A fault assessment method and system for power equipment based on AI large model

Abstract

The present invention discloses a fault assessment method and system for electric power equipment based on an AI large model, and relates to the technical field of fault assessment for electric power equipment. The fault assessment method for electric power equipment based on an AI large model comprises the following steps: collecting and processing electronic transformer fault data, and cleaning, denoising and extracting features of the electronic transformer fault data; selecting a recurrent neural network model for sequence data processing, and using a labeled fault data set to train the recurrent neural network model for identifying the working state of the transformer; and alternately using labeled training sets and validation sets to evaluate the generalization ability of the recurrent neural network model. The present invention achieves the effect of improving the reliability of electronic transformer fault assessment through a fault assessment method and system for electric power equipment based on an AI large model, and solves the problem of insufficient reliability of electronic transformer fault assessment in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of fault assessment of electric power equipment, and in particular to a fault assessment method and system for electric power equipment based on an AI large model. Background Art

[0002] The background technology of power equipment failure mainly involves the basic principles of the power system and the characteristics and operation mechanism of the power equipment. The electronic transformer is an important equipment in the power system, mainly used for the measurement of voltage and current, as well as the protection and control of the power system. The fault assessment of electronic transformers is a comprehensive technical field, involving the fields of power systems, electronic technology, communication technology, computer technology, etc. These technologies need to be comprehensively applied to achieve effective assessment and processing of transformer faults. With the development of technology, electronic transformers have been significantly improved in terms of accuracy, stability, and anti-interference ability, but there are still some technical problems.

[0003] The existing fault assessment method and system for power equipment based on AI large models collects electronic transformer data, cleans and removes noise from the collected electronic transformer data, determines the recurrent neural network model, and uses the cleaned data to train the recurrent neural network model. During the training period, the recurrent neural network model continuously adjusts its parameters to improve the accuracy of the prediction. After the recurrent neural network model is trained, it needs to be evaluated and debugged to determine the performance of the recurrent neural network model. The debugging process will also adjust the hyperparameters of the recurrent neural network model to improve the performance of the recurrent neural network model.

[0004] For example, the patent application with publication number: CN116227788A discloses a method and system for fault diagnosis and evaluation of power equipment in a substation, including: collecting status data of power equipment in the substation, judging based on the fault coefficient obtained by analyzing the collected status data, and then performing fault diagnosis and evaluation on the power equipment; performing secondary verification on the fault diagnosis and evaluation result of step one, one verification method is to analyze from the perspective of actual operation probability value, and the other verification method is to analyze from the perspective of actual score value; when analyzing from the perspective of actual operation probability value, analyzing the infrared characteristic image corresponding to the abnormal heating area, reducing the impact of the abnormal heating area on the surrounding area, thereby improving the accuracy of the data and the authenticity of the verification result; judging whether the verification results of the two methods are normal by verifying the interactive operation, and providing information feedback on the fault diagnosis and evaluation result of the power equipment in step one by means of feedback signal.

[0005] For example, a method for online diagnosis of voltage transformer faults disclosed in a patent application with publication number CN116243230A includes: in a healthy state, connecting a standard voltage transformer, and using the least squares method to obtain an input-output mathematical model of the voltage transformer under test; in a healthy state, obtaining a threshold value of a voltage transformer deviation index; in an operating state, connecting a standard voltage transformer, using the output of the standard voltage transformer as an input matrix of the input-output mathematical model of the voltage transformer under test, substituting the output of the standard voltage transformer into the mathematical model to solve the theoretical output value of the voltage transformer under test; in an operating state, collecting the actual output value of the voltage transformer under test; comparing the theoretical output value and the actual output value of the voltage transformer under test, and calculating the output deviation index of the voltage transformer; comparing the voltage transformer deviation index with the threshold value, if the deviation index is less than the threshold value, it is judged that the voltage transformer is in a normal state; if the deviation index is greater than the threshold value, it is judged that the voltage transformer is in a fault state.

[0006] However, in the process of implementing the technical solution of the invention in the embodiments of the present application, the present application found that the above technology has at least the following technical problems:

[0007] In the prior art, when an electronic transformer fails, the protection device cannot correctly measure the fault current or voltage, which results in the failure of the protection action and the inability to cut off the fault circuit in time, affecting the safe and stable operation of the system. Although electronic transformers are usually designed with high reliability in mind, in actual operation, there is a problem of insufficient reliability in fault assessment of electronic transformers. Summary of the invention

[0008] The embodiments of the present application solve the problem of insufficient reliability of electronic transformer fault assessment in the prior art by providing a fault assessment method and system for power equipment based on an AI large model, thereby achieving the effect of improving the reliability of electronic transformer fault assessment.

