A power failure detection system and method

Abstract

The present invention belongs to the technical field of power electronics. The present invention discloses a power supply fault detection system and method, including a data monitoring module for acquiring power supply operation data, power supply status data, and magnetic field characteristic data; a data processing module for preprocessing the power supply operation data and power supply status data to obtain an operation characteristic data set and a status characteristic data set, performing a correlation analysis on the operation characteristic data set and the status characteristic data set to obtain a power supply comprehensive characteristic data set; a model training module for training a power supply fault prediction model, inputting the power supply comprehensive characteristic data set into the power supply fault prediction model, and predicting a power supply failure rate index; a fault detection module for comparing the predicted power supply failure rate index with a preset power supply failure rate index threshold to determine whether the power supply fails; if a failure occurs, automatically trigger a power-off protection mechanism and generate a fault warning message; ensuring the continuous operation of the power supply system and production efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of power electronics, and more specifically, to a power supply fault detection system and method. Background Art

[0002] The power supply system is the basis for the normal operation of various devices and systems, and any power supply fault may lead to serious service interruptions and property losses.

[0003] The patent with the application publication number CN117375245A discloses a power supply control system and a power supply control method based on the Internet of Things. The present invention detects the faults of the main power supply by collecting the current and temperature data in real time when the main power supply is supplying power, and can obtain the corresponding fault types. When a fault is detected, an alarm message containing the fault type and location will be generated and uploaded to the power supply monitoring platform, and the power supply monitoring platform will generate a maintenance work order based on the alarm message and send it to the operation and maintenance end, so as to achieve rapid repair of the fault; at the same time, when the backup power supply is working, the present invention can calculate the charge of the backup power supply in real time and generate the corresponding power curve graph; based on this, the present invention not only realizes the operation and maintenance linkage of the fault, but also enables the operation and maintenance personnel to carry out targeted fault repair preparation work according to the fault type. In this way, the repair efficiency can be greatly improved; in addition, the present invention can also realize the power supply monitoring of the backup power supply. Therefore, it is suitable for large-scale applications in the field of power supply control.

[0004] Traditional power supply monitoring methods mainly rely on manual inspections and single-point sensor-based monitoring, and there are the following main problems:

[0005] When performing power supply fault detection, when there are a wide variety of collected data types, there is a lack of automatic trigger processing for feature extraction of the data. When processing the power supply operation data and power supply status data, no trigger conditions are considered, which consumes a lot of time and human resources, resulting in a decrease in data processing efficiency; in the absence of automatic trigger processing, feature extraction may be affected by subjective human factors, resulting in the extracted features being less objective and accurate; the processing of power supply operation data and power supply status data does not consider setting trigger conditions, resulting in missing key information or ignoring important data features; in the prior art, the correlation between the operation feature dataset and the status feature dataset is ignored, resulting in inaccurate analysis results; this will lead to a significant decrease in the accuracy of model prediction, and further affect the reliability of the prediction of the power supply failure rate index;

[0006] The prior art only considers power supply - related data when detecting power supply faults, lacking the analysis of magnetic field distribution data and magnetic field change data. As a result, abnormal magnetic field distribution and current anomalies in the power supply system cannot be detected in time, delaying the time of fault diagnosis and repair and affecting the stability and reliability of the power supply system. If the magnetic field distribution data, magnetic field change data, and electromagnetic compatibility test data are not comprehensively analyzed, it may lead to poor system maintenance and fault prevention effects. The lack of comprehensive data analysis may cause important problems to be overlooked, affecting the reliability and stability of the system. The failure to use advanced positioning tools such as directional antennas and near - field probes for interference source location results in a lack of accurate interference source location information when solving electromagnetic interference problems, increasing the difficulty and time of troubleshooting, and thus increasing the maintenance and replacement costs.

[0007] In view of this, the present invention proposes a power supply fault detection system and method to solve the above problems. Summary of the Invention

[0008] To overcome the above - mentioned defects of the prior art and to achieve the above - mentioned purpose, the present invention provides the following technical solution: A power supply fault detection system, comprising:

[0009] A data monitoring module, configured to obtain power supply operation data, power supply status data, and magnetic field characteristic data;

[0010] A data processing module, configured to pre - process the power supply operation data and power supply status data to obtain an operation characteristic data set and a status characteristic data set, and perform a correlation analysis on the operation characteristic data set and the status characteristic data set to obtain a comprehensive characteristic data set;

[0011] A model training module, configured to train a power supply fault prediction model, input the comprehensive characteristic data set into the power supply fault prediction model, and predict the power supply failure rate index;

[0012] A fault detection module, configured to compare the predicted power supply failure rate index with a preset power supply failure rate index threshold to determine whether the power supply fails; if a failure occurs, automatically trigger a power - off protection mechanism and generate a fault alarm message;

[0013] A fault control module, the power control terminal identifies the fault alarm message, detects the specific fault location of the power supply through the collected magnetic field intensity data, and sends a fault handling instruction to the operation and maintenance personnel to promptly handle the fault location of the power supply;

[0014] Each module is connected by wired and / or wireless means to achieve data transmission between modules.

[0015] Further, the power operation data includes voltage data, current data, power data, temperature data, frequency data, and harmonic data of the power supply; the power supply status data includes the remaining service life of the power supply, aging and wear indicators, fault history record data, environmental condition data, and user operation behavior data.

[0016] Further, the method for preprocessing the power operation data and the power supply status data to obtain an operation feature data set and a status feature data set includes:

[0017] Perform data cleaning on the obtained power operation data and power supply status data, perform standard deviation normalization processing on the constructed comprehensive smoke feature data set, convert it into a standard normal distribution with a mean of 0 and a standard deviation of 1, and eliminate the influence of the dimension between data;

[0018] Construct an operation-status association model, preset the power operation data and the power supply status data as independent variables of the operation-status association model, and the condition trigger function as the dependent variable of the model; when the probability that the dependent variable of the operation-status association model is 1 reaches the condition trigger threshold, run the condition trigger function;

[0019] Preset the condition trigger function. When the running condition trigger function reaches the preset influence coefficient threshold of the operation-status association model, automatically trigger the principal component analysis to preprocess the power operation data and the power supply status data;

[0020] Match the dimension-reduced operation feature data set and status feature data set to ensure that the number of data types in the operation feature data set and the status feature data set is the same, and obtain the processed operation feature data set and status feature data set;

[0021] The operation-status association model is:

[0022] ;

