Rapid positioning method and system for excitation system power cabinet fan fault

Abstract

The invention provides a rapid positioning method and system for excitation system power cabinet draught fan faults, and the method comprises the steps: collecting the operation parameters of an excitation system power cabinet draught fan in real time in a preset time period, and the operation parameters comprise the draught fan rotating speed, the bearing temperature, the vibration frequency, the current value and the voltage value; analyzing the operation parameters by using a threshold value judgment method and a trend analysis method, and judging whether the excitation system power cabinet fan has a fault or not; when the excitation system power cabinet draught fan breaks down, sound signals of the draught fan are collected, spectral analysis is conducted on the sound signals of the draught fan, and a spectral analysis result of the draught fan is obtained; and determining the fault type and the fault position of the fan based on the spectral analysis result of the fan. According to the technical scheme provided by the invention, the fan fault monitoring precision and efficiency are improved.

Description

Technical Field

[0001] This application relates to the field of electrical equipment fault detection technology, and in particular to a method and system for rapid location of faults in the power cabinet fan of an excitation system. Background Technology

[0002] The excitation system power cabinet fan is a core component in the heat dissipation system of large synchronous motors, bearing the important responsibility of ensuring the stable operation of the motor. In large synchronous motors, the excitation system power cabinet controls the magnetic field strength by adjusting the excitation current. The power devices inside generate a lot of heat when operating under high load. If the heat is not dissipated in time, the equipment temperature will rise sharply, directly affecting the lifespan of the power devices and the reliability of the motor operation. This fan, through its forced convection design, precisely directs cooling air to the key components of the power cabinet, effectively reducing their operating temperature and improving equipment stability and lifespan.

[0003] Existing fault monitoring technologies for excitation system power cabinet fans primarily rely on single-sensor threshold alarm mechanisms. These mechanisms trigger fault warnings by setting preset temperature or vibration thresholds. However, in industrial environments, instantaneous fluctuations in temperature or vibration can easily trigger these alarms, leading to frequent responses to ineffective alerts by maintenance personnel. This not only wastes manpower but may also mask the true fault signals. Furthermore, significant location delays severely restrict fault handling efficiency. Current technologies require manual reproduction of fault phenomena and point-by-point troubleshooting to pinpoint the fault source, a time-consuming process that can delay optimal maintenance, especially in complex electromechanical systems, increasing the risk of equipment damage. Therefore, there is an urgent need to propose a fault location solution that can improve monitoring efficiency and accuracy. Summary of the Invention

[0004] This application provides a method and system for rapid location of faults in the power cabinet fan of the excitation system, so as to at least solve the technical problems of high false alarm rate, serious location delay and low operation and maintenance efficiency caused by reliance on a single sensor threshold alarm mechanism in the existing excitation system power cabinet fan fault monitoring technology.

[0005] The first aspect of this application proposes a method for rapid location of faults in the power cabinet fan of an excitation system, the method comprising: The operating parameters of the power cabinet fan of the excitation system are collected in real time within a preset time period. The operating parameters include: fan speed, bearing temperature, vibration frequency, current value and voltage value. The operating parameters are analyzed using threshold judgment and trend analysis methods to determine whether the power cabinet fan of the excitation system is faulty. When the fan in the power cabinet of the excitation system fails, the sound signal of the fan is collected and the sound signal of the fan is subjected to spectrum analysis to obtain the spectrum analysis result of the fan. The fault type and location of the wind turbine are determined based on the spectrum analysis results.

[0006] Preferably, the fan speed is acquired using a Hall sensor installed on the fan main shaft; The bearing temperature is collected using a temperature sensor installed on the motor stator; The vibration frequency is collected using an ICP sensor or MEMS vibration sensor installed on the fan casing; The current and voltage values ​​were collected using a Hall current sensor and a voltage transformer connected to the wind turbine power supply line.

[0007] Preferably, the step of analyzing the operating parameters using threshold judgment and trend analysis methods to determine whether the excitation system power cabinet fan is faulty includes: Obtain the preset upper and lower threshold ranges for each of the aforementioned operating parameters; The threshold judgment method is used to determine whether each of the operating parameters is within its corresponding upper and lower threshold range. If so, the trend analysis method is used to determine whether the excitation system power cabinet fan is faulty. Otherwise, the excitation system power cabinet fan is determined to be faulty. Alternatively, obtain the preset deviation threshold for each of the aforementioned operating parameters; Based on the real-time acquisition of the operating parameters of the excitation system power cabinet fan from time 1 to time t-1 within a preset time period, the predicted values ​​of each operating parameter at time t are determined by the trend analysis method. Determine the absolute value of the difference between the predicted value of each operating parameter at time t and its corresponding real-time collected value at time t. Determine whether the absolute value of each difference is less than its corresponding deviation threshold. If so, use the threshold judgment method to determine whether the power cabinet fan of the excitation system is faulty; otherwise, determine that the power cabinet fan of the excitation system is faulty.

[0008] Furthermore, the step of collecting the sound signal of the fan and performing spectral analysis on the sound signal of the fan to obtain the spectral analysis results of the fan includes: The sound signal of the wind turbine is collected using a pre-deployed microphone array; The sound signal of the fan is framed using the Hanning window function, and the DC component is removed from the framed sound signal to obtain the processed sound signal. The processed audio signal is converted into a frequency domain signal using a fast Fourier transform algorithm, and a spectrum is generated based on the frequency domain signal. The spectrum diagram is used as the spectrum analysis result of the wind turbine.

[0009] Furthermore, determining the fault type and location of the wind turbine based on the spectral analysis results includes: The spectrogram is normalized and logarithmically compressed. The processed spectrum is compared with a pre-built feature library to determine the fault type and location of the wind turbine. The feature library includes: spectrum templates for the normal state of the wind turbine and fault types such as bearing wear, blade breakage, and imbalance.

[0010] Furthermore, comparing the processed spectrogram with a pre-built feature library to determine the fault type of the wind turbine includes: When the real-time collected bearing temperature value is greater than the preset temperature threshold, and the vibration kurtosis value in the spectrum is greater than the preset kurtosis threshold, the fault type of the fan is determined to be bearing overheating with local impact fault. When the percentage of total harmonic distortion of current in the spectrum is greater than the preset harmonic distortion fault threshold, and the main frequency of the vibration signal in the spectrum is equal to the passing frequency of the blade, the fault type of the wind turbine is determined to be blade breakage fault. When the vibration energy in the spectrum is concentrated in the frequency domain within a preset frequency band of the rotor's natural frequency, the fault type of the fan is determined to be dynamic balance failure.

[0011] Furthermore, the location of the fault in the wind turbine is determined through three-dimensional fault visualization; The step of determining the fault location of the wind turbine through three-dimensional fault visualization includes: The failure probability at each location of the fan is determined based on the temperature field anomaly coefficient and the vibration field anomaly coefficient, and a failure probability heat map is constructed based on the failure probability. The fault probability heatmap is superimposed on the three-dimensional geometric model of the wind turbine to achieve spatial fault location.