[0009] The present application provides a method for fault assessment of electric power equipment based on an AI large model, comprising the following steps:

[0010] Collect and process electronic transformer fault data, clean, remove noise and extract features from the collected electronic sensor fault data to obtain a marked fault data set;

[0011] By selecting a recurrent neural network model for sequence data processing, the recurrent neural network model is trained using a marked fault data set, the working state of the transformer is identified through the recurrent neural network model, and the marked fault data monitored in real time is predicted and processed;

[0012] The labeled fault data set is divided into a training set and a validation set, and the labeled training set and validation set are used alternately to evaluate the generalization ability of the recurrent neural network model. The generalization ability of the recurrent neural network model is evaluated, and the overfitting or underfitting problems are identified through the generalization ability of the recurrent neural network model. Fault warning of the electronic transformer is performed based on the evaluation results of the recurrent neural network model.

[0013] Furthermore, the specific collection process for collecting and processing electronic transformer fault data is: collecting electronic transformer fault data through sensors, the electronic transformer fault data including electronic transformer temperature fault data, electronic transformer humidity fault data and electronic transformer noise signal fault data, and removing invalid data, filling missing values ​​and processing abnormal values ​​from the electronic transformer fault data.

[0014] Furthermore, the operating data of the electronic transformer is arranged in a time series to obtain sequence data, and the sequence data is converted into a format input into a recurrent neural network model, and the converted sequence data is marked and indicated at the beginning and end of the time series.

[0015] Furthermore, a fault prediction model of the electronic transformer is established based on the marked fault data monitored in real time and the input data of the electronic transformer, and the fault prediction model of the electronic transformer is used to predict the fault of the electronic transformer to obtain the fault data of the electronic transformer.

[0016] Furthermore, the labeled fault data set is divided into different training sets and validation sets, and the performance of the recurrent neural network model is evaluated using different training sets and validation sets. By comparing the performance of the recurrent neural network model under different combinations of training sets and validation sets, one or more training sets and validation sets that are not used in the training process are selected to evaluate the generalization ability of the recurrent neural network model.

[0017] Furthermore, the sequence data is scaled within a range, the elements in the sequence are converted into one-hot encoding, the shorter sequence data are padded, the sequence data is divided into segments of the same length, features are extracted according to the characteristics of the sequence, the converted sequence data is obtained, and the recurrent neural network model is trained in batches using the converted sequence data.

[0018] Furthermore, a specific evaluation process of using the fault prediction model of the electronic transformer to predict the fault of the electronic transformer to obtain the fault evaluation data of the electronic transformer is as follows: the temperature fault data of the electronic transformer is evaluated to obtain the temperature fault evaluation coefficient of the electronic transformer, the humidity fault data of the electronic transformer is evaluated to obtain the humidity fault evaluation coefficient of the electronic transformer, the noise signal fault data of the electronic transformer is evaluated to obtain the noise signal fault evaluation coefficient of the electronic transformer, and the temperature fault coefficient of the electronic transformer, the humidity fault coefficient of the electronic transformer and the noise signal fault coefficient of the electronic transformer are evaluated to obtain the fault evaluation coefficient of the electronic transformer.

[0019] Furthermore, a specific early warning process for fault warning of the electronic transformer is provided according to the evaluation result of the recurrent neural network model: a fault assessment is performed on the electronic transformer in the recurrent neural network model, and the fault assessment is performed on three aspects of temperature, humidity and noise signal to obtain a fault assessment coefficient of the electronic transformer, and an allowable threshold is set for the fault assessment coefficient of the electronic transformer. If the fault assessment coefficient of the electronic transformer falls outside the allowable threshold, the electronic transformer will issue an early warning.

[0020] Furthermore, the specific analysis formula of the fault assessment coefficient of the electronic transformer is:

[0021]

[0022] In the formula, The fault evaluation coefficient of the electronic transformer is used to evaluate the comprehensive impact of the temperature, humidity and noise signal of the electronic transformer on the fault of the electronic transformer. Indicates the temperature failure coefficient of the electronic transformer, Indicates the humidity failure coefficient of the electronic transformer, Indicates the noise signal fault coefficient of the electronic transformer.

[0023] The embodiment of the present application provides a fault assessment system for electric power equipment based on an AI large model, comprising:

[0024] Data collection and processing module: used to collect and process electronic transformer fault data, clean, remove noise and extract features from the collected electronic sensor fault data to obtain a marked fault data set;

[0025] Model selection and training module: used to process sequence data by selecting a recurrent neural network model, train the recurrent neural network model using a marked fault data set, identify the working state of the transformer through the recurrent neural network model, and perform predictive processing on the marked fault data monitored in real time;

[0026] Model evaluation and early warning module: used to divide the labeled fault data set into multiple parts, and alternately use the labeled training set and validation set to evaluate the generalization ability of the recurrent neural network model. The generalization ability of the recurrent neural network model is evaluated to identify overfitting or underfitting problems.

[0027] An embodiment of the present application provides a computer-readable storage medium for storing a program, which, when executed by a processor, implements a fault assessment method for power equipment based on an AI large model.

[0028] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0029] 1. The electronic transformer fault data is collected and processed through the data collection and processing module, and the collected electronic transformer fault data is cleaned, denoised and features are extracted to obtain marked fault data, thereby improving the accuracy of the electronic transformer fault data and effectively solving the problem of insufficient accuracy of the electronic transformer fault data in the prior art.