[0023] Wherein, is the dependent variable of the operation-status association model; is the probability that the dependent variable of the operation-status association model is 1; is the value of the dependent variable when the independent variable of the operation-status association model is 0; is the independent variable of the operation-status association model; is the coefficient of the independent variable; is the number of coefficients of the independent variable; is the number of independent variables;

[0024] The condition trigger function is:

[0025] Q = ln ∑ [ ( ( x 1 − x ^ 1 ) 2 + ( x 2 − x ^ 2 ) 2 ) 2 ] ;

[0026] Among them, is the power operation data obtained in real time; is the preset standard power operation data; is the power status data obtained in real time; is the preset standard power status data;

[0027] For example, the preset condition trigger threshold is 0.1. When reaches a probability of 0.1, the operation condition trigger function is triggered; the influence coefficient threshold of the preset operation - status association model is 0.5. When the operation condition trigger function reaches 0.5, the principal component analysis is automatically triggered to preprocess the power operation data and the power status data;

[0028] The method for automatically triggering the principal component analysis to preprocess the power operation data and the power status data includes:

[0029] There is a preset data set containing samples and features , and the data set is composed of the power operation data and the power status data; the data set is rows column matrix; the data set is:

[0030] O = [ x ′ 11 x ′ 12 ⋅⋅⋅ x ′ 1 q x ′ 21 x ′ 22 ⋅⋅⋅ x ′ 2 q ⋅⋅⋅ ⋅⋅⋅ ⋅⋅⋅ ⋅⋅⋅ x ′ p 1 x ′ p 2 ⋅⋅⋅ x ′ pq ] ;

[0031] Among them, is the th sample of the power operation data and the power status data th feature point; is the number of rows of the data set ; is the number of columns of the data set ;

[0032] Perform standardization processing on the entire data set to calculate the mean vector and the covariance matrix ; the mean vector is:

[0033] ;

[0034] Among them, is the number of samples; is the number of samples; is the th sample's feature point;

[0035] The covariance matrix is:

[0036] ;

[0037] wherein is a vector with characteristics; is transpose;

[0038] Calculate the eigenvectors and eigenvalues of the dataset , select the first eigenvectors to form the transformation matrix ; Transform the dataset through the transformation matrix for dimensionality reduction to obtain the dimensionality-reduced dataset ;

[0039] The method of matching the dimensionality-reduced operating feature dataset and the state feature dataset to ensure that the number of data types in the operating feature dataset and the state feature dataset is the same includes: presetting the number of data types in the dimensionality-reduced operating feature dataset as ; The number of data types in the dimensionality-reduced state feature dataset is ; Construct a data type matching model to make the number of data types in the operating feature dataset and the state feature dataset consistent;

[0040] The type matching model is:

[0041] ;

[0042] wherein is the number of data types to be discarded in the dimensionality-reduced operating feature dataset; is the number of data types to be discarded in the dimensionality-reduced state feature dataset.

[0043] For example, it is preset that there is a dataset containing 15 samples and 11 features , the dataset is composed of power operation data and power state data; the dataset can be represented as a 15×13 matrix; First, standardize the entire dataset, calculate the mean vector and covariance matrix; then calculate the eigenvectors and eigenvalues of the dataset, and select the first 11 eigenvectors to form the transformation matrix; finally, perform dimensionality reduction on the dataset through the transformation matrix to obtain the dimensionality-reduced dataset;

[0044] The reduced - dimensional dataset consists of 11 feature vectors. Among them, the operating feature dataset consists of 6 types of data, and the state feature dataset consists of 5 types of data. Through the type - matching model, it can be known that at this time, the number of data types in the reduced - dimensional operating feature dataset is greater than that in the reduced - dimensional state feature dataset. Therefore, the operating feature dataset needs to discard 1 type of data to make the number of its data types consistent with that of the state feature dataset.

[0045] Furthermore, the method for performing correlation analysis on the operating feature dataset and the state feature dataset to obtain the comprehensive feature dataset includes:

[0046] Preset the operating feature dataset as ; where is the th feature of the operating feature data; is the number of feature types in the operating feature dataset;

[0047] Preset the state feature dataset as ; where is the th feature of the state feature data; is the number of feature types in the state feature dataset;

[0048] For each pair of features , calculate their mutual information ; where is the th feature of the operating feature and the state feature. The mutual - information formula is used to measure the dependence relationship between two random feature variables and ;

[0049] The mutual - information formula is:

[0050] ;

[0051] where is a random feature variable in the operating feature dataset; is a random feature variable in the state feature dataset; and are the values of the random feature variables and ; is the joint probability distribution that the random feature variables and simultaneously take the values of and ; is the probability that the random feature variable takes the value of The marginal probability distribution; is a random feature variable takes values The marginal probability distribution;

[0052] A preset mutual information threshold. When the mutual information value is greater than or equal to the mutual information threshold, it is determined that the feature pair has a significant correlation. Filter out the feature pairs with mutual information values greater than or equal to the threshold, and construct a significant correlation feature subset; fuse the selected significant correlation feature subsets to form a comprehensive feature dataset.

[0053] For example, it is preset that the operation feature dataset has 5 features and the status feature dataset has 5 features; for each pair of features, calculate their mutual information; the preset mutual information threshold is 0.1. When the mutual information value is greater than the mutual information threshold, it is determined that the feature pair has a significant correlation. At this time, the feature pairs with mutual information values greater than the threshold can be filtered out, a significant correlation feature subset can be constructed, and the selected significant correlation feature subsets can be fused to form a comprehensive feature dataset.

[0054] Furthermore, the training method of the power failure prediction model includes;

[0055] Use the comprehensive feature dataset and the corresponding output label power failure rate as the sample set; use the sample set as the dataset of the model, and divide the dataset into a training set, a validation set, and a test set; select a gradient boosting machine framework LightGBM suitable for regression tasks; initialize the LightGBM parameters, and the LightGBM parameters include the maximum number of leaf nodes in each tree, the learning step size, the number of trees, and the maximum depth of the tree; use the historical comprehensive feature dataset as the input data and the corresponding power failure rate index as the output label to train the power failure prediction model; the power failure prediction model is a gradient boosting machine model;

[0056] Use the training set data to train the model. During the training process, use the validation set for early stopping to prevent overfitting; evaluate the model performance on the test set, and use the root mean square error as the loss function to measure the error of the model prediction;

[0057] The root mean square error loss function is:

[0058] ;

[0059] where is the number of the sample set; is the th actual value of the sample; is the th predicted value of the sample;

[0060] Update the parameters of the LightGBM model through the backpropagation algorithm to minimize the loss function; use the validation set to evaluate the performance of the power failure prediction model, and calculate the precision metric to measure the performance of the model;

[0061] Select SGD as the optimizer to find the parameter combination that minimizes the loss function; according to the performance feedback of the validation set precision, adjust the structure of the model and the LightGBM parameters to optimize the model; finally evaluate the performance of the model in the prediction task through the test set, and stop the test when the performance of the model in the prediction task reaches the preset performance threshold to obtain the power failure prediction model.