[0012] Furthermore, after determining the fault type and location of the wind turbine based on the spectrum analysis results, the process further includes: When the failure probability is greater than the first failure rate threshold, an emergency shutdown command is triggered. When the failure probability is less than or equal to the first failure rate threshold and greater than the second failure rate threshold, the standby fan is started and the load distribution is dynamically adjusted.

[0013] Furthermore, the method also includes: When a dynamic balance failure occurs, a self-healing control algorithm is used to determine the optimal speed command value for the fan. The fan speed is adjusted based on the optimized value of the fan speed command to achieve self-healing linkage.

[0014] A second aspect of this application provides a rapid location system for faults in the power cabinet fan of an excitation system, comprising: The acquisition module is used to acquire the operating parameters of the power cabinet fan of the excitation system in real time within a preset time period. The operating parameters include: fan speed, bearing temperature, vibration frequency, current value and voltage value. The judgment module is used to analyze the operating parameters using threshold judgment method and trend analysis method to determine whether the power cabinet fan of the excitation system is faulty; The analysis module is used to collect the sound signal of the fan and perform spectrum analysis on the sound signal of the fan when the fan of the power cabinet of the excitation system fails, so as to obtain the spectrum analysis result of the fan. The determination module is used to determine the fault type and fault location of the wind turbine based on the spectrum analysis results of the wind turbine.

[0015] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the method described in the first aspect embodiment.

[0016] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method described in the first aspect.

[0017] The technical solutions provided by the embodiments of this application bring at least the following beneficial effects: This application proposes a method and system for rapid fault location of the power cabinet fan in an excitation system. The method includes: real-time acquisition of operating parameters of the power cabinet fan in the excitation system within a preset time period, wherein the operating parameters include: fan speed, bearing temperature, vibration frequency, current value, and voltage value; analysis of the operating parameters using a threshold judgment method and a trend analysis method to determine whether the power cabinet fan in the excitation system is faulty; when the power cabinet fan in the excitation system is faulty, acquisition of the fan's sound signal and spectral analysis of the fan's sound signal to obtain the spectral analysis result of the fan; and determination of the fault type and fault location of the fan based on the spectral analysis result of the fan. The technical solution proposed in this application improves the accuracy and efficiency of fan fault monitoring.

[0018] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0019] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1This is a flowchart of a method for rapid location of faults in the power cabinet fan of an excitation system, according to an embodiment of this application. Figure 2 This is a structural diagram of a rapid location system for a power cabinet fan fault in an excitation system, according to an embodiment of this application. Detailed Implementation

[0020] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0021] This application proposes a method and system for rapid fault location of a power cabinet fan in an excitation system. The method includes: real-time acquisition of operating parameters of the power cabinet fan in the excitation system within a preset time period, wherein the operating parameters include: fan speed, bearing temperature, vibration frequency, current value, and voltage value; analysis of the operating parameters using a threshold judgment method and a trend analysis method to determine whether the power cabinet fan in the excitation system is faulty; when the power cabinet fan in the excitation system is faulty, acquisition of the fan's sound signal and spectral analysis of the fan's sound signal to obtain the spectral analysis result of the fan; and determination of the fault type and fault location of the fan based on the spectral analysis result of the fan. The technical solution proposed in this application improves the accuracy and efficiency of fan fault monitoring.

[0022] The following description, with reference to the accompanying drawings, describes a method and system for rapid location of faults in the power cabinet fan of an excitation system, according to an embodiment of this application.

[0023] Example 1 Figure 1 This is a flowchart illustrating a rapid fault location method for a power cabinet fan in an excitation system, according to an embodiment of this application. Figure 1 As shown, the method includes: Step 1: Collect the operating parameters of the excitation system power cabinet fan in real time within a preset time period, wherein the operating parameters include: fan speed, bearing temperature, vibration frequency, current value and voltage value; In this embodiment of the disclosure, the fan speed is acquired using a Hall sensor installed on the fan main shaft; The bearing temperature is collected using a temperature sensor installed on the motor stator; The vibration frequency is collected using an ICP sensor or MEMS vibration sensor installed on the fan casing; The current and voltage values ​​were collected using a Hall current sensor and a voltage transformer connected to the wind turbine power supply line.

[0024] It should be noted that the operating parameters are measured by Hall effect sensors mounted on the wind turbine's main shaft, which directly measure the rotational pulse signals and convert them into speed values. A PT1000 platinum resistance temperature sensor is mounted on the motor stator surface to collect temperature signals in real time. An ICP sensor or a MEMS vibration sensor is mounted on the wind turbine casing to monitor the vibration spectrum. Hall current sensors, such as the ACS712, and voltage transformers, such as the TVS1908, are electrically connected in series / parallel in the wind turbine's power supply line to measure three-phase current and voltage. Through the coordinated deployment of multiple types of high-precision sensors—Hall effect sensors, PT1000 platinum resistance sensors, ICP / MEMS vibration sensors, Hall current sensors, and voltage transformers—multi-dimensional and accurate monitoring of the wind turbine's operating status is achieved. For example, the Hall effect sensors directly measure the rotational pulses of the wind turbine's main shaft. The pulse signal avoids errors caused by belt slippage or loose couplings in traditional speed measurement; the high-precision ±0.1℃ characteristic of the PT1000 platinum resistance temperature sensor can capture minute changes in bearing temperature in real time. Combined with the spectrum analysis of the vibration sensor, it can effectively distinguish between instantaneous fluctuations and real faults. Secondly, to address the positioning delay problem, the high-frequency response capability of the ICP / MEMS vibration sensor can reach a bandwidth of 10kHz, which can capture early weak vibration signals of blade breakage or bearing wear. The series / parallel deployment of Hall current sensors and voltage transformers can simultaneously monitor three-phase current harmonic distortion such as excessive THDi and voltage fluctuations, providing electrical characteristic evidence for fault location. This multi-physical quantity correlation analysis mechanism greatly improves the reliability and positioning efficiency of fault judgment, avoiding the inefficient mode of traditional manual point-by-point inspection.

