A Transformer Fault Simulation Detection System

Abstract

The invention relates to the technical field of transformer fault simulation, and discloses a transformer fault simulation detection system. Firstly, transformer fault data is obtained from a monitoring system, fault report and maintenance record channel of an electric power system; and corresponding parameters of the transformer during normal operation are collected for comparison; missing data are processed, wherein the parameters are cleaned and integrated by a linear interpolation method, and data standardization is performed; fault simulation data is generated, and a transformer physical model is established; fault identification is performed on the simulated data in transformer fault simulation software, specifically including extraction of electrical fault features and non-electrical fault features; a fault diagnosis model is constructed and trained; and fault data is output. The invention can more accurately judge the fault type and avoid misjudgment that may occur in single feature diagnosis.

Description

Technical Field

[0001] The invention relates to the technical field of transformer simulation detection, and in particular to a transformer fault simulation detection system. Background Art

[0002] The transformer fault simulation detection system is a comprehensive software system that aims to generate transformer fault simulation data by combining historical fault data of the power system, and use this data for fault detection and diagnosis. The system integrates multiple functional modules such as data processing, model simulation, feature extraction and fault diagnosis, providing effective technical support for the safe operation of transformers in power systems.

[0003] Although the system attempts to combine historical fault data for simulation, the complexity of actual transformer faults is difficult to fully simulate. For example, in actual operation, the transformer may be affected by a variety of complex environmental factors (such as electrical factors, non-electrical factors, transformer oil temperature, gas, etc.) and fail. However, it is difficult to accurately take all these factors into account in the simulation system, resulting in a deviation between the generated fault simulation data and the actual fault situation.

[0004] For some rare fault types, it is difficult to accurately establish fault models for simulation due to the lack of historical data. For example, for faults in transformer cores made of certain special materials, there is no corresponding simulation to support the extraction of fault characteristics and the simulation of fault scenarios.

[0005] Therefore, in order to solve the above problems, a transformer fault simulation detection system is needed. Summary of the invention

[0006] The purpose of the present invention is to provide a transformer fault simulation detection system. The present invention can take both electrical and non-electrical data into account, reduce the deviation between the generated fault simulation data and the actual fault situation, and can more accurately determine the fault type, avoiding the misjudgment that may occur in single feature diagnosis.

[0007] The present invention is implemented in this way. The present invention provides a transformer fault simulation detection method which is specifically performed according to the following steps:

[0008] S1: First, obtain transformer fault data from the power system monitoring system, fault report and maintenance record channels, including fault time and fault type, specifically including short circuit, insulation breakdown and overheating data;

[0009] Electrical parameters during fault: voltage U, current I, power factor and non-electrical parameters such as oil temperature T, dissolved gas composition in oil; and at the same time, the corresponding parameters of the transformer during normal operation are collected for comparison;

[0010] S2: Processing of missing data, where parameters are cleaned and integrated using linear interpolation, and data standardization;

[0011] When processing electrical parameters, the data points before and after the missing point are assumed to be (t1, U1) and (t2, U2), and the missing time is t; the voltage U is obtained by interpolation, as shown in formula (3);

[0012]

[0013] The calculated voltage is interpolated to fill in the missing voltage data; and the abnormal data is identified by setting the threshold;

[0014] Perform data standardization, as shown in formula (4);

[0015]

[0016] Among them, x is the standardization parameter, μ is the mean, and σ is the standard deviation.

[0017] Further, S3: generate fault simulation data and establish a transformer physical model, such as formula (1)-formula (2);

[0018] Z1=E1+I1R1+jI1X1 Formula (1)

[0019] Z=E2-I2R2-jI2X2 Formula (2)

[0020] Among them, Z1 and Z2 are the voltages on the primary and secondary sides of the transformer; I1 and I2 are the currents on the primary and secondary sides of the transformer; R1 and R2 are the turns of the primary and secondary windings of the transformer, E1 and E2 are the induced electromotive force on the primary and secondary sides, and X1 and X2 are the leakage reactance on the primary and secondary sides;

[0021] S4: Perform fault injection and simulation to simulate short-circuit faults; specifically, in transformer fault simulation software, simulate short-circuit faults by modifying inductance parameters to obtain simulation data of voltage and current parameters under short-circuit faults;

