Operation data analysis method and system for brushless exciter

Abstract

The invention discloses an operation data analysis method and system for a brushless exciter. The method comprises the following steps of 1, collecting multi-dimensional original data in the process of the brushless exciter through a multi-source sensor; 2, performing multi-dimensional preprocessing on the acquired data; 3, based on the preprocessed data, the operation states of the brushless exciter are classified through a preset fault classification algorithm, and classification comprises a normal operation state and at least one fault type; 4, generating corresponding data analysis output according to a classification result, wherein the output comprises but is not limited to a fault type identifier, a fault severity level and a repair suggestion; according to the implementation of the method, the fault diagnosis precision and the system reliability of the brushless exciter are remarkably improved through a comprehensive technical scheme of multi-dimensional data acquisition, deep learning classification and dynamic threshold adjustment.

Description

Technical Field

[0001] The present invention relates to the technical field of brushless exciters, and particularly to an operation data analysis method and system for a brushless exciter. Background Art

[0002] Brushless exciters (such as three-phase synchronous generators, AC excitation systems, etc.) are widely used in fields such as power systems, ship power, and industrial automation. Their core advantages include no contact wear, high reliability, and maintenance-free characteristics. However, the existing technologies still have the following deficiencies in operation data analysis and fault diagnosis:

[0003] Limitations in data acquisition and processing: Traditional methods rely on a single sensor (such as voltage, current) or local features (such as vibration frequency), and cannot comprehensively cover multi-dimensional parameters such as rotor current, stator voltage, and temperature field distribution, resulting in incomplete fault identification.

[0004] Insufficient accuracy in fault classification: Existing technologies mostly use fixed threshold methods or simple pattern recognition algorithms, and it is difficult to handle fault modes (such as rotor eccentricity, rectifier diode open circuit, etc.) under complex working conditions, and it is easy to have misjudgment or missed judgment.

[0005] Lack of dynamic adaptability: The operating environment of the brushless exciter (such as load fluctuation, temperature change) is complex, and the existing systems lack the ability to dynamically adjust thresholds and classification models, resulting in a decrease in long-term operation reliability.

[0006] Insufficient real-time performance and scalability: Traditional systems mostly rely on offline analysis or manual intervention, and cannot achieve real-time response to faults and cloud data retrospective analysis, which limits the improvement of operation and maintenance efficiency.

[0007] In summary, there is a need for an operation data analysis method and system for a brushless exciter to solve the deficiencies existing in the prior art. Summary of the Invention

[0008] In view of the deficiencies of the prior art, the present invention provides an operation data analysis method and system for a brushless exciter, aiming to solve the above problems.

[0009] To achieve the above object, the present invention provides the following technical solution: An operation data analysis method for a brushless exciter, comprising the following steps:

[0010] Step 1: Collect multi-dimensional raw data during the process of the brushless exciter through multi-source sensors;

[0011] Step 2: Perform multi-dimensional preprocessing on the collected data;

[0012] Step 3: Based on the preprocessed data, classify the operating state of the brushless exciter through a preset fault classification algorithm, and the classification includes a normal operating state and at least one fault type;

[0013] Step 4: Generate corresponding data analysis outputs according to the classification results, and the outputs include but are not limited to fault type identifiers, fault severity levels, and repair suggestions;

[0014] Step 5: Feed back the data analysis outputs to the monitoring system or control terminal to achieve real-time monitoring and fault response of the brushless exciter. Through the comprehensive technical solution of multi-dimensional data acquisition, deep learning classification, and dynamic threshold adjustment, the fault diagnosis accuracy and system reliability of the brushless exciter are significantly improved.

[0015] Further, in the said Step 1, the multi-dimensional raw data specifically includes:

[0016] Three-phase voltage and current waveform data;

[0017] Vibration acceleration and frequency data of rotating components;

[0018] Real-time temperature field distribution data of the exciter winding;

[0019] DC voltage and current data output by the rectifier;

[0020] Speed control signals and feedback signals of the control system.

[0021] Further, in the said Step 2, the multi-dimensional preprocessing includes:

[0022] Perform wavelet transform denoising on the voltage and current waveform data;

[0023] Perform Fourier transform on the vibration frequency data and extract frequency domain features;

[0024] Perform moving average filtering on the temperature field distribution data to eliminate instantaneous fluctuations;

[0025] Convert the normalized data into standardized feature vectors to adapt to subsequent classification algorithms.

