Method, device, electronic device and storage medium for evaluating the operating status of a generator set

Abstract

The present application belongs to the field of electrical equipment detection technology, and in particular, relates to a method, device, electronic device and storage medium for evaluating the operating status of a generator set. The solution provided by the present application obtains n reference time period data streams of the generator set within a time window; obtains m data streams of the time period to be detected; and updates based on the n reference time period data streams and the m data streams of the time period to be detected to ensure that n data are always maintained. Then, the normalized score of the anomaly detection is determined through the anomaly detection model, and the operating status of the generator set is evaluated based on the normalized score. The present application captures short-term pattern changes to avoid the waste of computing resources caused by excessive data accumulation, and realizes accurate monitoring of the operating status of the generator set by mining the correlation of a large amount of complex and numerous generator-related operating data, and deducing whether the operating status of the unit is abnormal.

Description

Technical Field

[0001] The present application belongs to the technical field of electrical equipment detection, and in particular, relates to a method, device, electronic equipment and storage medium for evaluating the operating status of a generator set. Background Art

[0002] The generator sets of modern hydroelectric power plants are already equipped with many monitoring and surveillance systems. Monitoring the operating status of the generator sets of hydroelectric power plants plays a vital role. By monitoring the operating status of the generator sets, it is possible to promptly find out whether the equipment is in normal operation and whether the relevant parameters need to be modified. This helps to ensure that the generator sets can carry out normal production work within a certain time range, avoid large wear and tear caused by long-term overload operation, thereby extending the service life of the generator sets and maintaining a balance between production efficiency and equipment costs.

[0003] The operation of various monitoring systems of the generator sets of hydropower plants will generate a large number of complex time series data, and there are complex correlations between the data. In the process of monitoring the operating status, the number of data collection points is huge and the collection frequency is high, resulting in a huge amount of time series data, and there are complex correlations between the data. These data are basically only used for inspection and monitoring by operators. Each system is self-contained, and there is a lack of intelligent comprehensive analysis methods between the data. In addition, processing and analyzing a large amount of time series data requires a lot of computing resources, including CPU, memory, and disk I / O. This may lead to a shortage of computing resources and affect the normal operation of other business systems.

[0004] Therefore, how to monitor and correlate time series data in real time and reduce the complexity of calculations to achieve accurate monitoring of the operating status of generator sets has become an urgent problem that needs to be solved. Summary of the invention

[0005] The present application provides a method, device, electronic device and storage medium for evaluating the operating status of a generator set, which monitor and associate time series data in real time, reduce the complexity of calculations, and achieve accurate monitoring of the operating status of the generator set.

[0006] In a first aspect, an embodiment of the present application provides a method for evaluating the operating status of a generator set, the method comprising: obtaining n reference time period data streams of the generator set within a time window; based on the n reference time period data streams, determining a reference covariance matrix corresponding to the n reference time period data streams; obtaining m time period data streams to be detected, wherein m ≤ n; based on the reference covariance matrix, the n reference time period data streams and the m time period data streams to be detected, determining an updated data stream, wherein the updated data stream is: deleting the first m data of the reference time period data stream in the n reference time period data streams, and adding m time period data streams to be detected, and determining the covariance matrix to be detected according to the updated data stream; based on the reference covariance matrix and the covariance matrix to be detected, determining the eigenvalue decomposition results corresponding to the reference covariance matrix and the covariance matrix to be detected respectively through eigenvalue decomposition, wherein the eigenvalue decomposition results include The invention discloses a method for determining a normalized score of abnormality detection based on the eigenvalue decomposition results, comprising a characteristic matrix and a diagonal matrix corresponding to a benchmark covariance matrix and a characteristic matrix and a diagonal matrix corresponding to a covariance matrix to be detected; based on each of the eigenvalue decomposition results, the principal component matrix, the principal component score matrix and the eigenvalues ​​corresponding to the covariance matrix to be detected and the benchmark covariance matrix are determined through a dimensionality reduction process; based on each of the principal component matrices, each of the principal component score matrices, each of the eigenvalues, the covariance matrix to be detected and the benchmark covariance matrix, the normalized score of abnormality detection is determined through an abnormality detection model, and the normalized score is used to evaluate the operating status of the generator set.

[0007] In a possible implementation, the method of determining the principal component matrix, principal component score matrix and eigenvalues ​​corresponding to the covariance matrix to be detected and the benchmark covariance matrix through dimensionality reduction processing based on the eigenvalue decomposition results includes: determining the rank k corresponding to the benchmark covariance matrix and the covariance matrix to be detected through the proportion of principal component information based on the benchmark covariance matrix and the diagonal matrix corresponding to the covariance matrix to be detected; determining the principal component matrix and eigenvalues ​​corresponding to the benchmark covariance matrix and the covariance matrix to be detected respectively based on each k, each characteristic matrix and each diagonal matrix; determining the principal component score matrix corresponding to the benchmark covariance matrix and the covariance matrix to be detected based on each principal component matrix, the benchmark time period data stream and the updated data stream.

