Hydroelectric generating set vibration signal feature extraction method, system and equipment

Abstract

The invention discloses a hydroelectric generating set vibration signal feature extraction method, system and device, and relates to the field of feature extraction, and the method comprises the steps: obtaining a vibration signal of a hydroelectric generating set; based on the conditional bubble entropy, performing hierarchical decomposition on the time sequence of the vibration signal by adopting a moving average and moving difference method to obtain a matrix form of a corresponding operator of each layer; constructing a vector of a non-negative integer e, obtaining a matrix form of the vector in a corresponding operator of each layer, and obtaining a hierarchical component at a node e of each layer through a product of the matrix form of the corresponding operator of each layer and the time sequence; calculating a CBE value of each layer, and defining a CBE value set of all components as an improved hierarchical condition bubble entropy HCBE; the optimal parameter of the HCBE is searched, and when the non-negative integer e is smaller than a threshold value, IHCBE is output to serve as the characteristic for representing the fault. According to the method, the hierarchical condition bubble entropy is improved, so that the accuracy of hierarchical bubble entropy calculation is improved while the obtained information is more detailed.

Description

Technical Field

[0001] The present invention relates to the field of feature extraction, and particularly to a method, system and device for extracting vibration signal features of a hydropower unit. Background Art

[0002] The complex operating environment and frequently changing working conditions of hydropower units make the vibration signals contain a large amount of noise signals. Therefore, effectively expressing the non-linear and non-stationary signals generated by such non-linear systems as hydropower units is an important and difficult problem in the field of feature extraction.

[0003] Bubble entropy is a mathematical index for measuring signal complexity. Compared with other entropy theory-based feature extraction methods, it is more suitable for analyzing and extracting features in non-linear systems. However, the state information of the vibration of the hydropower unit contained in the bubble entropy value is insufficient and cannot fully express the state information of the hydropower unit. Summary of the Invention

[0004] The purpose of the present invention is to provide a feature extraction method, system and device based on improved hierarchical conditional bubble entropy to solve the problems in the prior art for the deficiencies of the above-mentioned existing technologies.

[0005] The present invention specifically provides the following technical solutions. A method for extracting vibration signal features of a hydropower unit includes: Obtaining the vibration signal of the hydropower unit; Based on the conditional bubble entropy CBE, using the methods of moving average and moving difference to hierarchically decompose the time series of the vibration signal to obtain the matrix form of the corresponding operator for each layer; the operator includes the low-frequency component and the high-frequency component of the time series; Constructing a vector representing non-negative integers e and obtaining the matrix form of the vector corresponding to each layer of the operator. By multiplying the matrix form of the corresponding operator for each layer by the time series, the hierarchical component at each layer node e is obtained; Based on the hierarchical components at each layer node e , calculating the conditional bubble entropy CBE value for each layer, and defining the set of conditional bubble entropy CBE values of all components as the improved hierarchical conditional bubble entropy value HCBE; Using a multi-strategy fusion Harris hawk algorithm to find the optimal parameters of HCBE, and when the non-negative integer e is less than the threshold, outputting the final improved hierarchical conditional bubble entropy value IHCBE as the feature for characterizing faults in different states of the hydropower unit.

[0006] Preferably, the process of obtaining the conditional bubble entropy CBE is specifically as follows: Based on the bubble entropy BE, the conditional entropy is used to replace the Rényi entropy to calculate the vector sorting distribution, and the conditional bubble entropy CBE is obtained; Among them, the conditional entropy replacing the Rényi entropy The expression of is: ; Among them, the specific expression of the conditional bubble entropy CBE is: ; Among them, The embedding dimension set for the conditional bubble entropy, Is the i Conditional probability of the row vector appearing, n Is the time series The maximum number of rows of, i Is Natural number of, Indicates that the conditional coefficient is 2 and the embedding dimension is Conditional entropy of, Indicates that the conditional coefficient is 2 and the embedding dimension is Conditional entropy of.

