Generator speed measuring brake system

Abstract

The present invention discloses a generator speed measurement and braking system, which relates to the technical field of generators, and includes a speed measurement module, a brake module, a signal processing module, and a control module, wherein the signal processing module receives the signal of the speed measurement module, processes and analyzes it, the brake module receives the instruction of the control module, controls the input and output of the brake, the signal processing module sends the braking control instruction to the control module, and the signal processing module receives the data information of the control module. The data collected by the mechanical speed measurement unit and the residual pressure speed measurement unit of the present invention are processed by a specific filter to eliminate noise and interference. The application of RC filters and low-pass filters effectively smooths the signal and reduces the influence of high-frequency noise, making the subsequent signal analysis more accurate, thereby achieving higher-precision speed measurement in the entire system.

Description

Technical Field

[0001] The invention relates to the technical field of generators, and in particular to a generator speed measuring and braking system. Background Art

[0002] In the field of hydro-turbine power generation, accurate speed measurement is crucial to ensure the safe and efficient operation of the generator set. Traditional speed measurement methods are mainly divided into two types: mechanical speed measurement and residual pressure speed measurement. Mechanical speed measurement directly measures the rotor speed through a contact probe, while residual pressure speed measurement indirectly measures the speed by analyzing the frequency of the generator output voltage. However, both methods have their own limitations: mechanical speed measurement may be affected by wear and contamination, while residual pressure speed measurement may be affected by electrical interference. When there is a problem with speed measurement, the unit will stop at high speed during shutdown, causing damage to the generator rotor, brake, or detachment of metal parts.

[0003] Especially in an environment with more complex water level environment, the amount of data for residual pressure speed measurement is large, and due to more interference, its accuracy and measurement efficiency cannot meet the actual needs. In the existing technology, although there are multiple speed measurement methods used in parallel, they are all separate speed measurements, and data fusion is not achieved. It is impossible to achieve accurate and efficient speed measurement through complementary methods, and the safety and stability of the system cannot be guaranteed. Summary of the invention

[0004] The present invention aims to provide a generator speed measuring braking system to solve the above-mentioned problems.

[0005] The present invention is achieved through the following technical solutions:

[0006] The generator speed measurement and braking system includes a speed measurement module, a brake module, a signal processing module, and a control module, wherein the signal processing module receives the signal of the speed measurement module for processing and analysis, the brake module receives the instruction of the control module to control the input and output of the brake, the signal processing module sends the braking control instruction to the control module, and the signal processing module receives the data information of the control module.

[0007] Preferably, in the signal processing module, data processing is achieved through the following steps:

[0008] S1: Data preprocessing, preprocessing the data collected by the mechanical speed measuring unit and the data collected by the residual pressure speed measuring unit, wherein the data collected by the mechanical speed measuring unit is filtered by an RC filter, and the data collected by the residual pressure speed measuring unit is filtered by a low-pass filter;

[0009] The data collected by the mechanical speed measuring unit and the residual pressure speed measuring unit of the present invention are processed through a specific filter to eliminate noise and interference. The application of the RC filter and the low-pass filter effectively smoothes the signal and reduces the influence of high-frequency noise, making the subsequent signal analysis more accurate. This preprocessing not only improves the signal-to-noise ratio of the data, but also lays a solid foundation for subsequent signal conditioning and digital-to-analog conversion, thereby achieving higher-precision speed measurement in the entire system.

[0010] S2: Signal conditioning and digital-to-analog conversion, adjusting the amplitude and impedance of the filtered signal, then converting the analog signal into a digital signal, and finally further filtering it through a digital low-pass filter and a band-pass filter;

[0011] The filtered signal of the present invention is adjusted in amplitude and impedance, and then converted into a digital signal through an analog-to-digital converter. This process not only improves the stability of the signal, but also enables the signal to be processed by a digital system, thereby improving the flexibility and accuracy of signal processing. The further application of digital low-pass filters and band-pass filters further removes unnecessary frequency components, ensures the purity of the signal, and provides a high-quality data basis for feature extraction and subsequent analysis.

[0012] S3: Feature extraction: for the data collected by the mechanical speed measuring unit, the features related to the speed calculation in the signal are extracted, including the rotation direction, time mark, state mark, and pulse data of the generator rotor;

[0013] For the data collected by the residual pressure speed measurement unit, extract the features related to the speed calculation, including the residual pressure frequency, the accumulated running time, and the time mark;

[0014] The present invention conducts in-depth analysis on the data collected by the mechanical speed measuring unit and the residual pressure speed measuring unit, and the system can extract key features closely related to the speed calculation. These features include the rotation direction, time mark, state mark and pulse data of the generator rotor, as well as the residual pressure frequency and the accumulated running time. The extraction of these features not only improves the accuracy of the speed calculation, but also enhances the system's ability to monitor the operating status of the generator, allowing the system to respond more effectively to various operating conditions, thereby improving the reliability and safety of the entire system.