[0026] Further, in the said Step 3, the fault classification algorithm includes:

[0027] Composite detection based on hardware and software thresholds, where the hardware thresholds include overcurrent, overvoltage, undervoltage, and overtemperature standards, and the software thresholds include pattern recognition of open phase, rotor eccentricity, and rectifier diode open circuit;

[0028] Fault pattern recognition based on deep learning, where the deep learning model uses a one-dimensional deep convolutional neural network to classify fault types by training the deep features of the excitation current waveform.

[0029] Further, the fault classification algorithm further includes:

[0030] Evaluating the confidence level of the fault type, where the confidence level is calculated based on the output probability distribution of the classification model;

[0031] Screening the classification result according to a preset confidence level threshold, and triggering a secondary verification process when the confidence level is lower than the threshold.

[0032] Further, in step 4, corresponding data analysis outputs are generated according to the classification result, and the data analysis outputs include:

[0033] Outputting the fault type identifier in the form of a pulse signal, where different fault types correspond to pulse sequences with different frequencies;

[0034] Visualizing the waveform characteristics of the three-phase voltage and current in the form of an RGB image, where the image is generated by mapping the three-phase data to the red, green, and blue channels respectively;

[0035] Outputting fault repair suggestions in the form of text or voice, where the fault repair suggestions are based on a preset fault-repair knowledge base.

[0036] Further, the data analysis output further includes:

[0037] Dynamically adjusting the preset hardware and software thresholds, where the adjustment is based on the statistical distribution of real-time operation data and a machine learning model of historical fault cases;

[0038] Establishing an association rule base between fault types and repair suggestions, where the rule base is jointly constructed by expert knowledge and data-driven methods.

[0039] Further, in step 1, the multi-source sensor includes a high-frequency current probe, a voltage sensor, a distributed optical fiber temperature measurement array, and a three-axis MEMS vibration sensor. The high-frequency current probe is used to monitor the rotor current of the brushless exciter, the voltage sensor is used to step down the stator three-phase voltage to a measurable range, and then collect digital signals through an analog-to-digital converter. The three-axis MEMS vibration sensor is used to monitor the vibration spectrum, and the distributed optical fiber temperature measurement array is used to obtain the temperature field distribution data of the brushless exciter. The data acquisition period is dynamically adjusted according to the exciter speed.

[0040] An operating data analysis system for a brushless exciter, which is used to execute a method as described above, includes:

[0041] A data acquisition module, which is used to obtain multi-dimensional raw data of the brushless exciter in real time;

[0042] A data preprocessing module for filtering, normalizing, and extracting features from raw data;

[0043] A fault classification module for classifying the operating state based on a preset algorithm;

[0044] An output generation module for generating fault identifiers, visualization images, and repair suggestions according to the classification results;

[0045] A feedback control module for feeding back the data analysis results to the monitoring system or the control terminal

[0046] Furthermore, the fault classification module includes:

[0047] A hardware threshold detection unit for quickly judging basic faults such as overcurrent and overvoltage;

[0048] A deep learning unit for identifying complex fault patterns through a 1D-DCNN model;

[0049] A confidence evaluation unit for calculating the confidence of the classification results and triggering a secondary verification process;

[0050] A dynamic threshold adjustment unit for updating the classification threshold according to real-time data and historical cases.

[0051] The substantial effects of the present invention:

[0052] 1. In the present invention, by synchronously collecting key parameters such as rotor current, stator voltage, vibration spectrum, and temperature field distribution, and combining algorithms such as wavelet transform and 1D-DCNN model, more than 12 typical fault types including overcurrent, open phase, and bearing faults can be identified. The fault identification accuracy is improved significantly. And through the collaborative work of the hardware threshold and the deep learning algorithm, the limitations of a single method are solved. For example, the "turn-to-turn short circuit of the rotor winding" fault is judged by combining the frequency domain characteristics of the vibration spectrum and the harmonic distortion of the rotor current, and the false alarm rate is reduced.

[0053] 2. In the present invention, through a machine learning model based on real-time operating data and historical cases, the thresholds for overvoltage, overheating, etc. can be adaptively updated, so that the system can still maintain high reliability under load fluctuations or environmental changes, and the maintenance cost is effectively reduced. And through the multi-modal output of pulse signals, visualization images, and text suggestions, a millisecond-level response to faults is achieved, avoiding equipment damage or downtime caused by delays, and the operation and maintenance efficiency is improved.

[0054] 3. In the present invention, through early fault warning and accurate repair suggestions, unplanned outages and over-maintenance are reduced, the service life of the equipment is extended, the excitation control strategy is optimized, reactive power loss is reduced, the system efficiency is improved, and it conforms to the development trend of green energy. Description of the Drawings

[0055] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0056] Figure 1 It is a schematic flowchart of the present invention.