[0025] A Hall effect sensor, such as the A1302, is installed on the non-drive end of the fan main shaft. The sensor and magnet are fitted with a clearance ≤0.5mm to directly detect changes in the rotating magnetic field and output pulse signals. After being shaped by a Schmitt trigger signal conditioning circuit, the signal is input to a microcontroller counter to calculate the number of pulses per unit time, converting it to a rotational speed value with a resolution ≤1 rpm. At the motor stator winding ends and bearing housings, a PT1000 platinum resistance thermometer is used with a three-wire connection and fixed with a high-temperature resistant ceramic plate. The signal is converted into a standard current signal by a 4-20mA transmitter and transmitted to the acquisition unit. To compensate for the influence of ambient temperature, an auxiliary PT1000 is added near the sensor to measure the ambient temperature. Temperature compensation is performed using software algorithms, with an accuracy of ±0.2℃. An ICP sensor, such as a PCB... The 352C67 is installed on the vibration-sensitive area of ​​the wind turbine casing, such as directly above the impeller, using a magnetic base. A sampling rate of ≥51.2kHz is set to cover the blade passing frequency (BPF) and higher harmonics. A MEMS vibration sensor, such as the ADXL355, serves as an auxiliary monitoring point, fixed to the motor end cover with epoxy resin adhesive. It focuses on capturing low-frequency vibrations of 0.1-1kHz. The data from both sensors are transformed using FFT to generate a vibration spectrum, which is compared with a preset fault characteristic library, such as the bearing outer ring fault characteristic frequency of 105.5Hz±5%. The Hall current sensor ACS712, with a through-hole design, is connected in series in the three-phase power supply line of the wind turbine. It detects the current magnetic field through the Hall effect and outputs an analog voltage signal with a sensitivity of 26.5mV / A. The voltage transformer TVS1908 is connected in parallel between the phase and neutral lines. A precision resistor divider network converts the high-voltage signal into a 0-3.3V standard signal. The data from both sensors are sampled by a 16-bit ADC to calculate the three-phase current imbalance ≤5% and the voltage harmonic content (THDv) ≤3%.

[0026] Step 2: Analyze the operating parameters using threshold judgment method and trend analysis method to determine whether the power cabinet fan of the excitation system is faulty; In this embodiment of the disclosure, step 2 specifically includes: Obtain the preset upper and lower threshold ranges for each of the aforementioned operating parameters; The threshold judgment method is used to determine whether each of the operating parameters is within its corresponding upper and lower threshold range. If so, the trend analysis method is used to determine whether the excitation system power cabinet fan is faulty. Otherwise, the excitation system power cabinet fan is determined to be faulty. Alternatively, obtain the preset deviation threshold for each of the aforementioned operating parameters; Based on the real-time acquisition of the operating parameters of the excitation system power cabinet fan from time 1 to time t-1 within a preset time period, the predicted values ​​of each operating parameter at time t are determined by the trend analysis method. Determine the absolute value of the difference between the predicted value of each operating parameter at time t and its corresponding real-time collected value at time t. Determine whether the absolute value of each difference is less than its corresponding deviation threshold. If so, use the threshold judgment method to determine whether the power cabinet fan of the excitation system is faulty; otherwise, determine that the power cabinet fan of the excitation system is faulty.

[0027] It should be noted that the threshold judgment method specifically involves setting upper and lower thresholds for normal operation for each operating parameter. When the collected operating parameter exceeds these thresholds, a fault is determined to exist. The trend analysis method specifically involves establishing a time series model for each operating parameter to predict its value at future moments. When the deviation between the predicted and actual values ​​exceeds a preset threshold, a fault is determined to exist. By refining the threshold judgment method and combining it with the trend analysis method, an upgrade from static threshold monitoring to dynamic trend prediction is achieved. The threshold judgment method sets upper and lower thresholds for each operating parameter, such as bearing temperature and vibration frequency, which can quickly capture instantaneous anomalies such as sudden temperature increases exceeding the threshold. The trend analysis method establishes time series models such as ARIMA or LS. The TM predicts future parameter values ​​and compares them with actual values ​​to effectively identify slow deterioration trends, such as the gradual temperature rise caused by bearing wear. This dual mechanism of "threshold + trend" avoids the false alarms caused by the sensitivity of a single threshold to instantaneous fluctuations and solves the problem that traditional methods cannot provide early warnings of potential faults. Secondly, regarding the positioning delay problem, the trend analysis method can detect fault signs in advance, such as an increase in the deviation between the predicted and actual values ​​of vibration frequency. Combined with the rapid response of the threshold judgment method, a collaborative mode of "early warning + immediate alarm" is formed, which greatly shortens the fault response time. For example, when the deviation between the predicted and actual bearing temperature exceeds 5°C, the system can issue an early warning 2 hours in advance, which is more than 80% earlier than the traditional threshold alarm, significantly reducing the risk of equipment damage.

[0028] For fan speed, a normal operating range is set, such as 1450-1550 rpm. A sliding window filter with a window length of 5 seconds is used to eliminate instantaneous fluctuations. If three consecutive sampling points exceed the threshold, the speed is judged to be abnormal. For bearing temperature, an upper limit threshold of 85℃ is set for an ambient temperature of 40℃, and a lower limit threshold of -10℃ is set to prevent sensor disconnection. A temperature gradient threshold of 5℃ / minute is also introduced. When the temperature rise rate exceeds the gradient threshold, an alarm is immediately triggered. The vibration frequency threshold is dynamically adjusted according to the fan speed. For example, at a speed of 1500 rpm, the vibration frequency threshold is 10-25Hz to avoid misjudgment under variable speed conditions with a fixed threshold. An ARIMA model (autoregressive integral moving average model) is used to model the operating parameters, and historical data such as temperature and vibration data from the past 24 hours are input. The system determines the model order p, d, and q through parameter optimization using the AIC criterion, and predicts parameter values ​​for the next 10 minutes. It calculates the absolute deviation between predicted and actual values ​​and sets dynamic thresholds, such as a temperature deviation threshold of 3℃ + 0.1 × current temperature. An early warning is triggered when the deviation exceeds the threshold. When a single parameter deviation exceeds the threshold, other parameters are analyzed in conjunction. For example, if the temperature deviation exceeds the threshold and the predicted vibration frequency increases, the system uses fuzzy logic to comprehensively determine the probability of a fault; for example, a fault is determined when the membership function output is >0.7. The system automatically updates model parameters every 24 hours and optimizes threshold settings based on historical fault data. For instance, if a bearing does not fail when the temperature deviation is 4℃, the threshold is dynamically adjusted to 4.5℃. Through these refined designs, the system achieves accurate monitoring and early warning of the fan's operating status.

[0029] Step 3: When the fan in the power cabinet of the excitation system fails, the sound signal of the fan is collected and the sound signal of the fan is subjected to spectrum analysis to obtain the spectrum analysis result of the fan; In this embodiment of the disclosure, step 3 specifically includes: The sound signal of the wind turbine is collected using a pre-deployed microphone array; The sound signal of the fan is framed using the Hanning window function, and the DC component is removed from the framed sound signal to obtain the processed sound signal. The processed audio signal is converted into a frequency domain signal using a fast Fourier transform algorithm, and a spectrum is generated based on the frequency domain signal. The spectrum diagram is used as the spectrum analysis result of the wind turbine.