[0008] In a possible implementation, the anomaly detection model includes: a principal component change detection model, a reconstruction error detection model, a correlation deviation detection model and a statistical distribution estimation detection model; the determining of the normalized score of anomaly detection based on each of the principal component matrices, each of the principal component score matrices, each of the eigenvalues, the covariance matrix to be detected and the benchmark covariance matrix through the anomaly detection model includes: based on each of the eigenvalues, through the principal component change detection model, determining the normalized score of the principal component change detection; based on each of the principal component score matrices and each of the principal component matrices, through The reconstruction error detection model determines the normalized score of reconstruction error detection; based on the covariance matrix to be detected and the benchmark covariance matrix, the correlation deviation detection normalized score is determined through the correlation deviation detection model; based on each of the principal component score matrices, the statistical distribution estimation detection normalized score is determined through the statistical distribution estimation detection model; based on the principal component change detection normalized score, the reconstruction error detection normalized score, the correlation deviation detection normalized score and the statistical distribution estimation detection normalized score, the abnormality detection normalized score is determined through a preset formula.

[0009] In a possible implementation, after determining the normalized score of anomaly detection based on the principal component matrices, the principal component score matrices, the eigenvalues, the covariance matrix to be detected and the benchmark covariance matrix through an anomaly detection model, the method further includes: determining the operating state of the generator set based on the normalized score and a preset threshold; when the normalized score is greater than the preset threshold, determining that the operating state of the generator set is abnormal and that abnormal data exists in the covariance matrix to be detected; determining the source of the abnormality based on the principal component matrix corresponding to the benchmark covariance matrix and the principal component matrix corresponding to the covariance matrix to be detected.

[0010] In a possible implementation, determining the source of the abnormality based on the principal component matrix corresponding to the benchmark covariance matrix and the principal component matrix corresponding to the covariance matrix to be detected includes: determining a benchmark load matrix and a load matrix to be detected based on the principal component matrix corresponding to the benchmark covariance matrix and the principal component matrix corresponding to the covariance matrix to be detected; determining abnormal change elements based on benchmark elements in the benchmark load matrix and elements to be detected in the load matrix to be detected; and determining the source of the abnormality based on the abnormal change elements.

[0011] In a possible implementation, determining the source of the abnormality based on the principal component matrix corresponding to the benchmark covariance matrix and the principal component matrix corresponding to the covariance matrix to be detected also includes: determining a benchmark load matrix and a load matrix to be detected based on the principal component matrix corresponding to the benchmark covariance matrix and the principal component matrix corresponding to the covariance matrix to be detected; determining the projection change of the benchmark elements based on the benchmark elements in the benchmark load matrix, and determining the projection change of the elements to be detected based on the elements to be detected in the load matrix to be detected; determining the source of the abnormality based on the projection change of the benchmark elements and the projection change of the elements to be detected.

[0012] In a possible implementation, the preset formula is:

[0013]

[0014] Among them, S represents the normalized score, w1, w2, w3, w4 represent adjustable parameters, represents the normalized score of principal component change detection, represents the normalized score of reconstruction error detection, represents the normalized score of correlation deviation detection, Represents the statistical distribution estimated detection normalized score.

[0015] In a second aspect, an embodiment of the present application provides an evaluation device for the operating status of a generator set, the device comprising: an acquisition module for acquiring n reference time period data streams of the generator set within a time window; a determination module for determining a reference covariance matrix corresponding to the n reference time period data streams based on the n reference time period data streams; the acquisition module is also used to acquire m time period data streams to be detected, wherein m ≤ n; the determination module is also used to determine an updated data stream based on the reference covariance matrix, the n reference time period data streams and the m time period data streams to be detected, wherein the updated data stream is: deleting the first m data of the reference time period data stream in the n reference time period data streams, and adding m time period data streams to be detected, and determining the covariance matrix to be detected according to the updated data stream; the determination module is also used to determine the eigenvalue decomposition results corresponding to the reference covariance matrix and the covariance matrix to be detected respectively through eigenvalue decomposition based on the reference covariance matrix and the covariance matrix to be detected, wherein the eigenvalue decomposition results The results include characteristic matrices and diagonal matrices corresponding to the benchmark covariance matrix and characteristic matrices and diagonal matrices corresponding to the covariance matrix to be detected; the determination module is also used to determine the principal component matrix, principal component score matrix and eigenvalues ​​corresponding to the covariance matrix to be detected and the benchmark covariance matrix through dimensionality reduction processing based on each of the eigenvalue decomposition results; the detection module is used to determine the normalized score of anomaly detection through an anomaly detection model based on each of the principal component matrices, each of the principal component score matrices, each of the eigenvalues, the covariance matrix to be detected and the benchmark covariance matrix, and the normalized score is used to evaluate the operating status of the generator set.

[0016] In a third aspect, an embodiment of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method described in the first aspect or any one of the implementation methods thereof is implemented.

[0017] In a fourth aspect, an embodiment of the present application 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, the method described in the first aspect or any one of the implementation methods thereof is implemented.

[0018] In a fifth aspect, an embodiment of the present application provides a computer program product. When the computer program product runs on an electronic device, the electronic device executes the method described in the first aspect or any one of the implementation methods thereof.