[0007] Preferably, the method of moving average and moving difference is used to hierarchically decompose the time series of the vibration signal to obtain the matrix form of the corresponding operator for each layer, specifically: For the time series , Define the operators And , The specific expression is: ; Among them, Is the time series length; And Respectively represent the low-frequency component and the high-frequency component of the time series, Is the Time point, Is the Time point; The method of moving average and moving difference is used to hierarchically decompose the time series of the vibration signal to obtain the matrix form of the corresponding operator for each layer, the specific expression is: ; Among them, Is the decomposition layer number, Is the operator corresponding to the Layer, .

[0008] Preferably, the vector representing the non-negative integer e The specific expression is: ; Among them, when a given integer e exists, there is a unique vector , where , represents the th vector value.

[0009] Preferably, the hierarchical component at each layer node e is obtained by multiplying the matrix form of the corresponding operator of each layer by the time series, and the specific expression is: ; Among them, is the hierarchical component at the th layer node , is the matrix form of the vector corresponding to each layer of the operator.

[0010] Preferably, the conditional bubble entropy CBE value of each layer is calculated, and the set of conditional bubble entropy CBE values of all components is defined as the improved hierarchical conditional bubble entropy value HCBE, specifically: ; Among them, h represents the level, k represents the number of decomposition layers, e represents a non-negative integer, is the embedding dimension set for the conditional bubble entropy, represents the vector of non-negative integer e with the conditional bubble entropy calculation level number k at the embedding dimension.

[0011] Preferably, the multi-strategy fusion Harris hawk optimization algorithm is used to find the optimal parameters of HCBE, and when the non-negative integer e is less than the threshold, the final improved hierarchical conditional bubble entropy value IHCBE is output, specifically: When setting the hierarchical conditional bubble entropy using the multi-strategy fusion Harris hawk optimization algorithm, the entropy value of HCBE is input into the kernel extreme learning machine for classification, and the classification accuracy of the test set is obtained; Using the multi-strategy fusion Harris hawk optimization algorithm to make the hierarchical conditional bubble entropy , and inputting the obtained entropy value into the kernel extreme learning machine for classification, the classification accuracy of the test set is obtained. If , then the current value is output. At this time, the value is the optimal number of levels, and the final improved hierarchical conditional bubble entropy value IHCBE is obtained.

[0012] The present invention provides a vibration signal feature extraction system for a hydropower unit, including: A data acquisition module for acquiring vibration signals of a hydro-generator unit; A matrix acquisition module for hierarchically decomposing the time series of vibration signals by using moving average and moving difference methods based on the conditional bubble entropy (CBE) to obtain the matrix form of the corresponding operator for each layer; the operator includes the low-frequency component and the high-frequency component of the time series; A hierarchical component module for constructing a vector representing non-negative integers e and obtaining the matrix form of the vector corresponding to each layer of the operator, and obtaining the hierarchical component at each layer node e by multiplying the matrix form of the corresponding operator for each layer by the time series; An entropy value calculation module for calculating the CBE value of each layer based on the hierarchical component at each layer node e and defining the set of CBE values of all components as the improved hierarchical conditional bubble entropy value (HCBE); A feature extraction module for using a multi-strategy fusion Harris hawk optimization algorithm to find the optimal parameter of HCBE, and when the non-negative integer e is less than the threshold, outputting the final improved hierarchical conditional bubble entropy value (IHCBE) as the feature characterizing the fault.

[0013] The present invention provides a computer device, including a memory and a processor. When a program stored in the memory is executed by the processor, the processor executes the steps of the above-mentioned method for extracting features of vibration signals of a hydro-generator unit.