[0030] 2. Through the model selection and training module, considering the complexity of the recurrent neural network model architecture, the number of parameters, the training time and the prediction speed, the recurrent neural network model is selected for sequence data processing, so as to predict and process the marked fault data monitored in real time, thereby filling the gap of the AI ​​large model in the fault assessment of electronic transformers.

[0031] 3. The marked fault data set is divided into multiple parts through the model evaluation and early warning module, and multiple parts of the marked fault data set are alternately used as training sets and validation sets to evaluate the generalization ability of the recurrent neural network model, thereby identifying the overfitting or underfitting of the recurrent neural network model, thereby achieving the effect of improving the reliability of electronic transformer fault assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A flow chart of a fault assessment method for power equipment based on an AI large model provided in an embodiment of the present application;

[0033] Figure 2 A schematic diagram of the structure of a fault assessment system for power equipment based on an AI large model provided in an embodiment of the present application; DETAILED DESCRIPTION

[0034] The embodiment of the present application solves the problem of insufficient reliability of electronic transformer fault assessment in the prior art by providing a fault assessment method and system for power equipment based on an AI large model. The electronic transformer fault data is collected and processed by a data collection and processing module. The model selection and training module selects a recurrent neural network model and uses a labeled fault data set to train the recurrent neural network model. The generalization ability of the recurrent neural network model is evaluated by a model evaluation and early warning module to identify overfitting or underfitting problems, thereby achieving the effect of improving the reliability of electronic transformer fault assessment.

[0035] The technical solution in the embodiment of the present application is to solve the above-mentioned problem of insufficient reliability of electronic transformer fault assessment. The overall idea is as follows:

[0036] By collecting and processing electronic transformer fault data, the collected electronic sensor fault data is cleaned, denoised and features are extracted to obtain a labeled fault data set. A recurrent neural network model is selected for sequence data processing, and the labeled fault data set is used to train the recurrent neural network model. The labeled training set and validation set are used alternately to evaluate the generalization ability of the recurrent neural network model, thereby achieving the effect of improving the reliability of electronic transformer fault assessment.

[0037] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0038] like Figure 1 As shown, it is a flow chart of a fault assessment method for electric power equipment based on an AI large model provided in an embodiment of the present application. The method is applied to a fault assessment system for electric power equipment based on an AI large model. The method includes the following steps:

[0039] Collect and process electronic transformer fault data, clean, remove noise and extract features from the collected electronic sensor fault data to obtain a marked fault data set;

[0040] By selecting a recurrent neural network model for sequence data processing, the recurrent neural network model is trained using a marked fault data set, the working state of the transformer is identified through the recurrent neural network model, and the marked fault data monitored in real time is predicted and processed;

[0041] The labeled fault data set is divided into a training set and a validation set, and the labeled training set and validation set are used alternately to evaluate the generalization ability of the recurrent neural network model. The generalization ability of the recurrent neural network model is evaluated, and the overfitting or underfitting problems are identified through the generalization ability of the recurrent neural network model. Fault warning of the electronic transformer is performed based on the evaluation results of the recurrent neural network model.

[0042] Furthermore, electronic transformer fault data is collected through sensors, the electronic transformer fault data includes electronic transformer temperature fault data, electronic transformer humidity fault data and electronic transformer noise signal fault data, and invalid data is removed from the electronic transformer fault data, missing values ​​are filled and abnormal values ​​are processed.

[0043] In this embodiment, the temperature fault data of the electronic transformer includes the heat dissipation area of ​​the electronic transformer, the thermal expansion coefficient of the electronic transformer, the heat transfer resistance of the electronic transformer, the operating temperature of the electronic transformer, the ambient temperature of the electronic transformer and the thermal stress of the electronic transformer;

[0044] The humidity fault data of the electronic instrument transformer includes the insulation resistivity of the electronic instrument transformer, the resonant frequency of the dielectric constant of the electronic instrument transformer, the corrosion rate of the electronic instrument transformer, the capacitance value of the dielectric constant of the electronic instrument transformer and the total pollution value of the electronic instrument transformer;

[0045] The electronic transformer noise signal fault data includes the signal-to-noise ratio of the electronic transformer, the noise power spectrum of the electronic transformer, the noise density of the electronic transformer, the total harmonic distortion of the electronic transformer and the bit error rate of the electronic transformer.

[0046] Furthermore, the operating data of the electronic transformer is arranged in a time series to obtain sequence data, and the sequence data is converted into a format input into a recurrent neural network model, and the converted sequence data is marked and indicated at the beginning and end of the time series.

[0047] In this embodiment, the operation data of the electronic transformer is arranged in time series to obtain sequence data, the sequence data is converted into a format suitable for input to the recurrent neural network model, the time step sequence needs to be considered when processing the data, the sequence data is converted into an input format suitable for the recurrent neural network, the sequence is divided into batches of fixed length, and the sequence is truncated or padded;

[0048] To help the recurrent neural network better understand the starting and ending points of the sequence, markers are used to indicate the beginning and end of the sequence. In time series analysis, these markers need to be defined based on the actual situation of the sequence.