[0022] Simulate winding short circuit, assuming that the ratio of short-circuited turns to total turns is v, then the equivalent inductance after short circuit is as shown in formula (5);

[0023] L′=(1-v)L Formula (5)

[0024] Where, L is the original inductance;

[0025] Simulate the inter-turn short circuit fault, specifically by adjusting the resistance and inductance parameters to simulate and obtain the corresponding inter-turn short circuit fault data; as shown in formula (6);

[0026] R′=(1+m)R Formula (6)

[0027] Where m is the resistance variation coefficient;

[0028] Simulate the core fault, according to the magnetization curve of the transformer, specifically simulate the core fault by changing the magnetic permeability of the core; as shown in formula (7);

[0029] B=f(H) Formula (7)

[0030] Where B is the magnetic flux density and H is the magnetic field strength;

[0031] Combine different fault types and degrees, including simulating turn-to-turn short circuit and core overheating faults at the same time, and specifically change the value range of fault parameters, including the turn-to-turn short circuit degree from mild to severe, specifically k from 0.01-0.1), and the resistance change coefficient m caused by core overheating from 0.1-0.5, to generate simulation data for multiple fault combinations.

[0032] S5: performing fault identification on the simulated data in the transformer fault simulation software, specifically including extracting electrical fault features and non-electrical fault features;

[0033] Please follow the steps below:

[0034] S5.1: wherein the extraction of electrical fault characteristics includes current harmonic analysis, voltage unbalance calculation, and voltage unbalance calculation;

[0035] S5.2: Wherein the current harmonic analysis includes performing Fourier transform on the current signal i(t), such as equations (8)-(9);

[0036]

[0037] Where N is the number of sampling points, k is the harmonic order;

[0038]

[0039] Among them, I k is the subharmonic amplitude, I1 is the fundamental amplitude, and THD is the harmonic distortion rate, which is used to measure the degree of distortion of the current waveform as a fault feature;

[0040] S5.3: wherein the unbalance degree of the three-phase transformer voltage is calculated, as shown in equations (10) to (12);

[0041]

[0042] Among them, U1 is the positive sequence voltage component, U2 is the negative sequence voltage component, ∈ is the balance degree, U a , U b and U cThey are the three-phase voltages of the three-phase transformer respectively.

[0043] Furthermore, non-electrical characteristics specifically include oil temperature change rate and dissolved gas analysis in oil;

[0044] The oil temperature change rate is as shown in formula (13);

[0045]

[0046] Where, T(t) is the function of oil temperature changing with time, and r is the rate of change of oil temperature in the time interval [t1, t2];

[0047] The analysis of dissolved gases in oil specifically determines the fault type based on the content ratio of hydrogen H2, methane CH4, ethane C2H6, ethylene C2H4 and acetylene C2H2 gases in the oil.

[0048] S6: construct and train a fault diagnosis model;

[0049] Construct and train the fault diagnosis model and train the model through a neural network. The specific steps are as follows:

[0050] S6.1: Use the grid search method to adjust the hyperparameters of the neural network, including setting the hyperparameters of the neural network model learning rate, number of iterations, and batch size;

[0051] S6.2: Input the training data set into the constructed model for training. During the training process, the parameters of the model are updated according to the Adam adaptive learning rate optimization algorithm; as shown in formula (14):

[0052]

[0053] Among them, m t and v t are the first-order moment estimate and the second-order moment estimate, β1 and β2 are the decay rates, g t is the current gradient, and is the corrected moment estimate, α is the learning rate, ∈ is a small constant to prevent the denominator from being zero, and θ t are the parameters of the model. During the training process, if the model training meets the iterative convergence conditions, the model parameters will be stopped from being updated;

[0054] And use the forward propagation of the neural network to train the data, as shown in equations (15)-(16);

[0055] a (l) =f(z (l) ) Formula (15)

[0056] Among them, z (l) =W(l-1) a (l-1) +b (l-1) , l represents the number of layers, W is the weight matrix, b is the bias vector, a is the activation value, and f is the activation function;

[0057] The neural network is trained by historical fault data and simulation data, and the weights and biases are updated using the back propagation algorithm, as shown in formula (16);

[0058]

[0059] Where N is the number of samples, y i is the true label, is the predicted label.