[0014] Compared with the prior art, the present invention has the following remarkable advantages: Based on the conditional bubble entropy (CBE), the present invention hierarchically decomposes the original time series of vibration signals of a hydro-generator unit by using moving average and moving difference methods, making the obtained vibration signal information more detailed, providing data support for subsequent feature extraction. Also, by constructing a vector and hierarchical components representing non-negative integers e , obtaining the CBE value of each layer, and defining the set of CBE values of all components as the improved hierarchical conditional bubble entropy value (HCBE), and by finding the optimal parameter of HCBE, and when the non-negative integer e is less than the threshold, outputting the final improved hierarchical conditional bubble entropy value (IHCBE) as the extracted feature, that is, by the above-mentioned adaptive optimization algorithm to find the optimal number of layers, it can distinguish and express valuable information in the high-frequency and low-frequency component time series of the original signal, realizing that while making the obtained information more detailed through the optimization algorithm, improving the accuracy of hierarchical bubble entropy calculation, and being able to express different state information of the hydro-generator unit in a more detailed form, and better controlling the health state of the hydro-generator unit. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is the fault signal diagram of the hydraulic generator set provided in the embodiment of the present invention; among them Figure 1 (a) of it is the original signal, Figure 1 (b) of it is the noise signal, Figure 1 (c) of it is the original signal spectrogram, Figure 1 (d) of it is the noise signal spectrogram; Figure 2 This is the entropy mean distribution diagram under different decibel noises provided in the embodiment of the present invention; among them Figure 2 (a) of it is the diagram under HPE, Figure 2 (b) of it is the diagram under HBE, Figure 2 (c) of it is the diagram under IHCBE; Figure 3 This is the vibration signal waveform diagram provided in the embodiment of the present invention; among them Figure 3 (a) of it, Figure 3 (b) of it, Figure 3 (c) of it and Figure 3 (d) of it are respectively the vibration signal waveform diagrams under 4 different unit states; Figure 4 This is the different entropy visualization result diagram provided in the embodiment of the present invention; among them Figure 4 (a) of it is the diagram under IHCBE, Figure 4 (b) of it is the diagram under HBE, Figure 4 (c) of it is the diagram under HPE; Figure 5 This is the KELM classification result diagram provided in the embodiment of the present invention; among them Figure 5 (a) of it is the diagram under IHCBE-KELM, Figure 5 (b) of it is the diagram under HBE-KELM; Figure 6 This is an additional attached drawing of the KELM classification result provided in the embodiment of the present invention, which is the diagram under HPE-KELM; Figure 7 This is the overall flowchart of a method for extracting vibration signal characteristics of a hydro-generator set provided by the present invention; Figure 8 This is the flowchart of IHCBE-KELM provided by the present invention; Figure 9 This is the overall flowchart of a method for extracting vibration signal characteristics of a hydro-generator set provided by the present invention. Specific implementation manners

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts shall fall within the protection scope of the present invention.

[0017] As Figure 7 , Figure 8 and Figure 9 shown, a method for extracting the vibration signal characteristics of a hydropower unit in this embodiment includes the following steps: Step S1: Obtain the vibration signal of the hydropower unit.

[0018] Step S2: Based on the conditional bubble entropy (CBE), use the methods of moving average and moving difference to hierarchically decompose the time series of the vibration signal to obtain the matrix form of the corresponding operator for each layer; the operator includes the low-frequency component and the high-frequency component of the time series.

[0019] Bubble entropy is an algorithm based on permutation entropy. It sorts the vectors embedded in the space using permutation entropy. The metric used by bubble entropy is the number of steps of the sorting algorithm. In the sorting process, bubble sort is used to calculate the vector, and the number of exchanges required to sort it in ascending order. Then calculate the Rényi entropy of the vector sorting distribution to obtain the bubble entropy value. For a given time series signal , the calculation method of BE is as follows: (1); For BE, is set to 2, then under this condition, the Rényi entropy can be expressed as: (2); Use to represent the conditional Rényi entropy value, and represent the vector sorting distribution through it. Similar to approximate entropy, the expression of bubble entropy is: (3); Since bubble entropy only focuses on the sorting of the signal and cannot effectively transmit the signal information, in view of this problem, the present invention combines the conditional entropy and the bubble entropy theory and proposes the conditional bubble entropy. The conditional bubble entropy is that the permutation entropy in the bubble entropy is replaced by the conditional entropy, and the expression of the conditional entropy is as follows: (4); Among them, is the embedding dimension set for the conditional bubble entropy, is the conditional probability of the occurrence of the i th row vector, n is the time series 's maximum number of rows,i For natural numbers, represents the conditional entropy with a conditional coefficient of 2 and an embedding dimension of , represents the conditional entropy with a conditional coefficient of 2 and an embedding dimension of .