[0057] Figure 2 It is a schematic diagram for testing the brushless exciter system of the present invention. Detailed implementation manners

[0058] To facilitate the understanding of the present invention, the present invention will be described in more detail below in conjunction with the accompanying drawings and specific implementation manners. It should be noted that when an element is expressed as "fixed to" another element, it can be directly on the other element, or there can be one or more intermediate elements therebetween. When an element is expressed as "connected to" another element, it can be directly connected to the other element, or there can be one or more intermediate elements therebetween. The terms "vertical", "horizontal", "left", "right" and similar expressions used in this specification are only for the purpose of illustration.

[0059] Unless otherwise defined, all technical and scientific terms used in this specification 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 are only for the purpose of describing specific implementation manners and are not used to limit the present invention. The term "and / or" used in this specification includes any and all combinations of one or more of the related listed items.

[0060] As Figure 1 shown, a method for analyzing the operation data of a brushless exciter includes the following steps:

[0061] Step 1: Collect multi-dimensional raw data during the process of the brushless exciter through multi-source sensors;

[0062] Step 2: Perform multi-dimensional preprocessing on the collected data;

[0063] Step 3: Based on the preprocessed data, classify the operating state of the brushless exciter through a preset fault classification algorithm, and the classification includes a normal operating state and at least one fault type;

[0064] Step 4: Generate corresponding data analysis outputs according to the classification results, and the outputs include but are not limited to fault type identifiers, fault severity levels, and repair suggestions;

[0065] Step 5: Feed the data analysis output back to the monitoring system or the control terminal to achieve real-time monitoring and fault response of the brushless exciter.

[0066] As an implementation manner, in Step 1, the multi-dimensional raw data specifically includes:

[0067] Three-phase voltage and current waveform data;

[0068] Vibration acceleration and frequency data of the rotating components;

[0069] Real-time temperature field distribution data of the exciter winding;

[0070] DC voltage and current data output by the rectifier;

[0071] Speed regulation signals and feedback signals of the control system.

[0072] As an implementation manner, in Step 2, the multi-dimensional preprocessing includes:

[0073] Perform wavelet transform denoising on the voltage and current waveform data;

[0074] Perform Fourier transform on the vibration frequency data and extract frequency domain features;

[0075] Perform moving average filtering on the temperature field distribution data to eliminate instantaneous fluctuations;

[0076] Convert the normalized data into standardized feature vectors to adapt to subsequent classification algorithms.

[0077] As an implementation manner, in Step 3, the fault classification algorithm includes:

[0078] Composite detection based on hardware and software thresholds. The hardware thresholds include overcurrent, overvoltage, undervoltage, and overtemperature standards, and the software thresholds include pattern recognition of open phase, rotor eccentricity, and rectifier diode open circuit;

[0079] Fault pattern recognition based on deep learning. The deep learning model uses a one-dimensional deep convolutional neural network to classify fault types by training the deep features of the excitation current waveform.

[0080] As an implementation manner, the fault classification algorithm further includes:

[0081] Evaluate the confidence level of the fault type. The confidence level is calculated based on the output probability distribution of the classification model;

[0082] Screen the classification results according to the preset confidence threshold. When the confidence level is lower than the threshold, trigger the secondary verification process.

[0083] As an implementation, in step 4, corresponding data analysis outputs are generated according to the classification results. The data analysis outputs include:

[0084] Output the fault type identifier in the form of a pulse signal, where different fault types correspond to pulse sequences of different frequencies;

[0085] Visualize the waveform characteristics of the three-phase voltage and current in the form of an RGB image, and the image is generated by mapping the three-phase data to the red, green, and blue channels respectively;

[0086] Output fault repair suggestions in the form of text or voice, and the fault repair suggestions are based on a preset fault-repair knowledge base.

[0087] As an implementation, the data analysis outputs also include:

[0088] Dynamically adjust the preset hardware and software thresholds, and the adjustment is based on the statistical distribution of real-time operation data and a machine learning model of historical fault cases;

[0089] Establish an association rule base for fault types and repair suggestions, and the rule base is jointly constructed by expert knowledge and data-driven methods.

[0090] As an implementation, in step 1, the multi-source sensors include a high-frequency current probe (sampling rate ≥ 100 kHz), a voltage sensor, a distributed optical fiber temperature measurement array (spatial resolution ≤ 5 mm), and a three-axis MEMS vibration sensor. The high-frequency current probe is used to monitor the rotor current of the brushless exciter, the voltage sensor is used to step down the stator three-phase voltage to a measurable range, and then the digital signal is collected through an analog-to-digital converter. The three-axis MEMS vibration sensor is used to monitor the vibration spectrum, and the distributed optical fiber temperature measurement array is used to obtain the temperature field (such as stator windings, rotor surface, bearings, etc.) distribution data of the brushless exciter. The data acquisition period is dynamically adjusted according to the exciter speed.

