Fault diagnosis system construction method and device for hydroelectric generating set, equipment and medium

Abstract

The invention discloses a fault diagnosis system construction method and device for a hydroelectric generating set, equipment and a medium, and belongs to the technical field of electric power systems, and the method comprises the steps: constructing a sample matrix through collecting vibration signals, calculating a covariance matrix after standardization processing, extracting a principal component, generating a score matrix, and calculating a statistical magnitude. And a control threshold is set based on the health sample, and abnormity judgment and alarm are carried out on the monitoring sample. According to the method, key features of fault data of the hydroelectric generating set can be extracted, and redundant information is removed. By analyzing the key features, the fault type and the fault degree can be identified more accurately, so that the accuracy of fault diagnosis is improved. Meanwhile, the method also combines principal component space statistics and prediction residual sum of squares statistics, and can help a user to comprehensively evaluate the vibration state of the hydroelectric generating set.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a method, device, equipment and medium for constructing a fault diagnosis system for a hydroelectric generator set. Background Art

[0002] With the increasing complexity and scale of turbine units, any minor fault can quickly escalate, posing a serious threat to the daily operations of power plants and even the stability of the entire power system. The potential economic losses and social impacts are incalculable. Therefore, establishing a comprehensive, real-time fault diagnosis method for hydropower units and rapidly monitoring their fault status are prerequisites for ensuring safe and stable operation in complex and changing operating environments. This fault diagnosis model must be highly sensitive and accurate, capable of instantly capturing abnormal signals during unit operation. Using advanced data analysis techniques, it can quickly locate the source of problems and provide a scientific basis for maintenance decisions, effectively preventing major accidents, ensuring the continuity and reliability of power supply, and laying a solid foundation for the intelligent transformation and sustainable development of the hydropower industry. Summary of the Invention

[0003] In view of the above-mentioned problems, the present invention is proposed.

[0004] Therefore, the technical problem solved by the present invention is to solve the defects of existing hydropower unit fault diagnosis methods such as vector machines and random forests, which are easily overfitted and difficult to converge due to the interference of noise data.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for constructing a fault diagnosis system for a hydroelectric generator set, comprising:

[0006] Collect vibration signals of the hydroelectric generator set and construct an initial vibration sample matrix based on the vibration signals;

[0007] For each characteristic variable in the initial vibration sample matrix, calculate its mean and standard deviation respectively, and perform standardization based on them to obtain a standardized data matrix;

[0008] Construct a covariance matrix based on the standardized data matrix, and perform eigenvalue decomposition on the covariance matrix to obtain a set of principal component vectors consisting of eigenvalues ​​and their corresponding eigenvectors;

[0009] Select several principal component vectors corresponding to the largest eigenvalues ​​from the principal component vector set, and calculate the principal component score matrix by combining with the standardized data matrix;

[0010] Based on the principal component score matrix and the principal component vector set, the principal component space statistics and the sum of squares of the prediction residuals are calculated to characterize the sample characteristics;

[0011] Based on the healthy operation samples of hydroelectric generating units, the control thresholds of principal component spatial statistics and the sum of squares of prediction residuals are determined;

[0012] The monitoring samples were processed in the same way as the healthy samples to obtain the principal component spatial statistics and the sum of squares of the prediction residuals of the samples, and compared with the control thresholds.

[0013] When either the principal component spatial statistics or the sum of squares of the predicted residuals of the monitoring sample exceeds the corresponding control threshold, the operating status of the hydropower generating unit reflected by the sample is judged to be abnormal, and the diagnosis result and alarm information are output.

[0014] As a preferred solution of the method for constructing a fault diagnosis system for a hydroelectric generator set described in the present invention, the initial vibration sample matrix is ​​composed of multiple vibration samples collected at different time points, each vibration sample includes six characteristic variables: main shaft horizontal deviation, main shaft vertical deviation, water guide bearing horizontal deviation, water guide bearing vertical deviation, water guide bearing horizontal swing and water guide bearing vertical swing. Each row in the matrix corresponds to a vibration sample, and each column corresponds to a characteristic variable.

