Steam turbine generator shaft voltage data denoising and fault diagnosis system and method
Through the use of de -noise algorithms and deep learning modules in the axis voltage data of the turbine generator axis voltage data and the fault diagnosis system, the problem of de -noise data de -diagnosis of axis voltage data and fault diagnosis of complex background conditions is solved, and the axis of the turbine generator shaft is realized. The effective diagnosis of voltage data and the accurate diagnosis of faults ensure the safe and stable operation of the generator set.
Patent Information
- Application Number
- CN202411241214.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-05
- Publication Date
- 2025-01-28
AI Technical Summary
The prior art is difficult to accurately denoise the shaft voltage data of the steam turbine generator and perform fault diagnosis under complex background conditions, resulting in the inability to effectively prevent the threats caused by shaft voltage and affect the safe and stable operation of the generator set.
Provide a steam -generator axis voltage data noise and fault diagnosis system, including axis voltage measurement module, noise module, waveform database, and deep learning module. Construct and train generator status recognition models to achieve fault diagnosis of axial voltage waveform.
The accurate diagnosis of the axis voltage data of the turbine generator under complex background conditions and the accurate diagnosis of internal faults, the accuracy of the generator status recognition is improved, and the safe and stable operation of the generator set is ensured.
Smart Images
Figure CN119377661A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of turbine generator fault diagnosis, and in particular to a turbine generator shaft voltage data denoising and fault diagnosis system and method. Background Art
[0002] In the modern power industry, the majority of generators are powered by steam turbine generators. Compared with other prime movers, steam turbine generators have a series of advantages, such as large single-unit power, high thermal economy, safe and reliable operation, and low unit power cost. Steam turbine generators are not only the prime movers commonly used in modern thermal power generation and nuclear power industries, but are also widely used in metallurgy, chemical industry, shipping and other sectors.
[0003] During the design, manufacturing, installation, and operation of large steam turbine generators, a voltage of a certain amplitude will be generated between the two ends of the generator shaft or between the shaft and the bearing due to various reasons. This is called shaft voltage. If the shaft voltage is high during the operation of the generator, it may break through the oil film between the shaft and the bearing, generate shaft current, and cause damage to the bearings, bearings, shaft journals and other components. In severe cases, it may also cause component damage or even cause shutdown events. Therefore, accurately measuring the shaft voltage while the generator is running and taking effective protective measures can reduce or eliminate the threat posed by the shaft voltage and ensure the safe and stable operation of the generator set.
[0004] At present, the detection of turbine generator shaft voltage is usually carried out by manual inspection, and the shaft voltage on the surface of the main shaft is measured with a multimeter at a fixed time every day. This method has many disadvantages, such as being close to the high-speed rotating shaft of the generator, which is dangerous, and the multimeter can only measure the effective value of the shaft voltage, which is not very meaningful for judging the actual operation of the main shaft of the generator. There are no relevant regulations for the shaft voltage test during operation in the existing national standards and power industry standard systems, which leads to misunderstandings about the standards when conducting shaft voltage tests on site, bringing man-made risks to the operation of the unit. At present, there is no method for denoising the shaft voltage data and internal fault diagnosis of turbine generators under complex background conditions. Summary of the invention
[0005] The purpose of the present invention is to overcome the defects of the above-mentioned technology and provide a system and method for steam turbine generator shaft voltage data denoising and fault diagnosis, which is suitable for steam turbine generator shaft voltage data denoising and internal fault diagnosis under complex background conditions.
[0006] In order to achieve the above object, the present invention provides the following technical solutions:
[0007] In one aspect, the present invention provides a system for denoising and fault diagnosis of shaft voltage data of a steam turbine generator, comprising:
[0008] The steam turbine generator shaft voltage measurement module is used to measure the shaft voltage data of the steam turbine generator to be tested;
[0009] The steam turbine generator shaft voltage data denoising module is used to remove the influence of background interference on the shaft voltage measurement results;
[0010] Steam turbine generator shaft voltage waveform database, used to provide raw data for training generator state recognition model;
[0011] Deep learning module, used to train a model that meets recognition performance using database data;
[0012] The turbine generator fault diagnosis module is used to diagnose faults based on the input shaft voltage waveform.
[0013] In some embodiments, the steam turbine generator shaft voltage measurement module is connected to the steam turbine generator to be tested through a brush, the steam turbine generator shaft voltage measurement module is connected to the steam turbine generator shaft voltage data denoising module, the steam turbine generator shaft voltage waveform database is connected to the deep learning module, the steam turbine generator shaft voltage data denoising module is connected to the steam turbine generator fault diagnosis module, and the deep learning module is connected to the steam turbine generator fault diagnosis module.
[0014] In some embodiments, the measurement data of the on-site shaft voltage is obtained through the brush, and the bulk measurement noise and background electromagnetic interference noise of the shaft voltage under different brush surface states and contact pressures are obtained.
