Fault key information identification method and system of transformer substation, and storage medium

Abstract

The invention discloses a transformer substation fault key information identification method and system and a storage medium, so as to assist the fault diagnosis of a transformer substation, and the method comprises the steps: S1, obtaining the equipment fault data of the transformer substation, and carrying out the preprocessing of the equipment fault data; s2, constructing a fault matrix based on the preprocessed equipment fault data, and carrying out equalization processing on the fault matrix to obtain a fault mean matrix; s3, calculating a covariance matrix, characteristic values and characteristic vectors corresponding to the characteristic values based on the fault mean value matrix; s4, sorting the feature values and the feature vectors, and intercepting part of the feature vectors according to a preset threshold value to form a feature matrix; s5, constructing a dimension reduction matrix based on the feature matrix and the mean value matrix, and calculating the contribution rate of each component in the dimension reduction matrix; and S6, according to a preset contribution rate threshold value, determining equipment fault characteristics of the transformer substation, thereby filtering redundant information in the fault key information, and screening out effective key information, so that a fault diagnosis result is more accurate.

Description

Technical Field

[0001] The present application relates to the technical field of power system protection, and in particular to a method, system and storage medium for identifying critical fault information of a substation. Background Art

[0002] In recent years, high voltage direct current (HVDC) technology has achieved breakthrough development and has been widely used in the operation of power systems. As a very novel transmission technology, HVDC technology occupies an extremely important position in the new power system because of its unique advantages. Compared with traditional high voltage AC transmission technology, high voltage direct current transmission technology has many advantages such as low line cost, large transmission capacity, low loss, ability to complete asynchronous networking and limit short-circuit current of power system. Therefore, it is widely used in long-distance, high-power, asynchronous networking transmission occasions. The HVDC system is a very large and complex system. Various faults are inevitable during operation. When the system fails, it is necessary to identify the fault in time and organize personnel to repair it to avoid the bad situation of the entire system being shut down.

[0003] Existing fault diagnosis methods easily produce a lot of complex and redundant information, which affects the identification of key fault information and makes it difficult to effectively screen out valid information and then carry out targeted processing and solutions. Summary of the invention

[0004] In response to the defects existing in the prior art, the present application provides a method, system and storage medium for identifying key fault information of a substation to solve the problem that the existing fault diagnosis methods are difficult to identify key fault information, difficult to effectively screen out valid information, and then specifically handle and solve the fault.

[0005] To achieve the above objectives, the present application provides a method for identifying key fault information of a substation, comprising the following steps:

[0006] S1. Obtain equipment failure data of the substation and pre-process the equipment failure data;

[0007] S2. constructing a fault matrix based on the preprocessed equipment fault data, and performing mean processing on the fault matrix to obtain a fault mean matrix;

[0008] S3. Calculate a covariance matrix, eigenvalues, and eigenvectors corresponding to the eigenvalues ​​based on the fault mean matrix;

[0009] S4, sorting the eigenvalues ​​and the eigenvectors, and intercepting part of the eigenvectors according to a preset threshold value to form a feature matrix;

[0010] S5. Based on the feature matrix and the mean matrix, construct a dimension reduction matrix, and calculate the contribution rate of each component in the dimension reduction matrix;

[0011] S6. Determine equipment failure characteristics of the substation according to a preset contribution rate threshold.

[0012] Furthermore, the equipment fault data includes switch status, protection action signal, timing signal, inspection auxiliary equipment data, online detection signal and main equipment information.

[0013] Furthermore, S1, the step of obtaining equipment fault data of the substation and preprocessing the equipment fault data specifically includes:

[0014] Set the data collection time window;

[0015] When the delay time after the equipment failure of the substation meets the failure start threshold, the equipment failure data is acquired based on the data collection time window, and the equipment failure data is grouped;

[0016] The device fault data after being grouped is corrected.

[0017] Furthermore, the data collection time window is a scalable and adjustable time window.

[0018] Furthermore, grouping the equipment fault data includes:

[0019] Screening and filtering the alarm events within the multiple data collection time windows;

[0020] An alarm event group is formed based on each filtered alarm event;

[0021] If the events in the alarm event group are related, they are grouped into the same group; otherwise, they are grouped into different groups.