[0049] Designing the network architecture of a recurrent neural network includes determining the number of network layers, the number of neurons in each layer, and the activation function used. The choice of the number of network layers and neurons will affect the complexity and learning ability of the recurrent neural network model, while the activation function determines the nonlinear transformation ability of the network.

[0050] To train a recurrent neural network, set the appropriate learning rate, optimizer, and loss function. The learning rate controls the speed at which the network parameters are updated. The optimizer is responsible for adjusting the network parameters to minimize the loss function, which measures the difference between the predicted results of the recurrent neural network model and the actual results. When selecting these hyperparameters, the characteristics of the sequence data and the complexity of the task need to be considered to achieve the best recurrent neural network model performance.

[0051] Furthermore, a fault prediction model of the electronic transformer is established based on the marked fault data monitored in real time and the input data of the electronic transformer, and the fault prediction model of the electronic transformer is used to predict the fault of the electronic transformer to obtain the fault data of the electronic transformer.

[0052] In this embodiment, a recurrent neural network model is constructed to identify the normal and abnormal conditions of the input data of the electronic transformer. The recurrent neural network model can monitor the health status of the electronic transformer in real time, thereby issuing an early warning when the equipment shows signs of failure.

[0053] By deeply analyzing the input data characteristics of the electronic transformer, the recurrent neural network model will use historical fault data as training samples, learn the time series characteristics and patterns in the data through the recurrent neural network, and after data preprocessing and recurrent neural network model tuning, the recurrent neural network model can identify abnormal data patterns that may cause faults;

[0054] After the RNN model is established and verified, the electronic transformer is monitored in real time by continuously inputting the latest data and letting the RNN model analyze it. Once the RNN model identifies an abnormal pattern in the input data, it will generate an early warning to avoid potential equipment failures or reduce the impact of failures.

[0055] The recurrent neural network model continuously improves itself by constantly learning new data and fault cases, which enables the recurrent neural network model to not only provide reliable fault prediction in the initial stage, but also its prediction accuracy will continue to improve with the passage of time and the accumulation of data, thereby better detecting the health management and intelligent maintenance of electronic transformers.

[0056] Furthermore, the labeled fault data set is divided into different training sets and validation sets, and the performance of the recurrent neural network model is evaluated using different training sets and validation sets. By comparing the performance of the recurrent neural network model under different combinations of training sets and validation sets, one or more training sets and validation sets that are not used in the training process are selected to evaluate the generalization ability of the recurrent neural network model.

[0057] In this embodiment, the marked fault data set is divided into multiple non-overlapping independent parts. After the data set is divided, different training sets and validation sets can be used in turn to evaluate the generalization ability of the recurrent neural network model, and the generalization error of the recurrent neural network model is obtained. The upper bound of the generalization error is calculated by the empirical risk minimization function. The generalization error represents the performance of the recurrent neural network model on unknown data and is also one of the important indicators for evaluating the pros and cons of the recurrent neural network model. The training set and validation set that are not used in the training process are selected to comprehensively evaluate the generalization ability of the recurrent neural network model.

[0058] Furthermore, the sequence data is scaled within a range, the elements in the sequence are converted into one-hot encoding, the shorter sequence data are padded, the sequence data is divided into segments of the same length, features are extracted according to the characteristics of the sequence, the converted sequence data is obtained, and the recurrent neural network model is trained in batches using the converted sequence data.

[0059] In this embodiment, the sequence data is scaled to a uniform range to avoid the influence of values ​​in different ranges on the training of the recurrent neural network model, the scaled sequence data is divided into segments of appropriate lengths so that the recurrent neural network model can process the information in the sequence one by one, the elements in the sequence are converted into one-hot encoding reliability so that the machine can better understand and process the sequence data, and the shorter sequences are padded with specific padding characters or vectors so that all sequences have the same length.

[0060] Furthermore, a fault assessment of the electronic transformer is performed in the recurrent neural network model. The fault assessment is based on three aspects: temperature, humidity and noise signals. A fault assessment coefficient of the electronic transformer is obtained. An allowable threshold is set for the fault assessment coefficient of the electronic transformer. If the fault assessment coefficient of the electronic transformer falls outside the allowable threshold, the electronic transformer issues an early warning.

[0061] In this embodiment, fault assessment is performed on the electronic transformer in terms of temperature, humidity and noise signals in a recurrent neural network model to obtain a fault assessment coefficient of the electronic transformer. An allowable threshold value of the fault assessment coefficient is set according to the performance standard and safety requirements of the electronic transformer. When the fault assessment coefficient of the electronic transformer falls outside the allowable threshold value, the electronic transformer will immediately trigger an early warning mechanism and automatically adjust the operating parameters of the electronic transformer to ensure the normal operation of the equipment.