[0015] S4: Analyze and process the data of the mechanical speed measuring unit and the residual pressure speed measuring unit, denoise the speed of the residual pressure speed measuring unit and fuse the speed data, and output the speed information of the rotor by setting an output decision;

[0016] The present invention ensures the accuracy and reliability of data by receiving the signal of the speed measurement module and performing in-depth processing and analysis. This optimization not only improves the efficiency of signal processing, but also enhances the system's ability to respond to abnormal situations. Through precise data processing, the system can more accurately control the input and output of the brake, thereby ensuring the safe operation of the generator. In addition, the data interaction between the signal processing module and the control module is also smoother, making the decision-making and execution process of the entire system faster and more accurate.

[0017] Preferably, in S4, the analysis and processing comprises the following steps:

[0018] S4.1: Singular spectrum analysis;

[0019] Decompose the original time series data in the residual pressure speed measurement unit into two component sequences, specifically the residual pressure frequency signal sequence and the random noise sequence, and fuse and reconstruct the time series;

[0020] The trajectory matrix of the time series data of the residual pressure velocity measurement unit after fusion and reconstruction is obtained by the following formula:

[0021]

[0022] in It is a matrix composed of selected left singular vectors, representing the intensity change distribution of the original time series data at different frequencies;

[0023] It is a diagonal matrix composed of corresponding singular values, and the elements on its diagonal are singular values, where larger singular values ​​are related to real signals and smaller singular values ​​are related to noise signals. In this expression, by comparing the residual pressure frequency data of the residual pressure speed measuring unit in the time series with the speed data of the mechanical speed measuring unit in the time series, the threshold of the singular value is determined by the residual method to obtain the random noise sequence;

[0024] It is the conjugate transpose of the matrix composed of the selected right singular vectors, representing the projection of the residual voltage signal sequence data on the frequency and amplitude eigenvectors;

[0025] is the reconstructed trajectory matrix, which contains the main features of the original data of the residual pressure speed measurement unit, while removing the influence of noise, and through the conversion formula between residual pressure frequency and speed, , get the speed time series information, where is the rotor speed, is the residual voltage frequency, is the number of pole pairs of the rotating magnetic field on the rotor;

[0026] S4.2: using the decision output method, comprehensively extracting the original speed information of the residual pressure speed measuring unit, the original speed information of the mechanical speed measuring unit, and the fused speed information obtained in S4.1, and outputting the final decision;

[0027] The present invention effectively separates the residual pressure frequency signal and random noise by decomposing the data of the residual pressure speed measuring unit into two component sequences. This analysis method not only improves the accuracy of the speed data, but also enhances the system's ability to identify abnormal situations. By comparing the data of the residual pressure speed measuring unit and the mechanical speed measuring unit, the system can determine the threshold of the singular value, thereby identifying and removing the noise signal. This process not only improves the accuracy of the speed measurement, but also enhances the system's ability to monitor the operating status of the generator, providing a strong guarantee for the stable operation of the system.

[0028] Preferably, in S4.2, the decision output method comprises the following steps:

[0029] S4.2.1: Data alignment, by aligning the original speed information of the residual pressure speed measuring unit, the original speed information of the mechanical speed measuring unit and the fused speed information obtained in S4.1, the speeds of the three sources are aligned in time;

[0030] S4.2.2: Anomaly detection, through anomaly detection algorithm, outliers are marked;

[0031] S4.2.3: Priority decision, according to reliability and accuracy, set the priority of decision output, the fused speed information> the original speed information of the mechanical speed measuring unit> the original speed information of the residual pressure speed measuring unit;

[0032] S4.2.4: Output the decision result. According to the set priority, when the abnormality detection step does not detect an abnormality or detects an abnormality of low priority data, the speed information is output in the order of priority. When the highest priority abnormality is detected, the next priority data without abnormality is selected as the final output according to the priority. If all three data are abnormal, the data with the highest credibility is selected as the final output through comparison.

[0033] The decision output method in the present invention is the key to improving the robustness of the system. Through the steps of data alignment, anomaly detection, priority decision and output of decision results, the system can integrate speed information from multiple sources to ensure that the most reliable speed information can be output under different circumstances. Data alignment ensures the temporal consistency of speed information from different sources, the anomaly detection algorithm can mark outliers, and the priority decision sets the order of decisions based on reliability and accuracy. This comprehensive decision-making process not only improves the response speed of the system, but also enhances the stability and accuracy of the system when facing complex operating conditions.

[0034] Preferably, the speed measurement module further includes a generator physical information acquisition unit, wherein the vibration amplitude, vibration frequency, temperature, and internal pressure information of the generator are acquired through sensors;

[0035] The introduction of the generator physical information acquisition unit provides more data support for the speed measurement and braking system. In particular, when both the mechanical speed measurement and the residual pressure speed measurement fail, the system can still monitor the running speed of the generator by collecting the vibration amplitude, vibration frequency, temperature, and internal pressure information of the generator through sensors. These additional data not only enhance the accuracy of the speed measurement, but also improve the reliability and safety of the entire system.

[0036] Preferably, the steps of the signal processing module also include S5: fusing the vibration amplitude, vibration frequency, temperature, and internal pressure information of the generator with the speed information of the generator rotor through a neural network algorithm, and when the vibration amplitude, vibration frequency, temperature, and internal pressure information of the generator are input, the speed of the generator rotor is output.

