Hydropower station generator fault diagnosis method based on multi-dimensional empirical mode decomposition

Abstract

The invention discloses a hydropower station generator fault diagnosis method based on multi-dimensional empirical mode decomposition, and the method comprises the following steps: obtaining hydropower station generator signal data, carrying out the preprocessing, and constructing a generator time history signal data set; performing empirical mode decomposition on the generator time history signal data set to obtain a plurality of eigenmode function components, namely a first eigenmode function component, a second eigenmode function component and a third eigenmode function component; constructing a hydropower station generator fault diagnosis model comprising a feature extraction module and a time sequence processing module, and inputting the first eigenmode function component, the second eigenmode function component, the third eigenmode function component and the generator time history signal data set into the hydropower station generator fault diagnosis model for training. Obtaining a hydropower station generator fault diagnosis result; through the method, information of different scales can be fused, fault diagnosis can be effectively carried out on the hydropower station generator, and an accurate fault diagnosis result is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of generator fault diagnosis, and particularly to a method for diagnosing faults of a hydropower station generator based on multi-dimensional empirical mode decomposition. Background Art

[0002] The working principle of a hydropower station generator is mainly based on the process of converting water energy into mechanical energy and then converting mechanical energy into electrical energy through the principle of electromagnetic induction; specifically: Hydropower stations are usually built near reservoirs, and the stored water in the reservoir forms a certain water head (height of water) and flow rate (amount of water flow per unit time). This stored water has gravitational potential energy and kinetic energy; the water in the reservoir flows into the water turbine through pipelines (such as diversion tunnels, penstocks, etc.); when the water flow impacts the blades of the water turbine, it will push the runner to rotate. In this process, the gravitational potential energy and kinetic energy of the water are converted into the mechanical energy of the runner, and the rotation of the runner is transmitted to the generator through the shaft; when the generator rotor rotates, the magnetic field on it cuts the stator coils, thereby generating an alternating induced electromotive force in the stator coils; this process converts mechanical energy into electrical energy; the alternating current generated by the generator is stepped up by a transformer and then transmitted to various power consumption sites through transmission lines;

[0003] Therefore, hydropower station power generation utilizes the power of water flow, does not produce any harmful gases or waste, and has extremely little impact on the environment. Compared with traditional fossil energy power generation methods, hydropower station power generation can effectively reduce air pollution, water pollution, and greenhouse gas emissions. This power generation method has the advantages of being clean and renewable, meeting the requirements of modern society for environmental protection and sustainable development, and thus has been widely used in water conservancy projects.

[0004] Generator sets are important components in the power system, and their operating status and maintenance conditions directly affect the stability and reliability of the power system; therefore, it is very important to strengthen the health online monitoring of hydropower station generator sets, timely master the operating status and performance parameters of the units, and promptly discover and handle potential faults to ensure the stable operation of the power system.

[0005] The condition monitoring and fault detection of hydropower station generator sets usually rely on the real-time monitoring and diagnosis of a series of key parameters, including parameters such as vibration and temperature, which can reflect the operating status and health condition of the generator sets, so as to promptly discover abnormal situations of the generator sets and take corresponding measures for fault detection and repair; however, there is currently a lack of effective diagnosis of important parameters of hydropower stations, and there are mainly the following problems:

[0006] 1) Currently, many diagnostic methods mainly start from perspectives such as the time domain and frequency domain, converting time history signals into one or more characteristic values and analyzing faults from a macroscopic perspective. However, this method may ignore the changes in the signal at the microscopic level (i.e., each moment).