[0062] Furthermore, the temperature fault data of the electronic transformer is evaluated to obtain the temperature fault evaluation coefficient of the electronic transformer; the humidity fault data of the electronic transformer is evaluated to obtain the humidity fault evaluation coefficient of the electronic transformer; the noise signal fault data of the electronic transformer is evaluated to obtain the noise signal fault evaluation coefficient of the electronic transformer; the temperature fault coefficient of the electronic transformer, the humidity fault coefficient of the electronic transformer and the noise signal fault coefficient of the electronic transformer are comprehensively evaluated to obtain the fault evaluation coefficient of the electronic transformer.

[0063] Furthermore, the specific constraint formula of the temperature fault evaluation coefficient of the electronic transformer is obtained from this analysis:

[0064]

[0065] In the formula, It represents the temperature fault assessment coefficient of the electronic instrument transformer, which is used to evaluate the influence of the electronic instrument transformer temperature on the electronic instrument transformer fault. A represents the heat dissipation area of ​​the electronic instrument transformer, B represents the thermal expansion coefficient of the electronic instrument transformer, C represents the heat transfer resistance of the electronic instrument transformer, D represents the operating temperature of the electronic instrument transformer, E represents the ambient temperature of the electronic instrument transformer, and F represents the thermal stress of the electronic instrument transformer.

[0066] In this embodiment, the heat dissipation area of ​​the electronic transformer mainly refers to the area of ​​the heat sink that the internal heating element contacts with the outside. The heat dissipation area of ​​the electronic transformer is determined by thermal simulation and testing, and the heat dissipation area ranges from a few square centimeters to tens of square centimeters.

[0067] The thermal expansion coefficient of an electronic transformer refers to the change in its size when the temperature of the electronic transformer changes. The thermal expansion coefficient of an electronic transformer can be measured by the expansion method, the resistance strain gauge method, the thermocouple method and the thermal expansion instrument method. The thermal expansion coefficient of an electronic transformer will change with the metal material of the electronic transformer.

[0068] The heat transfer resistance of an electronic transformer refers to the degree of resistance to heat transfer caused by the internal materials and structures of the transformer when the temperature changes. The heat transfer resistance of an electronic transformer is measured by heat conduction testing and thermal performance simulation. The heat transfer resistance of an electronic transformer is usually 1-100℃ / W.

[0069] The working temperature and ambient temperature of the electronic transformer refer to the temperature of the electronic transformer when it is working and stationary. The working temperature and ambient temperature of the electronic transformer can be directly measured by the sensor. The working temperature and ambient temperature of the electronic transformer are usually 20-40℃;

[0070] The thermal stress of an electronic transformer refers to the stress caused by the different thermal expansion coefficients of the internal and external materials of the transformer under the action of temperature changes. The thermal stress of an electronic transformer can be determined by thermal stress analysis and metallographic analysis. The thermal stress of an electronic transformer is usually 1-10N.

[0071] The temperature fault assessment coefficient of the electronic transformer can also be analyzed biomimetically using professional power system simulation software to simulate the working state of the electronic transformer at different temperatures and estimate the temperature fault assessment coefficient through simulation results.

[0072] Furthermore, the specific constraint formula of the humidity fault assessment coefficient of the electronic transformer is obtained from this analysis:

[0073]

[0074] In the formula, several time monitoring points are set and numbered in sequence, a0=1, 2, ..., a, a0 represents the number of the time monitoring point, a represents the total number of the time monitoring point, It represents the humidity fault assessment coefficient of the electronic transformer, which is used to assess the influence of humidity on the fault of the electronic transformer. Indicates the insulation resistance of the electronic transformer at the a0th time monitoring point, represents the resonant frequency of the dielectric constant of the electronic transformer at the a0th time monitoring point, represents the corrosion rate of the electronic transformer at the a0th time monitoring point, J represents the capacitance value of the dielectric constant of the electronic transformer, and I represents the total pollution value of the electronic transformer.

[0075] In this embodiment, the insulation resistivity of the electronic transformer refers to the resistance characteristic of the transformer insulation material. The insulation resistivity of the electronic transformer is a physical quantity used to measure the ability of the material to prevent the passage of current. The insulation resistivity of the electronic transformer can be measured by a high-voltage insulation resistance tester and a dielectric loss factor test. The insulation resistivity of the electronic transformer is usually 100-1000Ω·m;

[0076] The resonant frequency of the dielectric constant of an electronic transformer refers to the dependence of the dielectric constant of the dielectric material in the transformer on the frequency under specific resonant conditions. The resonant frequency of the dielectric constant of an electronic transformer can be measured by a dielectric property tester and a resonance method. The resonant frequency of the dielectric constant of an electronic transformer is usually 0.01-1Hz.

[0077] The corrosion rate of electronic transformer refers to the speed of material loss caused by chemical corrosion under specific environmental conditions. The corrosion rate of electronic transformer can be determined by electrochemical corrosion test and weight loss method. The corrosion rate of electronic transformer is usually 0.01-0.1mm / y.

