oil and gas data distribution law description index based on the training set is used as the characteristic variable of the input layer of the back propagation artificial neural network. The weight and bias of the back propagation artificial neural network are optimized by genetic algorithm. The determination coefficient, mean absolute error, mean absolute percentage error, mean square error, and root mean square error of the back propagation artificial neural network are evaluated based on the test set to obtain the abnormal operating condition analysis model of the UHV equipment, which specifically includes: Set the structure of the back-propagation artificial neural network, including the number of layers and neurons in each layer, network weights and biases, and initially encode the weights and biases as individuals of the genetic algorithm; The back propagation artificial neural network is trained through the training set, and the quantifiable oil and gas data distribution law description index is used as the characteristic variable of the input layer of the back propagation artificial neural network, and the error of the training result is used as the fitness evaluation standard of the individual in the genetic algorithm; According to the error, individuals with better performance are selected for reproduction, the selected individuals are paired and recombined to produce new offspring, and the weights and biases of the offspring are mutated to introduce new genetic diversity; Calculate the error between the output of the back-propagation artificial neural network and the expected output. If the error meets the set termination condition, stop the optimization process, output the final weights and biases, and evaluate the coefficient of determination, mean absolute error, mean absolute percentage error, mean square error, and root mean square error of the back-propagation artificial neural network through the test set; If the set termination condition is not met, the selection operation, crossover operation, and mutation operation will continue to be performed under the set optimal weight threshold until the set termination condition is met; The final weights and biases are substituted into the structure of the back-propagation artificial neural network to obtain the abnormal operating condition analysis model of UHV equipment.

7. The method for fault monitoring of UHV equipment based on machine learning according to claim 1, characterized in that: The collecting of concentration signals and ratio signals of multiple gas components of the oil chromatogram of the UHV equipment, inputting the concentration signals and ratio signals into the abnormal operating condition analysis model of the UHV equipment, determining whether the concentration signals and ratio signals are abnormal and reporting the abnormal signals specifically includes: The concentration signals and proportion signals of H2, CO, CO2, CH4, C2H6, C2H4, C2H2 and total hydrocarbons in the oil chromatogram of the UHV equipment are collected for evaluation. If the collected concentration signals and proportion signals meet the set integrity, validity and stability ranges, they are valid concentration signals and proportion signals; The effective concentration signal and the proportion signal are input into the abnormal operating condition analysis model of the UHV equipment to determine whether the effective concentration signal and the proportion signal are abnormal and report the abnormal signal.

8. A fault monitoring system for UHV equipment based on machine learning, characterized in that: include: A database establishment module is used to form a comprehensive diagnosis database of the operation status of UHV equipment based on the fault-related data during the historical operation of UHV equipment; An index generation module is used to establish quantifiable oil and gas data distribution law description indicators based on the fault feature data in the comprehensive diagnosis database of the UHV equipment operation status; A model building module, used to train an artificial neural network based on the fault-related data and quantifiable oil and gas data distribution law description indicators to establish an abnormal operating condition analysis model for UHV equipment; The fault detection module is used to collect the concentration signals and proportion signals of multiple gas components of the oil chromatogram of the ultra-high voltage equipment, input the concentration signals and proportion signals into the abnormal operating condition analysis model of the ultra-high voltage equipment, determine whether the concentration signals and proportion signals are abnormal and report the abnormal signals.

9. A computer storage medium, wherein the computer readable storage medium stores a computer program, characterized in that: When the computer program is executed by a processor, the fault monitoring method for UHV equipment based on machine learning as described in any one of claims 1 to 7 is implemented.

10. An electronic device, characterized in that: It includes a memory and a processor: the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the fault monitoring method for ultra-high voltage equipment based on machine learning as described in any one of claims 1 to 7 is implemented.

Images