[0019] The solution provided by the present application is to obtain n reference period data streams of the generator set within the time window; based on the n reference period data streams, determine the covariance matrix corresponding to the n reference period data streams; obtain m to-be-detected period data streams, where m ≤ n; based on the reference covariance matrix, n reference period data streams and m to-be-detected period data streams, determine the updated data stream, the updated data stream is: delete the first m data of the reference period data stream in the n reference period data streams, and add m to-be-detected period data streams, and determine the to-be-detected covariance matrix according to the updated data stream; the present application captures short-term pattern changes to avoid excessive data accumulation leading to waste of computing resources. The data of the time window is incrementally updated to ensure real-time performance and adapt to the dynamic changes of streaming data. Furthermore, based on the benchmark covariance matrix and the covariance matrix to be detected, the eigenvalue decomposition results corresponding to the benchmark covariance matrix and the covariance matrix to be detected are determined by eigenvalue decomposition, and the eigenvalue decomposition results include the characteristic matrix and diagonal matrix corresponding to the benchmark covariance matrix and the characteristic matrix and diagonal matrix corresponding to the covariance matrix to be detected; based on each eigenvalue decomposition result, the principal component matrix, principal component score matrix and eigenvalue corresponding to the covariance matrix to be detected and the benchmark covariance matrix are determined by dimensionality reduction processing, which significantly reduces the computational complexity and is suitable for high-dimensional and large-scale data monitoring. Further, through the anomaly detection model, the normalized score of anomaly detection is determined, and the operating status of the generator set is evaluated according to the normalized score. In this process, by mining the correlation of a large number of complex and numerous generator-related operating data, and deducing whether the operating status of the unit is abnormal, the source of the data anomaly can be accurately determined, and the operating status of the generator set can be accurately monitored. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 A schematic flow chart of a method for evaluating the operating status of a generator set provided in one embodiment of the present application;

[0022] Figure 2 A structural block diagram of a device for evaluating the operating status of a generator set provided in one embodiment of the present application;

[0023] Figure 3 A schematic diagram of the structure of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0024] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0025] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.

[0026] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0027] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0028] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0029] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0030] The generator sets of modern hydroelectric power plants are already equipped with many monitoring and surveillance systems, which will generate a large amount of complex and numerous generator-related operating data, and there are complex correlations between the data. However, these data are basically only used for inspection and monitoring by operators, and each system is self-contained, and there is a lack of intelligent comprehensive analysis methods between the data. Mining the correlation of these data, and judging whether the operating status of the unit is abnormal by deduction, and on this basis reducing the complexity of calculations to achieve accurate monitoring of the operating status of the generator set has become a problem to be solved.

[0031] Figure 1 A flow chart of a method for evaluating the operating status of a generator set provided in one embodiment of the present application, the method comprising:

[0032] S110, obtaining n reference time period data streams of the generator set within the time window.

[0033] For example, there are many types of variables related to hydroelectric generators, with different dimensions (such as current, vibration, and temperature). First, obtain the cleaned original baseline period data stream , the size is ,express groups of samples, each group of samples contains Features:

[0034]

[0035] Among them, the initial time window size is defined as , that is, the data of the data flow during the reference period is consecutive time points.

[0036] S120, based on the n reference period data streams, determining a reference covariance matrix corresponding to the n reference period data streams.

[0037] Exemplarily, after obtaining the data streams of n reference periods within the time window, it is necessary to uniformly perform calculation and analysis, specifically including:

[0038] Normalize to unit variance:

[0039]

[0040] in is the mean vector of each feature, with dimension :

[0041]

[0042] in is the standard deviation vector of each feature, It can be calculated by the following formula:

[0043]

[0044] Furthermore, the standardized base covariance matrix C is calculated based on the standardization to unit variance:

[0045]

[0046] C is a A matrix where each element C ij represents the covariance of features i and j. Since the data has been standardized, this baseline covariance matrix is ​​equivalent to the correlation matrix.

[0047] The covariance matrix C1 corresponding to the data stream in the reference period can be determined by the above formula.

[0048] S130, obtaining m data streams in time periods to be detected, where m ≤ n.

[0049] Exemplarily, m data streams for the time period to be detected are obtained, wherein the m data streams for the time period to be detected are used to update the data stream for the reference time period. Each time the data streams for the time period to be detected arrive, they are added to the time window and the earliest m groups of old data are deleted. The window always keeps the latest Group samples.

[0050] S140, based on the benchmark covariance matrix, n benchmark time period data streams and m to-be-detected time period data streams, determine the updated data stream, the updated data stream is: delete the first m data of the benchmark time period data stream in the n benchmark time period data streams, and add m to-be-detected time period data streams, and determine the to-be-detected covariance matrix according to the updated data stream.

[0051] For example, the operation data related to the motor group is a continuously input data stream, so the data range must be limited to a certain extent to prevent the data volume from being too large and exceeding the computing capacity. Therefore, through the time window, each calculation is performed on the standardized reference period data stream within the time window. When new sample data arrives, it is added to the time window and the oldest data is deleted so that the window size remains unchanged and the updated data stream is determined.

[0052] Exemplarily, the reference period data flow within the time window Total samples, there are m data streams to be detected Arrival, while deleting the base period data flow The oldest data in , so the total number of samples in the time window is still n, Transformed into updated data stream For real-time monitoring, the number m of sample data updated each time can range from 1 to n, which can be selected according to the characteristics of the data and the application scenario.

[0053] Further, the data flow in the reference period can be calculated by the formula in step S120: The mean vector of , m oldest data The mean vector of , data flow during the detection period The mean vector of , The covariance matrix of Then the updated data flow The mean vector of for:

[0054]

[0055] Updated data flow The corresponding covariance matrix for:

[0056]

[0057] As another example, when m = n, that is, each batch of data completely replaces the previous data, the entire window is completely updated without using the covariance matrix calculated last time.