[0015] As a preferred solution of the method for constructing a fault diagnosis system for a hydroelectric generator set described in the present invention, the standardization processing includes respectively calculating the mean and standard deviation of the characteristic variables in each column of the original vibration data matrix, and subtracting the mean of the characteristic variable from each data value and dividing it by its standard deviation to obtain a standardized data matrix.

[0016] As a preferred solution of the method for constructing a fault diagnosis system for a hydroelectric generator set described in the present invention, the method of obtaining a set of principal component vectors composed of eigenvalues ​​and their corresponding eigenvectors includes constructing a covariance matrix based on a standardized data matrix, and performing eigenvalue decomposition on the covariance matrix to obtain a number of eigenvalues ​​and their corresponding eigenvectors, wherein the eigenvalues ​​are arranged in a diagonal matrix, and the eigenvectors constitute an orthogonal matrix.

[0017] As a preferred solution of the method for constructing a fault diagnosis system for hydropower generating units described in the present invention, the principal component space statistic is a quantitative indicator calculated by the ratio of the principal component score to the corresponding eigenvalue, which is used to characterize the distribution of samples in the principal component space, and its control threshold is determined in combination with the set number of principal components, number of samples and statistical confidence level.

[0018] As a preferred solution of the method for constructing a fault diagnosis system for a hydroelectric generator set described in the present invention, the sum of squares of the predicted residuals is represented as the residual information of the sample vector outside the principal component space, which is calculated by the difference between the original value and the reconstructed value of the sample, and is used to evaluate whether the sample has abnormal features that are not covered by the principal component space.

[0019] As a preferred solution of the method for constructing a fault diagnosis system for hydropower generating units described in the present invention, the control threshold of the sum of squares of the predicted residuals is derived based on multiple power statistics of the eigenvalues ​​corresponding to the unselected principal components, and an intermediate variable is introduced to establish a correlation relationship with the critical value of the standard normal distribution to form a dynamic judgment standard for the degree of abnormality in the residual space.

[0020] Another object of the present invention is to provide a fault diagnosis system construction system for a hydroelectric generator set.

[0021] To solve the above technical problems, the present invention provides the following technical solutions: a system for constructing a fault diagnosis system for a hydroelectric generator set, comprising: a vibration signal acquisition module for acquiring vibration signals of the hydroelectric generator set and constructing a vibration initial sample matrix based on the acquired vibration signals;

[0022] A data standardization module is used to calculate the mean and standard deviation of each characteristic variable in the vibration initial sample matrix, and perform standardization processing accordingly to obtain a standardized data matrix;

[0023] A principal component extraction module is used to construct a covariance matrix based on the standardized data matrix and perform eigenvalue decomposition to obtain a principal component vector set consisting of eigenvalues ​​and their corresponding eigenvectors;

[0024] A principal component score calculation module is used to select a number of principal component vectors corresponding to the maximum eigenvalue from the principal component vector set, and calculate a principal component score matrix in combination with the standardized data matrix;

[0025] A statistics analysis module, configured to calculate principal component spatial statistics and a sum of squares of prediction residuals based on the principal component score matrix and the principal component vector set;

[0026] A threshold generation module is used to determine the control thresholds of the principal component spatial statistics and the sum of squares of the prediction residuals based on the healthy operation samples of the hydropower generating units;

[0027] An operating status determination module is used to obtain statistics for the monitoring samples according to the same processing flow as the healthy samples, and compare them with the control threshold;

[0028] The abnormality judgment and alarm module is used to judge it as abnormal when any of the principal component space statistics or the sum of squares of the predicted residuals of the monitoring sample exceeds the control threshold, and output the diagnosis result and alarm information.

[0029] The present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program and is characterized in that when the processor executes the computer program, the steps of the method for constructing a fault diagnosis system for a hydroelectric generator set are implemented.

[0030] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of a method for constructing a fault diagnosis system for a hydroelectric generator set.

[0031] The present invention has the beneficial effect of extracting key features from hydroelectric generator fault data and removing redundant information. By analyzing these key features, the fault type and severity can be more accurately identified, thereby improving the accuracy of fault diagnosis. Furthermore, the present method combines principal component space statistics with prediction residual sum of squares statistics to help users comprehensively assess the vibration status of hydroelectric generators. Once an anomaly is detected, the system immediately triggers an alarm signal and sends it to the status monitoring interface, enabling rapid response and accurate judgment of the fault status. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions of 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 paying any creative work.