[0037] If the fused speed information obtained in S4, the original speed information of the mechanical speed measuring unit, and the original speed information data of the residual pressure speed measuring unit are all abnormal, the speed output of S4 is cut off in this step, and the speed information obtained in S5 is output separately.

[0038] Preferably, the neural network algorithm in S5 specifically includes the following steps:

[0039] S5.1 construct the input layer;

[0040] Construct the node feature matrix: ,in is the number of nodes, including all characteristic nodes and target speed nodes, is the characteristic dimension of each node, where the characteristic vector includes vibration amplitude, vibration frequency, temperature, internal pressure, and R is a real number set. Represents a Line and A real matrix of columns;

[0041] Construct the connection matrix: , represents the connection structure of the graph;

[0042] S5.2: Construct the first layer of Fourier graph transport symbols;

[0043] Node feature matrix Perform a discrete Fourier transform: , where DFT is discrete Fourier transform, Represents the node feature matrix Perform discrete Fourier transform;

[0044] Introducing a learnable weight matrix: ,in refers to the weight matrix of the first layer;

[0045] Performing element-wise multiplication in Fourier space and then applying the inverse Fourier transform (IDFT) yields:

[0046]

[0047] in is the output of the first layer, Represents the adjacency matrix Perform a discrete Fourier transform, is element-wise multiplication;

[0048] S5.3: Construct subsequent layers;

[0049] For each layer k, repeat S5.1-S5.2: ,in is the learnable weight matrix of the kth layer, where k=2,3,...,K;

[0050] S5.4: get the output layer;

[0051] By integrating S5.1-S5.3, we get the output of the last layer for:

[0052]

[0053] in , Input to the first layer of the model;

[0054] is the Fourier transform of the feature representation of the k-1th layer node;

[0055] S5.5: Model training;

[0056] The weight matrix is ​​optimized by minimizing the loss function between the predicted output and the true speed. .

[0057] Preferably, the model training specifically includes the following steps:

[0058] S5.5.1: Define loss function;

[0059] Define a loss function to measure the difference between the model prediction result and the actual speed. Its mathematical expression is:

[0060]

[0061] in, is the model's predicted probability for the sample. For positive samples, is the probability of predicting a positive; for negative samples, is the complement of the probability of predicting negative;

[0062] It is a parameter to adjust the weight of positive and negative samples;

[0063] It is a parameter to adjust the weight of difficult and easy samples;

[0064] S5.5.2: Forward propagation;

[0065] Establish a training set to obtain characteristic parameters at different speeds, specifically vibration amplitude, vibration frequency, temperature, and internal pressure parameters. Then propagate these characteristic parameters forward as the output layer, calculate the output of each node, and finally output the speed.

[0066] S5.5.3: Calculate losses;

[0067] Use the loss function of S5.5.1 to calculate the difference between the output speed and the actual speed. This difference is the value of the loss function.

[0068] S5.5.4: Back propagation;

[0069] Calculate the loss function with respect to the weight matrix This process is achieved through the chain rule, that is, by calculating the loss of the weight matrix The partial derivatives (gradients) of , and pass these gradients back to each layer of the network;

[0070] S5.5.5: Parameter update and iterative training;

[0071] Update the parameters in the network according to the gradient information so that the value of the loss function continues to decrease, and repeat S5.5.1-S5.5.4 until the performance of the model on the validation set no longer improves.

[0072] Steps S5.1 to S5.5 describe the construction and training process of the neural network in detail. The input layer, Fourier graph transport, subsequent layers, and output layer are constructed, and the model is trained to optimize the weight matrix and improve the accuracy of speed prediction. This process not only improves the generalization ability of the model, but also enhances the system's ability to predict the generator speed under different operating conditions. By minimizing the loss function between the predicted output and the actual speed, the system can continuously optimize the model parameters, improve the accuracy of the prediction and the stability of the system.

[0073] Preferably, the control module includes a braking control unit, a communication interface unit and an output instruction unit, wherein the communication interface unit receives the fused rotational speed in S4 and the braking instruction issued by the signal processing module. When the braking instruction is issued, the braking control unit compares whether the real-time rotor speed exceeds the braking threshold. If it does not exceed the threshold, a braking instruction is issued to the output instruction unit. If it exceeds the braking threshold, the braking control unit issues an error report instruction to the signal processing module. At this time, the signal processing module needs to confirm whether to brake for the second time. When the braking control unit receives the braking instruction again, regardless of whether it exceeds the braking threshold, a braking instruction is issued to the braking module through the output instruction unit.

[0074] The coordinated work of the brake control unit, the communication interface unit and the output command unit in the present invention realizes the precise control of the brake command. The communication interface unit receives the fused speed and brake command, and the brake control unit determines whether braking is required based on the real-time rotor speed. This process not only improves the response speed of the system, but also enhances the safety of the system. When the rotor speed exceeds the braking threshold, the system can issue an error report command in time, which requires secondary confirmation from the signal processing module to ensure the stable operation of the system.