[0058] S150, based on the benchmark covariance matrix and the covariance matrix to be detected, determine the eigenvalue decomposition results corresponding to the benchmark covariance matrix and the covariance matrix to be detected respectively through eigenvalue decomposition, the eigenvalue decomposition results include the characteristic matrix and diagonal matrix corresponding to the benchmark covariance matrix and the characteristic matrix and diagonal matrix corresponding to the covariance matrix to be detected.

[0059] For example, the eigenvalue decomposition formula is:

[0060]

[0061] in yes The characteristic matrix (each column is an eigenvector), the eigenvectors define new coordinate axes (principal components), forming a new vector matrix corresponding to the data stream X. yes A diagonal matrix with eigenvalues ​​on the diagonal , the eigenvalue represents the variance explained by each principal component. Arrange in descending order, that is, .

[0062] The characteristic matrix and diagonal matrix corresponding to the reference covariance matrix C1 and the characteristic matrix and diagonal matrix corresponding to the covariance matrix to be detected C2 are determined by the above formulas.

[0063] S160, based on the eigenvalue decomposition results, determine the principal component matrix, the principal component score matrix and the eigenvalues ​​corresponding to the covariance matrix to be detected and the reference covariance matrix through dimensionality reduction processing.

[0064] In one possible implementation, based on the diagonal matrices corresponding to the baseline covariance matrix and the covariance matrix to be detected, the rank k corresponding to the baseline covariance matrix and the covariance matrix to be detected is determined by the proportion of principal component information; based on each k, each characteristic matrix and each diagonal matrix, the principal component matrices and eigenvalues ​​corresponding to the baseline covariance matrix and the covariance matrix to be detected are determined respectively; based on each principal component matrix, the data stream of the baseline time period and the data stream of the time period to be detected, the principal component score matrix corresponding to the baseline covariance matrix and the covariance matrix to be detected is determined.

[0065] Exemplarily, the rank k corresponding to the reference covariance matrix and the covariance matrix to be tested is determined by the proportion of the principal component information, wherein the formula for the proportion of the principal component information is:

[0066]

[0067] in It is eigenvalues, for example, let RT i = 99%, and the i that satisfies the above formula is deduced. Let rank k = i to determine the rank k corresponding to the benchmark covariance matrix and the covariance matrix to be tested.

[0068] As another example, the calculation formula for determining the rank k corresponding to the reference covariance matrix and the covariance matrix to be detected is:

[0069]

[0070] Among them, k is the rank corresponding to C, and C is the covariance matrix.

[0071] Furthermore, according to k, , extract the eigenvectors corresponding to the first k non-zero eigenvalues ​​to form The matrix .

[0072] Further, according to step S150, The diagonal matrix , and the diagonal is arranged in descending order, that is, .definition is a A diagonal matrix with the eigenvalues ​​arranged in descending order on the diagonal ,Right now yes The reduced-dimensional diagonal matrix can be used to determine the eigenvalues ​​corresponding to the reference covariance matrix and the covariance matrix to be tested. .

[0073] According to the characteristic matrix determined in step S150 is a The matrix only retains the eigenvectors corresponding to the first k larger eigenvalues, then its dimension reduction matrix is a Only the eigenvector matrix of the principal component is retained, so The principal component matrix is ​​used as the principal component matrix, thereby determining the principal component matrix corresponding to the benchmark covariance matrix and the covariance matrix to be detected.

[0074] Project the original data X into the principal component space:

[0075]

[0076] in Is the size of The principal component score matrix is ​​a principal component score matrix with each column being a principal component score. is the data stream of the reference period or the updated data stream, is the benchmark covariance matrix principal component matrix or the covariance matrix principal component matrix to be tested.

[0077] S170, based on each principal component matrix, each principal component score matrix, each eigenvalue, the covariance matrix to be detected and the reference covariance matrix, a normalized score of anomaly detection is determined through an anomaly detection model, and the normalized score is used to evaluate the operating status of the generator set.

[0078] In a possible implementation, the anomaly detection model includes: a principal component change detection model, a reconstruction error detection model, a correlation deviation detection model and a statistical distribution estimation detection model; based on each principal component matrix, each principal component score matrix, each eigenvalue, the covariance matrix to be detected and the benchmark covariance matrix, determining the normalized score of anomaly detection through the anomaly detection model includes: based on each eigenvalue, determining the normalized score of principal component change detection through the principal component change detection model; based on each principal component score matrix and each principal component matrix, determining the normalized score of reconstruction error detection through the reconstruction error detection model; based on the covariance matrix to be detected and the benchmark covariance matrix, determining the normalized score of correlation deviation detection through the correlation deviation detection model; based on each principal component score matrix, determining the normalized score of statistical distribution estimation detection through the statistical distribution estimation detection model; based on the normalized score of principal component change detection, the normalized score of reconstruction error detection, the normalized score of correlation deviation detection and the normalized score of statistical distribution estimation detection, determining the normalized score of anomaly detection through a preset formula.

[0079] Principal component change detection normalized score

[0080] Normally, the rank of the covariance matrix calculated for each batch should be the same. If the rank changes, it indicates that there may be an anomaly, and the increased or decreased eigenvectors need to be checked to see if there is an anomaly.

[0081] If the eigenvalues ​​of the covariance matrix C change significantly, it indicates that an abnormality may have occurred. This can be judged by the changes in the principal component contribution rate and the proportion of principal component information.