[0022] Furthermore, the step of correcting the grouped device fault data specifically includes:

[0023] Numbering different groups in the alarm event group, and obtaining a protection action time correction value for each group;

[0024] The time of the main and auxiliary protection devices of each group is corrected in turn until the correction of all groups is completed.

[0025] Furthermore, the feature vector and the preset threshold satisfy the following relationship:

[0026]

[0027] Among them, λi is the eigenvalue of the covariance matrix, i is the number of eigenvalues ​​of the covariance matrix, k is the number of eigenvectors of the intercepted part, n is the number of fault features, and t is a preset threshold.

[0028] Furthermore, the contribution rate and eigenvalue of each component in the dimension reduction matrix satisfy the following relationship:

[0029]

[0030] Among them, λ i is the eigenvalue of the covariance matrix, i is the number of eigenvalues ​​of the covariance matrix, n is the number of fault features, y i is the contribution rate of the i-th fault feature.

[0031] The present application also provides a fault critical information identification system for a substation, the fault critical information identification system comprising:

[0032] An acquisition module, used for acquiring equipment fault data of the substation and preprocessing the equipment fault data;

[0033] A fault matrix construction module is used to construct a fault matrix for the preprocessed equipment fault data, and perform mean value processing to obtain a fault mean value matrix;

[0034] A calculation and analysis module, used to calculate a covariance matrix, eigenvalues ​​and eigenvectors corresponding to the eigenvalues ​​based on the fault mean matrix, sort the eigenvalues ​​and the eigenvectors, and intercept part of the eigenvectors according to a preset threshold to form a characteristic matrix;

[0035] The fault feature determination module is used to construct a dimension reduction matrix based on the feature matrix and the mean matrix, calculate the contribution rate, and determine the equipment fault features of the substation according to a preset contribution rate threshold.

[0036] The present application also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, a fault diagnosis method for a substation as described in any one of the above items is implemented.

[0037] Compared with the prior art, the advantages of this application are:

[0038] The present application constructs a fault matrix, performs mean processing on the fault matrix, calculates the covariance matrix, eigenvalues ​​and eigenvectors corresponding to the eigenvalues, and constructs a feature matrix. Based on the feature matrix and the mean matrix, a dimensionality reduction matrix is ​​constructed, and the contribution rate of each component in the dimensionality reduction matrix is ​​calculated, thereby filtering out redundant information in the key fault information and screening out effective key information, which can make the fault diagnosis result more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0040] Figure 1 A schematic diagram of a method for identifying critical fault information of a substation according to an embodiment of the present application;

[0041] Figure 2 A schematic diagram of a scalable and adjustable time window according to an embodiment of the present application;

[0042] Figure 3 A flowchart of equipment fault data correction according to an embodiment of the present application;

[0043] Figure 4 A flowchart of a PCA process according to an embodiment of the present application; and

[0044] Figure 5 This is a system block diagram of a fault critical information identification system for a substation according to an embodiment of the present application.

[0045] Description of reference numerals:

[0046] 100, acquisition module; 200, fault matrix construction module; 300, calculation and analysis module; 400, fault feature determination module. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0048] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0049] In the description of the present application, the terms "first / second" involved are merely used to distinguish similar objects and do not represent a specific order for the objects. It can be understood that "first / second" can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0050] It should be understood that the orientation or position relationship is based on the orientation or position relationship shown in the drawings. These orientation terms are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present application.

[0051] Power system fault diagnosis refers to the use of knowledge of the power system and its protection devices and information such as relay protection to accurately identify the location, type and malfunction of faulty components, thereby laying the foundation for taking measures to repair and improve existing deficiencies and improving the operation level of the power system.

[0052] In order to help the operation and maintenance personnel in the substation quickly judge and locate the cause of the fault, the automatic identification technology of the key information of the equipment fault is studied. The key to the identification of key fault information is feature extraction. The most relevant features are extracted from the fault phenomenon, and quantitative and qualitative analysis is performed on them to establish a feature library, which involves various disciplines such as electronic technology, statistical mathematics, modern control theory, signal processing and pattern recognition, computer science, artificial intelligence, etc.