[0060] S7: Output the fault data and compare it with the fault data collected in step S1 to determine the simulation accuracy of the transformer fault simulation detection system;

[0061] Please follow the steps below:

[0062] S7.1: Use the test data set to test the performance of the model and use the accuracy indicator to evaluate the model; as shown in formula (17);

[0063]

[0064] Among them, TP represents true positives, that is, the number of samples that are actually positive and predicted as positive by the model; TN represents true negatives, that is, the number of samples that are actually negative and predicted as negative by the model; FP represents false positives, that is, the number of samples that are actually negative but predicted as positive by the model; FN represents false negatives, that is, the number of samples that are actually positive but predicted as negative by the model;

[0065] S7.2: Optimize the model based on the evaluation results, including adjusting hyperparameters, increasing the amount of data, and improving the model structure until the model reaches the set performance.

[0066] S8: According to the comparison result, return to step S6 to continuously train the fault diagnosis model until the simulation accuracy of the set transformer fault simulation detection system is reached.

[0067] Further, the present invention provides a transformer fault simulation detection system, including a data collection submodule, for obtaining transformer historical fault data and normal operation data from data sources including a power system database and a fault record document, including electrical parameters, voltage, current, power factor and non-electrical parameters, oil temperature, and gas composition dissolved in oil, and recording fault type and occurrence time information;

[0068] Data cleaning submodule, checks the integrity of data and handles missing values ​​and outliers;

[0069] The data standardization submodule standardizes the data, normalizes the parameter data of different magnitudes to the same range, and performs Z-score standardization on the voltage, current and other data;

[0070] The fault simulation data generation module injects corresponding faults into the transformer model according to the fault types in the historical fault data;

[0071] A feature extraction module, which extracts electrical features and non-electrical features that can characterize the fault from the generated fault simulation data and historical fault data;

[0072] The machine learning diagnosis module takes the extracted fault features as input and the fault type as output to build a neural network learning model;

[0073] Model evaluation and optimization module is used to evaluate the performance of the fault detection model.

[0074] Furthermore, a computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any one of the above-mentioned transformer fault simulation detection methods when executing the computer program.

[0075] Furthermore, the present invention provides a computer storable medium, which includes a processing system and a stored program, and controls the execution of the steps of the above-mentioned transformer fault simulation detection method when the processing system control program runs.

[0076] Compared with the prior art, the present invention has the following beneficial effects:

[0077] 1. The present invention can generate multiple fault types and combined scenarios of different fault severity. For example, it can not only simulate a single turn-to-turn short circuit fault, but also simulate turn-to-turn short circuit and core overheating faults at the same time, and generate rich fault samples by adjusting relevant parameters, such as turn-to-turn short circuit ratio, core temperature rise amplitude, etc.

[0078] 2. Comprehensively consider electrical characteristics such as voltage harmonic content, current harmonic content, zero-sequence current, voltage imbalance, etc. and non-electrical characteristics such as oil temperature change rate, dissolved gas content in oil, etc. to perform fault diagnosis. For example, when a partial discharge fault occurs in a transformer, the content of hydrogen and acetylene in the dissolved gas in the oil will increase significantly, and may be accompanied by slight changes in electrical parameters. By analyzing these characteristics at the same time, the fault type can be judged more accurately, avoiding the misjudgment that may occur in single feature diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It is understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0080] Figure 1 is a flow chart of the method of the present invention;

[0081] Figure 2 It is a system structure diagram of the present invention. DETAILED DESCRIPTION

[0082] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the invention claimed for protection, but is only for selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.