[0082] If the rank of the covariance matrix is ​​k, then the information of these k principal components accounts for RT k that is:

[0083]

[0084] in, represents the eigenvalue, k is the rank, and the contribution rate of k principal components and the proportion of the total principal component information are calculated. If any feature has changed significantly compared to a previous time point, an early warning should be issued. The previous time point can be selected according to the actual operation of the generator set, such as one hour ago, one day ago, or one week ago.

[0085] If the principal component information corresponding to the data stream in the reference period accounts for RT kB As a benchmark, the proportion of the principal component information corresponding to the period to be detected is , then its normalized score for anomaly detection is:

[0086]

[0087] In a possible implementation, if the information proportion of k principal components is required to be at least 80%, the above formula is simplified to:

[0088]

[0089] In another possible implementation, if the information proportion of k principal components is required to be at least 90%, the above formula is simplified to:

[0090]

[0091] Note that the above formula may be greater than 1, which means that the main component has changed beyond the maximum limit and the generator set may be abnormal. Just set it to 1.

[0092] Reconstruction error detection normalized score:

[0093] Reconstruct the principal component matrix back to the space where the original data X is located:

[0094]

[0095] Calculate the squared residual to measure the projection error of the data in the principal component space:

[0096]

[0097] in, is the reconstructed value of the principal component data in the original data space, Represents the square of the Euclidean norm. is the residual of the reconstructed matrix, i.e. the reconstruction error. Set the threshold ,if , then it means that the current input data deviates from the principal component space and anomalies may occur. You need to make a choice based on the actual data situation.

[0098] As an example, Approximately follows a chi-square distribution with (n - k) degrees of freedom:

[0099]

[0100] Significance level It is the probability of being wrong when rejecting the null hypothesis in a hypothesis test. For example, when the significance level is set to 0.05, it means that there is a 5% probability of incorrectly rejecting the null hypothesis even if it is true. It has a great influence on the calculation results and needs to be selected according to the actual situation of the data. Generally, it is 0.05 or 0.01.

[0101] As another example, The significance level is , the confidence level is , then the threshold for calculating the residual space score is:

[0102]

[0103] in, is the inverse cumulative distribution function of the chi-square distribution, representing the confidence level , the critical value of the chi-square distribution with (n - k) degrees of freedom can be obtained by looking up the table or calculating by the program. is the chi-squared cumulative distribution function The inverse function of , which determines the corresponding chi-square statistic using a given probability and degrees of freedom.

[0104] As an example, if you use represents the maximum allowable value of the reconstruction error of the sample in the residual space. Then, the normalized score is:

[0105]

[0106] It should be noted that the reconstruction error detection determines the normalized score through its own data.

[0107] Note that the above formula may be greater than 1, indicating that the main component has undergone a large abnormal change, and the operating status of the generator set may be abnormal. Just set it to 1.

[0108] Correlation bias detection normalized score:

[0109] Updated data flow The corresponding covariance matrix is , the covariance matrix of the data stream during the reference period is ,So and The Difference It can be expressed as:

[0110]

[0111] In the formula, is the Euclidean norm or modulus of the matrix. If it is too large, it means that the current variable relationship is significantly different from the normal state, and the generator set may be in an abnormal operating state. Set the threshold ,if , then it means that the current input data deviates from the earlier data and an anomaly may have occurred. You need to make a choice based on the actual data situation.

[0112] As an example, Represents the maximum allowable value of the correlation deviation of the sample, then the normalized score is:

[0113]

[0114] Note that the above formula may be greater than 1, indicating that the main component has undergone a large abnormal change, and the operating status of the generator set may be abnormal. Just set it to 1.

[0115] Furthermore, if correlation bias is required The maximum value does not exceed the reference value h%, then the above formula becomes:

[0116]

[0117] For example, if the maximum deviation is required not to exceed 10%, then the above formula becomes:

[0118]

[0119] Statistical distribution estimation detection normalized score:

[0120] Based on the principal component score matrix , define Hotelling's test):

[0121]

[0122] in, is the principal component score matrix, is the corresponding principal component eigenvalue matrix, if If the value exceeds a certain threshold, it indicates an abnormality. Set the threshold ,if , there may be anomalies.

[0123] based on Probability distribution function, the threshold can be calculated Assuming the significance level is , then the threshold It can be expressed as:

[0124]

[0125] Among them, n is the number of samples, k is the number of principal components, that is, the rank of the principal component matrix, is the critical value of the F distribution with (k, n - k) degrees of freedom.

[0126] The F distribution, also known as the Fisher-Snedecor distribution, is a continuous probability distribution in statistics that is used to compare the variances of two samples. It is defined as the ratio of the variances of two independent chi-square distributions. The formula and calculation of the probability density function of the F distribution are relatively complex and are generally obtained by table lookup. There are also special programs or functions that can calculate it. If the table lookup method is used, at the significance level Once determined, the critical value of the F distribution can be obtained by querying a table with degrees of freedom (k, n - k).

[0127] Significance level It is the probability of being wrong when rejecting the null hypothesis in a hypothesis test. For example, when the significance level is set to 0.05, it means that there is a 5% probability of incorrectly rejecting the null hypothesis even if it is true. Usually, the significance level is set in advance. In general, The smaller the value, the higher the critical value of the F distribution, and the calculated threshold The higher the It has a great influence on the calculation results and needs to be selected according to the actual situation of the data. Generally, it is 0.05 or 0.01.