[0053] According to the differences in fault feature extraction sources and manifestations, the existing fault key information identification technologies can be divided into the following three categories: mathematical model-based methods, empirical knowledge-based methods, and data-driven methods.

[0054] The method based on mathematical model is to understand the variation law of equipment parameters under different working conditions and the range that should be achieved in theory by establishing an accurate mathematical model of the equipment. By understanding the relationship between various abnormal conditions in the process and the model parameters and states, it can be used to judge the real-time operating status of the equipment.

[0055] The method based on empirical knowledge mainly refers to the expert system method, which refers to a program system with expert-level problem-solving capabilities in a specific field. It uses a large number of rules to express the experience and professional knowledge accumulated by experts over the years, and gradually uses rules for reasoning to determine the operating status of equipment. It is an intelligent computer program.

[0056] Data-driven methods are more capable of fully extracting the fault signs, causal logic and other relationships implicit in the data. They do not need to establish precise mathematical models and require less prior knowledge. They can be divided into three categories: methods based on statistical analysis, signal analysis and artificial intelligence.

[0057] Statistical analysis mainly includes PCA principal component analysis, Fisher analysis discriminant method (FDA) and PLS partial least squares method. PCA principal component analysis is a commonly used feature extraction method that can eliminate data correlation and reduce data dimension. Fisher analysis discriminant method (FDA) and PCA have some similarities. They are both linear dimensionality reduction methods that can project high-dimensional training data into a few vector directions to characterize the system operation characteristics in the data. However, FDA takes into account the category information of the sample, so it can be used to identify the category to which the individual belongs. In addition, partial least squares (PLS) is a multivariate correction method based on factor analysis, and its mathematical basis is similar to PCA. However, PLS goes a step further than PCA. It introduces a new matrix Y, which contains the fault category information corresponding to the sample matrix X. Unlike PCA, the PLS method simultaneously performs principal component decomposition on the fault category matrix Y and the sample matrix X, and maximizes their covariance.

[0058] Signal analysis is often more suitable for the identification of shaft rubbing faults, and mainly includes wavelet transform, Hilbert-Huang transform (HHT) and spectral analysis. Wavelet transform belongs to time-frequency analysis. Compared with Fourier transform, it is more suitable for analyzing non-stationary signals because it can provide time and frequency information at the same time without losing time or frequency resolution. HHT transform is mainly used to analyze nonlinear and non-stationary signals, especially when the signal frequency changes with time. Spectral analysis is used to analyze the frequency domain characteristics of signals. Its basic idea is to convert the signal from the time domain (time domain) to the frequency domain (frequency domain) in order to better understand the frequency components and spectral structure of the signal. Spectral analysis can help identify the periodicity or frequency components in the signal and reveal the implicit laws and characteristics in the signal.

[0059] The AI-based methods mainly utilize the research results in the field of AI, including fuzzy logic, neural networks and other methods. Some commonly used methods for feature extraction include autoencoders and convolutional neural networks. Autoencoders are an artificial neural network that is often used for feature extraction and data reconstruction. The basic idea is to train the network to learn how to encode and decode the input data, thereby achieving feature extraction. Convolutional neural networks are a type of neural network specifically designed for processing grid data (such as images). It extracts the spatial features of the input data through convolutional layers and pooling layers, and then performs classification or regression tasks through fully connected layers and activation functions. During the training process, CNN will automatically learn the features in the data, thereby achieving the purpose of feature extraction.

[0060] Figure 1 A schematic diagram of a method for identifying critical fault information of a substation according to an embodiment of the present application; Figure 2 A schematic diagram of a scalable and adjustable time window according to an embodiment of the present application; Figure 3 A flowchart of equipment fault data correction according to an embodiment of the present application; Figure 4 Flow chart of a PCA process according to an embodiment of the present application. Figure 1-Figure 4 As shown, the embodiment of the present application provides a method for identifying critical fault information of a substation, comprising the following steps:

[0061] S1. Obtain equipment failure data of the substation and pre-process the equipment failure data;

[0062] S2. construct a fault matrix based on the preprocessed equipment fault data, and perform mean processing on the fault matrix to obtain a fault mean matrix;

[0063] S3. Calculate the covariance matrix, eigenvalues ​​and eigenvectors corresponding to the eigenvalues ​​based on the fault mean matrix;

[0064] S4, sorting the eigenvalues ​​and eigenvectors, and intercepting part of the eigenvectors according to a preset threshold value to form a feature matrix;

[0065] S5. Based on the feature matrix and the mean matrix, a dimension reduction matrix is ​​constructed, and the contribution rate of each component in the dimension reduction matrix is ​​calculated;

[0066] S6. Determine the equipment failure characteristics of the substation according to the preset contribution rate threshold.