[0062] Furthermore, the method of adjusting the structure of the model and the LightGBM parameters to optimize the model includes:

[0063] Use Bayesian optimization to optimize the LightGBM parameters in the power failure prediction model. The specific steps are:

[0064] Determine the number of trees and the maximum depth parameter of the tree in the power failure prediction model that need to be optimized;

[0065] Prepare a set of initial data training set, and evaluate the performance metric of the model under the given parameter configuration by calculating the accuracy metric; the accuracy calculation formula is:

[0066] ;

[0067] where, is the power failure rate metric with correct prediction; is the total number of power failure rate metrics predicted;

[0068] Use the initial training set to establish a Gaussian process initial probability model between the parameter configuration and the performance metric; start the iterative optimization process, and each iteration includes the following steps:

[0069] S61. According to the current probability model, use Gaussian process sampling to select the next parameter configuration;

[0070] S62. Use the selected parameter configuration to calculate its performance metric by calculating the accuracy;

[0071] S63. Add the new parameter configuration and performance metric to the training set;

[0072] S64. Update the probability model, add the new training set to the model training process to update the mapping relationship between the parameters and the performance;

[0073] S65. Repeat steps S61 - S64 until the predetermined number of iterations is reached;

[0074] After the iterative optimization is completed, according to the results of the evaluation function, select the parameter configuration with the best performance index as the final model configuration to obtain the trained power failure prediction model, and use the trained power failure prediction model to predict the current comprehensive feature data set to obtain the power failure rate index.

[0075] Furthermore, the method for comparing the predicted power failure evaluation index with the preset power failure evaluation index threshold to determine whether the power supply fails includes:

[0076] If the predicted power failure rate index is greater than or equal to the preset power failure rate index threshold, it is determined that the power supply has failed;

[0077] If the predicted power failure rate index is less than the preset power failure rate index threshold, it is determined that the power supply has not failed.

[0078] Furthermore, the evaluation method for the power failure evaluation index includes:

[0079] Evaluate the power failure evaluation index through the power failure evaluation mathematical model.

[0080] The power failure evaluation mathematical model is:

[0081] ;

[0082] Wherein, is the power failure evaluation index; is the average value of the operation feature data set; is the average value of the power operation data obtained in real time; is the average value of the state feature data set; is the average value of the power state data obtained in real time; is the average value of the preset standard magnetic field feature data; is the average value of the magnetic field feature data obtained in real time; , and are the influencing factors of the power failure evaluation index.

[0083] Furthermore, the method for detecting the specific fault location of the power supply through the collected magnetic field feature data includes:

[0084] The magnetic field feature data includes magnetic field distribution data, magnetic field change data, and electromagnetic compatibility test data;

[0085] Analyze the magnetic field feature data through the Biot-Savart model, and the Biot-Savart model is:

[0086] ;

[0087] Wherein, is the tiny magnetic field generated by the current element at the observation point ; is the magnetic permeability of vacuum, and its value is approximately ; is the current element; is the magnitude of the current; is the tiny length vector in the current direction; is the unit vector pointing from the current element to the observation point ; is the distance from the current element to the observation point ;

[0088] Analysis of magnetic field distribution data: By measuring the magnetic field intensity at multiple points around the power supply system, a magnetic field distribution map is obtained; for a specific region, the spatial distribution of the magnetic field intensity is represented by integrating the total magnetic field obtained from the Biot - Savart model;

[0089] Use the Biot - Savart model to calculate the magnetic field distribution of the power supply system under normal operating conditions, and compare it with the magnetic field distribution of the actually measured power supply system to identify the abnormal magnetic field distribution in the power supply system;

[0090] Analysis of magnetic field change data: Regularly measure the magnetic field intensity at the same observation point of the power supply part, and record the change of the magnetic field over time; a rapid or irregular magnetic field change indicates an abnormal current, and the current is also changing rapidly. Infer the specific part of the power supply system failure through the position of the magnetic field change;

[0091] Analysis of electromagnetic compatibility test data:

[0092] Use a spectrum analyzer and electromagnetic interference receiver equipment to conduct a comprehensive electromagnetic compatibility test on the power supply system; the test contents include radiation interference, conduction interference, and anti - interference ability;

[0093] Record the test environmental conditions, equipment configuration, and the operating state of the power supply system; the environmental conditions include temperature, humidity, and background noise level; the equipment configuration includes receiver settings, antenna type and position; the operating state of the power supply system includes operating frequency and load conditions;

[0094] Through spectrum analysis software, analyze the test results, identify the abnormal signal frequencies and intensities that deviate from the normal operating mode; analyze the source and generation mechanism of the abnormal signals, and judge whether they cause functional failure or performance degradation;

[0095] Compare the current test results with the preset test result threshold, evaluate the change trend of the electromagnetic interference level, and identify the faults of the power supply;

[0096] Using a directional antenna and a near-field probe positioning tool, combined with the results of spectrum analysis, gradually narrow down the search range and accurately locate the position of the interference source; for the power supply system, use time-domain reflectometry to improve the accuracy of positioning; identify the coupling mechanism between the interference source and the power supply components affected by interference, and find out the specific parts where the interference signal affects the power supply.

[0097] A power supply fault detection method, comprising:

[0098] S1. Obtain power supply operation data, power supply status data, and magnetic field characteristic data;

[0099] S2. Preprocess the power supply operation data and power supply status data to obtain an operation feature data set and a status feature data set, perform a correlation analysis on the operation feature data set and the status feature data set, and obtain a comprehensive feature data set;

[0100] S3. Train a power supply fault prediction model, input the comprehensive feature data set into the power supply fault prediction model, and predict to obtain a power supply fault evaluation index;

[0101] S4. Compare the predicted power supply fault evaluation index with a preset power supply fault evaluation index threshold to determine whether the power supply has a fault; if a fault occurs, automatically trigger a power-off protection mechanism and generate a fault alarm message;

[0102] S5. The power supply control terminal identifies the fault alarm message, detects the specific fault location of the power supply through the collected magnetic field characteristic data, sends a fault handling instruction to the operation and maintenance personnel, and timely processes the power supply fault location.