[0007] 2) The intrinsic mode function is the core concept of the empirical mode decomposition method. It reflects the local characteristics of the signal at different scales. Each sequence of intrinsic mode functions is single-component, which means that there is only one instantaneous frequency at each point of the sequence without the superposition of other frequency components. However, currently, when diagnosing faults in hydropower generator units, there is a lack of analysis of the different scale information of the signal from the perspective of the intrinsic mode function, resulting in the neglect of the local information of the signal at different scales, thus affecting the accuracy of the diagnosis results. Summary of the Invention

[0008] In view of the above-mentioned prior art, the present invention aims to provide a fault diagnosis method for hydropower generators based on multi-dimensional empirical mode decomposition, mainly to solve the technical problems existing in the above-mentioned background technology.

[0009] To achieve the above object, the technical solution of the embodiment of the present invention is realized as follows:

[0010] A fault diagnosis method for hydropower generators based on multi-dimensional empirical mode decomposition includes the following steps:

[0011] Obtain the signal data of the hydropower generator, perform preprocessing, and construct a time-history signal dataset of the generator;

[0012] Perform empirical mode decomposition on the time-history signal dataset of the generator to obtain multiple intrinsic mode function components, namely the first intrinsic mode function component, the second intrinsic mode function component, and the third intrinsic mode function component;

[0013] Construct a fault diagnosis model for hydropower generators including a feature extraction module and a time series processing module, and input the first intrinsic mode function component, the second intrinsic mode function component, the third intrinsic mode function component, and the time-history signal dataset of the generator into the fault diagnosis model for hydropower generators for training to obtain the fault diagnosis result of the hydropower generator.

[0014] Optionally, it further includes performing Fourier transform on the time-history signal dataset of the generator to obtain a frequency-domain signal dataset, and then inputting the frequency-domain signal dataset into the fault diagnosis model for hydropower generators for training.

[0015] Optionally, the feature extraction module includes multiple feature extraction units, and the multiple feature extraction units are used to process the corresponding intrinsic mode function component signals and the frequency-domain signal dataset. The specific processing process is as follows:

[0016] For the intrinsic mode function component signals and the frequency-domain signal dataset, at least one convolutional layer is used to smooth the data, and a pooling layer and a normalization layer are used to process the data to reduce the computational amount and reduce redundant information.

[0017] Optionally, a channel attention mechanism is introduced into the hydropower station generator fault diagnosis model, and at least two different intrinsic mode function component signals and the frequency domain signal are fused into a specified intrinsic mode function component signal to obtain a fused feature component.

[0018] Optionally, the number of convolutional layers is two, and each layer uses a 1*3 convolutional kernel.

[0019] Optionally, the time series processing module is used to process the fused feature component to capture the time dependence relationship in the fused feature component. The time series processing module includes two TCN layers, a fully connected layer, and a Dropout layer;

[0020] The two TCN layers are used to capture the signal information in the time series of the fused feature component;

[0021] The Dropout layer is used to avoid overfitting of data.

[0022] Optionally, the hydropower station generator fault diagnosis model further includes optimization using a loss function. The loss function is a cross-entropy function, and the expression of the cross-entropy function is:

[0023] -[y log(p)+(1-y)log(1-p)]

[0024] where y is the true value, log(p) represents the logarithm of p, and p is the probability of correct prediction.

[0025] Optionally, it further includes median filtering processing on the preprocessed data. The expression of the median filtering is:

[0026] y[t]=1 / 2*x[t]+1 / 4*(x[t-1]+x[t+1])

[0027] where t is a certain moment, x[t] is the input value at moment t, and y[t] is the output value at the corresponding moment t.

[0028] Optionally, the expression for performing empirical mode decomposition on the generator time history signal dataset is:

[0029] S i (x)=a i +b i (x-x i )+c i (x-x i ) 2 +d i (x-x i ) 3 ,i=0,1,…,n-1

[0030] wherein, a i 、b i 、c i 、d i are constants, S i is the output value, x is a variable, and x i is the i-th input value.