[0067] Specifically, the fault data of various equipment in the substation, such as current, voltage, power, temperature, etc., are collected through sensors, protection devices and other monitoring equipment. Then the collected fault data is preprocessed, including filtering, denoising, normalization and other operations to eliminate interference and outliers in the data and improve the accuracy of fault diagnosis. In step S2, a fault matrix is ​​constructed based on the preprocessed equipment fault data. For example, assuming that there are m types of equipment in the substation and each type of equipment has n fault characteristics, the fault matrix is ​​an m×n matrix, and the fault matrix is ​​averaged to obtain a fault mean matrix. For example, the average processing method can be to calculate the row mean or column mean of the matrix, and then replace each element with the mean of the corresponding row or column. In particular, the equipment fault data includes switch status, protection action signal, timing signal, inspection auxiliary equipment data, online detection signal and main equipment information.

[0068] The covariance matrix is ​​calculated based on the fault mean matrix. The covariance matrix can reflect the correlation between the fault features. The eigenvalues ​​and the eigenvectors corresponding to the eigenvalues ​​are calculated, which can be obtained by solving the characteristic equation of the covariance matrix. The eigenvalues ​​and eigenvectors respectively represent the importance and direction of the fault features. In step S4, the eigenvalues ​​and eigenvectors are sorted, and part of the eigenvectors are intercepted according to the preset threshold to form a feature matrix. The sorting can be performed according to the size of the eigenvalues. The larger the eigenvalue, the more important the corresponding eigenvector. Intercepting part of the eigenvector can reduce the dimension of the feature matrix and reduce the complexity of fault diagnosis. Based on the feature matrix and the mean matrix, a dimensionality reduction matrix is ​​constructed. For example, the construction method of the dimensionality reduction matrix can be a linear dimensionality reduction method, such as principal component analysis (PCA), linear discriminant analysis (LDA), etc. The contribution rate of each component in the dimensionality reduction matrix is ​​calculated, which can be obtained by analyzing the relationship between the dimensionality reduction matrix and the feature matrix. The contribution rate can represent the importance of each component in fault diagnosis.

[0069] In step S6, the equipment fault characteristics of the substation are determined according to the preset contribution rate threshold. The contribution rate threshold can be set according to actual application requirements. The larger the threshold, the more obvious the fault characteristics are. By determining the equipment fault characteristics, fault type identification, fault location location and other fault diagnosis tasks can be performed.

[0070] It should be noted that the use of fault matrix and feature matrix can effectively represent the fault characteristics of the equipment. At the same time, through dimension reduction matrix and contribution rate analysis, the complexity of fault diagnosis can be reduced, the accuracy of diagnosis can be improved, the contribution rate threshold can be set according to actual needs, and the fault diagnosis results can be flexibly adjusted, thereby filtering out redundant information in the key fault information and screening out effective information, making the fault diagnosis results more accurate.

[0071] In some embodiments, S1, the step of obtaining equipment fault data of a substation and preprocessing the equipment fault data specifically includes:

[0072] Set the data collection time window;

[0073] When the delay time after the equipment failure of the substation meets the failure start threshold, the equipment failure data is obtained based on the data collection time window, and the equipment failure data is grouped;

[0074] Correct the equipment failure data after grouping.