[0091] Rotor current preprocessing: Wavelet denoising, normalization processing, and feature extraction;

[0092] Wavelet denoising (removing high-frequency noise): ,

[0093] Among them, is the wavelet basis function, is the wavelet coefficient, and the main energy components are retained through the threshold truncation method;

[0094] Normalization processing (feature standardization):

[0095] ,

[0096] Among them, and are the mean and standard deviation of the rotor current, respectively;

[0097] Feature extraction (extracting time-domain and frequency-domain features):

[0098] ,

[0099] where FFT is the fast Fourier transform, and the main harmonic components (such as the amplitudes of the 2nd, 4th, and 6th harmonics) are extracted;

[0100] The preprocessing of the stator three-phase voltage includes moving average filtering, normalization, and phase correction

[0101] Moving average filtering (eliminating instantaneous fluctuations):

[0102] ,

[0103] where N is the length of the moving window (e.g., N = 10);

[0104] Normalization and phase correction:

[0105] ,

[0106] where is the reference voltage, is the allowable voltage fluctuation range;

[0107] The preprocessing of the vibration spectrum includes Fourier transform to extract frequency-domain features and extraction of key frequency components;

[0108] Fourier transform to extract frequency-domain features:

[0109] ,

[0110] where is the vibration acceleration time-domain signal, is the spectrum amplitude;

[0111] Extraction of key frequency components:

[0112] ,

[0113] where f1 and f2 are the fault-related frequency bands (such as the bearing fault characteristic frequency range);

[0114] The preprocessing of the temperature field distribution includes spatial averaging, outlier removal, and gradient feature extraction;

[0115] Spatial averaging and outlier removal:

[0116] ,

[0117] where T iis the temperature value of the i-th sensor, and M is the number of sensors;

[0118] Gradient feature extraction:

[0119] ,

[0120] used to detect local overheating or sudden changes in temperature difference;

[0121] The preprocessing of the excitation signal and the control command includes the normalization of the excitation signal and the timing alignment of the control command;

[0122] Normalization of the excitation signal:

[0123] ,

[0124] where I nom is the rated excitation current;

[0125] Timing alignment of the control command:

[0126] ,

[0127] to ensure consistency with the sensor sampling frequency;

[0128] On the other hand, as Figure 2 shown, this embodiment provides an operating data analysis system for a brushless exciter, which is used to execute the method as described above. The system includes:

[0129] A data acquisition module for real-time acquisition of multi-dimensional raw data of the brushless exciter;

[0130] A data preprocessing module for filtering, normalizing, and feature extraction of the raw data;

[0131] A fault classification module for classifying the operating state based on a preset algorithm;

[0132] An output generation module for generating a fault identifier, a visualization image, and a repair suggestion according to the classification result;

[0133] A feedback control module for feeding back the data analysis result to the monitoring system or the control terminal.

[0134] As an implementation manner, the fault classification module includes:

[0135] A hardware threshold detection unit for quickly judging basic faults such as overcurrent and overvoltage;

[0136] A deep learning unit for identifying complex fault patterns through a 1D-DCNN model;

[0137] A confidence evaluation unit for calculating the confidence of the classification result and triggering a secondary verification process;

[0138] A dynamic threshold adjustment unit for updating the classification threshold according to real-time data and historical cases.

[0139] It should be noted that the description and drawings of the present invention give preferred embodiments of the present invention. However, the present invention can be implemented in many different forms and is not limited to the embodiments described in this specification. These embodiments are not additional limitations to the content of the present invention. The purpose of providing these embodiments is to make the understanding of the disclosed content of the present invention more thorough and comprehensive. Moreover, the above technical features continue to be combined with each other to form various embodiments not listed above, which are all regarded as within the scope described in the specification of the present invention; further, for those of ordinary skill in the art, improvements or transformations can be made according to the above description, and all such improvements and transformations should fall within the protection scope of the appended claims of the present invention.

Claims

1. A method for analyzing operating data of a brushless exciter, characterized in that: It includes the following steps: Step 1: Collect multi-dimensional raw data during the brushless exciter process through multi-source sensors; Step 2: Perform multi-dimensional preprocessing on the collected data; Step 3: Based on the preprocessed data, classify the operating state of the brushless exciter through a preset fault classification algorithm, and the classification includes normal operating state and at least one fault type; Step 4: Generate corresponding data analysis outputs according to the classification results, and the outputs include but are not limited to fault type identification, fault severity level, and repair suggestions; Step 5: Feed back the data analysis outputs to the monitoring system or control terminal to achieve real-time monitoring and fault response of the brushless exciter.