[0033] Figure 1 An overall flow chart of a method for constructing a fault diagnosis system for a hydroelectric generator set provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0034] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0035] Example 1, with reference to Figure 1, is an embodiment of the present invention, which provides a method for constructing a fault diagnosis system for a hydroelectric generator set, comprising:

[0036] Step 1: The data acquisition module collects the vibration signal of the hydropower unit through the sensor. The vibration signal includes the following six characteristic variables: main shaft horizontal deviation, main shaft vertical deviation, water guide bearing horizontal deviation, water guide bearing vertical deviation, water guide bearing horizontal swing, and water guide bearing vertical swing.

[0037] The resulting vibration initial sample matrix Y is as follows:

[0038]

[0039] Where Y∈Rm×n, m represents the characteristic variables, including six vibration characteristic variables, and n is the number of samples collected.

[0040] The vibration health status samples and vibration monitoring samples of the hydropower unit are represented by Y1 and Y2 respectively.

[0041] Step 2: Apply principal component analysis to analyze and process the data in the data processing module;

[0042] The specific implementation is as follows: standardize the collected vibration characteristic variable matrix Y to eliminate the dimensional difference of the characteristic parameters and avoid misjudgment of the model due to scale deviation, and obtain the standardized data matrix X;

[0043] The covariance matrix of the standardized data matrix X is calculated to reflect the relationship between the collected characteristic variables (spindle horizontal deviation, spindle vertical deviation, water guide bearing horizontal deviation, water guide bearing vertical deviation, water guide bearing horizontal swing, water guide bearing vertical swing);

[0044] Then, the eigenvectors and eigenvalues ​​of the principal components are obtained by decomposing the covariance matrix. By discarding the principal components corresponding to low eigenvalues ​​(usually noise or redundant information), the data dimension (including noise) is significantly reduced, achieving data dimensionality reduction and information compression.

[0045] By calculating T2 and SPE, the T2 statistic is used to monitor the overall offset in the principal component space (such as the systematic offset caused by magnetic pull imbalance), and the SPE statistic captures local anomalies in the residual space. If the T2 and SPE calculated for a data point exceed either of these two thresholds, it can be considered an anomaly.

[0046] The implementation process is as follows: To mitigate the negative impact of data bias, unit dimension differences, and local outliers on analysis results, we perform preprocessing on the raw data. This includes standardization and zero-mean processing. Standardization aims to eliminate dimensional differences between different features and bring the data to a uniform scale; zero-mean processing ensures that the average value of the data is zero, preventing bias in the analysis results.

[0047] Standardize the data and obtain the standardized matrix X by standardizing the matrix Y. By calculating the mean and standard deviation of each characteristic variable and standardizing the original data, the dimensional differences between different characteristic variables are eliminated, making the data more concentrated, which is conducive to the model to better capture the intrinsic relationship between the data, thereby improving the accuracy of fault diagnosis.

[0048] The specific steps are as follows: Calculate the mean: Calculate the mean of each feature in matrix Y. For feature Y i , whose mean μ i The calculation formula is:

[0049]

[0050] Where n is the number of samples, Y ij is the value of the jth sample on the i-th feature.

[0051] Calculate standard deviation: Calculate the standard deviation σ of each feature i :

[0052]

[0053] Standardized data: Standardize each feature to obtain standardized data X ij ,

[0054]

[0055] The normalized matrix X is obtained by normalizing the initial vibration sample matrix Y:

[0056]

[0057] Calculate the covariance matrix R of the standardized sample

[0058]

[0059] Covariance matrix R is used to describe the linear correlation between feature variables; R ij is the element in the i-th row and j-th column of the covariance matrix, n is the number of samples, X ki is the element in the kth row and ith column of the normalized matrix X, X kj is the element in the kth row and jth column of the normalized matrix X.