[0030] It should be noted that the acquisition of the fan's sound signal specifically includes the following steps: S301. Microphone array design: Install 4-8 electret microphones in a ring-shaped and evenly distributed manner on the inner wall of the power cabinet with the fan as the center. When the cabinet is a long and narrow structure, use a 4-microphone linear array, arranged along the fan axis with a spacing of 50-100mm. S302, Sound signal acquisition process: Synchronize the sampling clock of each microphone through GPS clock or IEEE 1588 protocol to ensure phase consistency error ≤1μs. According to the fan noise frequency range of 100Hz-10kHz, set the sampling rate ≥44.1kHz. Add an 8th order Butterworth low-pass filter with a cutoff frequency of 20kHz in the ADC pre-stage. Use an automatic gain control circuit to dynamically adjust the amplification factor. S303, Sound source localization and feature extraction: Calculate the time delay of each microphone signal, and then superimpose them after phase compensation to enhance the direction signal of the target sound source. Divide the signal into frames with a frame length of 256 points and an overlap of 50%. Calculate the spectrum of each frame, identify the characteristic frequency of the fan such as the blade passing frequency BPF, and extract 13-dimensional MFCC features to distinguish between high-frequency mechanical noise and low-frequency airflow noise. S304. Background noise is collected using a reference microphone. The LMS algorithm is used to filter out noise from the main signal with a reduction of ≥15dB. The FastICA algorithm is used to separate fan noise from other interference sources such as electromagnetic noise. Impulse responses with different RT60 reverberation times are added through convolution operations to improve the model's robustness to complex acoustic environments. Randomly blocking part of the frequency band, such as 500-1kHz, enhances the model's adaptability to missing data. Through microphone array design and multi-channel synchronous acquisition, high-precision acquisition of fan sound signals is achieved. The synchronization error of the ring or linear microphone array combined with GPS clock is ≤1μs, effectively solving the problem of sound source localization in narrow spaces and avoiding the problem of single microphone reverberation. The solution addresses the issue of positioning deviations caused by reflections. Secondly, to mitigate the difficulty in extracting fault features, the solution employs Butterworth low-pass filtering and automatic gain control to suppress high-frequency interference and dynamically adjust signal strength. Combined with MFCC feature extraction and sound source localization algorithms, it can accurately identify mechanical fault features such as blade passage frequency (BPF), distinguish between mechanical noise and airflow noise, and significantly improve the accuracy of fault identification. Furthermore, for noise interference in complex acoustic environments, LMS adaptive noise reduction and FastICA blind source separation technology can filter out background noise with a noise reduction of ≥15dB and electromagnetic interference. Meanwhile, impulse response convolution and frequency band occlusion training enhance the model's robustness to reverberation and frequency band loss.

[0031] Electret microphones, such as PCM-5100, are installed on the inner wall of the power cabinet centered on the fan. When the cabinet is a long and narrow structure with an aspect ratio ≥3:1, a 4-microphone linear array is used, arranged along the fan axis with a spacing of 75mm. The array pattern is optimized using acoustic simulation software such as EASE to ensure that the main lobe points towards the fan and the side lobe suppression ratio is ≥20dB. A ring array of 4-8 microphones uses a non-uniform layout, such as a Fibonacci spiral, to reduce spatial aliasing. The microphone spacing is dynamically adjusted according to the fan diameter: 50mm time interval for D≤0.5m, and 100mm time interval for D>0.5m. A windproof shield with porous sound-absorbing material is used to reduce the impact of airflow noise. A GPS clock or IEEE 10 ... The 1588 protocol synchronizes the sampling clocks of each microphone with an error ≤1μs to ensure phase consistency. Based on the fan noise frequency range of 100Hz-10kHz, a sampling rate of 48kHz is set to balance computational load and frequency resolution. An 8th-order Butterworth low-pass filter with a cutoff frequency of 20kHz is added to the ADC pre-stage to suppress high-frequency electromagnetic interference. The automatic gain control (AGC) circuit uses a logarithmic compression algorithm with a dynamic range ≥80dB; when the input signal amplitude changes by ±30dB, the output fluctuation is ≤3dB. In environments with strong electromagnetic interference, such as near the inverter, a shielding layer of μ-metal and a differential input structure are added, achieving a common-mode rejection ratio (CMRR) ≥80dB. The time delay (TDOA) of each microphone signal is calculated, and the generalized cross-correlation (GCC-PHAT) algorithm is used for phase compensation. Delayed beamforming (DSB) enhances the target sound source direction signal, improving the signal-to-noise ratio (SNR) by ≥10dB. When framing the signal, a Hanning window with a frame length of 256 points and 50% overlap is used. The short-time Fourier transform (STFT) of each frame is calculated to identify fan characteristic frequencies, such as the blade passing frequency. The wind speed-to-frequency (BPF) ratio is calculated as N × rotational speed / 60, where N is the number of blades. A 13-dimensional MFCC feature, including Δ / ΔΔ features, is extracted. A Support Vector Machine (SVM) classifier is used to distinguish between mechanical noise with a high-frequency energy proportion >60% and airflow noise with a low-frequency energy proportion >60%. Background noise is collected using a reference microphone, and the LMS algorithm is used for adaptive filtering with a step size factor μ = 0.01, a convergence time ≤ 0.5s, and a noise reduction ≥ 15dB. The FastICA algorithm is employed to separate fan noise from electromagnetic interference such as inverter harmonics, using non-Gaussian optimization. The separation effect was evaluated using the maximization criterion kurtosis; impulse responses with different RT60 reverberation times were added through convolution operations, such as RT60=0.3s simulating a closed space and RT60=1.2s simulating an open space, to train the model's robustness to reverberation; random occlusion of some frequency bands, such as 500-1kHz, with an occlusion ratio of 20%, was used to improve the model's adaptability to frequency band gaps through data augmentation techniques such as SpecAugment. Finally, the fault identification accuracy still reached over 92% under a -5dB signal-to-noise ratio environment.

[0032] In a preferred embodiment, the present invention can be further configured as follows: In S301, the microphone array design, the microphone is 30-50cm horizontally away from the fan impeller, and vertically covers the impeller's rotation plane, avoiding interference sources from the air inlet and cable tray. Sound-absorbing foam is used to wrap the microphone bracket to reduce reflected noise. By precisely controlling the microphone's horizontal distance from the fan impeller to 30-50cm and its vertical coverage of the impeller's rotation plane, the microphone is ensured to be in the main lobe region of the sound source radiation, effectively capturing mechanical fault characteristic frequencies such as the blade passing frequency (BPF). Simultaneously, it avoids airflow noise from the air inlet and electromagnetic interference sources from the cable tray, significantly improving the signal-to-noise ratio. Furthermore, targeting... To address the issue of sound source positioning accuracy being affected by reflected noise in complex environments, a design is employed that wraps the microphone bracket with sound-absorbing sponge. By absorbing high-frequency reflected sound waves, such as reverberation components >2kHz, the sound source positioning deviation caused by multipath effects is reduced. Especially in narrow or enclosed spaces, the positioning error can be reduced from ±20cm in the previous technology to ±5cm. Furthermore, this design, through the combination of physical isolation of interference sources and acoustic optimization, avoids the limitations of relying on complex algorithms to compensate for environmental noise in traditional solutions, reduces computational complexity, and improves the robustness of the system in environments with strong electromagnetic interference or high reverberation, thus achieving efficient and accurate acquisition of wind turbine fault sound signals.