[0031] The beneficial effects of the present invention are as follows: The method for diagnosing faults of a hydropower station generator based on multi-dimensional empirical mode decomposition provided by the present invention preprocesses the acquired signal data of the hydropower station generator, removes invalid and duplicate data, and constructs a time history signal dataset of the generator. Then, the time history signal dataset of the generator is subjected to empirical mode decomposition to obtain characteristic information of different frequencies and different time scales. Moreover, the Fourier transform is performed on the original time history signal dataset to obtain frequency information. The characteristic information obtained through empirical mode decomposition is fused with the frequency information obtained through Fourier transform to obtain characteristics containing information of different scales. Then, this data information is input into a time series processing module for further feature extraction, and the parameters of the fault diagnosis model of the hydropower station generator are optimized using a loss function. Finally, a fault diagnosis result is obtained;

[0032] Through this method, it is possible to combine time-domain characteristic information and frequency-domain characteristic information, thereby fusing information of different scales, effectively diagnosing faults of the hydropower station generator, and obtaining an accurate fault diagnosis result. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a schematic flow chart of the method for diagnosing faults of a hydropower station generator based on multi-dimensional empirical mode decomposition provided in an embodiment of the present invention;

[0034] Figure 2 is a block diagram of the method for diagnosing faults of a hydropower station generator based on multi-dimensional empirical mode decomposition provided in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0035] The following clearly and completely describes the technical solutions in the embodiments of the present invention with reference to the accompanying drawings of the specification and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. In the following description, the expression "some embodiments" describes a subset of all possible embodiments. However, it should be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict.

[0036] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the present invention. However, it will be apparent to one of ordinary skill in the art that the present invention may be practiced without one or more of these specific details. In other instances, well-known features have not been described in order to avoid obscuring the invention.

[0037] It should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. On the contrary, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. The terms used herein are for the purpose of describing particular embodiments only and are not intended to be limiting of the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the terms "comprising" and / or "including", when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. As used herein, the term "and / or" includes any and all combinations of the associated listed items.

[0038] In the description of the embodiments of the present invention, it should be noted that, unless otherwise clearly defined and limited, the terms "mounted", "connected" and "coupled" should be interpreted in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be a direct connection or an indirect connection through an intermediate medium, and it may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the embodiments of the present invention can be understood according to specific circumstances.

[0039] To thoroughly understand the present invention, detailed structures will be presented in the following description to illustrate the technical solutions proposed by the present invention. The alternative embodiments of the present invention are described in detail below. However, in addition to these detailed descriptions, the present invention may have other embodiments.

[0040] Embodiment

[0041] Please refer to the attached Figure 1 and the attached Figure 2 , the present application provides a hydropower generator fault diagnosis method based on multi-dimensional empirical mode decomposition, including the following steps:

[0042] Obtain the hydropower generator signal data, perform preprocessing, and construct a generator time history signal dataset;

[0043] Perform empirical mode decomposition on the generator time - history signal dataset to obtain multiple intrinsic mode function components, namely the first intrinsic mode function component, the second intrinsic mode function component, and the third intrinsic mode function component;

[0044] Construct a hydropower generator fault diagnosis model including a feature extraction module and a time - series processing module. Input the first intrinsic mode function component, the second intrinsic mode function component, the third intrinsic mode function component, and the generator time - history signal dataset into the hydropower generator fault diagnosis model for training to obtain the hydropower generator fault diagnosis result.

[0045] Specifically, by acquiring hydropower generator signal data, the hydropower generator signal data usually includes various operating state parameters such as temperature, current, voltage, vibration, swing, displacement, pressure, etc. These data can be collected in real - time through various sensors installed on the hydropower generator; then pre - process the collected hydropower generator signal data.

[0046] The pre - processing steps include data cleaning and data segmentation; data cleaning can remove invalid, abnormal or duplicate data. Exemplarily, data outside the measurement range or obviously unreasonable data needs to be excluded or corrected; at the same time, for missing data, the mean, median or interpolation method can be used to fill it.

