Non-contact monitoring and early warning method for power equipment state based on multi-source data

Abstract

The present application relates to the technical field of equipment maintenance, and specifically relates to a power equipment state non-contact monitoring and early warning method based on multi-source data, which constructs a composite abnormality index of fused multi-dimensional information by collecting vibration data, oil tank temperature data, narrowband voiceprint data and acoustic kurtosis data, and establishes a hierarchical progressive monitoring and early warning mechanism, and effectively filters out false abnormalities of the cooler coupling through jump ratio, coherence coefficient, local maximum value number, kurtosis difference value, load correction and harmonic ratio change, etc.; at the same time, early failure of the fan bearing, slight loosening of the winding and other similar faults are accurately traced and distinguished; finally, through adaptive weight coefficient correction and threshold dynamic adjustment mechanism, the system has self-learning and continuous optimization capabilities. The scheme proposed by the present application effectively improves the accuracy, reliability and practicability of the mechanical state monitoring of the power equipment, and provides technical support for the safe and stable operation of the power equipment.

Description

Technical Field

[0001] This invention relates to the field of equipment maintenance technology, specifically to a non-contact monitoring and early warning method for the status of power equipment based on multi-source data. Background Technology

[0002] During long-term operation, core equipment such as power transformers are subjected to a combination of factors such as electromagnetic force, thermal stress, and mechanical vibration, which gradually lead to mechanical defects such as winding loosening, core loosening, cooling system mechanical failure, and structural resonance. If these defects are not detected in time, they may develop into winding deformation, insulation wear, or even short-circuit faults. Therefore, continuous, online, non-contact monitoring of these equipment has significant engineering value. However, existing monitoring methods have two main problems: First, monitoring a single physical quantity cannot distinguish between "pseudo-anomalies" caused by cooler vibration coupling and genuine mechanical faults in the equipment itself. When the cooler vibration frequency is close to the mechanical modal frequency of the oil tank, vibration coupling occurs, significantly increasing the vibration amplitude on the oil tank surface. Relying solely on vibration amplitude parameters can easily lead to false alarms. Second, the characteristics of multi-source mechanical faults highly overlap in the frequency and time domains. For example, the 100Hz vibration spike caused by cooler coupling, the acoustic characteristics near 200Hz caused by early fan bearing failure, and the nonlinear increase in 100Hz vibration with temperature due to slight winding loosening are all similar and occur simultaneously. Traditional methods struggle to effectively determine the causes of these situations, making it difficult to accurately identify the causes of mechanical anomalies in power equipment. When an anomaly is detected, a significant amount of time is often wasted on troubleshooting. Therefore, there is an urgent need for a system that can accurately trace the source of defects with similar characteristics while intelligently filtering out pseudo-anomalies, thereby effectively improving the accuracy of power equipment fault diagnosis.

[0003] Chinese patent CN111190065A discloses a non-contact online monitoring system for power equipment. The system includes a wireless communication module, a temperature and humidity sensor module, a visible light camera module, an infrared imaging temperature measurement module, a network communication module, a UHF partial discharge signal acquisition module, a PoE power supply module, and a backend terminal. The wireless communication module connects to the temperature and humidity sensor module, the infrared imaging temperature measurement module, the network communication module, and the UHF partial discharge signal acquisition module via a communication interface, and to the visible light camera module via a video interface. It also communicates wirelessly or via wired connection to the backend terminal through a wireless transmitter or the network communication module. This solution, with its multi-functional modules, enables real-time online monitoring of multiple parameters of power equipment and performs calculations in the backend, allowing users to promptly understand the equipment's operating status and providing equipment fault warnings and alerts. However, this solution only achieves simple superposition and independent analysis of multiple parameters, lacking a decoupling and separation mechanism for multi-source mechanical fault characteristics and failing to solve the problem of distinguishing between "pseudo-anomalies" and real faults. Therefore, there is still a lack of a solution that can effectively decouple overlapping features and filter out false anomalies by making deep use of multi-source heterogeneous data such as vibration, acoustic patterns, and temperature, so as to achieve accurate monitoring and tracing of mechanical defects in power equipment. Summary of the Invention

[0004] This invention provides a non-contact monitoring and early warning method for the status of power equipment based on multi-source data, which overcomes the problem that existing non-contact power equipment status monitoring methods are unable to distinguish between highly overlapping mechanical fault characteristics and false anomalies, resulting in insufficient monitoring accuracy and difficulty in accurately locating the cause of anomalies.