[0075] Preferably, when the original speed information of the mechanical speed measuring unit is abnormal, it is necessary to notify the operation and maintenance personnel to calibrate and repair it in time. If the residual pressure speed measuring unit is also abnormal during the period without maintenance, the generator information is collected in real time through the generator physical information collection unit, and the generator speed is obtained through the neural network, providing multi-layer speed information protection for braking.

[0076] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0077] 1. The present invention conducts in-depth analysis on the data collected by the mechanical speed measuring unit and the residual pressure speed measuring unit, and the system can extract key features that are closely related to the speed calculation. These features include the rotation direction, time mark, state mark and pulse data of the generator rotor, as well as the residual pressure frequency and the accumulated running time. The extraction of these features not only improves the accuracy of the speed calculation, but also enhances the system's ability to monitor the operating status of the generator, allowing the system to respond more effectively to various operating conditions, thereby improving the reliability and safety of the entire system;

[0078] 2. The decision output method in the present invention is the key to improving the robustness of the system. Through the steps of data alignment, anomaly detection, priority decision and output of decision results, the system can integrate the speed information from multiple sources to ensure that the most reliable speed information can be output under different circumstances. Data alignment ensures the temporal consistency of speed information from different sources, the anomaly detection algorithm can mark outliers, and the priority decision sets the order of decisions based on reliability and accuracy. This comprehensive decision-making process not only improves the response speed of the system, but also enhances the stability and accuracy of the system when facing complex operating conditions;

[0079] 3. Steps S5.1 to S5.5 describe the construction and training process of the neural network in detail. The input layer, Fourier graph transport, subsequent layer, and output layer are constructed, and the model is trained to optimize the weight matrix and improve the accuracy of the speed prediction. This process not only improves the generalization ability of the model, but also enhances the system's ability to predict the generator speed under different operating conditions. By minimizing the loss function between the predicted output and the actual speed, the system can continuously optimize the model parameters, improve the accuracy of the prediction and the stability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0081] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0082] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0083] Example

[0084] like Figure 1 As shown, this embodiment includes: a generator speed measurement and braking system, including a speed measurement module, a brake module, a signal processing module, and a control module, wherein the signal processing module receives the signal of the speed measurement module for processing and analysis, the brake module receives the instruction of the control module to control the input and output of the brake, the signal processing module sends a braking control instruction to the control module, and the signal processing module receives data information from the control module.

[0085] In this embodiment, in the signal processing module, data processing is implemented through the following steps:

[0086] S1: Data preprocessing, preprocessing the data collected by the mechanical speed measuring unit and the data collected by the residual pressure speed measuring unit, wherein the data collected by the mechanical speed measuring unit is filtered by an RC filter, and the data collected by the residual pressure speed measuring unit is filtered by a low-pass filter;

[0087] The data collected by the mechanical speed measuring unit and the residual pressure speed measuring unit of the present invention are processed through a specific filter to eliminate noise and interference. The application of the RC filter and the low-pass filter effectively smoothes the signal and reduces the influence of high-frequency noise, making the subsequent signal analysis more accurate. This preprocessing not only improves the signal-to-noise ratio of the data, but also lays a solid foundation for subsequent signal conditioning and digital-to-analog conversion, thereby achieving higher-precision speed measurement in the entire system.

[0088] S2: Signal conditioning and digital-to-analog conversion, adjusting the amplitude and impedance of the filtered signal, then converting the analog signal into a digital signal, and finally further filtering it through a digital low-pass filter and a band-pass filter;

[0089] The filtered signal of the present invention is adjusted in amplitude and impedance, and then converted into a digital signal through an analog-to-digital converter. This process not only improves the stability of the signal, but also enables the signal to be processed by a digital system, thereby improving the flexibility and accuracy of signal processing. The further application of digital low-pass filters and band-pass filters further removes unnecessary frequency components, ensures the purity of the signal, and provides a high-quality data basis for feature extraction and subsequent analysis.

[0090] S3: Feature extraction: for the data collected by the mechanical speed measuring unit, the features related to the speed calculation in the signal are extracted, including the rotation direction, time mark, state mark, and pulse data of the generator rotor;

[0091] For the data collected by the residual pressure speed measurement unit, extract the features related to the speed calculation, including the residual pressure frequency, the accumulated running time, and the time mark;

[0092] The present invention conducts in-depth analysis on the data collected by the mechanical speed measuring unit and the residual pressure speed measuring unit, and the system can extract key features closely related to the speed calculation. These features include the rotation direction, time mark, state mark and pulse data of the generator rotor, as well as the residual pressure frequency and the accumulated running time. The extraction of these features not only improves the accuracy of the speed calculation, but also enhances the system's ability to monitor the operating status of the generator, allowing the system to respond more effectively to various operating conditions, thereby improving the reliability and safety of the entire system.

[0093] S4: Analyze and process the data of the mechanical speed measuring unit and the residual pressure speed measuring unit, denoise the speed of the residual pressure speed measuring unit and fuse the speed data, and output the speed information of the rotor by setting an output decision;

[0094] The present invention ensures the accuracy and reliability of data by receiving the signal of the speed measurement module and performing in-depth processing and analysis. This optimization not only improves the efficiency of signal processing, but also enhances the system's ability to respond to abnormal situations. Through precise data processing, the system can more accurately control the input and output of the brake, thereby ensuring the safe operation of the generator. In addition, the data interaction between the signal processing module and the control module is also smoother, making the decision-making and execution process of the entire system faster and more accurate.