[0103] The technical effects and advantages of a power supply fault detection system and method of the present invention:

[0104] The present invention constructs an operation-status correlation model and presets a condition trigger function. When the running condition trigger function reaches the influence coefficient threshold of the preset operation-status correlation model, it automatically triggers principal component analysis to preprocess the power supply operation data and power supply status data, and matches the reduced-dimensional operation feature data set and status feature data set to ensure that the number of data types in the operation feature data set and the status feature data set is the same, so as to facilitate subsequent correlation analysis of the operation feature data set and the status feature data set by the mutual information method. This automated processing not only improves the response speed of the system, but also reduces the need for human intervention and improves the data processing efficiency; at the same time, valuable correlation information is extracted from the operation feature data set and the status feature data set, providing strong support for fault prediction;

[0105] Through the analysis of magnetic field distribution data using the Biot-Savart model, the abnormal magnetic field distribution in the power supply system can be effectively identified. By analyzing the magnetic field change data, the current abnormality in the power supply system can be detected in a timely manner, thereby inferring the specific location of the system fault, which is of great value for the maintenance and fault prevention of the power supply system. The analysis of electromagnetic compatibility test data can help identify potential electromagnetic interference problems in the power supply system. By analyzing the source and generation mechanism of abnormal signals, the electromagnetic compatibility of the power supply system can be effectively improved, and the impact of interference on system performance can be reduced. Using advanced positioning tools such as directional antennas and near-field probes, combined with the results of spectrum analysis, the location of the interference source can be efficiently and accurately determined, which is of great significance for solving electromagnetic interference problems and ensuring the stable operation of the power supply system. BRIEF DESCRIPTION OF THE DRAWINGS

[0106] Figure 1 FIG. is a schematic structural diagram of a power supply fault detection system according to the present invention;

[0107] Figure 2 FIG. is a schematic flowchart of a power supply fault detection method according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0108] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0109] Embodiment 1

[0110] Please refer to Figure 1 and Figure 2 As shown, a power supply fault detection system in this embodiment includes:

[0111] A data monitoring module for acquiring power operation data, power status data, and magnetic field characteristic data;

[0112] A data processing module for preprocessing the power operation data and power status data to obtain an operation characteristic data set and a status characteristic data set, and performing a correlation analysis on the operation characteristic data set and the status characteristic data set to obtain a comprehensive characteristic data set;

[0113] A model training module for training a power supply fault prediction model, inputting the comprehensive characteristic data set into the power supply fault prediction model, and predicting the power supply failure rate index;

[0114] A fault detection module is used to compare the predicted power failure rate index with a preset power failure rate index threshold to determine whether the power supply has a fault; if a fault occurs, it automatically triggers a power-off protection mechanism and generates a fault alarm message.

[0115] A fault control module. The power control terminal identifies the fault alarm message, detects the specific fault location of the power supply through the collected magnetic field intensity data, and sends a fault handling instruction to the operation and maintenance personnel to promptly handle the fault location of the power supply.

[0116] Each module is connected by wired and / or wireless means to achieve data transmission between modules.

[0117] Power operation data includes voltage data, current data, power data, temperature data, frequency data, and harmonic data of the power supply; power status data includes the remaining service life of the power supply, aging and wear indicators, fault history record data, environmental condition data, and user operation behavior data.

[0118] The voltage data of the power supply is obtained through voltage sensors set around the power supply area; the current data is obtained through current sensors set around the power supply area; the power data is obtained by multiplying the voltage data and the current data; the temperature data is obtained through temperature sensors set around the power supply area; the frequency data is obtained through multifunctional meters with frequency measurement functions set around the power supply area; the harmonic data is obtained through harmonic analyzers set around the power supply area.

[0119] The remaining service life, aging and wear indicators of the power supply are obtained through long-term monitoring of the power operation data; the fault history record is obtained through the power management terminal; the environmental condition data is obtained through environmental sensors set around the power supply area; the user operation behavior data is recorded through the user interface log.

[0120] The methods for preprocessing the power operation data and the power status data to obtain the operation feature dataset and the status feature dataset include:

[0121] Perform data cleaning on the obtained power operation data and power status data, perform standard deviation normalization processing on the constructed comprehensive smoke feature dataset, convert it into a standard normal distribution with a mean of 0 and a standard deviation of 1, and eliminate the influence of the dimension between data.

[0122] Construct an operation-status association model, preset the power operation data and the power status data as the independent variables of the operation-status association model, and the condition trigger function as the dependent variable of the model; when the probability that the dependent variable of the operation-status association model reaches the condition trigger threshold, run the condition trigger function.

[0123] The preset condition trigger function automatically triggers the principal component analysis to preprocess the power operation data and power status data when the running condition trigger function reaches the influence coefficient threshold of the preset operation-status association model;

[0124] Match the dimension-reduced operation feature dataset and status feature dataset to ensure that the number of data types in the operation feature dataset and status feature dataset is the same, and obtain the processed operation feature dataset and status feature dataset;

[0125] The operation-status association model is:

[0126] ;

[0127] Where, is the dependent variable of the operation-status association model; is the probability that the dependent variable of the operation-status association model is 1; is the value of the dependent variable when the independent variable of the operation-status association model is 0; is the independent variable of the operation-status association model; is the coefficient of the independent variable; is the number of coefficients of the independent variable; is the number of independent variables;

[0128] The condition trigger function is:

[0129] Q = ln ∑ [ ( ( x 1 − x ^ 1 ) 2 + ( x 2 − x ^ 2 ) 2 ) 2 ] ;

[0130] Where, is the power operation data obtained in real time; is the preset standard power operation data; is the power status data obtained in real time; is the preset standard power status data;

[0131] The method for automatically triggering the principal component analysis to preprocess the power operation data and power status data includes:

[0132] Preset a dataset containing samples and features , and the dataset is composed of power operation data and power status data; the dataset is rows column matrix; the dataset is:

[0133] O = [ x ′ 11 x ′ 12 ⋅⋅⋅ x ′ 1 q x ′ 21 x ′ 22 ⋅⋅⋅ x ′ 2 q ⋅⋅⋅ ⋅⋅⋅ ⋅⋅⋅ ⋅⋅⋅ x ′ p 1 x ′ p 2 ⋅⋅⋅ x ′ pq ] ;

[0134] Where, is the th feature point of the th sample in the power operation data and power status data; is the number of rows of the dataset ; is the number of columns of the dataset

[0135] Normalize the entire dataset to calculate the mean vector and the covariance matrix ; The mean vector is: ; where is the number of samples; is the number of samples; is the th feature point of the

[0136] The covariance matrix is:

[0137] ;

[0138] where is a vector with features; is transpose of;

[0139] Calculate the eigenvectors and eigenvalues of the dataset and select the first eigenvectors to form the transformation matrix ; Reduce the dimension of the dataset through the transformation matrix to obtain the dimension-reduced dataset ;

[0140] The method of matching the dimension-reduced operation feature dataset and the status feature dataset to ensure that the number of data types in the operation feature dataset and the status feature dataset is the same includes: presetting the number of data types in the dimension-reduced operation feature dataset to be ; the number of data types in the dimension-reduced status feature dataset is ; making the number of data types in the operation feature dataset and the status feature dataset consistent by constructing a data type matching model;

[0141] The type matching model is:

[0142] ;

[0143] where The number of data types to be discarded from the reduced operation feature dataset; The number of data types to be discarded from the reduced state feature dataset.