[0078] The capacitance value of the dielectric constant of an electronic transformer refers to the ratio between the charge stored in the dielectric material in the transformer and the electric field strength under a given electric field strength. The capacitance value of the dielectric constant of an electronic transformer can be determined by the parallel plate capacitor method and the open capacitance method. The capacitance value of the dielectric constant of an electronic transformer is usually 10-100 pico-farads.

[0079] The total pollution value of the electronic transformer refers to the deposition of external pollutants on the surface of the transformer or the insulating material. The total pollution value of the electronic transformer can be measured by the pollution layer thickness measurement and leakage current test. The total pollution value of the electronic transformer is usually 1000-10000.

[0080] The appropriate fault assessment coefficient of the electronic transformer can also be estimated by using professional simulation software to simulate the working state of the transformer under different humidity conditions, and the humidity fault assessment coefficient can be estimated by the simulation results.

[0081] Furthermore, the specific constraint formula of the noise signal fault evaluation coefficient of the electronic transformer is obtained from this analysis:

[0082]

[0083] In the formula, It represents the noise signal fault assessment coefficient of the electronic transformer, which is used to assess the influence of the electronic transformer noise signal on the electronic transformer fault. L represents the signal-to-noise ratio of the electronic transformer, Q represents the noise power spectrum of the electronic transformer, M represents the noise density of the electronic transformer, N represents the total harmonic distortion of the electronic transformer, and P represents the bit error rate of the electronic transformer.

[0084] In this embodiment, the signal-to-noise ratio of the electronic transformer refers to the signal-to-noise ratio of the electronic transformer. The signal-to-noise ratio of the electronic transformer is confirmed by the transformer performance index. The signal-to-noise ratio of the electronic transformer is usually 60-100 dB / Hz.

[0085] The noise power spectrum of an electronic transformer refers to the characteristic that the power distribution of the transformer output noise signal changes with frequency. The noise power spectrum of an electronic transformer can be measured by a spectrum analyzer and a signal oscilloscope. The noise power spectrum of an electronic transformer is usually between 1 and 10V. 2 / Hz;

[0086] The noise density of an electronic transformer refers to the noise power within a unit frequency range. The noise density of an electronic transformer can be measured through digital signal processing and a signal generator. The noise density of an electronic transformer is usually 1-10pA / Hz.

[0087] The total harmonic distortion of an electronic transformer refers to the degree of distortion between the harmonic component and the fundamental component in the transformer output signal. The total harmonic distortion of a sub-transformer can be determined by calculating the amplitude of the harmonic component and resonance analysis. The total harmonic distortion of an electronic transformer is usually 10-100dB.

[0088] The bit error rate of an electronic transformer refers to the proportion or frequency of errors that occur during digital communication or data transmission. The bit error rate of an electronic transformer can be determined through error detection and analysis. The bit error rate of an electronic transformer is usually 10-20%.

[0089] The noise signal fault evaluation coefficient of the electronic transformer can also be obtained through laboratory testing under controlled noise conditions. The performance data of the transformer under different noise levels can be obtained. From these data, the noise signal fault evaluation coefficient can be calculated.

[0090] Furthermore, the specific constraint formula of the fault assessment coefficient of the electronic transformer is obtained from this analysis:

[0091]

[0092] In the formula, The fault evaluation coefficient of the electronic transformer is used to evaluate the comprehensive impact of the temperature, humidity and noise signal of the electronic transformer on the fault of the electronic transformer. Indicates the temperature failure coefficient of the electronic transformer, Indicates the humidity failure coefficient of the electronic transformer, Indicates the noise signal fault coefficient of the electronic transformer.

[0093] In this embodiment, the temperature fault assessment coefficient of the electronic transformer is used to assess the influence of the electronic transformer temperature on the electronic transformer fault, the humidity fault assessment coefficient of the electronic transformer is used to assess the influence of the electronic transformer humidity on the electronic transformer fault, and the noise signal fault assessment coefficient of the electronic transformer is used to assess the influence of the electronic transformer noise signal on the electronic transformer fault.

[0094] The fault assessment coefficient of the electronic transformer can also be obtained through laboratory tests under controlled conditions, which can obtain the performance data of the transformer under different fault conditions. From these data, the fault assessment coefficient can be calculated.

[0095] like Figure 2As shown, it is a structural schematic diagram of a fault assessment system for electric power equipment based on an AI large model provided in an embodiment of the present application. The fault assessment system for electric power equipment based on an AI large model provided in an embodiment of the present application includes:

[0096] Data collection and processing module: used to collect and process electronic transformer fault data, clean, remove noise and extract features from the collected electronic sensor fault data to obtain a marked fault data set;

[0097] Model selection and training module: used to process sequence data by selecting a recurrent neural network model, train the recurrent neural network model using a marked fault data set, identify the working state of the transformer through the recurrent neural network model, and perform predictive processing on the marked fault data monitored in real time;

[0098] Model evaluation and early warning module: used to divide the labeled fault data set into multiple parts, and alternately use the labeled training set and validation set to evaluate the generalization ability of the recurrent neural network model. The generalization ability of the recurrent neural network model is evaluated to identify overfitting or underfitting problems.