[0015] In some embodiments, the obtained measurement data is input into a steam turbine generator shaft voltage data denoising module, and a denoising algorithm is used to remove the influence of background interference on the shaft voltage measurement result, and the shaft voltage signal is extracted.
[0016] In some embodiments, a generator state recognition model is constructed, the prepared steam turbine generator shaft voltage waveform database is sent to a deep learning module, the constructed model is trained, the data is divided into a training set and a validation set, the trained model is evaluated using an independent test set, and the generator state recognition model is trained.
[0017] In another aspect, the present invention provides a method for denoising and diagnosing a shaft voltage data of a steam turbine generator, comprising:
[0018] Step 1: Construct a shaft voltage waveform database under different generator body and excitation mechanism working conditions;
[0019] Step 2: Use the brush to obtain the measurement data of the shaft voltage on site, obtain the main measurement noise and background electromagnetic interference noise of the shaft voltage under different brush surface conditions and contact pressures, and use the corresponding denoising algorithm to remove the influence of background interference on the shaft voltage measurement results;
[0020] Step 3: Use deep learning algorithm to train the generator state recognition model;
[0021] Step 4: Identify and diagnose the generator status based on the input shaft voltage waveform.
[0022] In step 1, the parameter range of the generator body and the excitation mechanism, as well as the type and control method of the excitation mechanism, are determined. According to the determined parameter range, a series of experimental schemes are designed. According to the experimental schemes, the experiments are performed one by one and the shaft voltage waveform data are recorded. The measured shaft voltage waveform data are sorted and archived, and the sorted shaft voltage waveform data are analyzed. A machine learning model is established according to the waveform characteristics and rules under different working conditions. According to the analysis and modeling results, a shaft voltage waveform database is constructed.
[0023] In step 2, the brush is installed at the set position and the contact pressure is ensured. The shaft voltage data during the measurement is recorded, and the data is divided into a signal part and a noise part. The denoising algorithm is used for processing to extract the shaft voltage signal.
[0024] In step 3, a deep learning algorithm is used to build a generator state recognition model, and the constructed model is trained using the turbine generator shaft voltage waveform database. The data is divided into a training set and a validation set. The training set is used to optimize the model parameters, and the validation set is used to evaluate and adjust the model performance.
[0025] In step 4, the shaft voltage signal extracted on site is input into the generator state recognition model, and a corresponding fault decision is made according to the input shaft voltage characteristics.
[0026] Compared with the prior art, the steam turbine generator shaft voltage data denoising and fault diagnosis system and method provided by the present invention has the following beneficial effects:
[0027] The present invention can realize denoising of steam turbine generator shaft voltage data and internal fault diagnosis under complex background conditions, and the steam turbine generator shaft voltage database can be continuously expanded, and the accuracy of motor status recognition also increases accordingly. During the inspection of steam turbine generators by on-site inspection personnel, the diagnosis results of the steam turbine generator shaft voltage can be used to timely discover internal faults of the steam turbine generator, thereby ensuring the safe operation of related electrical equipment. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for the technical description.
[0029] Figure 1 A schematic diagram of the structure of the steam turbine generator shaft voltage data denoising and fault diagnosis system provided by the present invention;
[0030] Figure 2The present invention provides a flow chart of the method for steam turbine generator shaft voltage data denoising and fault diagnosis.
[0031] Description of reference numerals:
[0032] 1. Steam turbine generator to be tested; 2. Steam turbine generator shaft voltage measurement module; 3. Steam turbine generator shaft voltage data denoising module; 4. Steam turbine generator shaft voltage waveform database; 5. Deep learning module; 6. Steam turbine generator fault diagnosis module. DETAILED DESCRIPTION
[0033] The following is further explained in detail through specific implementation methods.
[0034] like Figure 1 As shown, the present invention provides a steam turbine generator shaft voltage data denoising and fault diagnosis system, including a steam turbine generator to be tested 1, a steam turbine generator shaft voltage measurement module 2, a steam turbine generator shaft voltage data denoising module 3, a steam turbine generator shaft voltage waveform database 4, a deep learning module 5, and a steam turbine generator fault diagnosis module 6.
[0035] The steam turbine generator shaft voltage measurement module 2 measures the shaft voltage data of the steam turbine generator to be tested, the steam turbine generator shaft voltage data denoising module 3 removes the influence of background interference on the shaft voltage measurement results, the steam turbine generator shaft voltage waveform database 4 provides the original data for training the generator state recognition model, the deep learning module 5 uses the database data to train a model that meets the recognition performance, and the steam turbine generator fault diagnosis module 6 diagnoses the fault according to the input shaft voltage waveform.