[0083] See also Figure 1-Figure 2 ,The present invention provides a transformer fault simulation detection system, S1: firstly, transformer fault data including fault time and fault type, specifically including short circuit, insulation breakdown, and overheating data, are obtained from the monitoring system, fault report, and maintenance record channels of the power system;

[0084] Electrical parameters during fault: voltage U, current I, power factor and non-electrical parameters such as oil temperature T, dissolved gas composition in oil; and at the same time, the corresponding parameters of the transformer during normal operation are collected for comparison;

[0085] S2: Processing of missing data, where parameters are cleaned and integrated using linear interpolation, and data standardization;

[0086] When processing electrical parameters, the data points before and after the missing point are assumed to be (t1, U1) and (t2, U2), and the missing time is t; the voltage U is obtained by interpolation, as shown in formula (3);

[0087]

[0088] The calculated voltage is interpolated to fill in the missing voltage data; and the abnormal data is identified by setting the threshold;

[0089] Perform data standardization, as shown in formula (4);

[0090]

[0091] Among them, x is the standardization parameter, μ is the mean, and σ is the standard deviation.

[0092] Further, S3: generate fault simulation data and establish a transformer physical model, such as formula (1)-formula (2);

[0093] Z1=E1+I1R1+jI1X1 Formula (1)

[0094] Z2=E2-I2R2-jI2X2 Formula (2)

[0095] Among them, Z1 and Z2 are the voltages on the primary and secondary sides of the transformer; I1 and I2 are the currents on the primary and secondary sides of the transformer; R1 and R2 are the turns of the primary and secondary windings of the transformer, E1 and E2 are the induced electromotive force on the primary and secondary sides, and X1 and X2 are the leakage reactance on the primary and secondary sides;

[0096] S4: Perform fault injection and simulation to simulate short-circuit faults; specifically, in transformer fault simulation software, simulate short-circuit faults by modifying inductance parameters to obtain simulation data of voltage and current parameters under short-circuit faults;

[0097] Simulate winding short circuit, assuming that the ratio of short-circuited turns to total turns is v, then the equivalent inductance after short circuit is as shown in formula (5);

[0098] L′=(1-v)L Formula (5)

[0099] Where, L is the original inductance;

[0100] Simulate the inter-turn short circuit fault, specifically by adjusting the resistance and inductance parameters to simulate and obtain the corresponding inter-turn short circuit fault data; as shown in formula (6);

[0101] R′=(1+m)R Formula (6)

[0102] Where m is the resistance variation coefficient;

[0103] Simulate the core fault, according to the magnetization curve of the transformer, specifically simulate the core fault by changing the magnetic permeability of the core; as shown in formula (7);

[0104] B=f(H) Formula (7)

[0105] Where B is the magnetic flux density and H is the magnetic field strength;

[0106] Combine different fault types and degrees, including simulating turn-to-turn short circuit and core overheating faults at the same time, and specifically change the value range of fault parameters, including the turn-to-turn short circuit degree from mild to severe, specifically k from 0.01-0.1), and the resistance change coefficient m caused by core overheating from 0.1-0.5, to generate simulation data for multiple fault combinations.

[0107] S5: performing fault identification on the simulated data in the transformer fault simulation software, specifically including extracting electrical fault features and non-electrical fault features; specifically, performing the following steps:

[0108] S5.1: wherein the extraction of electrical fault characteristics includes current harmonic analysis, voltage unbalance calculation, and voltage unbalance calculation;

[0109] S5.2: Wherein the current harmonic analysis includes performing Fourier transform on the current signal i(t), such as equations (8)-(9);

[0110]

[0111] Where N is the number of sampling points, k is the harmonic order;

[0112]

[0113] Among them, I k is the subharmonic amplitude, I1 is the fundamental amplitude, and THD is the harmonic distortion rate, which is used to measure the degree of distortion of the current waveform as a fault feature;

[0114] S5.3: wherein the unbalance degree of the three-phase transformer voltage is calculated, as shown in equations (10) to (12);

[0115]

[0116]

[0117] Among them, U1 is the positive sequence voltage component, U2 is the negative sequence voltage component, ∈ is the balance degree, U a , U b and U c They are the three-phase voltages of the three-phase transformer respectively.

[0118] In this embodiment, the non-electrical characteristics specifically include oil temperature change rate and dissolved gas analysis in the oil;

[0119] The oil temperature change rate is as shown in formula (13);

[0120]

[0121] Where, T(t) is the oil temperature changing function over time, and r is the oil temperature changing rate in the time interval [t1, t2];

[0122] The analysis of dissolved gas in oil determines the fault type by the ratio of hydrogen H2, methane CH4, ethane C2H6, ethylene C2H4 and acetylene C2H2 in the oil. Specifically, it includes the ratio range corresponding to different fault types.