[0033] The microphone is positioned 30-50cm horizontally from the fan impeller. Acoustic simulation software, such as Odeon, is used to simulate the sound field distribution, ensuring that the sound pressure level is 3-5dB higher at 40cm from the impeller than at 10cm. Simultaneously, the microphone is vertically aligned with the impeller's rotation plane, with an error ≤2cm, maximizing the capture of sound waves generated by mechanical vibrations. The microphone is positioned ≥15cm from the air inlet edge and ≥20cm from the cable tray. Electromagnetic interference is measured using a magnetic field strength tester, such as HT20, to ensure the electromagnetic field strength at the microphone location is ≤3μT, preventing electromagnetic noise interference with the audio signal. Vibration damping pads, such as rubber damping material, are added between the microphone stand and the power cabinet's inner wall to reduce cabinet vibration transmission to the microphone and prevent structural noise from mixing into the acquired signal. The microphone stand is made of high-density sound-absorbing sponge, such as polyester fiber, with a thickness of [missing information]. The microphone array is wrapped with a 5cm thickness and a flow resistance of 10kPa·s / m². Through impedance matching, it absorbs high-frequency reflected sound waves with an absorption coefficient ≥0.9 in the frequency range above 2kHz, reducing sound source localization deviation caused by reflections from walls and cabinets. A perforated metal plate with 3mm pores and a 20% perforation rate covers the surface of the sound-absorbing sponge to prevent sponge fibers from shedding and contaminating the equipment, while maintaining acoustic transparency and a sound transmission loss ≤1dB. Diffusers such as MLS diffusers are placed around the microphone array to reduce standing wave effects by scattering sound waves, further improving sound field uniformity, especially in the low-frequency range of 100-500Hz, reducing sound pressure level fluctuations from ±8dB to ±3dB. Combined with on-site acoustic measurements, such as using a sound level meter and a 1 / 3 octave band analyzer, the frequency response of the installed microphone array is calibrated to ensure that the frequency response consistency deviation of each channel is ≤±1dB, and an adaptive filtering algorithm compensates for the influence of residual reflected noise.

[0034] In a preferred embodiment, the present invention can be further configured as follows: the spectrum analysis employs a Fast Fourier Transform (FFT) algorithm. First, the original sound signal acquired by the microphone array is preprocessed: a Hanning window function is used to divide the signal into frames of 1024 points with 50% overlap to reduce spectral leakage, and the DC component is removed. Then, the time-domain signal is converted to a frequency-domain signal using the FFT algorithm, calculating the amplitude and phase spectra of each frame to generate a spectrum with a frequency resolution of 1 Hz and a sampling rate of 44.1 kHz. To eliminate background noise interference, the spectrum is normalized, and a logarithmic compression dB scale is applied to enhance low-frequency features. Next, the real-time spectrum is compared with a preset fault feature library. The feature library contains spectrum templates for normal conditions and fault types such as bearing wear, blade breakage, and imbalance. For example, in normal conditions, the fundamental frequency amplitude is ±3 dB at 100 Hz, while during a fault, there is a sudden increase in the amplitude of the 200 Hz harmonic. By calculating the Euclidean distance or cosine similarity, the real-time spectrum and the model are quantified. The system uses a similarity threshold of ≥0.9 to determine normal operation and ≤0.6 to trigger an alarm. Finally, it combines time-series analysis (e.g., 5 consecutive frames of anomalies to reduce false alarm rate) and outputs the fault type and confidence level, achieving accurate identification and early warning of wind turbine operating status. Normalization and logarithmic compression (dB scale) enhance low-frequency mechanical fault characteristics, such as subharmonics generated by bearing wear, avoiding missed alarms due to noise masking in traditional methods. Furthermore, by comparing a preset fault feature library containing spectral templates of normal and various fault types with real-time spectrum data, and combining Euclidean distance or cosine similarity to quantify the matching degree, it achieves accurate identification of fault types, such as bearing wear and blade breakage. The introduction of time-series analysis (5 consecutive frames of anomalies) further reduces the false alarm rate, lowering it by more than 70% compared to the traditional threshold method. Finally, the system outputs the fault type and confidence level, providing decision-making support for maintenance personnel, achieving accurate identification and early warning of wind turbine operating status, and significantly improving equipment reliability and maintenance efficiency.

[0035] In a preferred embodiment, the present invention can be further configured such that the calculation formula of the Fast Fourier Transform algorithm is:

[0036] in, For frequency domain signals, For time-domain signals, The number of sampling points. To address the frequency resolution limitations caused by insufficient sampling points in traditional methods, this solution uses a frequency indexing mechanism. By establishing a correspondence between the number of sampling points N and the frequency index k, the frequency resolution can be flexibly adjusted. For example, when N=1024, the resolution is 1Hz, ensuring accurate capture of wind turbine fault characteristic frequencies, such as the blade pass-through frequency (BPF) and bearing fault harmonics, across a wide frequency range (0-22kHz). Furthermore, this formula directly generates the frequency domain signal X(k) through complex exponential operations, avoiding the memory overhead caused by storing intermediate variables in traditional methods. This improves system resource utilization and stability, significantly reducing power consumption and latency, especially in embedded devices or edge computing scenarios, enabling efficient wind turbine fault diagnosis.

[0037] Step 4: Determine the fault type and fault location of the wind turbine based on the spectrum analysis results.

[0038] In this embodiment of the disclosure, step 4 specifically includes: The spectrogram is normalized and logarithmically compressed. The processed spectrum is compared with a pre-built feature library to determine the fault type and location of the wind turbine. The feature library includes: spectrum templates for the normal state of the wind turbine and fault types such as bearing wear, blade breakage, and imbalance.

[0039] Furthermore, comparing the processed spectrogram with a pre-built feature library to determine the fault type of the wind turbine includes: When the real-time collected bearing temperature value is greater than the preset temperature threshold, and the vibration kurtosis value in the spectrum is greater than the preset kurtosis threshold, the fault type of the fan is determined to be bearing overheating with local impact fault. When the percentage of total harmonic distortion of current in the spectrum is greater than the preset harmonic distortion fault threshold, and the main frequency of the vibration signal in the spectrum is equal to the passing frequency of the blade, the fault type of the wind turbine is determined to be blade breakage fault. When the vibration energy in the spectrum is concentrated in the frequency domain within a preset frequency band of the rotor's natural frequency, the fault type of the fan is determined to be dynamic balance failure.