[0047] The data obtained after processing constitutes the generator time - history signal dataset, and then perform data segmentation on the generator time - history signal dataset. Specifically, divide the dataset into a training set and a test set according to a ratio of 8:2. The training set is used for training when constructing the hydropower generator fault diagnosis model to obtain a trained hydropower generator fault diagnosis model, and the test set is used to test the trained hydropower generator fault diagnosis model to test the accuracy and generalization ability of the hydropower generator fault diagnosis model.

[0048] In the obtained training set, perform empirical mode decomposition on the generator time - history signal dataset to obtain the first intrinsic mode function component, the second intrinsic mode function component, and the third intrinsic mode function component; the first intrinsic mode function component represents the highest - frequency component in the generator time - history signal, which contains some rapid changes or high - frequency noises during the operation of the generator. Through the first intrinsic mode function component, the operating characteristics of the generator in the high - frequency band can be understood; the second intrinsic mode function component represents the intermediate - frequency component in the generator time - history signal, which contains mechanical changes and electromagnetic interference during the operation of the generator, and the operating conditions of the generator during this process can be understood; the third intrinsic mode function component represents the low - frequency component in the generator time - history signal, which contains the long - term operation trend or periodic changes of the generator, and the overall operating conditions and performance of the generator can be understood.

[0049] It should be noted that the empirical mode decomposition method decomposes a complex signal into a series of intrinsic mode functions, and each intrinsic mode function represents the characteristics of the signal at different frequencies or time scales; to extract the intrinsic mode functions, it is first necessary to find the upper envelope and the lower envelope of the signal. These two envelopes pass through the local maximum and minimum points of the signal respectively and are fitted using a cubic spline interpolation function; calculate the mean of the upper envelope and the lower envelope; subtract this mean line L(t) from the original signal X(t), and the new signal obtained is an intrinsic mode function component, which represents the local characteristics of the signal at the current time scale; to obtain all the intrinsic mode function components, the above process needs to be repeated for the new signal obtained in the previous step until the stopping condition is met. Through the above iterative process, three different intrinsic mode function components can be obtained;

[0050] Therefore, the generator time-history signal is decomposed into three different intrinsic mode function components by the empirical mode decomposition method. The decomposed intrinsic mode function components contain local characteristic information of different time scales. First, the upper envelope of the original data is fitted using a cubic spline interpolation function based on all the maximum points of the time-history signal X(t), and so on, find all the minimum points, and fit all the minimum points into the lower envelope using a cubic spline interpolation function. The mean of the upper envelope and the lower envelope is denoted as L, and the original data signal X(t) is subtracted by L(t), thus obtaining a new time-history signal, which is an intrinsic mode function; continue the above steps until the decomposition is carried out three times;

[0051] Performing empirical mode decomposition on the generator time-history signal dataset can obtain characteristic information of different time scales. Then, the three intrinsic mode function components are input into the hydropower station generator fault diagnosis model for training, so as to improve the accuracy of signal processing, enhance the accuracy and reliability of the hydropower station generator fault diagnosis model for fault prediction. Moreover, since the three intrinsic mode function components have different characteristic information, the influence of different intrinsic mode function components on the fault diagnosis result can be considered, further optimizing the parameters of the hydropower station generator fault diagnosis model, training the hydropower station generator fault diagnosis model, and obtaining a trained hydropower station generator fault diagnosis model; finally, the trained hydropower station generator fault diagnosis model is tested through a test set, so as to evaluate the hydropower station generator fault diagnosis model.

[0052] As an optional implementation manner, it further includes performing Fourier transform on the generator time-history signal dataset to obtain a frequency-domain signal dataset, and then inputting the frequency-domain signal dataset into the hydropower station generator fault diagnosis model for training;

[0053] Specifically, Fourier transform is a mathematical tool for converting a signal from the time domain to the frequency domain. In the fault diagnosis of hydropower station generators, the time history signals generated by the generators often contain rich fault information. However, these signals may be difficult to directly observe and analyze in the time domain. Through Fourier transform, the time domain signals are converted into frequency domain signals, which can reflect the frequency distribution and energy distribution characteristics of the signals.