[0123] Low temperature overheating (below 150℃): Low temperature overheating (150-300℃): Medium temperature overheating (300-700℃):

[0124] High temperature overheating (above 700℃):

[0125] Partial discharge:

[0126] Low energy discharge:

[0127] High energy discharge:

[0128] The general principle for judging gas abnormality is that when the actual measured ratio exceeds the ratio interval corresponding to the normal operating range mentioned above, it can be judged as gas abnormality, indicating that there may be a corresponding type of fault. For example, if the measurement is According to the above standards, it can be judged as a partial discharge fault.

[0129] S6: construct and train a fault diagnosis model;

[0130] Construct and train the fault diagnosis model and train the model through a neural network. The specific steps are as follows:

[0131] S6.1: Use the grid search method to adjust the hyperparameters of the neural network, including setting the hyperparameters of the neural network model learning rate, number of iterations, and batch size;

[0132] S6.2: Input the training data set into the constructed model for training. During the training process, the parameters of the model are updated according to the Adam adaptive learning rate optimization algorithm; as shown in formula (14):

[0133]

[0134] Among them, m t and v t are the first-order moment estimate and the second-order moment estimate, β1 and β2 are the decay rates, gt is the current gradient, and is the corrected moment estimate, α is the learning rate, ∈ is a small constant to prevent the denominator from being zero, and θ t are the parameters of the model. During the training process, if the model training meets the iterative convergence conditions, the model parameters will be stopped from being updated;

[0135] And use the forward propagation of the neural network to train the data, as shown in equations (15)-(16);

[0136] a (l) =f(z (l) ) Formula (15)

[0137] Among them, z (l) =W (l-1) a (l-1) +b (l-1) , l represents the number of layers, W is the weight matrix, b is the bias vector, a is the activation value, and f is the activation function;

[0138] The neural network is trained by historical fault data and simulation data, and the weights and biases are updated using the back propagation algorithm, as shown in formula (16);

[0139]

[0140] Where N is the number of samples, y i is the true label, is the predicted label.

[0141] S7: Output the fault data and compare it with the fault data collected in step S1 to determine the simulation accuracy of the transformer fault simulation detection system;

[0142] Please follow the steps below:

[0143] S7.1: Use the test data set to test the performance of the model and use the accuracy indicator to evaluate the model; as shown in formula (17);

[0144]

[0145] Among them, TP represents true positives, that is, the number of samples that are actually positive and predicted as positive by the model; TN represents true negatives, that is, the number of samples that are actually negative and predicted as negative by the model; FP represents false positives, that is, the number of samples that are actually negative but predicted as positive by the model; FN represents false negatives, that is, the number of samples that are actually positive but predicted as negative by the model;

[0146] S7.2: Optimize the model based on the evaluation results, including adjusting hyperparameters, increasing the amount of data, and improving the model structure until the model reaches the set performance.

[0147] S8: According to the comparison result, return to step S6 to continuously train the fault diagnosis model until the simulation accuracy of the set transformer fault simulation detection system is reached.

[0148] In this embodiment, the present invention provides a transformer fault simulation detection system, including a data collection submodule, which is used to obtain transformer historical fault data and normal operation data from data sources including a power system database and a fault record document, including electrical parameters, voltage, current, power factor and non-electrical parameters, oil temperature, and dissolved gas composition in oil, and record fault type and occurrence time information;

[0149] Data cleaning submodule, checks the integrity of data and handles missing values ​​and outliers;

[0150] The data standardization submodule standardizes the data, normalizes the parameter data of different magnitudes to the same range, and performs Z-score standardization on the voltage, current and other data;

[0151] The fault simulation data generation module injects corresponding faults into the transformer model according to the fault types in the historical fault data;

[0152] A feature extraction module, which extracts electrical features and non-electrical features that can characterize the fault from the generated fault simulation data and historical fault data;

[0153] The machine learning diagnosis module takes the extracted fault features as input and the fault type as output to build a neural network learning model;

[0154] Model evaluation and optimization module is used to evaluate the performance of the fault detection model.