[0128] The Hotelling statistic test for the principal component score matrix is ,based on The probability distribution function calculates the threshold as . Then, the normalized score is:

[0129]

[0130] It should be noted that the statistical distribution estimation test determines the normalized score through its own data.

[0131] Note that the above formula may be greater than 1, indicating that the main component has undergone a large abnormal change, and the operating status of the generator set may be abnormal. Just set it to 1.

[0132] Furthermore, based on the normalized score of principal component change detection, the normalized score of reconstruction error detection, the normalized score of correlation deviation detection and the normalized score of statistical distribution estimation detection, a preset formula is used to determine the normalized score of anomaly detection.

[0133] The default formula is:

[0134]

[0135] Among them, S represents the normalized score, w1, w2, w3, w4 represent adjustable parameters, represents the normalized score of principal component change detection, represents the normalized score of reconstruction error detection, represents the normalized score of correlation deviation detection, Represents the statistical distribution estimated detection normalized score.

[0136] As an example, weights w1, w2, w3, and w4 are all set to 0.25, and are adjusted based on actual data conditions.

[0137] In a possible implementation, after determining the normalized score of anomaly detection through an anomaly detection model based on each principal component matrix, each principal component score matrix, each eigenvalue, the covariance matrix to be detected and the benchmark covariance matrix, it also includes: determining the operating state of the generator set based on the normalized score and a preset threshold; when the normalized score is greater than the preset threshold, determining that the operating state of the generator set is abnormal and there is abnormal data in the covariance matrix to be detected; determining the source of the abnormality based on the principal component matrix corresponding to the benchmark covariance matrix and the principal component matrix corresponding to the covariance matrix to be detected.

[0138] In one possible implementation, determining the source of the abnormality based on the principal component matrix corresponding to the benchmark covariance matrix and the principal component matrix corresponding to the covariance matrix to be detected includes: determining the benchmark load matrix and the load matrix to be detected based on the principal component matrix corresponding to the benchmark covariance matrix and the principal component matrix corresponding to the covariance matrix to be detected; determining the abnormal change elements based on the benchmark elements in the benchmark load matrix and the elements to be detected in the load matrix to be detected; and determining the source of the abnormality based on the abnormal change elements.

[0139] For example, when the abnormality indicator determines that there is an abnormality, it is necessary to trace the source of the abnormality, that is, to trace back to the original variable with the abnormality. To achieve this goal, it is necessary to calculate the reference load matrix and the load matrix to be tested based on the principal component matrix corresponding to the reference covariance matrix after dimensionality reduction and the principal component matrix corresponding to the covariance matrix to be tested. The principal component load matrix after dimensionality reduction can be calculated using the following formula:

[0140]

[0141] in, yes The diagonal matrix of yes The matrix of yes The size and sign of the loading matrix indicate how the original features affect the principal components.

[0142] Furthermore, assuming that the data flow during the reference period And the data flow during the period to be detected Perform principal component analysis to obtain the corresponding load matrix and :

[0143]

[0144]

[0145] Where p is the number of original variables and k is the number of principal components retained.

[0146] For each original variable j (the element in the jth row of the loading matrix), calculate its total degree of change on all principal components:

[0147]

[0148] , represents the total load change of variable j on the principal component. If the value is large, it means that the contribution of the variable to the principal component has changed significantly, and it may be an abnormal variable.

[0149] To more accurately determine the abnormal source of the original variable, we can analyze the load changes of the variable on a specific principal component i:

[0150]

[0151] if It is significantly greater than the changes in other principal components, indicating that the anomaly of variable j is mainly reflected in principal component i.

[0152] In a possible implementation, determining the source of the abnormality based on the principal component matrix corresponding to the benchmark covariance matrix and the principal component matrix corresponding to the covariance matrix to be detected also includes: determining the benchmark load matrix and the load matrix to be detected based on the principal component matrix corresponding to the benchmark covariance matrix and the principal component matrix corresponding to the covariance matrix to be detected; determining the projection change of the benchmark elements based on the benchmark elements in the benchmark load matrix, and determining the projection change of the elements to be detected based on the elements to be detected in the load matrix to be detected; determining the source of the abnormality based on the projection change of the benchmark elements and the projection change of the elements to be detected.

[0153] For example, according to the benchmark load matrix The reference elements and load matrix to be tested in For the variables to be tested, calculate the projection changes of the variables on the principal components.

[0154] For example, the reference period data flow or the to-be-detected period data flow The projection change on principal component i is:

[0155]

[0156]

[0157] Comparison of projection changes can be done by the following formula:

[0158]

[0159] Those variables that change significantly are potential sources of anomalies.