[0020] Among them, Figure 7 The enlarged views of HCBE and HPE in Figure 4 are shown as Figure 8 The enlargement of the test set in Figure 5 is shown specifically as

[0021] Based on the traditional CBE, HCBE uses the methods of moving average and moving difference to hierarchically decompose the original time series, and finds the optimal number of levels through the multi-strategy fusion Harris hawk optimization algorithm, making the obtained fault information more detailed and improving the accuracy of CBE calculation. The main steps of calculating the IHCBE of the given time series are as follows: For the original time series , define the operators and as: ; In the formula, is the length of the time series; and respectively represent the low-frequency component and the high-frequency component of the time series, is the th time point, is the th time point.

[0022] (2) The matrix form of the operator corresponding to the layer can be expressed as: ; In the formula, is the number of decomposition layers. If is too small, the frequency band division of the time series is not detailed enough to obtain sufficient low-frequency and high-frequency components. If is too large, the calculation efficiency is low. Considering comprehensively, is taken as 4.

[0023] Step S3: Construct a vector representing the non-negative integer e , and obtain the matrix form of the vector corresponding to each layer of operator. Obtain the hierarchical component at each layer node e through the product of the matrix form of the operator corresponding to each layer and the time series.

[0024] Construct a vector used to represent non - negative integers , where , as shown in the following formula: ; ; It can be seen that for a given integer , there is a unique corresponding to it, where , represents the th vector value.

[0025] The hierarchical component at the th - layer node can be obtained from the following formula: ; When the decomposition layer number is 4, the hierarchical decomposition process of the time series is as shown in Figure 1 .

[0026] Step S4: Based on the hierarchical components at each layer node e , calculate the conditional bubble entropy CBE value for each layer, and define the set of conditional bubble entropy CBE values of all components as the improved hierarchical conditional bubble entropy value HCBE.

[0027] Calculate the CBE value for each layer through equations (4) - (8), and define the set of CBE values of all components as HCBE. The specific process is shown in equation (9): (9); where h represents the hierarchy, k represents the decomposition layer number, e represents non - negative integers, is the embedding dimension set for the conditional bubble entropy, represents the vector of non - negative integer e with the conditional bubble entropy calculation hierarchy number k at the embedding dimension.

[0028] Step S5: Use the multi - strategy fusion Harris hawk optimization algorithm to find the optimal parameters of HCBE, and when the non - negative integer e is less than the threshold, output the final improved hierarchical conditional bubble entropy value IHCBE as the feature for characterizing faults in different states of the hydropower unit.

[0029] Since the feature extraction results of IHCBE, HBE, and HPE are all feature vectors and it is not intuitive to compare their data processing effects, TSNE is introduced to visually analyze the features they extract. TSNE is an embedding model that can map data in a high - dimensional space to a low - dimensional space and retain the local characteristics of the dataset. This algorithm is very common in papers and is mainly used for dimensionality reduction and visualization of high - dimensional data.

[0030] Algorithm verification and data classification: Verify the stability of IHCBE against different types of noise and its recognition ability respectively, and conduct comparative analysis using HBE and HPE. The experimental results show that IHCBE has the best stability, followed by HBE, and HPE has the worst stability among the three; IHCBE still has stability for different types of noise and has good anti-noise performance.

[0031] Furthermore, classify the extracted results using two classifiers, Extreme Learning Machine (ELM) and Kernel Extreme Learning Machine (KELM). The results show that IHCBE has the best feature extraction effect, and IHCBE-KELM has the highest diagnostic rate.

[0032] Classifiers: Extreme Learning Machine (ELM) and Kernel Extreme Learning Machine (KELM): The Extreme Learning Machine (ELM) does not require gradient-based backpropagation to adjust weights. Its structure is a new type of single-hidden-layer feedforward neural network, as Figure 2 shown. Compared with traditional neural networks, ELM has good generalization performance and extremely fast learning ability. The calculation process of ELM is as follows: ; The above formula can be abbreviated as: ; In the formula, is the number of hidden units; is the output of the network; is the number of samples; is the weight vector between the th hidden layer and the output; is the activation function; is the input vector; is the bias vector.