[0144] The method of performing correlation analysis on the operation feature dataset and the state feature dataset to obtain a comprehensive feature dataset includes:

[0145] Preset the operation feature dataset as ; where is the th feature of the operation feature data; is the number of feature types in the operation feature dataset;

[0146] Preset the state feature dataset as ; where is the th feature of the state feature data; is the number of feature types in the state feature dataset;

[0147] For each pair of features , calculate their mutual information ; where is the th feature of the operation feature and the state feature; measure the dependence relationship between two random feature variables and through the mutual information formula; the mutual information formula is: ;

[0148] where is a random feature variable in the operation feature dataset; is a random feature variable in the state feature dataset; and are the values of the random feature variables and ; is the joint probability distribution that the random feature variables and simultaneously take the values of and ; is the marginal probability distribution that the random feature variable takes the value of ; is the marginal probability distribution that the random feature variable takes the value of ;

[0149] Preset the mutual information threshold. When the mutual information value is greater than or equal to the mutual information threshold, it is determined that the feature pair has a significant correlation. Select the feature pairs with mutual information values greater than or equal to the threshold, and construct a significant correlation feature subset. When the mutual information is less than the mutual information threshold, it is determined that the feature pair does not have a significant correlation, and then the feature pair is removed. Fuse the selected significant correlation feature subsets to form a comprehensive feature dataset.

[0150] The training method of the power failure prediction model includes:

[0151] Use the comprehensive feature dataset and the corresponding output label power failure rate as the sample set. Use the sample set as the dataset of the model, and divide the dataset into a training set, a validation set, and a test set. Select a gradient boosting machine framework LightGBM suitable for regression tasks. Initialize the LightGBM parameters, where the LightGBM parameters include the maximum number of leaf nodes in each tree, the learning step size, the number of trees, and the maximum depth of the trees. Use the historical comprehensive feature dataset as the input data and the corresponding power failure rate index as the output label to train the power failure prediction model. The power failure prediction model is a gradient boosting machine model.

[0152] Use the training set data to train the model. During the training process, use the validation set for early stopping to prevent overfitting. Evaluate the model performance on the test set, and use the root mean square error as the loss function to measure the error of the model prediction.

[0153] The root mean square error loss function is:

[0154] ;

[0155] where is the number of the sample set; is the actual value of the th sample; is the th sample's predicted value;

[0156] Update the model LightGBM parameters through the backpropagation algorithm to minimize the loss function. Use the validation set to evaluate the performance of the power failure prediction model, and calculate the precision index to measure the performance of the model.

[0157] Select SGD as the optimizer to find the parameter combination that minimizes the loss function. According to the performance feedback of the validation set precision, adjust the structure of the model and the LightGBM parameters to optimize the model. Finally, evaluate the performance of the model in the prediction task through the test set. When the performance of the model in the prediction task reaches the preset performance threshold, stop the test to obtain the power failure prediction model.

[0158] Methods for optimizing the model by adjusting the model structure and LightGBM parameters include:

[0159] Use Bayesian optimization to tune the LightGBM parameters in the power failure prediction model. The specific steps are as follows:

[0160] Determine the number of trees and the maximum depth parameter of the tree in the power failure prediction model that need to be tuned;

[0161] Prepare a set of initial data training sets, and evaluate the performance metrics of the model under the given parameter configuration by calculating the accuracy metric; The accuracy calculation formula is:

[0162] ;

[0163] Among them, is the power failure rate metric predicted correctly; is the total number of power failure rate metrics predicted;

[0164] Use the initial training set to establish a Gaussian process initial probability model between the parameter configuration and the performance metric; Start the iterative optimization process. Each iteration includes the following steps:

[0165] S61. According to the current probability model, use Gaussian process sampling to select the next parameter configuration;

[0166] S62. Use the selected parameter configuration to calculate its performance metric by calculating the accuracy;

[0167] S63. Add the new parameter configuration and performance metric to the training set;

[0168] S64. Update the probability model, and add the new training set to the model training process to update the mapping relationship between the parameters and the performance;

[0169] S65. Repeat steps S61 - S64 until the predetermined number of iterations is reached;

[0170] After the iterative optimization is completed, according to the results of the evaluation function, select the parameter configuration with the best performance metric as the final model configuration to obtain the trained power failure prediction model, and use the trained power failure prediction model to predict the current comprehensive feature data set to obtain the power failure rate metric.

[0171] The method of comparing the predicted power failure evaluation metric with the preset power failure evaluation metric threshold to determine whether the power supply fails includes:

[0172] If the predicted power failure rate metric is greater than or equal to the preset power failure rate metric threshold, it is determined that the power supply has failed;

[0173] If the predicted power supply failure rate index is less than the preset power supply failure rate index threshold, it is determined that the power supply has not failed.

[0174] The evaluation method of the power supply failure evaluation index includes:

[0175] Evaluating the power supply failure evaluation index through a power supply failure evaluation mathematical model,

[0176] ;

[0177] Wherein, is the power supply failure evaluation index; is the average value of the operating characteristic data set; is the average value of the power supply operation data obtained in real time; is the average value of the state characteristic data set; is the average value of the power supply state data obtained in real time; is the average value of the preset standard magnetic field characteristic data; is the average value of the magnetic field characteristic data obtained in real time; is the influence factor of the average value of the power supply operation data obtained in real time; is the influence factor of the average value of the power supply state data obtained in real time; is the influence factor of the average value of the magnetic field characteristic data obtained in real time.

[0178] For example, if the average value of the operating characteristic data set is 10, the average value of the power supply operation data obtained in real time is 11, the average value of the state characteristic data set is 8, the average value of the power supply state data obtained in real time is 9, the average value of the preset standard magnetic field characteristic data is 20, and the average value of the magnetic field characteristic data obtained in real time is 18, 、 and are 30%, 30% and 40% respectively, then the power supply failure evaluation index is approximately 2.7.