[0005] This invention provides a non-contact monitoring and early warning method for the status of power equipment based on multi-source data, including data acquisition and preprocessing, real-time acquisition of power equipment vibration data, oil tank temperature data, narrowband acoustic fingerprint data, and acoustic kurtosis data, and preprocessing them, wherein the vibration data includes cooler vibration data and oil tank vibration data; A composite anomaly index is constructed by determining the vibration component, temperature component, and acoustic penalty term based on the vibration data, the oil tank temperature data, and the narrowband acoustic signature data, respectively, and by determining the composite anomaly index based on the weighted summation of the vibration component, temperature component, and acoustic penalty term. The system monitors and provides early warnings, determines whether the equipment has a suspected abnormality based on the composite anomaly index, and detects the online status of the cooler when a suspected abnormality is detected. When the cooler is online, it determines the actual mechanical status of the equipment and the cause of the abnormality based on the ratio of the average amplitude of the vibration data in the oil tank within two adjacent jump detection time windows before and after the last start of the cooler of the power equipment. It also corrects the preset composite anomaly index and the composite anomaly index construction parameters based on the cause of the abnormality, or outputs an abnormality warning. The acoustic penalty term refers to an additive correction term introduced during the construction of the composite anomaly index, based on the degree of abnormal increase in sound pressure level in the 195Hz to 215Hz frequency band relative to the historical baseline level in narrowband acoustic data and acoustic kurtosis data.

[0006] Furthermore, the process of determining whether a device exhibits a suspected anomaly based on the composite anomaly index includes: If the composite anomaly index is less than or equal to the preset composite anomaly index, the power equipment is determined to be operating normally. If the composite anomaly index is greater than the preset composite anomaly index, it is determined that the power equipment has a suspected anomaly, and the cooler is checked to see if it is online. If the cooler is online, the actual mechanical state of the equipment is determined based on the ratio of the average amplitude of the vibration data in the two adjacent jump detection time windows before and after the most recent start of the cooler. If the cooler is offline, the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic data within the superimposed detection time window is used to determine whether a suspected fault superposition has occurred. The jump detection time window refers to a continuous time period of a preset duration that is used to detect the average amplitude in the vibration data.

[0007] Furthermore, the process of determining the actual mechanical state of the equipment based on the ratio of the average amplitude of the vibration data within two adjacent jump detection time windows before and after the most recent start-up of the cooler includes: The jump ratio is determined based on the ratio of the average amplitude of the vibration data within two adjacent jump detection time windows before and after the most recent start-up of the cooler by the power equipment. If the jump ratio is less than or equal to the preset jump ratio, the suspected abnormality is determined to be a true abnormality. The number of local maxima in the 195Hz-215Hz frequency band of the narrowband voiceprint data within the superimposed detection time window is used to determine whether a suspected fault superposition occurs. If the jump ratio is greater than the preset jump ratio, the actual mechanical state of the power equipment is determined based on the cooler vibration data and the oil tank vibration data in the superimposed detection time window. The superimposed detection time window refers to a continuous time period that is cut back to the moment when a suspected anomaly is determined.

[0008] Furthermore, the process of determining the actual mechanical state of the power equipment based on the cooler vibration data and the oil tank vibration data within the superimposed detection time window includes: Based on the cooler vibration data and the oil tank vibration data in the superimposed detection time window, the cross power spectrum, cooler power spectrum and oil tank power spectrum are determined. The coherence coefficient is determined based on the cross power spectrum, the cooler power spectrum, and the oil tank power spectrum; The actual mechanical state of the power equipment is determined based on the coherence coefficient. If the coherence coefficient is less than or equal to the preset coherence coefficient, the suspected anomaly is determined to be a superposition of a true anomaly and a false anomaly coupled to the cooler. Based on the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic data within the superposition detection time window, it is determined whether a suspected fault superposition occurs. If the coherence coefficient is greater than the preset coherence coefficient, the suspected anomaly is determined to be a false anomaly of the cooler coupling, and the weighting coefficient of the vibration component is adjusted when constructing the composite anomaly index based on the ratio of the preset coherence coefficient to the coherence coefficient.

[0009] Furthermore, the process of modifying the weighting coefficients of the vibration components when constructing the composite anomaly index based on the ratio of the coherence coefficient to the preset coherence coefficient includes: The coherence coefficient ratio is determined based on the ratio of the coherence coefficient to the preset coherence coefficient, and the vibration component weighting coefficient is reduced based on the coherence coefficient ratio, wherein the reduction in the vibration component weighting coefficient is proportional to the coherence coefficient ratio.