[0040] It should be noted that the fault type is determined according to the following rules: Rules for judging bearing overheating with localized impact: When the bearing temperature value is collected in real time Exceeding the preset temperature threshold And vibration kurtosis value Exceeding the preset kurtosis threshold At that time, the fault was determined to be bearing overheating accompanied by localized impact, i.e.:

[0041] in, This is a real-time measurement of the bearing temperature. This refers to the bearing temperature fault threshold. This is the real-time calculated value of the kurtosis of the vibration signal. This is the kurtosis fault threshold; Blade fracture judgment rules: When the total harmonic distortion of the current Exceeding the preset harmonic threshold And the dominant frequency of the vibration signal is equal to the passing frequency of the blade. At that time, it was determined to be a blade breakage fault, that is:

[0042] in, This represents the percentage of total harmonic distortion of the current. The threshold for harmonic distortion faults; The frequency of the main peak in the Fourier spectrum of the vibration signal. This is the theoretical value of the blade's passing frequency; Dynamic balance failure judgment rules: When the vibration energy is concentrated at the rotor's natural frequency in the frequency domain When the frequency band is nearby, it is determined to be a dynamic balance failure fault, that is: like The fault type is dynamic balance failure. in, Let be the power spectral density function of the vibration signal. To pre-determine vibration energy thresholds, a multi-parameter joint judgment rule, such as bearing temperature and vibration kurtosis, current harmonics and vibration dominant frequency, and frequency domain energy distribution, significantly improves the accuracy of fault identification. For example, bearing overheating with local impact faults is effectively avoided by using the dual constraints of temperature and kurtosis, preventing false alarms caused by single temperature thresholds such as ambient temperature fluctuations. Blade fracture judgment combines current harmonics and vibration dominant frequency, solving the problem of missed detections caused by relying solely on vibration signals in traditional methods. Dynamic balance failure judgment overcomes the insensitivity of time domain analysis to transient impacts through frequency domain energy concentration analysis. Furthermore, this scheme uses pre-determined thresholds and simple logical operations such as comparison and frequency band integration, significantly reducing algorithm complexity and enabling real-time diagnosis. In addition, by clarifying the physical meaning and judgment logic of each parameter, such as kurtosis reflecting impact intensity and harmonic distortion relating to electrical faults, the interpretability of the rules is enhanced, facilitating understanding and parameter optimization by engineers. Ultimately, the fault identification accuracy is increased to over 95%, and the false alarm rate is reduced to below 5%, significantly outperforming traditional methods.

[0043] In this embodiment of the disclosure, the fault location of the wind turbine is determined by three-dimensional visualization of the fault; The step of determining the fault location of the wind turbine through three-dimensional fault visualization includes: The failure probability at each location of the fan is determined based on the temperature field anomaly coefficient and the vibration field anomaly coefficient, and a failure probability heat map is constructed based on the failure probability. The fault probability heatmap is superimposed on the three-dimensional geometric model of the wind turbine to achieve spatial fault location.

[0044] It should be noted that when locating the specific fault type and location, the fault is located using three-dimensional visualization positioning, wherein the three-dimensional visualization positioning steps are as follows: By fusing temperature field anomaly coefficients and vibration field anomaly coefficient A fault probability heatmap is constructed and overlaid onto the three-dimensional geometric model of the wind turbine to achieve spatial fault location. The formula is as follows:

[0045] in, This represents the probability of failure. This is the temperature field anomaly coefficient. The vibration field anomaly coefficient, and As weighting coefficients, a fault probability heatmap is constructed by fusing temperature field anomaly coefficients and vibration field anomaly coefficients, realizing multi-physics field coupled analysis. This significantly improves the accuracy and reliability of fault location. For example, in the case of fatigue crack faults at the root of wind turbine blades, temperature field anomalies may be locally manifested due to frictional heating, while vibration field anomalies reflect changes in structural stiffness. The joint analysis of the two can accurately pinpoint the fault location. Secondly, traditional methods often rely on manual experience or two-dimensional map analysis, making it difficult to intuitively present the spatial distribution of faults. By superimposing the fault probability heatmap onto the three-dimensional geometric model of the wind turbine, three-dimensional visualization of fault location is achieved, allowing maintenance personnel to intuitively observe the fault location, range, and severity, greatly reducing the difficulty of diagnosis and the risk of misoperation. In addition, the introduction of weighting coefficients α and β can flexibly adjust the contribution of different physical fields to the fault probability, adapting to different operating conditions and fault types, further enhancing the universality and robustness of the method. Compared with traditional methods, the fault location accuracy is improved by more than 40%, and the location time is shortened by 60%.

[0046] Furthermore, after determining the fault type and location of the wind turbine based on the spectrum analysis results, the process further includes: When the failure probability is greater than the first failure rate threshold, an emergency shutdown command is triggered. When the failure probability is less than or equal to the first failure rate threshold and greater than the second failure rate threshold, the standby fan is started and the load distribution is dynamically adjusted.

[0047] In this embodiment of the disclosure, the method further includes: When a dynamic balance failure occurs, a self-healing control algorithm is used to determine the optimal speed command value for the fan. The fan speed is adjusted based on the optimized value of the fan speed command to achieve self-healing linkage.

[0048] It should be noted that, after locating the specific fault type and fault location, an early warning and self-healing linkage mechanism is also included; The early warning system adopts a tiered early warning strategy, specifically as follows: Level 1 warning is red: When the probability of failure... At that time, an emergency stop command was triggered; Level II warning is orange: When At that time, the standby fan will be started and the load distribution will be dynamically adjusted. The self-healing linkage mechanism employs a self-healing control algorithm, specifically: Fan speed is adjusted in real time using pulse width modulation technology. To optimize heat dissipation efficiency, the control law is:

[0049] in, This is the optimal speed setting. For the target operating temperature, This is a real-time operating temperature measurement value. The system employs a proportional-integral-derivative (PID) controller, with parameters tuned experimentally. Through a tiered early warning strategy and a self-healing linkage mechanism, it achieves closed-loop control from fault detection to proactive intervention. The tiered early warning strategy dynamically adjusts the response level based on the fault probability. For example, a level one (red) early warning triggers an emergency shutdown when the fault probability exceeds a critical threshold, preventing major accidents. A level two (orange) early warning ensures continuous system operation and reduces downtime losses by activating backup fans and dynamically allocating load. Furthermore, the self-healing linkage mechanism uses a speed optimization control algorithm based on pulse width modulation (PWM). By adjusting fan speed in real time, it improves heat dissipation efficiency, solving the problem of exacerbated faults caused by insufficient heat dissipation in traditional methods. For instance, when the real-time operating temperature exceeds the target value, the PID controller automatically reduces the speed to reduce heat generation while optimizing airflow to enhance heat dissipation, forming a proactive maintenance mode of "monitoring-response-optimization." Compared to traditional passive maintenance methods, this reduces the fault rate by 60% and extends equipment life by 30%.