[0179] The method for detecting the specific failure location of the power supply through the collected magnetic field characteristic data includes:

[0180] The magnetic field characteristic data includes magnetic field distribution data, magnetic field change data and electromagnetic compatibility test data;

[0181] Analyzing the magnetic field characteristic data through the Biot-Savart model, and the Biot-Savart model is:

[0182] ;

[0183] Wherein, is the tiny magnetic field generated by the current element at the observation point ; is the magnetic permeability of vacuum, and its value is approximately ; is the current element; is the magnitude of the current; is the infinitesimal length vector in the current direction; is the unit vector pointing from the current element to the observation point ; is the distance from the current element to the observation point ;

[0184] Analysis of magnetic field distribution data: By measuring the magnetic field intensity at multiple points around the power supply system, a magnetic field distribution map is obtained; for a specific region, the spatial distribution of the magnetic field intensity is represented by integrating the total magnetic field obtained from the Biot - Savart model;

[0185] Use the Biot - Savart model to calculate the magnetic field distribution of the power supply system under normal operating conditions, and compare it with the magnetic field distribution of the actually measured power supply system to identify the abnormal magnetic field distribution in the power supply system;

[0186] Analysis of magnetic field change data: Regularly measure the magnetic field intensity at the same observation point of the power supply part, and record the change of the magnetic field over time; a rapid or irregular magnetic field change indicates an abnormal current, and the current is also changing rapidly. Infer the specific location of the power supply system fault through the position of the magnetic field change;

[0187] Analysis of electromagnetic compatibility test data:

[0188] Use a spectrum analyzer and an electromagnetic interference receiver device to conduct a comprehensive electromagnetic compatibility test on the power supply system; the test contents include radiation interference, conducted interference, and anti - interference ability;

[0189] Record the test environmental conditions, equipment configuration, and the operating state of the power supply system; environmental conditions include temperature, humidity, and background noise level; equipment configuration includes receiver settings, antenna type and position; the operating state of the power supply system includes operating frequency and load conditions;

[0190] Through spectrum analysis software, analyze the test results to identify the abnormal signal frequencies and intensities that deviate from the normal operating mode; analyze the source and generation mechanism of the abnormal signals, and determine whether they cause functional failures or performance degradation;

[0191] Compare the current test results with the preset test result threshold, evaluate the change trend of the electromagnetic interference level, and identify the faults of the power supply;

[0192] Using a directional antenna and a near-field probe positioning tool, combined with the spectrum analysis results, gradually narrow down the search range and accurately locate the position of the interference source; for the power supply system, use time-domain reflectometry to improve the positioning accuracy; identify the coupling mechanism between the interference source and the power supply components affected, and find out the specific parts where the interference signal affects the power supply.

[0193] The preset power failure rate index threshold is set by the staff. Different power failure rate indexes are collected through the power management terminal, and the average value of multiple power failure rate indexes is taken as the preset power failure rate index threshold; similarly, the preset performance threshold is set.

[0194] In this embodiment, through the constructed operation-state correlation model and the preset condition trigger function, when the running condition trigger function reaches the influence coefficient threshold of the preset operation-state correlation model, the principal component analysis is automatically triggered to preprocess the power supply operation data and the power supply state data, and the dimension-reduced operation feature dataset and state feature dataset are matched to ensure that the number of data types in the operation feature dataset and the state feature dataset is the same, so as to facilitate the subsequent correlation analysis of the operation feature dataset and the state feature dataset by the mutual information method. This automated processing not only improves the system response speed but also reduces the need for human intervention and improves the data processing efficiency; at the same time, valuable correlation information is extracted from the operation feature dataset and the state feature dataset, providing strong support for fault prediction;

[0195] Through the analysis of the magnetic field distribution data by the Biot-Savart model, the abnormal magnetic field distribution in the power supply system can be effectively identified; through the analysis of the magnetic field change data, the current abnormality in the power supply system can be detected in a timely manner, thereby inferring the specific part of the system fault, which has important value for the maintenance and fault prevention of the power supply system; the analysis of the electromagnetic compatibility test data can help identify the possible electromagnetic interference problems in the power supply system. By analyzing the source and generation mechanism of the abnormal signal, the electromagnetic compatibility of the power supply system can be effectively improved, and the influence of interference on the system performance can be reduced; using advanced positioning tools such as directional antennas and near-field probes, combined with the spectrum analysis results, the position of the interference source can be located efficiently and accurately, which is of great significance for solving electromagnetic interference problems and ensuring the stable operation of the power supply system.

[0196] Embodiment 2

[0197] Please refer to Figure 2 As shown, the parts not described in detail in this embodiment refer to the description content of Embodiment 1. A power failure detection method is provided, including:

[0198] S1. Obtain power supply operation data, power supply state data, and magnetic field characteristic data;

[0199] S2. Preprocess the power operation data and power status data to obtain an operation feature data set and a status feature data set, perform a correlation analysis on the operation feature data set and the status feature data set, and obtain a comprehensive feature data set;

[0200] S3. Train a power failure prediction model, input the comprehensive feature data set into the power failure prediction model, and predict the power failure evaluation index;

[0201] S4. Compare the predicted power failure evaluation index with the preset power failure evaluation index threshold to determine whether the power supply has failed; if a failure occurs, automatically trigger a power-off protection mechanism and generate a failure alarm message;

[0202] S5. The power control terminal identifies the failure alarm message, detects the specific failure location of the power supply through the collected magnetic field feature data, sends a failure handling instruction to the operation and maintenance personnel, and processes the power failure location in a timely manner.

[0203] Since the electronic device introduced in this embodiment is the electronic device used to implement the power failure detection system and method in the embodiments of the present application, based on the power failure detection system and method introduced in the embodiments of the present application, those skilled in the art can understand the specific implementation manners and various variations of the electronic device in this embodiment. Therefore, the specific implementation of how this electronic device implements the method in the embodiments of the present application will not be described in detail here. As long as those skilled in the art implement the electronic device used in the power failure detection system and method in the embodiments of the present application, it belongs to the scope protected by the present application.

[0204] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain a formula that is closest to the actual situation. The preset parameters and threshold selection in the formulas are set by those skilled in the art according to the actual situation.

[0205] The above are only the preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the idea of the present invention belong to the protection scope of the present invention. It should be noted that for those ordinary technical users in the technical field, several improvements and refinements made without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

[0206] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the protection scope of the present invention.