[0160] The solution provided by the present application obtains n reference period data streams of the generator set within the time window; based on the n reference period data streams, determines the reference covariance matrix corresponding to the n reference period data streams; obtains m time period data streams to be detected, m ≤ n; based on the benchmark covariance matrix, n benchmark period data streams and m to-be-detected period data streams, determine the updated data stream, the updated data stream is: delete the first m data of the benchmark period data stream in the n benchmark period data streams, and add m to-be-detected period data streams, and determine the to-be-detected covariance matrix according to the updated data stream; based on the benchmark covariance matrix and the to-be-detected covariance matrix, through eigenvalue decomposition, respectively determine the eigenvalue decomposition results corresponding to the benchmark covariance matrix and the to-be-detected covariance matrix, the eigenvalue decomposition results include the characteristic matrix and diagonal matrix corresponding to the benchmark covariance matrix and the characteristic matrix and diagonal matrix corresponding to the to-be-detected covariance matrix; based on each eigenvalue decomposition result, through dimensionality reduction processing, determine the principal component matrix, principal component score matrix and eigenvalue corresponding to the to-be-detected covariance matrix and the benchmark covariance matrix; based on each principal component matrix, each principal component score matrix, each eigenvalue, the to-be-detected covariance matrix and the benchmark covariance matrix, through the anomaly detection model, determine the normalized score of anomaly detection, and the normalized score is used to evaluate the operating status of the generator set. This application can simultaneously meet the needs of real-time data processing and the ability to capture short-term dynamic changes. The sliding window is used to limit the time range of interest (such as the last t period of time or the last N samples), capture short-term pattern changes, and quickly discover potential anomalies. Avoid excessive data accumulation and waste of computing resources. At the same time, the data in the sliding window is incrementally updated to extract the main components, ensure real-time performance, adapt to the dynamic changes of streaming data, and is suitable for high-dimensional and large-scale data monitoring systems.

[0161] Figure 2This is a structural block diagram of a generator set operating status evaluation device provided in an embodiment of the present application. For the sake of convenience, only the parts related to the embodiment of the present application are shown. Figure 2 The generator set operating status evaluation device 200 includes an acquisition module 201 , a determination module 202 , and a detection module 203 .

[0162] In one implementation, the apparatus 200 may be used to implement the above Figure 1 For example, the acquisition module 201 is used to implement S110 and S130, the determination module 202 is used to implement S120 and S140-S160, and the detection module 203 is used to implement S170.

[0163] The beneficial effects of the device provided by the present application and Figure 1 The beneficial effects of the methods shown are consistent and will not be repeated here.

[0164] Figure 3 This is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Figure 3 As shown, the electronic device 3 of this embodiment includes: at least one processor 30 ( Figure 3 Only one is shown in the figure) a processor, a memory 31, and a computer program 32 stored in the memory 31 and executable on the at least one processor 30, and when the processor 30 executes the computer program 32, the steps in any of the above method embodiments are implemented.

[0165] The electronic device 3 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The electronic device may include, but is not limited to, a processor 30 and a memory 31. Those skilled in the art will appreciate that Figure 3 It is only an example of the electronic device 3 and does not constitute a limitation on the electronic device 3. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, it may also include input and output devices, network access devices, etc.

[0166] The processor 30 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0167] In some embodiments, the memory 31 may be an internal storage unit of the electronic device 3, such as a hard disk or memory of the electronic device 3. In other embodiments, the memory 31 may also be an external storage device of the electronic device 3, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 3. Further, the memory 31 may also include both an internal storage unit and an external storage device of the electronic device 3. The memory 31 is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program. The memory 31 may also be used to temporarily store data that has been output or is to be output.

[0168] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0169] An embodiment of the present application also provides an electronic device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor implements the steps of any of the above-mentioned method embodiments when executing the computer program.

[0170] An embodiment of the present application further 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, the steps in the above-mentioned method embodiments can be implemented.

[0171] An embodiment of the present application provides a computer program product. When the computer program product runs on a mobile terminal, the mobile terminal can implement the steps in the above-mentioned method embodiments when executing the computer program product.

[0172] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium can at least include: any entity or device that can carry the computer program code to the camera / electronic device, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, RandomAccess Memory), electrical carrier signal, telecommunication signal and software distribution medium. For example, a USB flash drive, a mobile hard disk, a disk or an optical disk.

[0173] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0174] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0175] In the embodiments provided in the present application, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device embodiments described above are merely schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0176] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0177] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for evaluating the operating status of a generator set, characterized in that: The method comprises: Obtaining n reference period data streams of the generator set within the time window, wherein the reference period data streams include: current, vibration, and temperature; Based on the n reference period data streams, determining a reference covariance matrix corresponding to the n reference period data streams; Obtain m data streams for time periods to be detected, where m≤n; Based on the reference covariance matrix, the n reference time period data streams and the m to-be-detected time period data streams, an updated data stream is determined, wherein the updated data stream is: the first m data of the reference time period data stream are deleted from the n reference time period data streams, and the m to-be-detected time period data streams are added, and the to-be-detected covariance matrix is ​​determined according to the updated data stream; Based on the reference covariance matrix and the covariance matrix to be detected, determining the eigenvalue decomposition results corresponding to the reference covariance matrix and the covariance matrix to be detected respectively through eigenvalue decomposition, wherein the eigenvalue decomposition results include the characteristic matrix and diagonal matrix corresponding to the reference covariance matrix and the characteristic matrix and diagonal matrix corresponding to the covariance matrix to be detected; Based on each of the eigenvalue decomposition results, determining the principal component matrix, the principal component score matrix and the eigenvalues ​​corresponding to the covariance matrix to be detected and the reference covariance matrix through dimensionality reduction processing; Based on each of the principal component matrices, each of the principal component score matrices, each of the eigenvalues, the covariance matrix to be detected and the reference covariance matrix, a normalized score of anomaly detection is determined through an anomaly detection model, wherein the normalized score is used to evaluate the operating status of the generator set; The anomaly detection model includes: a principal component change detection model, a reconstruction error detection model, a correlation deviation detection model and a statistical distribution estimation detection model; The determining of the normalized score of anomaly detection based on each of the principal component matrices, each of the principal component score matrices, each of the eigenvalues, the covariance matrix to be detected and the reference covariance matrix through an anomaly detection model comprises: Based on each of the eigenvalues, determining a principal component change detection normalized score through the principal component change detection model; Based on each of the principal component score matrices and each of the principal component matrices, determining a reconstruction error detection normalized score through the reconstruction error detection model; Based on the covariance matrix to be detected and the reference covariance matrix, determining a correlation deviation detection normalized score through the correlation deviation detection model; Based on each of the principal component score matrices, determining a statistical distribution estimation detection normalized score through the statistical distribution estimation detection model; Based on the principal component change detection normalized score, the reconstruction error detection normalized score, the correlation deviation detection normalized score and the statistical distribution estimation detection normalized score, a preset formula is used to determine an anomaly detection normalized score; The preset formula is: Among them, S represents the normalized score, w1, w2, w3, w4 represent adjustable parameters, represents the normalized score of principal component change detection, represents the normalized score of reconstruction error detection, represents the normalized score of correlation deviation detection, Represents the statistical distribution estimated detection normalized score.