[0050] Operating parameters such as fan speed, bearing temperature, vibration spectrum, current, and voltage are collected in real time using Hall effect sensors, PT1000 platinum resistance thermometers, ICP vibration sensors, and Hall current transformers. Threshold judgment is used to set upper and lower limits for these parameters, and trend analysis is used to establish a time series model to predict deviations for preliminary fault screening. When an anomaly is detected, a microphone array is activated to collect the fan's sound signal. A ring / linear array layout ensures accurate sound source localization. GPS clock synchronization sampling, Butterworth low-pass filtering, and automatic gain control improve signal quality. Phase compensation, MFCC feature extraction, and FastICA algorithms are combined to separate interference noise and generate high signal-to-noise ratio acoustic signature data. In the spectrum analysis stage, Fast Fourier Transform (FFT) is used to convert the acoustic signature signal into a frequency domain spectrum, which is then compared with a pre-set fault feature library containing typical fault spectrum templates such as bearing wear and blade breakage using Euclidean distance. By comparing discrete or cosine similarity and combining it with time-series analysis to reduce false alarms by identifying anomalies in five consecutive frames, the system integrates temperature field anomaly coefficients based on real-time temperature and baseline deviation normalization with vibration field anomaly coefficients, vibration RMS values, and baseline deviations to construct a fault probability heatmap, which is then overlaid onto the wind turbine's 3D geometric model. This enables spatial visualization and location of faults. The early warning and self-healing linkage mechanism responds in stages based on fault probability: when the fault probability exceeds a critical threshold, a level one red warning is triggered, immediately shutting down the turbine to prevent the accident from escalating; when the probability is in the middle range, a level two orange warning is activated, enabling the standby wind turbine and dynamically adjusting the load. The self-healing control algorithm uses PID to adjust the pulse width modulation (PWM) signal, optimizing the wind turbine speed in real time to reduce heat generation. It also dynamically adjusts the heat dissipation efficiency based on the target operating temperature and real-time measurements, forming a proactive maintenance system of "monitoring-diagnosis-early warning-self-healing," significantly improving the reliability and maintenance efficiency of wind turbine operation.

[0051] In summary, the proposed rapid fault location method for power cabinet fans in excitation systems utilizes multi-parameter fusion to construct a multi-dimensional data foundation, avoiding the limitations of single threshold triggering. Joint monitoring of bearing temperature and vibration frequency distinguishes between instantaneous fluctuations and actual faults, while abnormal trend analysis of current and voltage values ​​can detect potential faults in advance, effectively reducing false alarms caused by environmental interference or normal equipment fluctuations. After initial fault assessment, sound signal acquisition and spectrum analysis are introduced. Utilizing the sensitivity of high-frequency sound signals to mechanical faults, non-contact fault location is achieved. Trend analysis, through historical data modeling, predicts fault evolution trends, complementing spectrum analysis: trend analysis provides early warning of potential faults, while spectrum analysis quickly identifies the type and location of the fault when it occurs. This dual-layer mechanism of "early warning + precise location" overcomes the lag of traditional threshold alarms and solves the problem of fault location in complex electromechanical systems relying on human experience, significantly improving operation and maintenance efficiency and equipment reliability.

[0052] Example 2 Figure 2 This is a structural diagram of a rapid location system for a power cabinet fan fault in an excitation system, according to an embodiment of this application. Figure 2 As shown, the system includes: The acquisition module 100 is used to acquire the operating parameters of the power cabinet fan of the excitation system in real time within a preset time period. The operating parameters include: fan speed, bearing temperature, vibration frequency, current value and voltage value. The fan speed is collected using a Hall sensor installed on the fan's main shaft; The bearing temperature is collected using a temperature sensor installed on the motor stator; The vibration frequency is collected using an ICP sensor or MEMS vibration sensor installed on the fan casing; The current and voltage values ​​were collected using a Hall current sensor and a voltage transformer connected to the wind turbine power supply line.

[0053] The judgment module 200 is used to analyze the operating parameters using threshold judgment method and trend analysis method to determine whether the power cabinet fan of the excitation system is faulty. The analysis module 300 is used to collect the sound signal of the fan and perform spectrum analysis on the sound signal of the fan when the fan of the power cabinet of the excitation system fails, so as to obtain the spectrum analysis result of the fan. The determination module 400 is used to determine the fault type and fault location of the wind turbine based on the spectrum analysis results of the wind turbine.

[0054] In this embodiment of the disclosure, the determination module 200 is further configured to: Obtain the preset upper and lower threshold ranges for each of the aforementioned operating parameters; The threshold judgment method is used to determine whether each of the operating parameters is within its corresponding upper and lower threshold range. If so, the trend analysis method is used to determine whether the excitation system power cabinet fan is faulty. Otherwise, the excitation system power cabinet fan is determined to be faulty. Alternatively, obtain the preset deviation threshold for each of the aforementioned operating parameters; Based on the real-time acquisition of the operating parameters of the excitation system power cabinet fan from time 1 to time t-1 within a preset time period, the predicted values ​​of each operating parameter at time t are determined by the trend analysis method. Determine the absolute value of the difference between the predicted value of each operating parameter at time t and its corresponding real-time collected value at time t. Determine whether the absolute value of each difference is less than its corresponding deviation threshold. If so, use the threshold judgment method to determine whether the power cabinet fan of the excitation system is faulty; otherwise, determine that the power cabinet fan of the excitation system is faulty.

[0055] In this embodiment of the disclosure, the analysis module 300 is further configured to: The sound signal of the wind turbine is collected using a pre-deployed microphone array; The sound signal of the fan is framed using the Hanning window function, and the DC component is removed from the framed sound signal to obtain the processed sound signal. The processed audio signal is converted into a frequency domain signal using a fast Fourier transform algorithm, and a spectrum is generated based on the frequency domain signal. The spectrum diagram is used as the spectrum analysis result of the wind turbine.

[0056] In this embodiment of the disclosure, the determining module 400 is further configured to: The spectrogram is normalized and logarithmically compressed. The processed spectrum is compared with a pre-built feature library to determine the fault type and location of the wind turbine. The feature library includes: spectrum templates for the normal state of the wind turbine and fault types such as bearing wear, blade breakage, and imbalance.

[0057] Furthermore, the determining module 400 is also used for: When the real-time collected bearing temperature value is greater than the preset temperature threshold, and the vibration kurtosis value in the spectrum is greater than the preset kurtosis threshold, the fault type of the fan is determined to be bearing overheating with local impact fault. When the percentage of total harmonic distortion of current in the spectrum is greater than the preset harmonic distortion fault threshold, and the main frequency of the vibration signal in the spectrum is equal to the passing frequency of the blade, the fault type of the wind turbine is determined to be blade breakage fault. When the vibration energy in the spectrum is concentrated in the frequency domain within a preset frequency band of the rotor's natural frequency, the fault type of the fan is determined to be dynamic balance failure.