[0075] Specifically, a data collection time window is set. For example, the data collection time window can be set to within 5 minutes after the fault occurs. When a substation device fails and meets the fault start threshold, the device fault data is obtained based on the data collection time window. In particular, the data collection time window is a scalable and adjustable time window (see Figure 2 ), for the data preprocessing part, the method of time correction of the acquisition time window is adopted. The first step is to select the acquisition window, because the size of the acquisition window involves the grouping of fault alarms. In order to avoid some errors in the identification results caused by using a fixed acquisition window that is too small or too large, a scalable time window is used, such as Figure 2 As shown in the figure, the scalable time window is equal to the minimum fixed time window (default 10 seconds) plus the floating time window. t1 represents the moment when the delay time after the equipment failure meets the failure start threshold. Then a fixed time window is collected until t2, and then it is the floating collection time. The floating collection time is calculated as follows: if Δalarm / Δt (the number of alarms in the t time interval) is greater than a certain threshold, the floating time window is extended, otherwise, the floating time window stops.

[0076] The switch status, protection action signal, timing signal and other data are obtained through the protection device, monitoring system and the like, and the acquired equipment fault data are grouped. For example, the data can be grouped according to the equipment type, fault type and the like. The grouped equipment fault data are corrected, and other pre-processing operations are performed on the corrected equipment fault data, and the pre-processed equipment fault data are used in subsequent fault diagnosis steps.

[0077] In some embodiments, grouping the device failure data includes:

[0078] Screen and filter alarm events within multiple data collection time windows;

[0079] An alarm event group is formed based on each filtered alarm event;

[0080] If the events in the alarm event group are related, they are divided into the same group, otherwise, they are divided into different groups.

[0081] Specifically, certain screening conditions are set, such as the severity and occurrence time of the alarm event, to screen out valid alarm events, and an alarm event group is formed based on each filtered alarm event. For example, alarm events occurring in the same device, the same fault type or the same time window can be classified into an alarm event group, and the correlation between them is determined by calculating indicators such as the similarity and correlation coefficient between the alarm events. If the events in the alarm event group are correlated, they are divided into the same group. For example, if multiple alarm events are caused by the same fault cause, they can be divided into the same group. If the events in the alarm event group are not correlated, they are divided into different groups. For example, if multiple alarm events are caused by different fault causes, they can be divided into different groups, and the divided alarm event groups are used for subsequent fault diagnosis steps. For example, each alarm event group can be analyzed to determine the fault type and fault cause.

[0082] In some embodiments, the step of correcting the grouped device fault data specifically includes:

[0083] Number different groups in the alarm event group and obtain the protection action time correction value of each group;

[0084] The time of the main and auxiliary protection devices of each group is corrected in turn until the correction of all groups is completed.

[0085] Specifically, different groups in the alarm event group are numbered to distinguish and manage each alarm event group. By analyzing the protection action time in each alarm event group, the protection action time correction value of each group is determined. The main and auxiliary protection device time of each group is corrected in turn. For example, the main protection device time of the first group can be corrected first, and then the auxiliary protection device time of the first group can be corrected, and so on. During the correction process, certain correction algorithms can be used, such as least squares method, Kalman filtering, etc., to improve the accuracy of time correction. After the correction is completed, the protection action time of each group is updated to obtain the corrected protection action time.

[0086] In some embodiments, the feature vector and the preset threshold satisfy the following relationship:

[0087]

[0088] Among them, λ i is the eigenvalue of the covariance matrix, i is the number of eigenvalues ​​of the covariance matrix, k is the number of eigenvectors of the intercepted part, n is the number of fault features, and t is the preset threshold.

[0089] The contribution rate and eigenvalue of each component in the dimension reduction matrix satisfy the following relationship:

[0090]

[0091] Among them, λ i is the eigenvalue of the covariance matrix, i is the number of eigenvalues ​​of the covariance matrix, n is the number of fault features, y i is the contribution rate of the i-th fault feature.

[0092] Specifically, the PCA method is applied to the substation, and the substation sample fault data of m substations with n fault characteristics are composed into a matrix X mn , then average the matrix X mn , we can get Then, based on the mean-valued matrix X, the covariance matrix C is calculated: Calculate the eigenvalue λ of the covariance matrix C i and the eigenvector ω i , select the eigenvectors ω1, ω2, ..., ω corresponding to the k largest eigenvalues k , the matrix W is formed by rows from top to bottom according to the corresponding eigenvalue size kn , the eigenvalue selection method refers to the following relationship,

[0093]

[0094] The larger t is, the more eigenvalues ​​are retained, which can be selected according to the complexity of substation fault information. The matrix after dimensionality reduction can be obtained by the following formula: km =W kn X mn ,

[0095] Calculate the contribution rate of all principal component vectors in turn, and the contribution rate of the principal component vector to the overall substation fault information y i for:

[0096]

[0097] All principal component vectors are sorted in descending order, and then the optimal fault characteristics of substation equipment are selected based on the principle that the cumulative contribution rate reaches 90%.