[0060] Calculate the eigenvalues ​​λ and eigenvectors u of the covariance matrix R:

[0061] By decomposing the covariance matrix, the eigenvectors and eigenvalues ​​of the principal components are obtained. By discarding the principal components corresponding to low eigenvalues ​​(usually noise or redundant information), the data dimension is significantly reduced, achieving data dimensionality reduction and information compression. This helps improve computational efficiency and reduces the impact of noise on fault diagnosis results. The specific steps are:

[0062] Perform eigenvalue decomposition on the covariance matrix: R = XX T =UΛV, where U is the eigenvector matrix (orthogonal matrix), V is the transposed matrix of U, and the column vectors u1,u2,…,u m Corresponding to the principal component direction; Λ is a diagonal matrix, and the diagonal elements are the eigenvalues ​​λ1,λ2,…,λ of R m (m is the total number of characteristic variables, which is 6 in the present invention), sorted from large to small, i.e. λ1≥λ2≥…≥λ m ≥0;λ i ≥0(i=1,2,...,m) represents an eigenvalue.

[0063] In a preferred embodiment of the present invention, the specific process of constructing a covariance matrix includes the following steps: first, for each characteristic variable in the collected vibration standardized data matrix, its standardized value at all sample points is extracted column by column; then, any two characteristic variables are selected in turn, and their standardized values ​​are multiplied in pairs according to the corresponding sample points to obtain a set of product values; then, the set of product values ​​is summed and divided by the number of samples minus one to calculate the covariance corresponding to the two characteristic variables; the above steps are repeated to calculate the covariance of all characteristic variable combinations, and finally a symmetric covariance matrix is ​​obtained.

[0064] For example, in the case of 6 vibration characteristic variables, the covariance matrix is ​​of 6×6 dimensions, where the (i, j)th item represents the covariance value between the i-th variable and the j-th variable, the elements on the main diagonal are the variance values ​​of each characteristic variable, and the off-diagonal elements represent the linear coupling strength between different variables.

[0065] The beneficial effects of this preferred technical solution are: through the above-mentioned covariance matrix construction method, it is possible to quantitatively reflect the correlation structure between the vibration characteristic variables, retain the statistical correlation information of the data, and provide a basis for subsequent principal component extraction, so that the model can capture the potential coupling patterns between variables in fault evolution, thereby improving the accuracy and discrimination ability of the overall modeling.

[0066] In an optional embodiment of the present invention, the covariance matrix can alternatively be constructed using a "correlation coefficient matrix" as a representation of the relationship between characteristic variables. Each matrix element represents the Pearson correlation coefficient between two variables, calculated as the covariance of the two variables divided by the product of their standard deviations. This approach can eliminate the effects of differences in the dimensions and scales of the variables and is suitable for analyzing operating data in scenarios where the spans of the characteristic units vary widely.

[0067] Principal component selection: Select the first a principal components (a≤m) according to the cumulative variance contribution rate (such as ≥85%), corresponding to the eigenvalues ​​λ1,…,λ a and eigenvectors u1,…,u a .

[0068] The calculation formula for cumulative variance contribution rate is:

[0069]

[0070] Compute principal component analysis statistics and their thresholds.

[0071] In a preferred embodiment of the present invention, the process of calculating the principal component score matrix includes the following specific steps: first, from the eigenvectors obtained by decomposing the covariance matrix, several principal component vectors are selected in descending order of eigenvalues ​​to construct a principal component loading matrix, in which each column is an eigenvector; then, the standardized data matrix is ​​subjected to matrix multiplication with the principal component loading matrix to obtain a principal component score matrix. Each row in the matrix corresponds to the projection value of a sample in each principal component direction, i.e., the principal component score, which reflects the representation result of the sample in the low-dimensional space. In practical applications, the number of principal components a is usually determined based on the first a eigenvalues ​​whose cumulative variance contribution rate reaches a preset threshold (such as 85%), and the score matrix T is constructed based on this, whose dimension is n×a, where n is the number of samples.

[0072] The beneficial effects of this preferred technical solution are: the principal component score matrix generated in this way can maximize the preservation of the most representative structural information in the original vibration data, while significantly reducing feature dimensions, redundancy, and noise, making subsequent statistical analysis and fault diagnosis more efficient and accurate. Furthermore, the score matrix has good visual interpretation, facilitating pattern recognition of abnormal behavior.

[0073] In an optional embodiment of the present invention, singular value decomposition (SVD) can be used to decompose the standardized data matrix when calculating the principal component score matrix, thereby indirectly obtaining the principal component directions and the score matrix. This method is particularly suitable for scenarios where the number of samples is much smaller than the number of characteristic variables, and can improve computational stability and convergence when working with high-dimensional, small-sample vibration data.