[0095] In this embodiment, in S4, the analysis process includes the following steps:

[0096] S4.1: Singular spectrum analysis;

[0097] Decompose the original time series data in the residual pressure speed measurement unit into two component sequences, specifically the residual pressure frequency signal sequence and the random noise sequence, and fuse and reconstruct the time series;

[0098] The trajectory matrix of the time series data of the residual pressure velocity measurement unit after fusion and reconstruction is obtained by the following formula:

[0099]

[0100] in It is a matrix composed of selected left singular vectors, representing the intensity change distribution of the original time series data at different frequencies;

[0101] It is a diagonal matrix composed of corresponding singular values, and the elements on its diagonal are singular values, where larger singular values ​​are related to real signals and smaller singular values ​​are related to noise signals. In this expression, by comparing the residual pressure frequency data of the residual pressure speed measuring unit in the time series with the speed data of the mechanical speed measuring unit in the time series, the threshold of the singular value is determined by the residual method to obtain the random noise sequence;

[0102] It is the conjugate transpose of the matrix composed of the selected right singular vectors, representing the projection of the residual voltage signal sequence data on the frequency and amplitude eigenvectors;

[0103] is the reconstructed trajectory matrix, which contains the main features of the original data of the residual pressure speed measurement unit, while removing the influence of noise, and through the conversion formula between residual pressure frequency and speed, , get the speed time series information, where is the rotor speed, is the residual voltage frequency, is the number of pole pairs of the rotating magnetic field on the rotor;

[0104] S4.2: using the decision output method, comprehensively extracting the original speed information of the residual pressure speed measuring unit, the original speed information of the mechanical speed measuring unit, and the fused speed information obtained in S4.1, and outputting the final decision;

[0105] The present invention effectively separates the residual pressure frequency signal and random noise by decomposing the data of the residual pressure speed measuring unit into two component sequences. This analysis method not only improves the accuracy of the speed data, but also enhances the system's ability to identify abnormal situations. By comparing the data of the residual pressure speed measuring unit and the mechanical speed measuring unit, the system can determine the threshold of the singular value, thereby identifying and removing the noise signal. This process not only improves the accuracy of the speed measurement, but also enhances the system's ability to monitor the operating status of the generator, providing a strong guarantee for the stable operation of the system.

[0106] In this embodiment, in S4.2, the decision output method includes the following steps:

[0107] S4.2.1: Data alignment, by aligning the original speed information of the residual pressure speed measuring unit, the original speed information of the mechanical speed measuring unit and the fused speed information obtained in S4.1, the speeds of the three sources are aligned in time;

[0108] S4.2.2: Anomaly detection, through anomaly detection algorithm, outliers are marked;

[0109] S4.2.3: Priority decision, according to reliability and accuracy, set the priority of decision output, the fused speed information> the original speed information of the mechanical speed measuring unit> the original speed information of the residual pressure speed measuring unit;

[0110] S4.2.4: Output the decision result. According to the set priority, when the abnormality detection step does not detect an abnormality or detects an abnormality of low priority data, the speed information is output in the order of priority. When the highest priority abnormality is detected, the next priority data without abnormality is selected as the final output according to the priority. If all three data are abnormal, the data with the highest credibility is selected as the final output through comparison.

[0111] The decision output method in the present invention is the key to improving the robustness of the system. Through the steps of data alignment, anomaly detection, priority decision and output of decision results, the system can integrate speed information from multiple sources to ensure that the most reliable speed information can be output under different circumstances. Data alignment ensures the temporal consistency of speed information from different sources, the anomaly detection algorithm can mark outliers, and the priority decision sets the order of decisions based on reliability and accuracy. This comprehensive decision-making process not only improves the response speed of the system, but also enhances the stability and accuracy of the system when facing complex operating conditions.

[0112] In this embodiment, the speed measurement module also includes a generator physical information acquisition unit, in which the vibration amplitude, vibration frequency, temperature, and internal pressure information of the generator are collected through sensors;

[0113] The introduction of the generator physical information acquisition unit provides more data support for the speed measurement and braking system. In particular, when both the mechanical speed measurement and the residual pressure speed measurement fail, the system can still monitor the running speed of the generator by collecting the vibration amplitude, vibration frequency, temperature, and internal pressure information of the generator through sensors. These additional data not only enhance the accuracy of the speed measurement, but also improve the reliability and safety of the entire system.

[0114] In this embodiment, the steps of the signal processing module also include S5: through a neural network algorithm, the vibration amplitude, vibration frequency, temperature, and internal pressure information of the generator are integrated with the speed information of the generator rotor. When the vibration amplitude, vibration frequency, temperature, and internal pressure information of the generator are input, the speed of the generator rotor is output.

[0115] If the fused speed information obtained in S4, the original speed information of the mechanical speed measuring unit, and the original speed information data of the residual pressure speed measuring unit are all abnormal, the speed output of S4 is cut off in this step, and the speed information obtained in S5 is output separately.