Claims

1. A power failure detection system, characterized in that: include: Data monitoring module, used to obtain power supply operation data, power supply status data and magnetic field characteristic data; A data processing module is used to pre-process the power supply operation data and the power supply status data to obtain an operation feature data set and a status feature data set, and to perform correlation analysis on the operation feature data set and the status feature data set to obtain a comprehensive feature data set; A model training module is used to train a power failure prediction model, input a comprehensive feature data set into the power failure prediction model, and predict a power failure rate indicator; A fault detection module is used to compare the predicted power supply failure rate index with a preset power supply failure rate index threshold to determine whether the power supply fails; If a fault occurs, the power-off protection mechanism will be automatically triggered and a fault alarm message will be generated; The fault control module and the power control terminal identify the fault alarm information, detect the specific fault location of the power supply through the collected magnetic field characteristic data, send fault handling instructions to the operation and maintenance personnel, and handle the fault location of the power supply in time; Each module is connected by wired and / or wireless means to achieve data transmission between modules; The power supply operation data includes the power supply voltage data, current data, power data, temperature data, frequency data and harmonic wave data; the power supply status data includes the power supply remaining service life, aging and wear indicators, fault history data, environmental condition data and user operation behavior data; The method of preprocessing the power supply operation data and the power supply state data to obtain the operation characteristic data set and the state characteristic data set comprises: The acquired power operation data and power status data are cleaned, and the standard deviation of the cleaned power operation data and power status data is normalized to convert them into a standard normal distribution with a mean of 0 and a standard deviation of 1, thereby eliminating the dimensional influence between the data; Construct an operation-status association model, preset power operation data and power status data as independent variables of the operation-status association model, and a conditional trigger function as the dependent variable of the model; when the probability that the dependent variable of the operation-status association model is 1 reaches a conditional trigger threshold, the conditional trigger function is executed; A preset condition trigger function is used. When the running condition trigger function reaches the influence coefficient threshold of the preset operation-state correlation model, the principal component analysis is automatically triggered to pre-process the power supply operation data and power supply state data. The running feature data set and the state feature data set after dimensionality reduction are matched to ensure that the number of data types in the running feature data set and the state feature data set are consistent, and the processed running feature data set and the state feature data set are obtained.

2. The power failure detection system according to claim 1, characterized in that: The operation-state association model is: Where Y is the dependent variable of the operation-status association model; P(Y=1) is the probability that the dependent variable of the operation-status association model is 1; b0 is the value of the dependent variable when the independent variable of the operation-status association model is 0; X1, X2, ···, X n are the independent variables of the operation-state association model; b1, b2, ···, b n is the coefficient of the independent variable; n is the number of independent variable coefficients; n″ is the number of independent variables; The conditional trigger function is: Among them, x1 is the power supply operation data obtained in real time; is the preset standard power supply operation data; x2 is the real-time power supply status data; It is preset standard power status data; The method of automatically triggering principal component analysis to pre-process power supply operation data and power supply status data includes: It is preset that there is a data set O containing p samples and q features, and the data set O is composed of power supply operation data and power supply status data; the data set O is a matrix with p rows and q columns; the data set is: Among them, x′ pq is the qth feature point of the pth sample in the power operation data and the power status data; p is the number of rows in the data set O; q is the number of columns in the data set O; The data set O is standardized, and the mean vector μ and the covariance matrix ω are calculated; the mean vector μ is: Among them, p is the number of samples; u is the number of samples; x′ u is the feature point of the u-th sample; The covariance matrix ω is: Where (x′ u -μ) is a vector with q features; (x′ u -μ) T is (x′ u -μ) transpose; Calculate the eigenvectors and eigenvalues ​​of the data set O, select the first k eigenvectors to form the transformation matrix V; perform dimensionality reduction processing on the data set O through the transformation matrix V to obtain the reduced-dimensional data set Z = OV; The method of matching the reduced-dimensional operation feature data set with the state feature data set to ensure that the number of data types of the operation feature data set and the state feature data set are consistent includes: presetting the number of data types of the reduced-dimensional operation feature data set to I SF ; The number of data types in the state feature data set after dimensionality reduction is I SG ; The number of data types in the operation feature data set and the state feature data set are kept consistent by building a data type matching model; The type matching model is: Among them, I′ SF The number of data types that need to be discarded for the running feature data set after dimensionality reduction; I′ SG The number of data types that need to be discarded for the state feature data set after dimensionality reduction.

3. The power failure detection system according to claim 2, characterized in that: The method of performing correlation analysis on the operation feature data set and the state feature data set to obtain a comprehensive feature data set includes: The preset running feature data set is X′={x″1,x″2,···,x″ e }; where x″ e is the e-th feature of the running feature data; e is the feature type of the running feature data set; The preset state feature data set is Y′={y″1,y″2,···,y″ r }; where y″ r is the rth feature of the state feature data; r is the feature type of the state feature data set; For each pair of features (x″ i ,y″ i ), calculate its mutual information L(x″ i ,y″ i ); where i is the i-th feature of the running feature and the state feature; the mutual information formula is used to measure the dependency between two random feature variables X″ and Y″; the mutual information formula is: Where X″ is a random feature variable in the running feature data set; Y″ is a random feature variable in the state feature data set; x″ and y″ are the values ​​of the random feature variables X″ and Y″; p′(x″, y″) is the joint probability distribution of the random feature variables X″ and Y″ taking the values ​​of x″ and y″ at the same time; p′(x″) is the marginal probability distribution of the random feature variable X″ taking the value of x″; p′(y″) is the marginal probability distribution of the random feature variable Y″ taking the value of y″; A mutual information threshold is preset. When the mutual information value is greater than or equal to the mutual information threshold, the feature pair is judged to have a significant correlation. The feature pairs whose mutual information values ​​are greater than or equal to the threshold are screened out to construct a significant correlation feature subset. The screened significant correlation feature subsets are fused to form a comprehensive feature data set.