[0074] The principal component matrix is ​​obtained by principal component analysis using characteristic variables, and then the principal component analysis statistics and its threshold are calculated. The vibration state of the hydropower unit is evaluated by analyzing whether the statistical value of the real-time monitoring data of the hydropower unit vibration exceeds the threshold set by the principal component analysis statistics of the healthy sample. The principal component space statistics T is used to calculate the principal component matrix. 2 and the predicted residual sum of squares statistic SPE a ,A data point can be considered as an outlier if it exceeds any of these two thresholds.

[0075] Using the selected eigenvectors u1,…,u a , calculate the principal component score matrix T:

[0076] T=X·[u1,…,u a ]

[0077] X is the standardized data matrix, u i is the i-th eigenvector of the covariance matrix R, and a is the number of principal components selected (satisfying the cumulative variance contribution rate ≥ 85%)

[0078] Principal component space T 2 The statistic is expressed as,

[0079]

[0080] Among them, t represents the sample of the principal component score matrix T, a represents the number of principal components selected, and t i represents the score vector of the i-th principal component (the i-th column of matrix T), λ i is the eigenvalue corresponding to the i-th principal component. 2 Control threshold of the statistic The calculation method is expressed as,

[0081]

[0082] Where n represents the number of sample points; a represents the number of principal components selected; and the significance level β (usually β = 0.05 or β = 0.01) represents the probability of misclassifying a healthy sample as abnormal. β (a, na) represents the critical value of the F distribution with (a, na) degrees of freedom and β significance level. This critical value is determined according to the confidence level, and F represents the probability density function of the F distribution.

[0083] If T of a sample point 2 The statistic is greater than the control threshold, which indicates that the point is significantly different from the main trend of the data set and is an outlier.

[0084] The prediction residual sum of squares statistic is a large statistic. The larger the prediction residual sum of squares statistic SPE is, the farther the data point is from the model fitting, and the data point should be an outlier.

[0085]

[0086] Among them, e is the residual matrix of the standard data matrix X, e T is the transposed matrix of the residual matrix e, J is the sample point dimension; Y j is the jth sample point of the original sample; and is the jth value after reconstruction by the model.

[0087] The threshold value of the SPE statistic is:

[0088]

[0089] Among them, h0 is an intermediate variable, θ i Represents the eigenvalues ​​of the covariance matrix for j = a + 1 to j = m The sum is calculated; θ1 is the sum of the eigenvalues ​​of the unselected principal components (from +1 to m); θ2 is the sum of the squares of the eigenvalues ​​of the unselected principal components; θ3 is the sum of the cubes of the eigenvalues ​​of the unselected principal components. m represents the total number of eigenvalues ​​of the covariance matrix; a represents the number of selected principal components; λ is the eigenvalue of the covariance matrix R; c α It represents the threshold value when the significance level of the standard normal distribution is α.

[0090] By calculating the principal component space statistic T² and the predicted residual sum of squares statistic SPE, we can effectively monitor the overall offset in the principal component space and local anomalies in the residual space. When the T² or SPE statistic of a data point exceeds a set threshold, it can be considered an outlier, enabling rapid detection of hydropower unit fault conditions.

[0091] In a preferred embodiment of the present invention, the control threshold of the principal component space statistic is set based on the T2 statistic. The calculation method is as follows: the number of principal components selected is a, the number of healthy samples is n, the significance level is set to β, the critical values ​​of the F distribution with degrees of freedom a and na are obtained by table lookup, and the T2 threshold is calculated according to the statistical formula. At the same time, the control threshold of the sum of squared prediction residuals (SPE) is calculated by calculating the multi-order power accumulation of the corresponding eigenvalues ​​of the unselected principal components, and the intermediate variable h0 is constructed from them. Finally, the control threshold of SPE is derived by combining the critical value ca of the standard normal distribution at the significance level α.

[0092] This calculation method uses healthy samples as a reference and comprehensively considers multiple factors such as sample size, number of principal components, residual eigenvalues, and statistical distribution form, so that the two thresholds have good theoretical support and distinguishing ability.