[0116] In this embodiment, the neural network algorithm in S5 specifically includes the following steps:

[0117] S5.1 construct the input layer;

[0118] Construct the node feature matrix: ,in is the number of nodes, including all characteristic nodes and target speed nodes, is the characteristic dimension of each node, where the characteristic vector includes vibration amplitude, vibration frequency, temperature, internal pressure, and R is a real number set. Represents a Line and A real matrix of columns;

[0119] Construct the connection matrix: , represents the connection structure of the graph;

[0120] In this embodiment, the node feature matrix is a The matrix of , where i includes all nodes and the target speed node, so i is 5, where the eigenvalues ​​of vibration frequency, temperature, internal pressure, and target speed are all one, and the vibration amplitude has 3 different measured values ​​in space, so j is 7, the node feature matrix is a 5×7 matrix.

[0121] S5.2: Construct the first layer of Fourier graph transport symbols;

[0122] Node feature matrix Perform a discrete Fourier transform: , where DFT is discrete Fourier transform, Represents the node feature matrix Perform discrete Fourier transform;

[0123] Introducing a learnable weight matrix: ,in refers to the weight matrix of the first layer;

[0124] Performing element-wise multiplication in Fourier space and then applying the inverse Fourier transform (IDFT) yields:

[0125]

[0126] in is the output of the first layer, Represents the adjacency matrix Perform a discrete Fourier transform, is element-wise multiplication;

[0127] S5.3: Construct subsequent layers;

[0128] For each layer k, repeat S5.1-S5.2: ,in is the learnable weight matrix of the kth layer, where k=2,3,...,K;

[0129] S5.4: get the output layer;

[0130] By integrating S5.1-S5.3, we get the output of the last layer for:

[0131]

[0132] in , Input to the first layer of the model;

[0133] is the Fourier transform of the feature representation of the k-1th layer node;

[0134] S5.5: Model training;

[0135] The weight matrix is ​​optimized by minimizing the loss function between the predicted output and the true speed. .

[0136] In this embodiment, the model training specifically includes the following steps:

[0137] S5.5.1: Define loss function;

[0138] Define a loss function to measure the difference between the model prediction result and the actual speed. Its mathematical expression is:

[0139]

[0140] in, is the model's predicted probability for the sample. For positive samples, is the probability of predicting a positive; for negative samples, is the complement of the probability of predicting negative;

[0141] It is a parameter to adjust the weight of positive and negative samples;

[0142] It is a parameter to adjust the weight of difficult and easy samples;

[0143] S5.5.2: Forward propagation;

[0144] Establish a training set to obtain characteristic parameters at different speeds, specifically vibration amplitude, vibration frequency, temperature, and internal pressure parameters. Then propagate these characteristic parameters forward as the output layer, calculate the output of each node, and finally output the speed.

[0145] S5.5.3: Calculate losses;

[0146] Use the loss function of S5.5.1 to calculate the difference between the output speed and the actual speed. This difference is the value of the loss function.

[0147] S5.5.4: Back propagation;

[0148] Calculate the loss function with respect to the weight matrix This process is achieved through the chain rule, that is, by calculating the loss of the weight matrix The partial derivatives (gradients) of , and pass these gradients back to each layer of the network;

[0149] S5.5.5: Parameter update and iterative training;

[0150] Update the parameters in the network according to the gradient information so that the value of the loss function continues to decrease, and repeat S5.5.1-S5.5.4 until the performance of the model on the validation set no longer improves.

[0151] Steps S5.1 to S5.5 describe the construction and training process of the neural network in detail. The input layer, Fourier graph transport, subsequent layers, and output layer are constructed, and the model is trained to optimize the weight matrix and improve the accuracy of speed prediction. This process not only improves the generalization ability of the model, but also enhances the system's ability to predict the generator speed under different operating conditions. By minimizing the loss function between the predicted output and the actual speed, the system can continuously optimize the model parameters, improve the accuracy of the prediction and the stability of the system.

[0152] In this embodiment, the control module includes a braking control unit, a communication interface unit and an output instruction unit, wherein the communication interface unit receives the fused speed in S4 and the braking instruction issued by the signal processing module. When the braking instruction is issued, the braking control unit compares whether the real-time rotor speed exceeds the braking threshold. If it does not exceed the threshold, the braking instruction is issued to the output instruction unit. If it exceeds the braking threshold, the braking control unit issues an error instruction to the signal processing module. At this time, the signal processing module needs to confirm whether to brake for the second time. When the braking control unit receives the braking instruction again, regardless of whether it exceeds the braking threshold, the braking instruction is issued to the braking module through the output instruction unit.

[0153] In this embodiment, the control module further includes a brake block and a brake fault monitoring unit, wherein there are a total of 8 brake blocks, wherein each brake block is provided with a switch signal monitor, the switch signal monitor is used to detect whether the brake block is correctly braked or de-braked according to the instruction of the control module, and the switch signal monitor sends the switch information of each brake block to the brake fault detection unit;

[0154] The brake fault detection unit performs fault analysis in the following ways:

[0155] First, it is determined whether the switch information of each brake block corresponds to the instruction of the output instruction unit in time. When the output instruction unit does not output information, but the switch information of a brake block changes, it is determined that the brake block is faulty.