2. The operation data analysis method for a brushless exciter according to claim 1, characterized in that, In the said Step 1, the multi-dimensional raw data specifically includes: Three-phase voltage and current waveform data; Vibration acceleration and frequency data of rotating components; Real-time temperature field distribution data of the exciter winding; DC voltage and current data output by the rectifier; Speed control signals and feedback signals of the control system.

3. The method for analyzing operating data of a brushless exciter according to claim 1, characterized in that: In the said Step 2, the multi-dimensional preprocessing includes: Perform wavelet transform denoising on voltage and current waveform data; Perform Fourier transform on vibration frequency data and extract frequency domain features; Perform moving average filtering on temperature field distribution data to eliminate instantaneous fluctuations; Convert the normalized data into standardized feature vectors to adapt to subsequent classification algorithms.

4. The operating data analysis method for a brushless exciter according to claim 1, wherein In the said Step 3, the fault classification algorithm includes: Composite detection based on hardware and software thresholds, where the hardware thresholds include overcurrent, overvoltage, undervoltage, and overtemperature standards, and the software thresholds include pattern recognition of open phase, rotor eccentricity, and rectifier diode open circuit; Fault pattern recognition based on deep learning, where the deep learning model uses a one-dimensional deep convolutional neural network to classify fault types by training the deep features of the excitation current waveform.

5. The operating data analysis method for a brushless exciter according to claim 4, wherein The fault classification algorithm further includes: Evaluate the confidence level of the fault type, and the confidence level is calculated based on the output probability distribution of the classification model; Screen the classification results according to a preset confidence threshold, and trigger a secondary verification process when the confidence level is lower than the threshold.

6. The operation data analysis method for a brushless exciter according to claim 1, characterized in that, In the said Step 4, corresponding data analysis outputs are generated according to the classification results, and the data analysis outputs include: Output the fault type identification in the form of a pulse signal, and different fault types correspond to pulse sequences with different frequencies; Visualize the waveform features of three-phase voltage and current in the form of an RGB image, and the image is generated by mapping three-phase data to the red, green, and blue channels respectively; Output fault repair suggestions in the form of text or voice, and the fault repair suggestions are based on a preset fault-repair knowledge base.

7. A method for analyzing the operation data of a brushless exciter according to claim 6, characterized in that, The data analysis outputs further include: Dynamically adjust preset hardware and software thresholds, and the adjustment is based on the statistical distribution of real-time operation data and a machine learning model of historical fault cases; Establish an association rule base for fault types and repair suggestions, and the rule base is jointly constructed by expert knowledge and data-driven methods.

8. A method for analyzing the operation data of a brushless exciter according to claim 1, characterized in that, In the said step 1, the multi-source sensor includes a high-frequency current probe, a voltage sensor, a distributed optical fiber temperature measurement array, and a three-axis MEMS vibration sensor. The high-frequency current probe is used to monitor the rotor current of the brushless exciter. The voltage sensor is used to step down the stator three-phase voltage to a measurable range, and then collect digital signals through an analog-to-digital converter. The three-axis MEMS vibration sensor is used to monitor the vibration spectrum. The distributed optical fiber temperature measurement array is used to obtain the temperature field distribution data of the brushless exciter. The data acquisition period is dynamically adjusted according to the exciter speed.

9. An operating data analysis system for a brushless exciter, characterized in that: Including: A data acquisition module, which is used to obtain the multi-dimensional raw data of the brushless exciter in real time; A data preprocessing module, which is used to filter, normalize, and extract features from the raw data; A fault classification module, which is used to classify the operating state based on a preset algorithm; An output generation module, which is used to generate a fault identifier, a visualization image, and a repair suggestion according to the classification result; A feedback control module, which is used to feedback the data analysis result to the monitoring system or the control terminal.

10. The operation data analysis system for a brushless exciter according to claim 9, characterized in that, The said fault classification module includes: A hardware threshold detection unit, which is used to perform a quick judgment on basic faults such as overcurrent and overvoltage; A deep learning unit, which is used to identify complex fault modes through a 1D-DCNN model; A confidence evaluation unit, which is used to calculate the confidence of the classification result and trigger a secondary verification process; A dynamic threshold adjustment unit, which is used to update the classification threshold according to real-time data and historical cases.

Images