[0036] The steam turbine generator 1 to be tested is connected to the steam turbine generator shaft voltage measurement module 2 through a brush, the steam turbine generator shaft voltage measurement module 2 is connected to the steam turbine generator shaft voltage data denoising module 3, the steam turbine generator shaft voltage waveform database 4 is connected to the deep learning module 5, the steam turbine generator shaft voltage data denoising module 3 is connected to the steam turbine generator fault diagnosis module 6, and the deep learning module 5 is connected to the steam turbine generator fault diagnosis module 6.
[0037] In the application, the inspection personnel first use the brush to obtain the measurement data of the shaft voltage on site, and obtain the intrinsic measurement noise and background electromagnetic interference noise of the shaft voltage under different brush surface conditions and contact pressures.
[0038] As an implementable method, the obtained measurement data is input into the turbine generator shaft voltage data denoising module 3, and denoising algorithms such as wavelet denoising and adaptive filtering are used to remove the influence of background interference on the shaft voltage measurement result, and the shaft voltage signal is extracted.
[0039] Construct a generator state recognition model, send the prepared turbine generator shaft voltage waveform database into the deep learning module 5, train the constructed model, divide the data into a training set and a validation set, and use an independent test set to evaluate the trained model. Finally, try to adjust the model structure, optimize hyperparameters, increase the amount of training data, etc. to improve the performance of the model and train the generator state recognition model.
[0040] Furthermore, the generator state recognition model is sent to the turbine generator fault diagnosis module 6. The generator state recognition model is specifically composed of Mel spectrum conversion and deep residual shrinkage networks (DRSN). Mel spectrum is a spectrum diagram under a specific scale of Mel. It maps the spectrum of each frame in the data to the logarithmic Mel scale through short-time Fourier transform (STFT), and then uses a certain number of filter bands to obtain a multi-dimensional feature vector; DRSN is a feature extraction network built on the basis of convolutional neural network (CNN). It adds an attention mechanism and a soft threshold processing that can realize denoising function on the basis of the residual block to further enhance the network performance. This model first converts the shaft voltage signal into a Mel spectrum and then sends it to the DRSN network in the form of a spectrum for feature extraction, and identifies the state of the generator through a fully connected layer.
[0041] Furthermore, the denoised shaft voltage waveform is input into the turbine generator fault diagnosis module 6, and the generator state recognition model trained by the deep learning module 5 is used to realize the recognition and diagnosis of the generator state.
[0042] In addition, if Figure 2 As shown, the present invention also provides a method for denoising and fault diagnosis of shaft voltage data of a steam turbine generator, comprising:
[0043] Step 1: Construct a shaft voltage waveform database under different generator body and excitation mechanism working conditions.
[0044] Determine the parameter range of the generator body and excitation mechanism, including the rated power, speed, electrical characteristics of the generator, and the type and control method of the excitation mechanism. Design a series of experimental plans based on the determined parameter range. According to the experimental plans, perform experiments one by one and record the shaft voltage waveform data. Organize and archive the measured shaft voltage waveform data. Use statistical methods and signal processing technology to analyze the organized shaft voltage waveform data. Establish a machine learning model based on the waveform characteristics and rules under different working conditions. Construct a shaft voltage waveform database based on the analysis and modeling results.
[0045] Step 2: Use the brush to obtain the measurement data of the shaft voltage on site, obtain the bulk measurement noise and background electromagnetic interference noise of the shaft voltage under different brush surface conditions and contact pressures, and use the corresponding denoising algorithm to remove the influence of background interference on the shaft voltage measurement results. Among them, the denoising algorithm uses wavelet packet transform to filter and denoise the signal data. Wavelet packet transform is a signal processing technology based on the concept of wavelet transform. It decomposes the signal in layers, not only decomposing the approximate coefficients, but also decomposing the detail coefficients, providing a richer representation of the signal, thus having higher flexibility and adaptability.
[0046] Install the brush in the appropriate position and ensure sufficient contact pressure. Record the shaft voltage data during the measurement, divide the data into signal part and noise part, and use denoising algorithms such as wavelet denoising and adaptive filtering to process and extract the shaft voltage signal.
[0047] Step 3: Use a deep learning algorithm to train the generator state recognition model. The deep learning algorithm used in the present invention is a deep shrinkage residual network, which was first used in 2019 for vibration-based fault diagnosis of rotating motors under strong background noise. It is improved on the basis of the traditional deep residual network ResNet, and the soft thresholding mechanism is introduced to improve the feature learning ability of the model when processing noisy or complex data. Based on this network, the present invention uses the shaft voltage time-frequency conversion diagram to retrain it, obtains a deep learning network suitable for generator state recognition, and applies it.