[0033] Using the least squares method, we can solve: ; where, is the regularization factor; is the diagonal matrix; is the expected output.

[0034] KELM is an algorithm that improves ELM by introducing a kernel function. It has higher stability and stronger performance while retaining the advantages of ELM. The KELM regression model can be expressed as: ; where, is the kernel matrix; As the kernel function, the radial basis function kernel RBF is selected in the present invention.

[0035] Set the hierarchical conditional bubble entropy using the multi-strategy fusion Harris hawk optimization algorithm When doing so, input the entropy value of HCBE into the kernel extreme learning machine for classification to obtain the classification accuracy of the test set ; Use the multi-strategy fusion Harris hawk optimization algorithm to make the hierarchical conditional bubble entropy be , input the obtained entropy value into the kernel extreme learning machine for classification to obtain the classification accuracy of the test set , if , then repeat , if , then output the current value. At this time value is the optimal number of levels, and the final improved hierarchical conditional bubble entropy value IHCBE is obtained. The present invention uses multi-class and multi-group data of a rotor vibration test bench to verify the feature extraction ability of the proposed method, and the comparative experiment and visualization results verify the effectiveness of the proposed method.

[0036] In one embodiment, the stability of IHCBE to different types of noise.

[0037] The complex operating environment of the hydropower unit and the frequently changing working conditions make the vibration signal contain a large amount of noise signals. This requires IHCBE, as a feature extraction tool, to have good anti-noise performance. The present invention simulates the vibration fault signal caused by uneven guide vane opening through formula (14): ; In the formula, is the simulated noisy vibration fault signal; is Gaussian white noise, and in the present invention, 45 dB, 50 dB, 55 dB, and 60 dB are taken. As Figure 1 shows, the vibration signal after being contaminated by noise is more complex, and it is more difficult to extract the fault information features.

[0038] By analyzing and comparing the distribution of IHCBE, HBE, and hierarchical permutation entropy (HPE) under different noises, the results are as Figure 2 shown.

[0039] It can be seen from Figure 2 that large fluctuations occur in the HBE and HPE of the vibration signal, indicating that the two entropy models of HBE and HPE are greatly affected by noise and have weak anti-noise performance. While at any scale, the overall distribution of the IHCBE of the vibration signal fluctuates less and is relatively stable, indicating that the IHCBE has better anti-noise performance.

[0040] Simulation experiment: Introduce HPE, HBE, and IHCBE for comparative experiments. Among them, the parameter settings are as follows: the number of levels = 4.

[0041] Rotor vibration test bench: The rotor of a hydropower unit has the characteristics of high-speed rotation and a complex operating environment, which makes it prone to damage, thus bringing great potential safety hazards to the operation and production of rotating machinery.

[0042] This test mainly uses the rotor test bench to simulate the normal state of the unit and three common faults of hydropower units, namely rubbing, imbalance, and misalignment (for the convenience of representing these four states, they are abbreviated as Nor, Rub, Imb, and Mis respectively). Among them, the Rub state is simulated by screwing the rubbing bolt into the rubbing thread box and contacting the rotating shaft; the Imb state is achieved by placing a mass block on the shaft; the Mis state is simulated by misplacing the positions of the two shafts on the coupling. When collecting signals, the unit speed is set to 1200 r / min and the sampling frequency is 2048 Hz. 100 groups of data are collected for each of the four unit states through the rotor test bench. Each group of data contains 2048 points. The vibration signals of the four unit states are as Figure 3 shown.

[0043] Use T-SNE to perform visual analysis on the feature information extracted by IHCBE, HBE, and HPE. The specific results are as Figure 4 shown. The states are normal, rubbing, imbalance, and misalignment respectively. It can be seen that the coupling features of different unit states extracted by IHCBE are not aliased; however, there are different degrees of aliasing between rubbing and misalignment in HBE; and there are different degrees of aliasing among the four states in HPE. It can be seen that IHCBE has better feature extraction ability.

[0044] Classifiers: ELM and KELM. To further compare the feature extraction effects of IHCBE, HBE, and HPE, the above feature extraction results are classified by using the extreme learning machine (ELM) and the kernel extreme learning machine. The results are as Figure 8 and Table 1.