[0010] Furthermore, the process of determining whether a suspected fault superposition occurs based on the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic data within the superposition detection time window includes: The number of maxima is determined based on the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic text data within the superimposed detection time window. If the maximum number is 0, the current anomaly is determined to be a normal anomaly, and the cause of the anomaly is determined based on preset rules; If the maximum value is greater than 0 and less than or equal to the preset maximum value, the speed of the suspected fan is reduced based on the speed test coefficient, and the uncorrected kurtosis and corrected kurtosis are determined based on the acoustic kurtosis data in the last verification cycle before correction and the second bearing fault verification window after correction, respectively. The anomaly type is determined based on the difference between the corrected kurtosis and the uncorrected kurtosis. If the maximum value is greater than the preset maximum value, the current anomaly is determined to be a fault superposition, and an early warning of fan bearing fault and winding loosening fault is output. The suspected fan refers to the cooling fan that generates abnormal sound sources, as determined based on the narrowband acoustic data.

[0011] Further, the corrected kurtosis is determined based on the acoustic kurtosis data in the second bearing fault verification window after correction, the uncorrected kurtosis is determined based on the acoustic kurtosis data in the last verification cycle before correction, and the kurtosis difference is determined based on the absolute value of the difference between the corrected kurtosis and the uncorrected kurtosis. The anomaly type is determined based on the kurtosis difference value; If the kurtosis difference is less than or equal to the preset kurtosis difference, the load of the power equipment is corrected based on the ratio of the composite anomaly index to the preset composite anomaly index. If the kurtosis difference value is greater than the preset kurtosis difference value, the current anomaly is determined to be an early failure of the fan bearing, a fault warning is output and the suspected fan speed is restored; The bearing fault verification window refers to a preset continuous time period used to determine the kurtosis of the acoustic kurtosis data.

[0012] Furthermore, the process of correcting the load of the power equipment based on the ratio of the composite anomaly index to the preset composite anomaly index includes, The composite anomaly index ratio is determined based on the ratio of the composite anomaly index to the preset composite anomaly index, and the load of the power equipment is reduced based on the composite anomaly index ratio, wherein the reduction in the load of the power equipment is proportional to the composite anomaly index.

[0013] Furthermore, the original amplitude ratio is determined based on the ratio of the amplitude of the 200Hz frequency component to the amplitude of the 100Hz frequency component in the tank vibration data within the first winding fault verification window before correction, and the corrected component amplitude ratio is determined based on the ratio of the amplitude of the 200Hz frequency component to the amplitude of the 100Hz frequency component in the tank vibration data within the second winding fault verification window. The absolute value of the amplitude difference is determined based on the absolute value of the difference between the original amplitude ratio and the amplitude ratio of the modified component; Based on determining the cause of the abnormal absolute value of the amplitude difference; If the absolute value of the amplitude difference is less than or equal to the set absolute value of the preset amplitude difference, it is determined that the preset composite anomaly index is set unreasonably, the load of the power equipment is restored, and the 100Hz vibration amplitude is determined based on the oil tank vibration data in the second winding fault verification window after restoration. The preset composite anomaly index is corrected based on the ratio of the 100Hz vibration amplitude to the set preset 100Hz vibration amplitude. If the absolute value of the amplitude difference is greater than the preset absolute value of the amplitude difference, the current anomaly is determined to be a winding loosening fault, a fault warning is output, and the load of the power equipment is restored. The winding fault verification window refers to a preset continuous time period used to determine the corrected 100Hz vibration amplitude.

[0014] Furthermore, the process of correcting the preset composite anomaly index based on the ratio of the 100Hz vibration amplitude to the preset 100Hz vibration amplitude includes, The vibration amplitude of 100Hz is determined based on the oil tank vibration data in the second winding fault verification window after recovery. The vibration amplitude ratio is determined based on the ratio of the 100Hz vibration amplitude to the preset 100Hz vibration amplitude. The preset composite anomaly index is increased based on the vibration amplitude ratio, and the increase of the preset composite anomaly index is proportional to the vibration amplitude ratio.