[0048] Use deep learning algorithms such as convolutional neural networks and recurrent neural networks to build a generator state recognition model, use the prepared turbine generator shaft voltage waveform database to train the constructed model, divide the data into training set and validation set, use the training set to optimize the model parameters, use the validation set to evaluate and adjust the model performance, and use an independent test set to evaluate the trained model. According to the evaluation results, improve the model performance by adjusting the model structure, optimizing hyperparameters, increasing the amount of training data, etc., until the model achieves satisfactory performance.
[0049] Step 4: Based on the input shaft voltage waveform, the generator status is identified and diagnosed. The shaft voltage signal extracted on site is input into the generator status identification model, and the corresponding fault decision is made based on the input shaft voltage characteristics.
[0050] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.
Claims
1. A steam turbine generator shaft voltage data denoising and fault diagnosis system, characterized in that: include: The steam turbine generator shaft voltage measurement module is used to measure the shaft voltage data of the steam turbine generator to be tested; The steam turbine generator shaft voltage data denoising module is used to remove the influence of background interference on the shaft voltage measurement results; Steam turbine generator shaft voltage waveform database, used to provide raw data for training generator state recognition model; Deep learning module, used to train a model that meets recognition performance using database data; The turbine generator fault diagnosis module is used to diagnose faults based on the input shaft voltage waveform.
2. The turbine generator shaft voltage data denoising and fault diagnosis system according to claim 1, characterized in that: The turbine generator shaft voltage measurement module is connected to the turbine generator to be tested through a brush, the turbine generator shaft voltage measurement module is connected to the turbine generator shaft voltage data denoising module, the turbine generator shaft voltage waveform database is connected to the deep learning module, the turbine generator shaft voltage data denoising module is connected to the turbine generator fault diagnosis module, and the deep learning module is connected to the turbine generator fault diagnosis module.
3. The steam turbine generator shaft voltage data denoising and fault diagnosis system according to claim 1 or 2, characterized in that: The measurement data of the on-site shaft voltage is obtained through the brush, and the bulk measurement noise and background electromagnetic interference noise of the shaft voltage under different brush surface conditions and contact pressures are obtained.
4. The steam turbine generator shaft voltage data denoising and fault diagnosis system according to claim 1, characterized in that: The obtained measurement data is input into the turbine generator shaft voltage data denoising module, and the denoising algorithm is used to remove the influence of background interference on the shaft voltage measurement result, and the shaft voltage signal is extracted.
5. The steam turbine generator shaft voltage data denoising and fault diagnosis system according to claim 1, characterized in that: Construct a generator state recognition model, feed the prepared turbine generator shaft voltage waveform database into the deep learning module, train the constructed model, divide the data into a training set and a validation set, use an independent test set to evaluate the trained model, and train the generator state recognition model.
6. A method for denoising and fault diagnosis of shaft voltage data of a steam turbine generator, characterized in that: include: Step 1: Construct a shaft voltage waveform database under different generator body and excitation mechanism working conditions; Step 2: Use the brush to obtain the measurement data of the shaft voltage on site, obtain the main measurement noise and background electromagnetic interference noise of the shaft voltage under different brush surface conditions and contact pressures, and use the corresponding denoising algorithm to remove the influence of background interference on the shaft voltage measurement results; Step 3: Use deep learning algorithm to train the generator state recognition model; Step 4: Identify and diagnose the generator status based on the input shaft voltage waveform.
7. The method for denoising and fault diagnosis of shaft voltage data of a steam turbine generator according to claim 6, characterized in that: In step 1, the parameter range of the generator body and the excitation mechanism, as well as the type and control method of the excitation mechanism, are determined. According to the determined parameter range, a series of experimental schemes are designed. According to the experimental schemes, the experiments are performed one by one and the shaft voltage waveform data are recorded. The measured shaft voltage waveform data are sorted and archived, and the sorted shaft voltage waveform data are analyzed. A machine learning model is established according to the waveform characteristics and rules under different working conditions. According to the analysis and modeling results, a shaft voltage waveform database is constructed.
8. The method for denoising and fault diagnosis of shaft voltage data of a steam turbine generator according to claim 6, characterized in that: In step 2, the brush is installed at the set position and the contact pressure is ensured. The shaft voltage data during the measurement is recorded, and the data is divided into a signal part and a noise part. The denoising algorithm is used for processing to extract the shaft voltage signal.
9. The method for denoising and fault diagnosis of shaft voltage data of a steam turbine generator according to claim 6, characterized in that: In step 3, a deep learning algorithm is used to build a generator state recognition model, and the constructed model is trained using the turbine generator shaft voltage waveform database. The data is divided into a training set and a validation set. The training set is used to optimize the model parameters, and the validation set is used to evaluate and adjust the model performance.
10. The method for steam turbine generator shaft voltage data denoising and fault diagnosis according to claim 6, characterized in that: In step 4, the shaft voltage signal extracted on site is input into the generator state recognition model, and a corresponding fault decision is made according to the input shaft voltage characteristics.