[0045] Figure 5 and Figure 6 are the confusion matrix diagrams of IHCBE-KELM, HBE-KELM, and HPE-KELM. It can be seen that the classification accuracy of IHCBE-KELM reaches 100%. Next, the comparison methods for other feature extraction and classification are shown in Table 1.

[0046] It can be seen that the diagnostic accuracy of IHCBE-ELM is 93.0, higher than 82.9 of HBE-ELM and 68.5 of HPE-ELM. The diagnostic accuracy of IHCBE-KELM is 100, higher than 92.5 of HBE-KELM and 89.5 of HPE-KELM, further proving the effectiveness of IHCBE feature extraction and indicating that the classification effect of KELM is better than that of ELM.

[0047] Table 1 Diagnostic accuracy of different methods

[0048] The present invention proposes a vibration signal feature extraction system for a hydropower unit, including: a data acquisition module, a matrix acquisition module, a hierarchical component module, an entropy value calculation module, and a feature extraction module.

[0049] Among them, the data acquisition module is used to acquire the vibration signal of the hydropower unit; the matrix acquisition module is used to hierarchically decompose the time series of the vibration signal by using the method of moving average and moving difference based on the conditional bubble entropy CBE to obtain the matrix form of the corresponding operator for each layer; the operator includes the low-frequency component and the high-frequency component of the time series; the hierarchical component module is used to construct a vector representing non-negative integers e and obtain the matrix form of the vector for the corresponding operator of each layer, and obtain the hierarchical component at each layer node e by multiplying the matrix form of the corresponding operator of each layer by the time series; the entropy value calculation module is used to calculate the conditional bubble entropy CBE value of each layer based on the hierarchical component at each layer node e and define the set of conditional bubble entropy CBE values of all components as the improved hierarchical conditional bubble entropy value HCBE; the feature extraction module is used to use the multi-strategy fusion Harris hawk optimization algorithm to find the optimal parameters of HCBE, and when the non-negative integer e is less than the threshold, output the final improved hierarchical conditional bubble entropy value IHCBE as the feature characterizing the faults in different states of the hydropower unit.

[0050] The above content is a further detailed description of the present invention in combination with specific preferred embodiments. For those skilled in the art of the present invention, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should all be regarded as belonging to the protection scope of the present invention.

Claims

1. A method for extracting vibration signal features of a hydropower unit, characterized in that: include: Obtain vibration signals of hydropower units; Based on the conditional bubble entropy CBE, the time series of the vibration signal is hierarchically decomposed by the moving average and moving difference methods to obtain the matrix form of the operator corresponding to each layer; the operator includes the low-frequency component and the high-frequency component of the time series; Construct a representation of non-negative integers e , and obtain the matrix form of the vector corresponding to the operator at each layer, and obtain the matrix form of each layer of nodes by the product of the matrix form of the corresponding operator at each layer and the time series. e The level component at Based on each layer of nodes e The hierarchical components at the position are used to calculate the conditional bubble entropy CBE value of each layer, and the set of conditional bubble entropy CBE values ​​of all components is defined as the improved hierarchical conditional bubble entropy value HCBE; The multi-strategy fusion Harris Eagle optimization algorithm is used to find the optimal parameters of HCBE, and the e When it is less than the threshold, the final improved hierarchical conditional bubble entropy value IHCBE is output as the characteristic of fault characterization of different states of hydropower units.

2. A method for extracting vibration signal characteristics of a hydropower unit according to claim 1, characterized in that: The specific process of obtaining the conditional bubble entropy CBE is as follows: Based on the bubble entropy BE, the conditional entropy is used instead of the Rényi entropy to calculate the vector sorting distribution, and the conditional bubble entropy CBE is obtained; Among them, the conditional entropy instead of the Rényi entropy The expression is: ； The specific expression of conditional bubble entropy CBE is: ； in, The embedding dimension set for the conditional bubble entropy, For the i The conditional probability of the row vector appearing, n For time series The maximum number of rows, i for The natural number of It means the conditional coefficient is 2 and the embedding dimension is The conditional entropy of It means the conditional coefficient is 2 and the embedding dimension is The conditional entropy of .