[0015] Compared with existing technologies, the beneficial effects of this invention are as follows: By collecting several types of multi-source heterogeneous data, including vibration data, tank temperature data, narrowband acoustic signature data, and acoustic kurtosis data, this invention constructs a composite anomaly index containing vibration components, temperature components, and acoustic penalty terms, thereby achieving a multi-dimensional comprehensive assessment of the mechanical condition of power equipment. Compared with monitoring a single physical quantity, this invention effectively overcomes the problem of misjudgment caused by occasional fluctuations. At the same time, through a hierarchical and progressive monitoring and early warning mechanism, it first identifies suspected anomalies based on the composite anomaly index, and then combines the online status judgment of the cooler with the ratio analysis of vibration data to achieve accurate differentiation between coupled false anomalies of the cooler and real faults of the equipment itself. This significantly improves the accuracy and reliability of non-contact monitoring and solves the problem in existing technologies of being unable to distinguish between highly overlapping mechanical fault characteristics and false anomalies.

[0016] Furthermore, by comparing the composite anomaly index with a preset composite anomaly index, a clear anomaly judgment threshold mechanism was established, enabling the system to maintain stable detection sensitivity under complex operating conditions. When a suspected anomaly is detected, the online status of the cooler is prioritized. By utilizing the key operating condition information of whether the cooler is running, the source of the anomaly is initially diverted. If the cooler is offline, the possibility of coupling interference is directly ruled out. If the cooler is online, further verification is performed by comparing the average amplitude within the jump detection time window. This hierarchical verification strategy effectively avoids the interference of vibration coupling during normal cooler operation on anomaly judgment, significantly reduces the false alarm rate of the system, and improves the engineering reliability of the monitoring results.

[0017] Furthermore, by calculating the ratio of the average amplitude of vibration data within two adjacent jump detection time windows before and after cooler startup, i.e., the jump ratio, the influence of the cooler startup event on the oil tank vibration amplitude is quantified. This parameter can intuitively reflect the coupling strength between oil tank vibration and cooler operation. When the jump ratio is less than or equal to the preset jump ratio, it indicates that the cooler startup has not caused a significant change in oil tank vibration, and the vibration mainly originates from the equipment itself. This allows for a rapid determination of a true anomaly, directly entering the fault superposition judgment process, avoiding unnecessary complex verification, optimizing the allocation of computing resources, and improving the system's real-time response capability.

[0018] Furthermore, by calculating the coherence coefficient between cooler vibration data and oil tank vibration data, the proportion of oil tank vibration energy originating from cooler transmission was precisely quantified, enabling accurate differentiation between two states: pure cooler-coupled false anomalies and coupled superimposed real faults. When the coherence coefficient is less than or equal to the preset coherence coefficient, it is determined to be a superposition of a real anomaly and a cooler-coupled false anomaly. In this case, not only is the vibration component weighting coefficient corrected, but further fault superposition judgment is also performed to ensure that no real faults are missed. When the coherence coefficient is greater than the preset coherence coefficient, it is determined to be a pure cooler-coupled false anomaly, and only the weighting coefficient needs to be corrected. This differentiated processing strategy ensures comprehensive diagnosis while avoiding excessive intervention in normal equipment operation, achieving an effective balance between monitoring accuracy and maintenance efficiency.

[0019] Furthermore, by determining the coherence coefficient ratio based on the ratio of the coherence coefficient to the preset coherence coefficient, and reducing the vibration component weighting coefficient according to this ratio, with the reduction magnitude proportional to the coherence coefficient ratio, adaptive matching between the correction magnitude and the cooler coupling strength is achieved. When the coupling strength is large, the reduction magnitude of the vibration component weighting coefficient increases accordingly, ensuring that the corrected composite anomaly index can truly reflect the mechanical state of the equipment body. When the coupling strength is small, the reduction magnitude decreases accordingly, avoiding over-correction that leads to a decrease in sensitivity to real faults. This dynamic correction mechanism effectively eliminates the problem of continuous false alarms caused by cooler coupling, while maintaining the system's ability to detect real faults in the equipment body, significantly improving the long-term stability and reliability of the monitoring system.

[0020] Furthermore, by detecting the number of local maxima in the 195Hz-215Hz frequency band of narrowband acoustic data within the superimposed detection time window, rapid triage and identification of two overlapping faults—early fan bearing faults and slight winding loosening—was achieved. When the number of maxima is 0, it is determined to be a common anomaly, and other mechanical faults are directly identified based on preset rules, avoiding unnecessary specialized verification processes. When the number of maxima is greater than 0 and less than or equal to the preset number of maxima, a clear causal relationship between the fault type and response characteristics is established by actively adjusting the speed of the suspected fan and comparing the kurtosis changes before and after the correction, thus achieving accurate differentiation between the two similar faults. When the number of maxima is greater than the preset number of maxima, it is directly determined to be a fault superposition state where both faults coexist, and a comprehensive warning is output. This three-level triage mechanism based on the number of local maxima significantly optimizes the diagnostic process, greatly reducing the consumption of computing resources and diagnostic time while ensuring diagnostic accuracy.