[0098] In order to better understand the method for identifying key fault information of a substation in the embodiment of the present application, Figures 2 to 4 Provide a detailed description.

[0099] The fault information identification technology adopted in the embodiment of the present application is designed in combination with specific application scenarios, and is aimed at extracting the features of key information of power equipment faults in substations. The main technical points used are the combination of scalable acquisition time windows and principal component analysis to extract fault features and establish key fault information screening rules, thereby providing support for the subsequent fault discrimination and analysis technology using deep learning.

[0100] Since most of the data in the data source are one-dimensional time series data, in the process of model design, it is necessary to extract data features more comprehensively and remove the redundant parts of the noise in the data, and pre-process the data. The acquisition time window correction method is adopted, and then the pre-processed data is subjected to feature extraction to obtain key information.

[0101] For the data preprocessing part, the method of time correction of the acquisition time window is adopted. The first step is to select the acquisition window, because the size of the acquisition window involves the grouping of fault alarms. In order to avoid some errors in the identification results caused by using a fixed acquisition window that is too small or too large, a scalable time window is adopted. The scalable time window is equal to the minimum fixed time window (default 10 seconds) plus the floating time window. t1 represents the moment when the delay time after the equipment failure meets the fault start threshold, and then a fixed time window is collected to reach the moment t2, and then it is a floating acquisition time. The floating acquisition time is calculated in the following way: if Δalarm / Δt (the number of alarms in the time interval t) is greater than a certain threshold, the floating time window is extended, otherwise, the floating time window stops.

[0102] Since the substation system is likely to have multiple related or unrelated faults in a short period of time, different events of the same fault may have different sources; different events of the same fault arrive at the master station in different orders, so it is obviously impossible to guarantee the strict accuracy of the results by using only one or several time windows. Therefore, a time window group is designed here. The so-called "group" refers to the remaining events after screening and filtering of all alarm events collected during this collection period, which form a "group". In this group, events with correlation are grouped into a "group" according to the model and real-time network topology. If there is a fault (single or complex) in the group, it forms a group. Otherwise, multiple groups are formed.

[0103] Establish a process for obtaining a unified time correction reference value for each group. Since the substation system may have multiple related or unrelated faults in a short period of time, the fault diagnosis module of the main station cannot use a certain time reference value to uniformly correct the time of all protection action events and switch action events SOE in this period. It is necessary to determine a unified time correction reference value for each group. If the startup criteria are met, start collecting data for this period, and then divide the collected data into groups and groups. The group numbers are named starting from 1, and then the correction values ​​of the i group time are obtained cyclically, and the time correction is also performed on the main and auxiliary protection devices of the i group. Then i points to the group number of the next group, and the above steps are repeated until the correction of all groups is completed.

[0104] The calculation based on the correction value is obtained by comprehensive comparison and analysis of three time values: the SOE time of the "start" event of the group of protection devices, the SOE time of the first action event of each protection device, and the reception time of each action event received by the master station.

[0105] For the feature extraction part, the PCA principal component analysis method is used, which is an unsupervised high-dimensional data dimensionality reduction method. It maximizes the variance of the original data by retaining multiple important components during the dimensionality reduction process, reducing the number of variables and making the results more interpretable. There are multiple influencing factors in the substation failure problem. Each type of failure and cause corresponds to a failure parameter. There is correlation and overlap between the parameters. Solving the substation failure problem requires extracting key data from these parameters.

[0106] The following case analysis is based on sample data of a power station equipment to reflect the characteristics of this application.