2. The method according to claim 1, characterized in that The determining of the principal component matrix, the principal component score matrix and the eigenvalues ​​corresponding to the covariance matrix to be detected and the reference covariance matrix by dimensionality reduction based on each of the eigenvalue decomposition results comprises: Based on the diagonal matrices corresponding to the reference covariance matrix and the covariance matrix to be detected, determining the rank k corresponding to the reference covariance matrix and the covariance matrix to be detected by the proportion of principal component information; Based on each of the k, each of the characteristic matrices and each of the diagonal matrices, respectively, determine the principal component matrices and eigenvalues ​​corresponding to the reference covariance matrix and the covariance matrix to be detected; Based on each of the principal component matrices, the reference period data stream and the updated data stream, a principal component score matrix corresponding to the reference covariance matrix and the covariance matrix to be detected is determined.

3. The method according to claim 1, characterized in that After determining the normalized score of anomaly detection through an anomaly detection model based on each of the principal component matrices, each of the principal component score matrices, each of the eigenvalues, the covariance matrix to be detected and the reference covariance matrix, the method further includes: Determining the operating state of the generator set based on the normalized score and a preset threshold; When the normalized score is greater than a preset threshold, it is determined that the operating state of the generator set is abnormal, and the covariance matrix to be detected has abnormal data; Based on the principal component matrix corresponding to the benchmark covariance matrix and the principal component matrix corresponding to the covariance matrix to be detected, the source of the abnormality is determined.

4. The method according to claim 3, characterized in that The determining of the abnormality source based on the principal component matrix corresponding to the benchmark covariance matrix and the principal component matrix corresponding to the covariance matrix to be detected includes: Determine a reference load matrix and a load matrix to be detected based on a principal component matrix corresponding to the reference covariance matrix and a principal component matrix corresponding to the covariance matrix to be detected; Determining abnormal change elements based on the reference elements in the reference load matrix and the elements to be detected in the load matrix to be detected; Based on the abnormal change element, the abnormal source is determined.

5. The method according to claim 3, characterized in that: The determining of the abnormal source based on the principal component matrix corresponding to the reference covariance matrix and the principal component matrix corresponding to the covariance matrix to be detected further includes: Determine a reference load matrix and a load matrix to be detected based on a principal component matrix corresponding to the reference covariance matrix and a principal component matrix corresponding to the covariance matrix to be detected; Determining a projection change of the reference element based on the reference element in the reference load matrix, and determining a projection change of the element to be detected based on the element to be detected in the load matrix to be detected; The abnormality source is determined based on the projection change of the reference element and the projection change of the element to be detected.

6. A device for evaluating the operating status of a generator set, the device being applied to the method according to claim 1, characterized in that: The device comprises: An acquisition module, used for acquiring n reference time period data streams of the generator set within a time window; A determination module, configured to determine a reference covariance matrix corresponding to the n reference period data streams based on the n reference period data streams; The acquisition module is further used to acquire m data streams in time periods to be detected, where m≤n; The determination module is further used to determine an updated data stream based on the reference covariance matrix, the n reference time period data streams and the m to-be-detected time period data streams, wherein the updated data stream is: deleting the first m data of the reference time period data stream from the n reference time period data streams, and adding m to-be-detected time period data streams, and determining the to-be-detected covariance matrix according to the updated data stream; The determination module is further used to determine the eigenvalue decomposition results corresponding to the benchmark covariance matrix and the covariance matrix to be detected respectively through eigenvalue decomposition based on the benchmark covariance matrix and the covariance matrix to be detected, wherein the eigenvalue decomposition results include the characteristic matrix and diagonal matrix corresponding to the benchmark covariance matrix and the characteristic matrix and diagonal matrix corresponding to the covariance matrix to be detected; The determination module is further used to determine the principal component matrix, the principal component score matrix and the eigenvalues ​​corresponding to the covariance matrix to be detected and the reference covariance matrix through dimensionality reduction processing based on each of the eigenvalue decomposition results; A detection module is used to determine a normalized score of anomaly detection based on each of the principal component matrices, each of the principal component score matrices, each of the eigenvalues, the covariance matrix to be detected and the benchmark covariance matrix through an anomaly detection model, and the normalized score is used to evaluate the operating status of the generator set.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images