[0156] Secondly, when the output command unit issues a braking or reset command to all brake pads, the switch information of all brake pads is judged. If the switch information of all brake pads is the same, there is no fault. If different switch information appears, all switch information is classified. Usually, the switch information is divided into two categories, namely braking information and reset information. Then the proportion of brake information brake blocks and the proportion of reset information brake blocks are calculated. Finally, the two proportions are compared. The information with a smaller proportion is fault information, and the brake pad corresponding to the fault information is judged to be faulty.

[0157] The coordinated work of the brake control unit, the communication interface unit and the output command unit in the present invention realizes the precise control of the brake command. The communication interface unit receives the fused speed and brake command, and the brake control unit determines whether braking is required based on the real-time rotor speed. This process not only improves the response speed of the system, but also enhances the safety of the system. When the rotor speed exceeds the braking threshold, the system can issue an error report command in time, which requires secondary confirmation from the signal processing module to ensure the stable operation of the system.

[0158] In this embodiment, when the original speed information of the mechanical speed measuring unit is abnormal, it is necessary to notify the operation and maintenance personnel to calibrate and repair it in time. If the residual pressure speed measuring unit is also abnormal during the period without maintenance, the generator information is collected in real time through the generator physical information collection unit, and the generator speed is obtained through the neural network, providing multi-layer speed information protection for braking.

[0159] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. Generator speed measuring brake system, characterized by: It includes a speed measuring module, a brake module, a signal processing module, and a control module, wherein the signal processing module receives the signal of the speed measuring module, processes and analyzes it, the brake module receives the instruction of the control module, controls the input and output of the brake, the signal processing module sends the control instruction of the brake to the control module, and the signal processing module receives the data information of the control module; In the signal processing module, data processing is achieved through the following steps: S1: Data preprocessing, preprocessing the data collected by the mechanical speed measuring unit and the data collected by the residual pressure speed measuring unit, wherein the data collected by the mechanical speed measuring unit is filtered by an RC filter, and the data collected by the residual pressure speed measuring unit is filtered by a low-pass filter; S2: Signal conditioning and digital-to-analog conversion, adjusting the amplitude and impedance of the filtered signal, then converting the analog signal into a digital signal, and finally further filtering it through a digital low-pass filter and a band-pass filter; S3: Feature extraction: for the data collected by the mechanical speed measuring unit, the features related to the speed calculation in the signal are extracted, including the rotation direction, time mark, state mark, and pulse data of the generator rotor; For the data collected by the residual pressure speed measurement unit, extract the features related to the speed calculation, including the residual pressure frequency, the accumulated running time, and the time mark; S4: Analyze and process the data of the mechanical speed measuring unit and the residual pressure speed measuring unit, denoise the speed of the residual pressure speed measuring unit and fuse the speed data, and output the speed information of the rotor by setting an output decision; In S4, the analysis process includes the following steps: S4.1: Singular spectrum analysis; Decompose the original time series data in the residual pressure speed measurement unit into two component sequences, specifically the residual pressure frequency signal sequence and the random noise sequence, and fuse and reconstruct the time series; The trajectory matrix of the time series data of the residual pressure velocity measurement unit after fusion and reconstruction is obtained by the following formula: ； in It is a matrix composed of selected left singular vectors, representing the intensity change distribution of the original time series data at different frequencies; It is a diagonal matrix composed of corresponding singular values, and the elements on its diagonal are singular values, where larger singular values ​​are related to real signals and smaller singular values ​​are related to noise signals. In this expression, by comparing the residual pressure frequency data of the residual pressure speed measuring unit in the time series with the speed data of the mechanical speed measuring unit in the time series, the threshold of the singular value is determined by the residual method to obtain the random noise sequence; It is the conjugate transpose of the matrix composed of the selected right singular vectors, representing the projection of the residual voltage signal sequence data on the frequency and amplitude eigenvectors; is the reconstructed trajectory matrix, which contains the main features of the original data of the residual pressure speed measurement unit, while removing the influence of noise, and through the conversion formula between residual pressure frequency and speed, , get the speed time series information, where is the rotor speed, is the residual voltage frequency, is the number of pole pairs of the rotating magnetic field on the rotor; S4.2: Using the decision output method, the original speed information of the residual pressure speed measuring unit, the original speed information of the mechanical speed measuring unit and the fused speed information obtained in S4.1 are comprehensively extracted, and the final decision is output.

2. The generator speed measuring and braking system according to claim 1 is characterized in that: In S4.2, the decision output method includes the following steps: S4.2.1: Data alignment, by aligning the original speed information of the residual pressure speed measuring unit, the original speed information of the mechanical speed measuring unit and the fused speed information obtained in S4.1, the speeds of the three sources are aligned in time; S4.2.2: Anomaly detection, through anomaly detection algorithm, outliers are marked; S4.2.3: Priority decision, according to reliability and accuracy, set the priority of decision output, the fused speed information> the original speed information of the mechanical speed measuring unit> the original speed information of the residual pressure speed measuring unit; S4.2.4: Output the decision result. According to the set priority, when the abnormality detection step does not detect an abnormality or detects an abnormality of low priority data, the speed information is output in order of priority. When the highest priority abnormality is detected, the next priority data without abnormality is selected as the final output according to the priority. If all three data are abnormal, the data with the highest credibility is selected as the final output through comparison.