[0054] As an optional implementation manner, the feature extraction module includes a plurality of feature extraction units, and the plurality of feature extraction units are used to process the corresponding intrinsic mode function component signals and the frequency domain signal dataset. The specific processing process is as follows:

[0055] For the intrinsic mode function component signals and the frequency domain signal dataset, at least one convolutional layer is used to smooth the data, and a pooling layer and a normalization layer are used to process the data to reduce the computational amount and reduce redundant information;

[0056] A channel attention mechanism is introduced into the hydropower station generator fault diagnosis model to fuse at least two different intrinsic mode function component signals and the frequency domain signal into a specified intrinsic mode function component signal to obtain a fused feature component.

[0057] The number of convolutional layers is two, and each layer uses a 1*3 convolutional kernel;

[0058] Specifically, the plurality of feature extraction units are respectively a first feature extraction unit, a second feature extraction unit, a third feature extraction unit, and a frequency domain feature extraction unit, which are respectively used to extract the features of the hydropower station generator signals from the first intrinsic mode function component signal, the second intrinsic mode function component signal, the third mode function component signal, and the frequency domain signal dataset to obtain a first feature component, a second feature component, a third feature component, and a frequency domain signal, and fuse the first feature component, the second feature component, the third feature component, and the frequency domain signal to obtain a fused feature component;

[0059] Among them, the first feature extraction unit, the second feature extraction unit, the third feature extraction unit, and the frequency domain feature extraction unit all include two 1x3 convolutional layers, a pooling layer, and a normalization layer; a channel attention mechanism is introduced after the first feature extraction unit, the second feature extraction unit, and the frequency domain feature extraction unit;

[0060] The first intrinsic mode function component obtained by empirical mode decomposition is input into the first feature extraction unit, the second intrinsic mode function component obtained by empirical mode decomposition is input into the second feature extraction unit, and the third intrinsic mode function component obtained by empirical mode decomposition is input into the third feature extraction unit. The first feature extraction unit, the second feature extraction unit, and the third feature extraction unit respectively process the first intrinsic mode function component, the second intrinsic mode function component, and the third intrinsic mode function component, extract features in high frequency, medium frequency, and low frequency respectively, and then fuse the features output by the first feature extraction unit, the second feature extraction unit, and the third feature extraction unit, so as to fuse the information of features at different scales and can more comprehensively reflect the operating state of the generator;

[0061] The time series signal dataset of the generator is subjected to Fourier transform to obtain a frequency domain signal dataset, and then the frequency domain signal dataset is input into the frequency domain feature extraction unit. The frequency domain feature extraction unit can extract features from the frequency domain signal dataset, and fuse the features obtained after being processed by the frequency domain feature extraction unit with the features processed by the first feature extraction unit, the second feature extraction unit, and the third feature extraction unit. Since the input feature of the frequency domain feature extraction unit is the frequency domain signal feature obtained by Fourier transform of the time series signal data, fusing the output features of the frequency domain feature extraction unit and the first feature extraction unit, the second feature extraction unit, and the third feature extraction unit can contain time domain and frequency domain feature information, can capture more information, and improve the diagnostic accuracy of the hydropower station generator fault diagnosis model;

[0062] The time domain signal may contain a large amount of noise components, and these noises will interfere with the extraction of fault features; while in the frequency domain, the noises and fault signals are often distributed in different frequency ranges, so the frequency domain signal can be extracted; therefore, in addition to the intrinsic mode function component, the frequency domain signal also provides the feature information of the generator signal in the frequency domain. Fusing the time domain features processed by the first feature extraction unit, the second feature extraction unit, and the third feature extraction unit and the frequency domain features processed by the frequency domain feature extraction unit can capture more fault features and improve the accuracy of fault diagnosis.