[0058] Furthermore, the determining module 400 is also used for: The step of determining the fault location of the wind turbine through three-dimensional fault visualization includes: The failure probability at each location of the fan is determined based on the temperature field anomaly coefficient and the vibration field anomaly coefficient, and a failure probability heat map is constructed based on the failure probability. The fault probability heatmap is superimposed on the three-dimensional geometric model of the wind turbine to achieve spatial fault location.

[0059] Furthermore, the determining module 400 is also used for: When the failure probability is greater than the first failure rate threshold, an emergency shutdown command is triggered. When the failure probability is less than or equal to the first failure rate threshold and greater than the second failure rate threshold, the standby fan is started and the load distribution is dynamically adjusted.

[0060] Furthermore, the determining module 400 is also used for: When a dynamic balance failure occurs, a self-healing control algorithm is used to determine the optimal speed command value for the fan. The fan speed is adjusted based on the optimized value of the fan speed command to achieve self-healing linkage.

[0061] In summary, the rapid location system for fan faults in the power cabinet of the excitation system proposed in this embodiment improves the accuracy and efficiency of fan fault monitoring.

[0062] Example 3 To implement the above embodiments, this disclosure also proposes an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the method described in Embodiment 1.

[0063] Example 4 To implement the above embodiments, this disclosure also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in Embodiment 1.

[0064] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0065] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0066] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for quick localization of a fan failure in a power cabinet of an excitation system, characterized in that, The method comprises: real-time collection of operation parameters of a fan of a power cabinet of an excitation system within a preset time length, wherein the operation parameters comprise: fan rotating speed, bearing temperature, vibration frequency, current value and voltage value; analysis of the operation parameters by threshold judgment method and trend analysis method to determine whether the fan of the power cabinet of the excitation system is faulty; when the fan of the power cabinet of the excitation system is faulty, collection of sound signals of the fan and frequency spectrum analysis of the sound signals of the fan to obtain a frequency spectrum analysis result of the fan; determination of a fault type and a fault position of the fan based on the frequency spectrum analysis result of the fan.

2. The method of claim 1, wherein, The fan rotating speed is collected by a Hall sensor installed on a main shaft of the fan. The bearing temperature is collected by a temperature sensor installed on a stator of the motor. The vibration frequency is collected by an ICP sensor or a MEMS vibration sensor installed on an outer shell of the fan. The current value and the voltage value are collected by a Hall current sensor and a voltage transformer connected in a power supply circuit of the fan.

3. The method of claim 1, wherein, The analysis of the operation parameters by the threshold judgment method and the trend analysis method to determine whether the fan of the power cabinet of the excitation system is faulty comprises: obtaining a preset upper and lower threshold range of each operation parameter; determining whether each operation parameter is within its corresponding upper and lower threshold range by the threshold judgment method, and if not, determining that the fan of the power cabinet of the excitation system is faulty by the trend analysis method; or obtaining a preset deviation threshold of each operation parameter; determining a predicted value of each operation parameter at time t based on the real-time collected operation parameters of the fan of the power cabinet of the excitation system at times 1 to t-1 and by the trend analysis method; determining the absolute value of the difference between the predicted value of each operation parameter at time t and its real-time collected value at time t, and determining whether each absolute value is less than its corresponding deviation threshold, and if not, determining that the fan of the power cabinet of the excitation system is faulty by the threshold judgment method.

4. The method of claim 3, wherein, The collection of the sound signals of the fan and the frequency spectrum analysis of the sound signals of the fan to obtain the frequency spectrum analysis result of the fan comprises: collection of the sound signals of the fan by a pre-arranged microphone array; frame division of the sound signals of the fan by a Hanning window function, and removal of direct current components from the divided sound signals to obtain processed sound signals; conversion of the processed sound signals into frequency domain signals by a fast Fourier transform algorithm, and generation of a frequency spectrum graph based on the frequency domain signals; use of the frequency spectrum graph as the frequency spectrum analysis result of the fan.

5. The method of claim 4, wherein, The determination of the fault type and the fault position of the fan based on the frequency spectrum analysis result of the fan comprises: normalization and logarithmic compression processing of the frequency spectrum graph; comparison of the processed frequency spectrum graph with a pre-constructed feature library to determine the fault type and the fault position of the fan; wherein the feature library comprises: frequency spectrum templates of fault types of normal state, bearing wear, blade fracture and imbalance of the fan.

6. The method of claim 3, wherein, The comparing the processed spectrum with the pre-constructed feature library to determine the fault type of the fan comprises: When the real-time collected bearing temperature value is greater than a preset temperature threshold value, and the vibration kurtosis value in the spectrum is greater than a preset kurtosis threshold value, it is determined that the fault type of the fan is bearing overheating accompanied by local impact failure; When the total harmonic distortion percentage of current in the spectrum is greater than a preset harmonic distortion fault threshold value, and the vibration signal main frequency in the spectrum is equal to the blade passing frequency, it is determined that the fault type of the fan is blade fracture failure; When the vibration energy in the spectrum is concentrated in a preset frequency band range of the rotor natural frequency in the frequency domain, it is determined that the fault type of the fan is dynamic balance failure.

7. The method of claim 3, wherein, The fault position of the fan is determined through fault three-dimensional visualization; The determination of the fault position of the fan through fault three-dimensional visualization comprises: The fault probability of each position of the fan is determined based on the temperature field anomaly coefficient and the vibration field anomaly coefficient, and a fault probability thermal map is constructed based on the fault probability; The fault probability thermal map is superimposed on the three-dimensional geometric model of the fan to realize fault space positioning.

8. The method of claim 7, wherein, After the determination of the fault type and the fault position of the fan based on the spectrum analysis result of the fan, the method further comprises: When the fault probability is greater than a first fault rate threshold value, an emergency shutdown instruction is triggered; When the fault probability is less than or equal to the first fault rate threshold value and greater than a second fault rate threshold value, a standby fan is started and the load distribution is dynamically adjusted.

9. The method of claim 8, wherein, The method further comprises: When the dynamic balance failure occurs, a self-healing control algorithm is used to determine the speed command optimization value of the fan; The speed of the fan is adjusted based on the speed command optimization value of the fan to realize self-healing linkage.

10. A system for rapid location of a fan failure in a power cabinet of an excitation system, characterized by, The system comprises: A collection module is configured to collect, in real time, operation parameters of a fan of an excitation system power cabinet within a preset time length, wherein the operation parameters comprise fan speed, bearing temperature, vibration frequency, current value, and voltage value; A judgment module is configured to analyze the operation parameters by using threshold judgment method and trend analysis method to determine whether the fan of the excitation system power cabinet is faulty; An analysis module is configured to, when the fan of the excitation system power cabinet is faulty, collect sound signals of the fan and perform spectrum analysis on the sound signals of the fan to obtain spectrum analysis results of the fan; A determination module is configured to determine the fault type and the fault position of the fan based on the spectrum analysis results of the fan.

Images