3. The generator speed measuring and braking system according to claim 2 is characterized in that: The speed measurement module also includes a generator physical information acquisition unit, which collects the vibration amplitude, vibration frequency, temperature, and internal pressure information of the generator through sensors.

4. The generator speed measuring and braking system according to claim 3 is characterized in that: The steps of the signal processing module also include S5: fusing the vibration amplitude, vibration frequency, temperature, and internal pressure information of the generator with the speed information of the generator rotor through a neural network algorithm, and outputting the speed of the generator rotor when the vibration amplitude, vibration frequency, temperature, and internal pressure information of the generator are input; If the fused speed information obtained in S4, the original speed information of the mechanical speed measuring unit, and the original speed information data of the residual pressure speed measuring unit are all abnormal, the speed output of S4 is cut off in this step, and the speed information obtained in S5 is output separately.

5. The generator speed measuring and braking system according to claim 4 is characterized in that: The neural network algorithm in S5 specifically includes the following steps: S5.1 construct the input layer; Construct the node feature matrix: ,in is the number of nodes, including all characteristic nodes and target speed nodes, is the characteristic dimension of each node, where the characteristic vector includes vibration amplitude, vibration frequency, temperature, internal pressure, and R is a real number set. Represents a Line and A real matrix of columns; Construct the connection matrix: , represents the connection structure of the graph; S5.2: Construct the first layer of Fourier graph transport symbols; Node feature matrix Perform a discrete Fourier transform: , where DFT is discrete Fourier transform, Represents the node feature matrix Perform discrete Fourier transform; Introducing a learnable weight matrix: ,in refers to the weight matrix of the first layer; Performing element-wise multiplication in Fourier space and then applying the inverse Fourier transform (IDFT) yields: ； in is the output of the first layer, Represents the adjacency matrix Perform a discrete Fourier transform, is element-wise multiplication; S5.3: Construct subsequent layers; For each layer k, repeat S5.1-S5.2: ,in is the learnable weight matrix of the kth layer, where k=2,3,...,K; S5.4: get the output layer; By integrating S5.1-S5.3, we get the output of the last layer for: ； in , Input to the first layer of the model; is the Fourier transform of the feature representation of the k-1th layer node; S5.5: Model training; The weight matrix is ​​optimized by minimizing the loss function between the predicted output and the true speed. .

6. The generator speed measuring and braking system according to claim 5 is characterized in that: In model training, the following steps are included: S5.5.1: Define loss function; Define a loss function to measure the difference between the model prediction result and the actual speed. Its mathematical expression is: ； in, is the model's predicted probability for the sample. For positive samples, is the probability of predicting a positive; for negative samples, is the complement of the probability of predicting negative; It is a parameter to adjust the weight of positive and negative samples; It is a parameter to adjust the weight of difficult and easy samples; S5.5.2: Forward propagation; Establish a training set to obtain characteristic parameters at different speeds, specifically vibration amplitude, vibration frequency, temperature, and internal pressure parameters. Then propagate these characteristic parameters forward as the output layer, calculate the output of each node, and finally output the speed. S5.5.3: Calculate losses; Use the loss function of S5.5.1 to calculate the difference between the output speed and the actual speed. This difference is the value of the loss function. S5.5.4: Back propagation; Calculate the loss function with respect to the weight matrix This process is achieved through the chain rule, that is, by calculating the loss of the weight matrix and pass these gradients back to each layer of the network; S5.5.5: Parameter update and iterative training; Update the parameters in the network according to the gradient information so that the value of the loss function continues to decrease, and repeat S5.5.1-S5.5.4 until the performance of the model on the validation set no longer improves.

7. The generator speed measuring and braking system according to claim 6 is characterized in that: The control module includes a braking control unit, a communication interface unit and an output instruction unit, wherein the communication interface unit receives the fused speed in S4 and the braking instruction issued by the signal processing module. When the braking instruction is issued, the braking control unit compares whether the real-time rotor speed exceeds the braking threshold. If it does not exceed the threshold, the braking instruction is issued to the output instruction unit. If it exceeds the braking threshold, the braking control unit issues an error instruction to the signal processing module. At this time, the signal processing module needs to confirm whether to brake for the second time. When the braking control unit receives the braking instruction again, regardless of whether it exceeds the braking threshold, the braking instruction is issued to the braking module through the output instruction unit.

8. The generator speed measuring and braking system according to claim 7 is characterized in that: When the original speed information of the mechanical speed measuring unit is abnormal, it is necessary to notify the operation and maintenance personnel for timely calibration and repair. If the residual pressure speed measuring unit is also abnormal during the period without maintenance, the generator information is collected in real time through the generator physical information collection unit, and the generator speed is obtained through the neural network to provide multi-layer speed information protection for braking.

Images