[0155] In this embodiment, a computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of any one of the above-mentioned transformer fault simulation detection methods are implemented.

[0156] In this embodiment, the present invention provides a computer storable medium, which includes a processing system and a stored program, and controls the execution of the steps of the above transformer fault simulation detection method when the processing system control program runs.

[0157] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention has various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A transformer fault simulation detection method, characterized in that: Please follow the steps below: S1: First, obtain transformer fault data from the power system monitoring system, fault report and maintenance record channels, including fault time and fault type, specifically including short circuit, insulation breakdown and overheating data; Electrical parameters during fault: voltage U, current I, power factor and non-electrical parameters such as oil temperature T, dissolved gas composition in oil; and at the same time, the corresponding parameters of the transformer during normal operation are collected for comparison; S2: Processing of missing data, where parameters are cleaned and integrated using linear interpolation, and data standardization; S3: Generate fault simulation data and establish a transformer physical model, such as formula (1)-formula (2); Z1=E1+I1R1+jI1X1 Formula (1) Z2=E2-I2R2-jI2X2 Formula (2) Among them, Z1 and Z2 are the voltages on the primary and secondary sides of the transformer; I1 and I2 are the currents on the primary and secondary sides of the transformer; R1 and R2 are functions of the number of turns of the primary and secondary windings of the transformer, E1 and E2 are the induced electromotive force on the primary and secondary sides, and X1 and X2 are the leakage reactance on the primary and secondary sides; S4: Perform fault injection and simulation to simulate short-circuit faults; specifically, in transformer fault simulation software, simulate short-circuit faults by modifying inductance parameters to obtain simulation data of voltage and current parameters under short-circuit faults; Simulate winding short circuit, assuming that the ratio of short-circuited turns to total turns is v, then the equivalent inductance after short circuit is as shown in formula (5); L′=(1-v)L Formula (5) Where, L is the original inductance; Simulate the inter-turn short circuit fault, specifically by adjusting the resistance and inductance parameters to simulate and obtain the corresponding inter-turn short circuit fault data; as shown in formula (6); R′=(1+m)R Formula (6) Where m is the resistance variation coefficient; Simulate the core fault, according to the magnetization curve of the transformer, specifically simulate the core fault by changing the magnetic permeability of the core; as shown in formula (7); B=f(H) Formula (7) Where B is the magnetic flux density and H is the magnetic field strength; Combine different fault types and degrees, including simulating turn-to-turn short circuit and core overheat faults at the same time, and specifically change the value range of fault parameters, including the turn-to-turn short circuit degree from mild to severe, specifically k from 0.01-0.1, and the resistance change coefficient m caused by core overheating from 0.1-0.5, to generate simulation data for multiple fault combinations; S5: performing fault identification on the simulated data in the transformer fault simulation software, specifically including extracting electrical fault features and non-electrical fault features, specifically performed in the following steps; S5.1: The extraction of electrical fault characteristics includes current harmonic analysis and voltage unbalance calculation; S5.2: Wherein the current harmonic analysis includes performing Fourier transform on the current signal i(t), such as equations (8)-(9); Where N is the number of sampling points, k is the harmonic order; Among them, I k is the subharmonic amplitude, I1 is the fundamental amplitude, and THD is the harmonic distortion rate, which is used to measure the degree of distortion of the current waveform as a fault feature; S5.3: wherein the unbalance degree of the three-phase transformer voltage is calculated, as shown in equations (10) to (12); Among them, U1 is the positive sequence voltage component, U2 is the negative sequence voltage component, ∈ is the balance degree, U a , U b and U c They are the three-phase voltages of the three-phase transformer; The non-electrical characteristics specifically include oil temperature change rate and dissolved gas analysis in oil; the oil temperature change rate is as shown in formula (13); Where, T(t) is the oil temperature changing function over time, and r is the oil temperature changing rate in the time interval [t1, t2]; The analysis of dissolved gas in oil specifically determines the fault type based on the content ratio of hydrogen H2, methane CH4, ethane C2H6, ethylene C2H4 and acetylene C2H2 in the oil; S6: construct and train a fault diagnosis model; S7: Output the fault data and compare it with the fault data collected in step S1 to determine the simulation accuracy of the transformer fault simulation detection system; S8: According to the comparison result, return to step S6 to continuously train the fault diagnosis model until the simulation accuracy of the set transformer fault simulation detection system is reached.