[0093] The beneficial effects of this preferred technical solution are: by dually modeling the statistical characteristics of the principal component space and the residual space, it can effectively define the reasonable fluctuation range of normal operating conditions. The mathematically rigorous and adjustable threshold setting can not only control the false alarm rate but also sensitively identify minor anomalies, thereby improving the accuracy of fault detection and the reliability of system response.

[0094] In an optional embodiment of the present invention, the control threshold can also be determined based on the empirical distribution of the T2 statistic and the SPE statistic in healthy samples, with the 95th or 99th percentile value selected as the corresponding control threshold, without relying on a theoretical distribution function. This approach is suitable for applications where the data distribution is unknown or fails to meet the normality assumption, and can achieve effective anomaly identification within a nonparametric statistical framework.

[0095] Step 3: Send the processed data results to the hydropower unit status display interface for user query.

[0096] Using the previously proposed feature extraction and sample construction methods, we obtain data on shaft horizontal deviation, main shaft vertical deviation, water-guided bearing horizontal deviation, water-guided bearing vertical deviation, water-guided bearing horizontal runout, and water-guided bearing vertical runout, and construct a vibration state sample matrix for the unit. We select two consecutive instances of normal unit operation as samples of the unit's vibration health status, construct a vibration health state matrix, and send it to the data processing module.

[0097] The data processing module uses the principal component analysis method to select the number of principal component vectors for the vibration health sample data, and calculates the principal component analysis statistics and its corresponding control threshold and healthy sample residual subspace.

[0098] For the collected real-time monitoring samples of vibration status, the same steps (1) and (2) are performed to calculate the principal component analysis statistics and the residual subspace of the monitoring samples;

[0099] From the SPE statistics and the principal component space T 2 The vibration status of the hydropower unit is evaluated from two aspects of statistics: judging whether the principal component analysis statistic of the monitoring sample exceeds the control threshold of the corresponding statistic of the healthy sample; if any of the above vibration status evaluation methods show abnormal results, it means that the vibration status of the hydropower unit is abnormal, and an alarm signal will be triggered and sent to the status monitoring interface for alarm.

[0100] Example 2 is an embodiment of the present invention, which provides a method for constructing a fault diagnosis system for a hydroelectric generator set. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0101] A hydropower station conducts real-time vibration monitoring on a hydropower unit with a rated power of 50MW. The data collection lasts for 10 minutes, with one set of data recorded every minute. A total of 10 samples (n=10) are obtained. Each sample contains 6 vibration characteristic variables (m=6), as shown in Table 1:

[0102] Table 1 Experimental data analysis table

[0103]

[0104]

[0105] The initial sample matrix is ​​expressed as:

[0106]

[0107] 2. Data processing stage (principal component analysis, PCA):

[0108] Step 2.1: Standardization:

[0109] Calculate the mean (taking feature V1 as an example):

[0110]

[0111] Calculate the standard deviation:

[0112]

[0113] Normalized data (taking V1 of sample 1 as an example):

[0114]

[0115] Similarly, all features are calculated to obtain the standardized matrix X.

[0116] Step 2.2: Calculate the covariance matrix R:

[0117]

[0118] (The eigenvalues ​​are: λ = [10, 8, 6, 4, 1, 0.5] and the corresponding eigenvector matrix V.) Step 2.3: Select the principal components:

[0119] Calculation of cumulative contribution rate:

[0120] The contribution rates of the first three principal components are:

[0121]

[0122] Cumulative contribution rate: 43.48% + 27.12% + 20.34% = 90.94% ≥ 85%, so the first three principal components (a = 3) are selected.

[0123] Step 2.4: Calculate the statistic T 2 and SPE:

[0124] Principal component score matrix:

[0125] T: T = X·V(:,1:3) (V(:,1:3) is the matrix composed of the first three eigenvectors)

[0126] T 2 Statistics (taking sample 1 as an example):

[0127]

[0128] Calculate T of sample 1 2 =2.5

[0129] Control threshold TThreshold2:

[0130]

[0131] Take the significance level α = 0.05, and look up the table to get F 0.05 (3,7)≈4.35, then:

[0132] Tthreshold2=10×73×9×11×4.35≈5.72;

[0133] SPE statistic (reconstruction error for sample 1):

[0134]

[0135] The calculated SPE is 1.2;

[0136] SPE threshold:

[0137]

[0138] Take c α =1.96 (corresponding to 95% confidence level), and the calculated SPE threshold is ≈3.85.