[0107] First, for data preprocessing technology, the time window group technology used can form a "group" of the remaining events after screening and filtering of all alarm events collected during the collection period. In this group, events with correlation are grouped into a "group" according to the model and real-time network topology, so that the data with correlation can be divided and analyzed together. For example, when a 500kV line fails, the line protection devices of the power stations on both sides are activated. Since they are associated with the same primary equipment, they are correlated. Since the primary equipment models associated with the switch protection devices on both sides are the switches on both sides, the switch and the line have connectivity or proximity according to the network topology relationship, so the switch protection devices on both sides are correlated with the line protection devices of the line.

[0108] For fault feature extraction, 8 types of representative data totaling 164 groups were selected, including 120 training samples and 44 test samples. The distribution of each sample data is shown in Table 1 below:

[0109] Table 1 Fault data

[0110] Fault number Substation status Training set Test Set D1 normal 21 8 D2 Circuit breaker tripping 18 7 D3 Control circuit disconnection fault 13 5 D4 Low oil (gas) pressure in the operating mechanism locks the switch-on and switch-off faults 17 5 D5 Hydraulic mechanism oil pump pressure timeout 16 7 D6 Circuit breaker arc chamber explosion 14 5 D7 Protection action circuit breaker refuses to open 12 4 D8 SF6 gas low pressure fault 9 3

[0111] The fault characteristics of all fault information are shown in Table 2:

[0112] Table 2 Fault characteristics table

[0113]

[0114]

[0115] By performing PCA principal component analysis on the fault feature group composed of all fault samples, the contribution rates of all principal component vectors are calculated in turn according to the corresponding formula and sorted in descending order. The results are shown in Table 3 below:

[0116] Table 3 Principal component contribution rate

[0117] PCA number Feature Number Contribution rate % Cumulative contribution rate% 1 A12 11.5906 11.5906 2 A7 9.7981 21.3887 3 A2 9.3085 30.6972 4 A3 8.8969 39.5941 5 A6 8.5567 48.1508 6 A1 8.3287 56.4795 7 A4 7.9624 64.4419 8 A14 7.7053 72.1472 9 A10 7.2627 79.4099 10 A13 6.8947 86.3046 11 A11 5.9493 92.2539 12 A5 2.3742 94.6281 13 A8 2.0847 96.7128 14 A9 1.7893 98.5021 15 A15 1.1227 99.6248

[0118] Finally, according to the 90% principle, the main fault features are screened out, and the fault features and types after dimension reduction are shown in Table 4 below:

[0119] Table 4 Fault characteristics and data types

[0120]

[0121]

[0122] By comparing the above analysis of the example data with the fault characteristics of the sample data, it is found that the fault key information identification technology of the embodiment of the present application can filter out redundant information and screen out effective key information.

[0123] Figure 5 A system block diagram of a fault diagnosis system for a substation according to an embodiment of the present application is shown in FIG. Figure 5As shown, the embodiment of the present application also provides a fault diagnosis system for a substation, and the fault diagnosis system includes an acquisition module 100, a fault matrix construction module 200, a calculation and analysis module 300, and a fault feature determination module 400. Among them, the acquisition module 100 is used to obtain the equipment fault data of the substation and preprocess the equipment fault data. The fault matrix construction module 200 is used to construct a fault matrix for the preprocessed equipment fault data and perform mean processing to obtain a fault mean matrix. The calculation and analysis module 300 is used to calculate the covariance matrix, the eigenvalues ​​and the eigenvectors corresponding to the eigenvalues ​​based on the fault mean matrix, and sort the eigenvalues ​​and the eigenvectors, and intercept part of the eigenvectors according to the preset threshold value to form a feature matrix. The fault feature determination module 400 is used to construct a dimensionality reduction matrix based on the feature matrix and the mean matrix, calculate the contribution rate, and determine the equipment fault characteristics of the substation according to the preset contribution rate threshold.

[0124] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned substation fault diagnosis method is implemented.

[0125] It should be noted that the storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a ROM (Read-Only Memory), a RAM (Random Access Memory), a magnetic disk or an optical disk.

[0126] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In a hardware implementation, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable storage medium, which may include a computer-readable storage medium (or a non-transitory medium) and a communication medium (or a temporary medium).

[0127] As is known to those of ordinary skill in the art, the term computer-readable storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer-readable storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is known to those of ordinary skill in the art that communication media typically contain computer-readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and may include any information delivery medium.