4. The power failure detection system according to claim 3, characterized in that: The training method of the power failure prediction model includes: The comprehensive feature data set and the corresponding output label power failure rate indicator are used as the sample set; the sample set is used as the data set of the model, and the data set is divided into a training set, a validation set and a test set; a gradient boosting machine framework LightGBM suitable for regression tasks is selected; the LightGBM parameters are initialized, and the LightGBM parameters include the maximum number of leaf nodes in each tree, the learning step size, the number of trees and the maximum depth of the tree; the historical comprehensive feature data set is used as the input data, and the corresponding power failure rate indicator is used as the output label to train the power failure prediction model; the power failure prediction model is a gradient boosting machine model; Use the training set data to train the model. During the training process, use the validation set for early stopping to prevent overfitting. Evaluate the model performance on the test set, using the root mean square error as the loss function to measure the error of the model prediction. The root mean square error loss function is: Where n′ is the number of sample sets; y i is the actual value of the i-th sample; is the predicted value of the i-th sample; Update the LightGBM model parameters through the back-propagation algorithm to minimize the loss function; use the validation set to evaluate the performance of the power failure prediction model and calculate the accuracy index to measure the performance of the model; SGD is selected as the optimizer to find the parameter combination that minimizes the loss function. According to the performance feedback of the accuracy of the validation set, the model structure and LightGBM parameters are adjusted to tune the model. The performance of the model in the prediction task is finally evaluated through the test set. When the performance of the model in the prediction task reaches the preset performance threshold, the test is stopped to obtain the power failure prediction model.

5. The power failure detection system according to claim 4, characterized in that: The method of adjusting the structure of the model and the LightGBM parameters to optimize the model includes: Use Bayesian optimization to tune the LightGBM parameters in the power failure prediction model. Specific steps: Determine the number of trees and the maximum depth of the trees in the power failure prediction model that need to be tuned; Prepare a set of initial data training sets and evaluate the performance indicators of the model under given parameter configuration by calculating the accuracy index; the accuracy calculation formula is: Among them, qj is the correctly predicted power failure rate index; sk is the total number of predicted power failure rate indexes; Using the initial training set, an initial Gaussian process probability model between parameter configuration and performance indicators is established; the iterative optimization process begins, and each iteration includes the following steps: S61. According to the current probability model, use Gaussian process sampling to select the next parameter configuration; S62, using the selected parameter configuration, calculating its performance index by calculating the accuracy; S63, adding new parameter configuration and performance indicators to the training set; S64, updating the probability model, adding the new training set to the model training process to update the mapping relationship between parameters and performance; S65, repeating steps S61-S64 until a predetermined number of iterations is reached; After the iterative optimization is completed, according to the result of the evaluation function, the parameter configuration with the best performance index is selected as the final model configuration to obtain a trained power supply fault prediction model. The trained power supply fault prediction model is used to predict the current comprehensive feature data set to obtain a power supply failure rate index.

6. The power failure detection system according to claim 5, characterized in that: The method of comparing the predicted power failure rate index with a preset power failure rate index threshold to determine whether a power failure occurs includes: If the predicted power failure rate index is greater than or equal to the preset power failure rate index threshold, it is determined that the power supply fails; If the predicted power supply failure rate index is less than a preset power supply failure rate index threshold, it is determined that the power supply has not failed.

7. The power failure detection system according to claim 6, characterized in that: The evaluation method of the power failure rate indicator includes: The power supply failure rate index is evaluated through the power supply failure rate mathematical model. The mathematical model of power failure rate is: Among them, ZSM is the power failure rate indicator; is the average value of the running feature data set; It is the average value of power supply operation data obtained in real time; is the average value of the state feature data set; is the average value of the power status data obtained in real time; is the average value of the preset standard magnetic field characteristic data; is the average value of the magnetic field characteristic data obtained in real time; α1, α2 and α3 are the influencing factors of the power supply failure rate index.

8. The power failure detection system according to claim 7, characterized in that: The method for detecting the specific fault location of the power supply by collecting magnetic field characteristic data includes: Magnetic field characteristic data include magnetic field distribution data, magnetic field change data and electromagnetic compatibility test data; The magnetic field characteristic data is analyzed by the Biot-Savart model, which is: Wherein, dB is the tiny magnetic field generated by the current element at the observation point U; μ0 is the magnetic permeability of vacuum, which is 4π×10 -7 N / A 2 ; Ids is the current element; I is the magnitude of the current; ds is the small length vector of the current direction; is the unit vector pointing from the current element to the observation point U; r 2 is the distance from the current element to the observation point U; Analysis of magnetic field distribution data: The magnetic field strength is measured at multiple points around the power system to obtain a magnetic field distribution map; for a specific area, the spatial distribution of the magnetic field strength is represented by the total magnetic field obtained by integrating the Biot-Savart model; The Biot-Savart model is used to calculate the magnetic field distribution of the power system under normal working conditions, and compared with the actual measured magnetic field distribution of the power system to identify abnormal magnetic field distribution in the power system. Analysis of magnetic field change data: regularly measure the magnetic field strength at the same observation point of the power supply, and record the change of the magnetic field over time; rapid or irregular magnetic field changes indicate abnormal current, and the current is also changing rapidly. The specific location of the power supply system fault can be inferred from the location of the magnetic field change; Analysis of electromagnetic compatibility test data: Use spectrum analyzer and electromagnetic interference receiver equipment to conduct comprehensive electromagnetic compatibility test on the power supply system; the test content includes radiation interference, conducted interference and anti-interference; Record the test's environmental conditions, equipment configuration, and power system operating status; environmental conditions include temperature, humidity, and background noise level; equipment configuration includes receiver settings, antenna type, and location; power system operating status includes operating frequency and load conditions; Analyze the test results through spectrum analysis software to identify the frequency and strength of abnormal signals that deviate from the normal working mode; analyze the source and generation mechanism of abnormal signals to determine whether they cause functional failure or performance degradation; Compare the current test results with the preset test result thresholds to evaluate the changing trend of the electromagnetic interference level and identify power supply failures; Directional antennas and near-field probe positioning tools are used, combined with spectrum analysis results, to gradually narrow the search range and accurately locate the interference source. For the power supply system, time domain reflection measurement is used to improve the accuracy of positioning. The coupling mechanism between the interference source and the interfered power supply components is identified to find out the specific parts of the power supply where the interference signal affects the interference signal.

9. A power failure detection method, applied to the power failure detection system according to any one of claims 1 to 8, characterized in that: include: S1. Obtain power supply operation data, power supply status data and magnetic field characteristic data; S2, preprocessing the power supply operation data and the power supply status data to obtain an operation feature data set and a status feature data set, and performing correlation analysis on the operation feature data set and the status feature data set to obtain a comprehensive feature data set; S3, training a power supply failure prediction model, inputting the comprehensive feature data set into the power supply failure prediction model, and predicting a power supply failure rate indicator; S4. Compare the predicted power failure rate index with the preset power failure rate index threshold to determine whether the power supply fails; if a failure occurs, automatically trigger the power-off protection mechanism and generate a fault alarm message; S5. The power control terminal identifies the fault alarm information, detects the specific fault location of the power supply through the collected magnetic field characteristic data, sends fault handling instructions to the operation and maintenance personnel, and handles the fault location of the power supply in a timely manner.

Images