[0139] 3. Fault diagnosis stage:

[0140] Real-time monitoring data: A new set of data is collected at the 10th minute and calculated after standardization: real-time

[0141]

[0142] Real-time SPE = 1.2 (< SPE threshold = 3.85);

[0143] Judgment logic: Since both the real-time T 2 and the real-time SPE do not exceed the threshold, no alarm signal is triggered, and there is no abnormal vibration in the hydro-generator unit.

[0144] Embodiment 3 is an embodiment of the present invention. This embodiment provides a system for constructing a fault diagnosis system for a hydro-generator unit, including:

[0145] A vibration signal acquisition module, configured to acquire the vibration signals of the hydro-generator unit and construct an initial vibration sample matrix based on the acquired vibration signals;

[0146] A data normalization module, configured to calculate the mean and standard deviation of each feature variable in the initial vibration sample matrix respectively, and perform normalization processing based on this to obtain a normalized data matrix;

[0147] A principal component extraction module, configured to construct a covariance matrix based on the normalized data matrix and perform eigenvalue decomposition to obtain a set of principal component vectors composed of eigenvalues and their corresponding eigenvectors;

[0148] A principal component score calculation module, configured to select several principal component vectors corresponding to the largest eigenvalues from the set of principal component vectors and calculate a principal component score matrix in combination with the normalized data matrix;

[0149] A statistic analysis module, configured to calculate the principal component space statistic and the predicted residual sum of squares based on the principal component score matrix and the set of principal component vectors;

[0150] A threshold generation module, configured to determine the control thresholds of the principal component space statistic and the predicted residual sum of squares based on the healthy operation samples of the hydro-generator unit;<000,0372>

[0151] An operating state determination module, configured to obtain the statistic of the monitoring samples according to the same processing flow as the healthy samples and compare it with the control threshold;

[0152] An abnormality determination and alarm module, configured to determine it as abnormal when either the principal component space statistic or the predicted residual sum of squares of the monitoring samples exceeds the control threshold, and output the diagnosis result and alarm information. <XXX00377>This embodiment also provides an electronic device applicable to the situation of a method for constructing a fault diagnosis system for a hydro-generator unit, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement a method for constructing a fault diagnosis system for a hydro-generator unit as proposed in the above embodiment.

[0154] This embodiment further provides a storage medium having a computer program stored thereon. When the program is executed by a processor, the method for constructing a fault diagnosis system for a hydroelectric generator set as proposed in the above embodiment is implemented.

[0155] The storage medium proposed in this embodiment and the method for constructing a fault diagnosis system for a hydroelectric generator set proposed in the above embodiment belong to the same inventive concept. The technical details not described in detail in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0156] Through the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented with the help of software and necessary general-purpose hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer's floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0157] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for constructing a fault diagnosis system for a hydroelectric generator set, characterized by: include, Collect vibration signals of the hydroelectric generator set and construct an initial vibration sample matrix based on the vibration signals; For each characteristic variable in the initial vibration sample matrix, calculate its mean and standard deviation respectively, and perform standardization based on them to obtain a standardized data matrix; Construct a covariance matrix based on the standardized data matrix, and perform eigenvalue decomposition on the covariance matrix to obtain a set of principal component vectors consisting of eigenvalues ​​and their corresponding eigenvectors; Select several principal component vectors corresponding to the largest eigenvalues ​​from the principal component vector set, and calculate the principal component score matrix by combining with the standardized data matrix; Based on the principal component score matrix and the principal component vector set, the principal component space statistics and the sum of squares of the prediction residuals are calculated to characterize the sample characteristics; Based on the healthy operation samples of hydroelectric generating units, the control thresholds of principal component spatial statistics and the sum of squares of prediction residuals are determined; The monitoring samples were processed in the same way as the healthy samples to obtain the principal component spatial statistics and the sum of squares of the prediction residuals of the samples, and compared with the control thresholds. When either the principal component spatial statistics or the sum of squares of the predicted residuals of the monitoring sample exceeds the corresponding control threshold, the operating status of the hydropower generating unit reflected by the sample is judged to be abnormal, and the diagnosis result and alarm information are output.