[0128] Exemplarily, the computer-readable storage medium may be an internal storage unit of the electronic device of the aforementioned embodiment, such as a hard disk or memory of the electronic device. The computer-readable storage medium may also be an external storage device of the electronic device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device.

[0129] The above is only a specific implementation of the present application, so that those skilled in the art can understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest range consistent with the principles and novel features applied for herein.

Claims

1. A method for identifying critical fault information of a substation, characterized in that: The steps include: S1. Obtain equipment failure data of the substation and pre-process the equipment failure data; S2. constructing a fault matrix based on the preprocessed equipment fault data, and performing mean processing on the fault matrix to obtain a fault mean matrix; S3. Calculate a covariance matrix, eigenvalues, and eigenvectors corresponding to the eigenvalues ​​based on the fault mean matrix; S4, sorting the eigenvalues ​​and the eigenvectors, and intercepting part of the eigenvectors according to a preset threshold value to form a feature matrix; S5. Based on the feature matrix and the mean matrix, construct a dimension reduction matrix, and calculate the contribution rate of each component in the dimension reduction matrix; S6. Determine the equipment failure characteristics of the substation according to a preset contribution rate threshold.

2. The fault diagnosis method for a substation according to claim 1, characterized in that: The equipment fault data includes switch status, protection action signal, timing signal, inspection auxiliary equipment data, online detection signal and main equipment information.

3. The method for identifying critical fault information of a substation according to claim 1, characterized in that: S1, the step of obtaining equipment failure data of the substation and preprocessing the equipment failure data specifically includes: Set the data collection time window; When the delay time after the equipment failure of the substation meets the failure start threshold, the equipment failure data is acquired based on the data collection time window, and the equipment failure data is grouped; The device fault data after being grouped is corrected.

4. The method for identifying critical fault information of a substation according to claim 3, characterized in that: The data collection time window is a scalable and adjustable time window.

5. The method for identifying critical fault information of a substation according to claim 3, characterized in that: Grouping the equipment failure data includes: Screening and filtering the alarm events within the multiple data collection time windows; An alarm event group is formed based on each filtered alarm event; If the events in the alarm event group are related, they are grouped into the same group; otherwise, they are grouped into different groups.

6. The method for identifying critical fault information of a substation according to claim 5, characterized in that: The step of correcting the device fault data after grouping specifically includes: Numbering different groups in the alarm event group, and obtaining a protection action time correction value for each group; The time of the main and auxiliary protection devices of each group is corrected in turn until the correction of all groups is completed.

7. The method for identifying critical fault information of a substation according to claim 1, characterized in that: The feature vector and the preset threshold satisfy the following relationship: Among them, λ i is the eigenvalue of the covariance matrix, i is the number of eigenvalues ​​of the covariance matrix, k is the number of eigenvectors of the intercepted part, n is the number of fault features, and t is a preset threshold.

8. The method for identifying critical fault information of a substation according to claim 7, characterized in that: The contribution rate and eigenvalue of each component in the dimension reduction matrix satisfy the following relationship: Among them, λ i is the eigenvalue of the covariance matrix, i is the number of eigenvalues ​​of the covariance matrix, n is the number of fault features, y i is the contribution rate of the i-th fault feature.

9. A fault critical information identification system for a substation, characterized in that: The critical fault information identification system comprises: An acquisition module, used for acquiring equipment fault data of the substation and preprocessing the equipment fault data; A fault matrix construction module is used to construct a fault matrix for the preprocessed equipment fault data, and perform mean value processing to obtain a fault mean value matrix; A calculation and analysis module, used to calculate a covariance matrix, eigenvalues ​​and eigenvectors corresponding to the eigenvalues ​​based on the fault mean matrix, sort the eigenvalues ​​and the eigenvectors, and intercept part of the eigenvectors according to a preset threshold to form a characteristic matrix; The fault feature determination module is used to construct a dimension reduction matrix based on the feature matrix and the mean matrix, calculate the contribution rate, and determine the equipment fault features of the substation according to a preset contribution rate threshold.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for identifying critical fault information of a substation according to any one of claims 1 to 8 is implemented.

Images