[0063] As an optional implementation manner, the time series processing module is used to process the fused feature components and capture the time dependence relationship in the fused feature components;

[0064] The time series processing module includes two TCN layers, a fully connected layer, and a Dropout layer;

[0065] The two TCN layers are used to capture the signal information in the time series of the fused feature components;

[0066] The fully connected layer is used to synthesize the temporal signal information captured by the two TCN layers to obtain more global feature information;

[0067] The Dropout layer is used to avoid overfitting of data;

[0068] Specifically, the ability of the TCN layer to process multiple time-step inputs in parallel enables better processing of the temporal information in the fused feature components through two TCN layers, accelerating the training speed of the hydropower generator fault diagnosis model; through the fully connected layer, the combination and transformation of the fused features can be achieved, thereby capturing higher-level abstract features. The fully connected layer can convert the data output by the two TCN layers into a feature representation suitable for hydropower generator fault diagnosis; the main role of the Dropout layer is to prevent overfitting and improve the generalization ability of the hydropower generator fault diagnosis model.

[0069] As an optional implementation manner, the hydropower generator fault diagnosis model further includes optimization using a loss function. The loss function is a cross-entropy function, and the expression of the cross-entropy function is:

[0070] -[y log(p)+(1-y)log(1-p)]

[0071] where y is the true value, log(p) represents the logarithm of p, and p is the probability of correct prediction.

[0072] Specifically, the cross-entropy loss function can quantify the difference between the fault categories predicted by the hydropower generator fault diagnosis model and the actual fault categories, thereby optimizing the hydropower generator fault diagnosis model. By minimizing the cross-entropy loss, the parameters of the hydropower generator fault diagnosis model can be continuously adjusted to obtain more accurate generator fault diagnosis results.

[0073] As an optional implementation manner, it further includes median filtering the preprocessed data. The expression of the median filtering is:

[0074] y[t]=1 / 2*x[t]+1 / 4*(x[t-1]+x[t+1])

[0075] where t is a certain moment, x[t] is the input value at moment t, and y[t] is the output value at the corresponding moment t.

[0076] Specifically, the median filtering is performed on the hydropower station generator signal data after data cleaning, specifically: smoothing the hydropower station generator signal data, introducing the values at adjacent times, making the value of the output signal more stable at each time, reducing sudden changes, and improving the quality of the obtained hydropower station generator signal data; while smoothing the signal, the local characteristics of the signal can be better retained; in the median filtering process, the weight of the current value is large, while the weight of the adjacent value is small, and the current value and the adjacent value are weighted and averaged to retain the low-frequency components and reduce the high-frequency components.

[0077] As an optional implementation manner, the expression for performing empirical mode decomposition on the generator time history signal dataset is:

[0078] S i (x) = a i + b i (x - x i ) + c i (x - x i ) 2 + d i (x - x i ) 3 , i = 0, 1, …, n - 1

[0079] Among them, a i , b i , c i , d i are constants, S i is the output value, x is the variable, and x i is the i-th input value.

[0080] After preprocessing the original hydropower station generator signal data, median filtering is performed to improve the quality of the hydropower station generator signal data. In this way, a generator time history signal dataset is constructed. The generator time history signal dataset is respectively subjected to empirical mode decomposition and frequency domain conversion. After empirical mode decomposition, the first intrinsic mode function component, the second intrinsic mode function component, and the third intrinsic mode function component are obtained. The first intrinsic mode function component, the second intrinsic mode function component, and the third intrinsic mode function component and the characteristic quantities obtained through frequency domain conversion are input into the hydropower station generator fault diagnosis model for fusion to obtain a fusion characteristic component, so as to combine the time domain signal and the frequency domain signal, thereby obtaining features including more scale information. Then, the fusion characteristic component is processed, and finally the fault diagnosis result of the hydropower station generator is obtained; through this method, an accurate fault diagnosis result of the hydropower station generator can be obtained.