[0099] The technical scheme in the above-mentioned embodiment of the present application has at least the following technical effects or advantages: relative to the method and system for fault diagnosis and evaluation of power equipment in a substation disclosed in the patent application with publication number: CN116227788A, the embodiment of the present application arranges the operation data of the electronic transformer in a time series to obtain sequence data, converts the sequence data into a format suitable for input into a recurrent neural network model, uses special markers to indicate the beginning and end of the sequence, determines the number of layers of the network, the number of neurons in each layer and the activation function, sets the learning rate, optimizer and loss function, thereby selecting a recurrent neural network model for sequence data processing, thereby achieving improved accuracy of electronic transformer fault data; relative to the patent application with publication number: CN116227788A, the embodiment of the present application arranges the operation data of the electronic transformer in a time series to obtain sequence data, converts the sequence data into a format suitable for input into a recurrent neural network model, uses special markers to indicate the beginning and end of the sequence, determines the number of layers of the network, the number of neurons in each layer and the activation function, sets the learning rate, optimizer and loss function, thereby selecting a recurrent neural network model for sequence data processing, and thereby achieving improved accuracy of electronic transformer fault data; relative to the patent application with publication number: CN116227788A, the embodiment of the present application arranges the operation data of the electronic transformer in a time series to obtain sequence data, converts the sequence data into a format suitable for input into a recurrent neural network model, uses special markers to indicate the beginning and end of the sequence, determines the number of layers of the network, the number of neurons in each layer and the activation function, sets the learning rate, optimizer and loss function, thereby achieving improved accuracy of electronic transformer fault data; Patent application CN116243230A discloses an online fault diagnosis method for voltage transformers. The embodiment of the present application divides a marked fault data set into multiple independent parts, and uses different training sets and validation sets in turn to evaluate the performance of a recurrent neural network model. The process is repeated multiple times using different training sets and validation sets. The performance of the recurrent neural network model under different combinations of training sets and validation sets is compared, and one or more training sets and validation sets that are not used in the training process are selected to evaluate the generalization ability of the recurrent neural network model, thereby alternately using marked training sets and validation sets to evaluate the generalization ability of the recurrent neural network model, thereby filling the gap in the AI ​​large model for electronic transformer fault assessment.

[0100] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0101] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0102] 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 including 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.

[0103] 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 in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0104] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0105] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A fault assessment method for power equipment based on AI large model, characterized in that: The following steps are involved: Collect and process electronic transformer fault data, clean, remove noise and extract features from the collected electronic sensor fault data to obtain a marked fault data set; By selecting a recurrent neural network model for sequence data processing, the recurrent neural network model is trained using a marked fault data set, the working state of the transformer is identified through the recurrent neural network model, and the marked fault data monitored in real time is predicted and processed; The labeled fault data set is divided into a training set and a validation set, and the labeled training set and validation set are used alternately to evaluate the generalization ability of the recurrent neural network model, and the generalization ability of the recurrent neural network model is evaluated. The overfitting or underfitting problem is identified through the generalization ability of the recurrent neural network model, and the fault warning of the electronic transformer is carried out according to the evaluation results of the recurrent neural network model; The specific prediction processing method for predicting and processing the marked fault data monitored in real time is as follows: According to the marked fault data monitored in real time and the input data of the electronic transformer, a fault prediction model of the electronic transformer is established, and the fault prediction model of the electronic transformer is used to predict the fault of the electronic transformer to obtain the fault data of the electronic transformer; The specific evaluation process of using the fault prediction model of the electronic transformer to predict the fault of the electronic transformer to obtain the fault data of the electronic transformer is as follows: The fault data of the electronic transformer includes temperature fault data of the electronic transformer, humidity fault data of the electronic transformer and noise signal fault data of the electronic transformer; Evaluating the temperature fault data of the electronic transformer to obtain a temperature fault evaluation coefficient of the electronic transformer; Evaluating the humidity fault data of the electronic transformer to obtain a humidity fault evaluation coefficient of the electronic transformer; Evaluating the noise signal fault data of the electronic transformer to obtain a noise signal fault evaluation coefficient of the electronic transformer; Comprehensively evaluating the temperature fault coefficient of the electronic instrument transformer, the humidity fault coefficient of the electronic instrument transformer, and the noise signal fault coefficient of the electronic instrument transformer to obtain a fault evaluation coefficient of the electronic instrument transformer; The specific analysis formula of the fault evaluation coefficient of the electronic transformer is: In the formula, The fault evaluation coefficient of the electronic transformer is used to evaluate the comprehensive impact of the temperature, humidity and noise signal of the electronic transformer on the fault of the electronic transformer. Indicates the temperature failure coefficient of the electronic transformer, Indicates the humidity failure coefficient of the electronic transformer, Indicates the noise signal fault coefficient of the electronic transformer; Specific constraint formula for the temperature fault evaluation coefficient of the electronic transformer: In the formula, It represents the temperature fault evaluation coefficient of the electronic transformer, which is used to evaluate the influence of the temperature of the electronic transformer on the fault of the electronic transformer. Indicates the heat dissipation area of ​​the electronic transformer. Indicates the thermal expansion coefficient of the electronic transformer, Indicates the heat transfer resistance of the electronic transformer. Indicates the operating temperature of the electronic transformer. Indicates the ambient temperature of the electronic transformer. Indicates the thermal stress of electronic transformer; Specific constraint formula for humidity fault assessment coefficient of electronic transformer: In the formula, several time monitoring points are set and numbered in sequence. , Indicates the number of the time monitoring point, Indicates the total number of time monitoring points. It represents the humidity fault assessment coefficient of the electronic transformer, which is used to assess the influence of humidity on the fault of the electronic transformer. Indicates The insulation resistance of the electronic transformer at each time monitoring point, Indicates The resonant frequency of the dielectric constant of the electronic transformer at each time monitoring point, Indicates The corrosion rate of the electronic transformer at each time monitoring point is The capacitance value representing the dielectric constant of the electronic transformer, Indicates the total pollution value of the electronic transformer; The specific constraint formula of the noise signal fault evaluation coefficient of the electronic transformer is: In the formula, It represents the noise signal fault evaluation coefficient of the electronic transformer, which is used to evaluate the influence of the electronic transformer noise signal on the electronic transformer fault. Represents the signal-to-noise ratio of the electronic transformer, represents the noise power spectrum of the electronic transformer, Represents the noise density of the electronic transformer, Indicates the total harmonic distortion of the electronic transformer. Indicates the bit error rate of the electronic transformer.