3. A method for extracting vibration signal characteristics of a hydropower unit according to claim 1, characterized in that: The method of moving average and moving difference is used to hierarchically decompose the time series of the vibration signal to obtain the matrix form of the corresponding operator of each layer, which is specifically: For time series , define the operator and , the specific expression is: ； in, is the length of the time series; and Respectively represent the low-frequency component and high-frequency component of the time series, For the A point in time, For the time point; The moving average and moving difference methods are used to hierarchically decompose the time series of the vibration signal to obtain the matrix form of the corresponding operator at each layer. The specific expression is: ； in, is the number of decomposition layers, For the The operator corresponding to the layer, .

4. A method for extracting vibration signal characteristics of a hydropower unit according to claim 1, characterized in that: The construct represents a non-negative integer e The specific expression is: ； where, given an integer e, there is a unique vector ,in , Indicates A vector value.

5. A method for extracting vibration signal characteristics of a hydropower unit according to claim 1, characterized in that: The nodes of each layer are obtained by multiplying the matrix form of the corresponding operator of each layer with the time series. e The hierarchical component at is expressed as: ； in, For the Layer Node The level component at It is the matrix form of the vector corresponding to the operator at each layer.

6. A method for extracting vibration signal characteristics of a hydroelectric unit according to claim 1, characterized in that: The conditional bubble entropy CBE value of each layer is calculated, and the conditional bubble entropy CBE value set of all components is defined as the improved hierarchical conditional bubble entropy value HCBE, specifically: ； Among them, h represents the level, k represents the number of decomposition levels, and e represents a non-negative integer. The embedding dimension set for the conditional bubble entropy, A vector of non-negative integers e representing the number of levels k at which the conditional bubble entropy is calculated in the embedding dimension.

7. A method for extracting vibration signal characteristics of a hydroelectric unit according to claim 1, characterized in that: The multi-strategy fusion Harris Eagle optimization algorithm is used to find the optimal parameters of HCBE, and in non-negative integer e When it is less than the threshold, the final improved hierarchical conditional bubble entropy value IHCBE is output, which is: Using multi-strategy fusion Harris Eagle optimization algorithm to set hierarchical conditional bubble entropy When , the entropy value of HCBE is input into the kernel extreme learning machine for classification, and the classification accuracy of the test set is obtained. ; Using multi-strategy fusion Harris Eagle optimization algorithm, the hierarchical conditional bubble entropy is , and input the obtained entropy value into the kernel extreme learning machine for classification, and obtain the classification accuracy of the test set ,like , then output the current Value, at this time The value is the optimal number of levels, and the final improved hierarchical conditional bubble entropy value IHCBE is obtained.

8. A hydropower unit vibration signal feature extraction system, characterized in that: include: A data acquisition module, used to acquire vibration signals of the hydropower unit; A matrix acquisition module is used to hierarchically decompose the time series of the vibration signal based on the conditional bubble entropy CBE by using the moving average and moving difference methods to obtain the matrix form of the operator corresponding to each layer; the operator includes the low-frequency component and the high-frequency component of the time series; Hierarchical component module for constructing representations of non-negative integers e , and obtain the matrix form of the vector corresponding to the operator at each layer, and obtain the matrix form of each layer of nodes by the product of the matrix form of the corresponding operator at each layer and the time series. e The level component at Entropy calculation module, used to calculate the entropy value based on each layer of nodes e The hierarchical components at the position are used to calculate the conditional bubble entropy CBE value of each layer, and the set of conditional bubble entropy CBE values ​​of all components is defined as the improved hierarchical conditional bubble entropy value HCBE; The feature extraction module is used to find the optimal parameters of HCBE using the multi-strategy fusion Harris Eagle optimization algorithm, and in non-negative integer e When it is less than the threshold, the final improved hierarchical conditional bubble entropy value IHCBE is output as the feature to characterize the fault.

9. A computer device, characterized in that: It comprises a memory and a processor, wherein a program is stored in the memory, and when the program is executed by the processor, the processor executes the steps of a method for extracting vibration signal features of a hydropower unit as described in any one of claims 1 to 7.

Images