2. A transformer fault simulation detection method according to claim 1, characterized in that: In step S2, when processing the electrical parameters, the data points before and after the missing point are assumed to be (t1, U1) and (t2, U2), and the missing time is t; the voltage U is obtained by interpolation, as shown in formula (3); The calculated voltage is interpolated to fill in the missing voltage data; and the abnormal data is identified by setting the threshold; Perform data standardization, as shown in formula (4); Among them, x is the standardization parameter, μ is the mean, and σ is the standard deviation.

3. A transformer fault simulation detection method according to claim 1, characterized in that: In step S6, a fault diagnosis model is constructed and trained, and the model training is performed through a neural network, which is specifically performed in the following steps: S6.1: Use the grid search method to adjust the hyperparameters of the neural network, including setting the hyperparameters of the neural network model learning rate, number of iterations, and batch size; S6.2: Input the training data set into the constructed model for training. During the training process, the parameters of the model are updated according to the Adam adaptive learning rate optimization algorithm; as shown in formula (14): Among them, m t and t are the first-order moment estimate and the second-order moment estimate, β1 and β2 are the decay rates, g t is the current gradient, and is the corrected moment estimate, α is the learning rate, ∈ is a small constant to prevent the denominator from being zero, and θ t are the parameters of the model. During the training process, if the model training meets the iterative convergence conditions, the model parameters will be updated; And use the forward propagation of the neural network to train the data, as shown in equations (15)-(16); a (l) =f(z (l) ) Formula (15) Among them, z (l) =W (l-1) a (l-1) +b (l-1) , l represents the number of layers, W is the weight matrix, b is the bias vector, a is the activation value, and f is the activation function; The neural network is trained by historical fault data and simulation data, and the weights and biases are updated using the back propagation algorithm, as shown in formula (16); Where N is the number of samples, y i is the true label, is the predicted label.

4. A transformer fault simulation detection method according to claim 1, characterized in that: In step S7, the following steps are specifically performed: S7.1: Use the test data set to test the performance of the model and use the accuracy indicator to evaluate the model; as shown in formula (17); Among them, TP represents true positives, that is, the number of samples that are actually positive and predicted as positive by the model; TN represents true negatives, that is, the number of samples that are actually negative and predicted as negative by the model; FP represents false positives, that is, the number of samples that are actually negative but predicted as positive by the model; FN represents false negatives, that is, the number of samples that are actually positive but predicted as negative by the model; S7.2: Optimize the model based on the evaluation results, including adjusting hyperparameters, increasing the amount of data, and improving the model structure until the model reaches the set performance.

5. A transformer fault simulation detection system, implementing a transformer fault simulation detection method as claimed in claim 1, characterized in that: It includes a data collection submodule, which is used to obtain the historical fault data and normal operation data of the transformer from data sources including the power system database and fault record documents, including electrical parameters, voltage, current, power factor and non-electrical parameters, oil temperature, and gas composition dissolved in the oil, and also records the fault type and occurrence time information; Data cleaning submodule, checks the integrity of data and handles missing values ​​and outliers; The data standardization submodule performs data standardization, normalizes parameter data of different magnitudes to the same range, and performs Z-score standardization on voltage and current data; The fault simulation data generation module injects corresponding faults into the transformer model according to the fault types in the historical fault data; A feature extraction module, which extracts electrical features and non-electrical features that can characterize the fault from the generated fault simulation data and historical fault data; The machine learning diagnosis module takes the extracted fault features as input and the fault type as output to build a neural network learning model; Model evaluation and optimization module is used to evaluate the performance of the fault detection model.

6. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of a transformer fault simulation detection method described in any one of claims 1-4 are implemented.

7. A computer storable medium, characterized in that: The computer storable medium includes a processing system and a stored program, and when the processing system control program runs, the steps of the transformer fault simulation detection method described in any one of claims 1 to 4 are controlled and executed.

Images