2. The fault assessment method for power equipment based on AI large model as claimed in claim 1, characterized in that: The specific collection process of collecting and processing the electronic transformer fault data is as follows: The electronic transformer fault data is collected by sensors, wherein the electronic transformer fault data includes electronic transformer temperature fault data, electronic transformer humidity fault data and electronic transformer noise signal fault data, and invalid data is removed, missing values ​​are filled and abnormal values ​​are processed for the electronic transformer fault data.

3. The fault assessment method for power equipment based on AI large model as claimed in claim 1, characterized in that: The specific processing method of selecting the recurrent neural network model to process sequence data is: The operating data of the electronic transformer is arranged in time series to obtain sequence data, which is then converted into a format input into a recurrent neural network model. The converted sequence data is marked and indicated at the beginning and end of the time series.

4. The fault assessment method for power equipment based on AI large model as claimed in claim 1, characterized in that: The specific evaluation process of alternating the use of labeled training sets and validation sets to evaluate the generalization ability of the recurrent neural network model is as follows: The labeled fault data set is divided into different training sets and validation sets, and the performance of the recurrent neural network model is evaluated using different training sets and validation sets. By comparing the performance of the recurrent neural network model under different combinations of training sets and validation sets, one or more training sets and validation sets that are not used in the training process are selected to evaluate the generalization ability of the recurrent neural network model.

5. The fault assessment method for power equipment based on AI large model as claimed in claim 3 is characterized in that: The specific conversion process of converting sequence data into a format that is input to a recurrent neural network model is as follows: Scale the sequence data within a range, convert the elements in the sequence into one-hot encoding, pad the shorter sequence data, divide the sequence data into segments of the same length, extract features based on the characteristics of the sequence, obtain the converted sequence data, and use the converted sequence data to batch train the recurrent neural network model.

6. The fault assessment method for power equipment based on AI large model as claimed in claim 1, characterized in that: The specific early warning process of the electronic transformer fault early warning according to the evaluation result of the recurrent neural network model is as follows: In the recurrent neural network model, the fault assessment of the electronic transformer is carried out on three aspects: temperature, humidity and noise signal. The fault assessment coefficient of the electronic transformer is obtained. An allowable threshold is set for the fault assessment coefficient of the electronic transformer. If the fault assessment coefficient of the electronic transformer falls outside the allowable threshold, the electronic transformer will issue an early warning.

7. A system using the fault assessment method for power equipment based on the AI ​​large model as described in any one of claims 1 to 6, characterized in that: Including data collection and processing module, model selection and training module and model evaluation and early warning module: Data collection and processing module: used to collect and process electronic transformer fault data, clean, remove noise and extract features from the collected electronic sensor fault data to obtain a marked fault data set; Model selection and training module: used to process sequence data by selecting a recurrent neural network model, train the recurrent neural network model using a marked fault data set, identify the working state of the transformer through the recurrent neural network model, and perform predictive processing on the marked fault data monitored in real time; Model evaluation and early warning module: used to divide the labeled fault data set into multiple parts, and alternately use the labeled training set and validation set to evaluate the generalization ability of the recurrent neural network model. The generalization ability of the recurrent neural network model is evaluated to identify overfitting or underfitting problems.

Images