[0081] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. The protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A hydropower station generator fault diagnosis method based on multi-dimensional empirical mode decomposition, characterized in that: The following steps are involved: Obtain the generator signal data of the hydropower station, perform preprocessing, and construct the generator time-history signal data set; Performing empirical mode decomposition on the generator time history signal data set to obtain a plurality of intrinsic mode function components, which are a first eigenmode function component, a second eigenmode function component, and a third eigenmode function component; A hydropower station generator fault diagnosis model including a feature extraction module and a timing processing module is constructed, and the first eigenmode function component, the second eigenmode function component, the third eigenmode function component, and the generator time-history signal data set are input into the hydropower station generator fault diagnosis model for training to obtain a hydropower station generator fault diagnosis result.

2. The method for fault diagnosis of a hydropower station generator based on multi-dimensional empirical mode decomposition according to claim 1 is characterized in that: The method also includes performing Fourier transform on the generator time-history signal data set to obtain a frequency domain signal data set, and then inputting the frequency domain signal data set into the hydropower station generator fault diagnosis model for training.

3. The method for fault diagnosis of a hydropower station generator based on multi-dimensional empirical mode decomposition according to claim 2 is characterized in that: The feature extraction module includes a plurality of feature extraction units, and the plurality of feature extraction units are used to process the corresponding intrinsic mode function component signal and the frequency domain component signal data set. The specific processing process is as follows: For the eigenmode function component signal and frequency domain signal data sets, at least one convolution layer is used to smooth the data, and the pooling layer and normalization layer are used to process the data to reduce the amount of calculation and reduce redundant information.

4. The method for diagnosing faults of hydropower station generators based on multi-dimensional empirical mode decomposition according to claim 3 is characterized in that: A channel attention mechanism is introduced into the hydropower station generator fault diagnosis model to fuse at least two different intrinsic mode function component signals and the frequency domain signal into a specified intrinsic mode function component signal to obtain a fused feature component.

5. The method for fault diagnosis of a hydropower station generator based on multi-dimensional empirical mode decomposition according to claim 3 is characterized in that: There are two convolutional layers, and each layer uses a 1*3 convolution kernel.

6. The method for fault diagnosis of a hydropower station generator based on multi-dimensional empirical mode decomposition according to claim 4 is characterized in that: The timing processing module is used to process the fused feature components to capture the time dependency in the fused feature components, and the timing processing module includes two TCN layers, a fully connected layer, and a Dropout layer; The two TCN layers are used to capture the signal information in the time sequence of the fused feature component; The Dropout layer is used to avoid overfitting of the data.

7. The method for fault diagnosis of a hydropower station generator based on multi-dimensional empirical mode decomposition according to claim 1, characterized in that: The hydropower station generator fault diagnosis model also includes optimization using a loss function, the loss function is a cross entropy function, and the expression of the cross entropy function is: -[ylog(p)+(1-y)log(1-p)] Where y is the true value, log(p) represents the logarithm of p, and p is the probability of correct prediction.

8. The method for fault diagnosis of a hydropower station generator based on multi-dimensional empirical mode decomposition according to claim 1, characterized in that: The method also includes performing median filtering on the pre-processed data, wherein the expression of the median filtering is: y[t]=1 / 2*x[t]+1 / 4*(x[t-1]+x[t+1]) Where t is a certain moment, x[t] is the input value at moment t, and y[t] is the output value at the corresponding moment t.

9. The method for fault diagnosis of a hydropower station generator based on multi-dimensional empirical mode decomposition according to claim 1, characterized in that: The expression for performing empirical mode decomposition on the generator time history signal data set is: S i (x)=a i +b i (x-x i )+c i (x-x i ) 2 +d i (x-x i ) 3 ,i=0,1,…,n-1 Among them, a i 、b i 、c i d i is a constant, S i is the output value, x is the variable, x i is the ith input value.

Images