2. The method for constructing a fault diagnosis system for a hydroelectric generator set according to claim 1, wherein: The vibration initial sample matrix is ​​composed of multiple vibration samples collected at different time points. Each vibration sample includes six characteristic variables: main shaft horizontal deviation, main shaft vertical deviation, water guide bearing horizontal deviation, water guide bearing vertical deviation, water guide bearing horizontal swing, and water guide bearing vertical swing. Each row in the matrix corresponds to a vibration sample, and each column corresponds to a characteristic variable.

3. The method for constructing a fault diagnosis system for a hydroelectric generator set according to claim 2, wherein: The standardization process includes respectively calculating the mean and standard deviation of each column of the characteristic variable in the original vibration data matrix, and subtracting the mean of the characteristic variable from each data value and dividing it by its standard deviation to obtain a standardized data matrix.

4. The method for constructing a fault diagnosis system for a hydroelectric generator set according to claim 3, wherein: The method of obtaining a set of principal component vectors consisting of eigenvalues ​​and their corresponding eigenvectors includes constructing a covariance matrix based on a standardized data matrix, and performing eigenvalue decomposition on the covariance matrix to obtain a plurality of eigenvalues ​​and their corresponding eigenvectors, wherein the eigenvalues ​​are arranged in a diagonal matrix and the eigenvectors constitute an orthogonal matrix.

5. The method for constructing a fault diagnosis system for a hydroelectric generator set according to claim 4, characterized in that: The principal component space statistic is a quantitative indicator calculated by the ratio of the principal component score to the corresponding eigenvalue, which is used to characterize the distribution of samples in the principal component space, and its control threshold is determined in combination with the set number of principal components, number of samples and statistical confidence level.

6. The method for constructing a fault diagnosis system for a hydroelectric generator set according to claim 4, characterized in that: The predicted residual sum of squares is expressed as the residual information of the sample vector outside the principal component space, which is calculated by the difference between the original value of the sample and the reconstructed value, and is used to evaluate whether the sample has abnormal features that are not covered by the principal component space.

7. The method for constructing a fault diagnosis system for a hydroelectric generator set according to claim 4, wherein: The control threshold of the sum of squares of the predicted residuals is derived based on multiple power statistics of the eigenvalues ​​corresponding to the unselected principal components, and an intermediate variable is introduced to establish an association relationship with the critical value of the standard normal distribution to form a dynamic judgment standard for the degree of abnormality in the residual space.

8. A system for constructing a fault diagnosis system for a hydroelectric generator set, applying the method for constructing a fault diagnosis system for a hydroelectric generator set according to any one of claims 1 to 7, characterized in that: include: A vibration signal acquisition module is used to acquire vibration signals of the hydroelectric generator set and construct an initial vibration sample matrix based on the acquired vibration signals; A data standardization module is used to calculate the mean and standard deviation of each characteristic variable in the vibration initial sample matrix, and perform standardization processing accordingly to obtain a standardized data matrix; A principal component extraction module is used to construct a covariance matrix based on the standardized data matrix and perform eigenvalue decomposition to obtain a principal component vector set consisting of eigenvalues ​​and their corresponding eigenvectors; A principal component score calculation module is used to select a number of principal component vectors corresponding to the maximum eigenvalue from the principal component vector set, and calculate a principal component score matrix in combination with the standardized data matrix; A statistics analysis module is used to calculate the principal component space statistics and the sum of squares of the prediction residuals based on the principal component score matrix and the principal component vector set; A threshold generation module is used to determine the control thresholds of the principal component spatial statistics and the sum of squares of the prediction residuals based on the healthy operation samples of the hydropower generating units; An operating status determination module is used to obtain statistics for the monitoring samples according to the same processing flow as the healthy samples, and compare them with the control threshold; The abnormality judgment and alarm module is used to judge it as abnormal when any of the principal component space statistics or the sum of squares of the predicted residuals of the monitoring sample exceeds the control threshold, and output the diagnosis result and alarm information.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for constructing a fault diagnosis system for a hydroelectric generator set according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a method for constructing a fault diagnosis system for a hydroelectric generator set according to any one of claims 1 to 7 are implemented.

Images