[0021] Furthermore, by calculating the kurtosis difference and comparing it with a preset kurtosis difference, a simple and clear binary judgment rule was established. When the kurtosis difference is greater than the preset kurtosis difference, it indicates that the impact of the acoustic signal is significantly reduced after the fan speed is reduced, which conforms to the physical law that the characteristic frequency of bearing failure is proportional to the speed. Thus, it is accurately determined to be an early failure of the fan bearing, and an early warning is output and the speed is restored. When the kurtosis difference is less than or equal to the preset kurtosis difference, it indicates that the impact characteristic does not change with the speed, ruling out the possibility of bearing failure. Instead, the winding looseness is verified by correcting the load. This verification method based on kurtosis change does not require precise extraction of the fault characteristic frequency, and the calculation is simple and efficient. Moreover, by selecting the second bearing failure verification window after correction for calculation, the transient interference during the speed adjustment transition period is effectively avoided, which significantly improves the accuracy of judgment and engineering practicality.

[0022] Furthermore, by determining the composite anomaly index ratio based on the ratio of the composite anomaly index to a preset composite anomaly index, and reducing the load on the power equipment according to this ratio, with the reduction magnitude proportional to the composite anomaly index ratio, an adaptive match between the load adjustment magnitude and the severity of the anomaly is achieved. When the composite anomaly index ratio is large, it indicates a more severe anomaly. In this case, a larger reduction in load causes a sufficiently significant decrease in electromagnetic force, thereby observing a significant change in vibration amplitude and fully verifying the existence of winding looseness. When the composite anomaly index ratio is small, only a smaller reduction in load is needed to observe the vibration response, avoiding unnecessary impact on the normal operation of the equipment. This design minimizes the impact of active intervention on the operation of the power system while ensuring the verification effect. Moreover, the load adjustment is a reversible operation, which can be restored immediately after verification, and has extremely high engineering application value.

[0023] Furthermore, by calculating the absolute value of the difference between the original amplitude ratio and the amplitude ratio of the corrected component, i.e., the absolute value of the amplitude difference, and determining the cause of the anomaly based on this parameter, the system accurately distinguishes between two situations: winding loosening faults and unreasonable preset composite anomaly index settings. When the absolute value of the amplitude difference is less than or equal to the absolute value of the preset amplitude difference, it indicates that the harmonic ratio change before and after load reduction is not significant, which does not conform to the physical characteristics of winding loosening. Therefore, it is determined that the preset composite anomaly index setting is unreasonable. At this time, the preset composite anomaly index is corrected based on the ratio of the recovered 100Hz vibration amplitude to the preset 100Hz vibration amplitude, so that the system threshold matches the actual operating state of the equipment. When the absolute value of the amplitude difference is greater than the absolute value of the preset amplitude difference, it indicates that the harmonic ratio decreases significantly with load reduction, which conforms to the characteristics of winding loosening. A fault warning is output and the load is restored. This judgment method based on changes rather than absolute values ​​eliminates the influence of individual equipment differences and operating condition differences, making the diagnostic rules more universal and robust, while realizing closed-loop optimization of the system's self-learning ability.

[0024] Furthermore, the vibration amplitude ratio is determined by comparing the 100Hz vibration amplitude with a preset 100Hz vibration amplitude. Based on this ratio, a preset composite anomaly index is increased, with the increase proportional to the vibration amplitude ratio, thus achieving adaptive correction of the preset composite anomaly index. When the vibration amplitude ratio is large, it indicates that after eliminating all specific faults, the current vibration level still deviates significantly from the normal baseline. In this case, the preset composite anomaly index is increased substantially to match the corrected threshold with the actual health status of the equipment. When the vibration amplitude ratio is small, a small increase in the preset composite anomaly index suffices. This dynamic correction mechanism ensures that the system threshold remains within a reasonable range, preventing frequent false alarms due to excessively low thresholds and missed alarms due to excessively high thresholds. This effectively improves the long-term stability and reliability of the system, achieving self-optimization and continuous improvement of the monitoring system. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating the non-contact monitoring and early warning method for power equipment status based on multi-source data in an embodiment of the present invention.

[0026] Figure 2 This is a flowchart illustrating how a composite anomaly index is used to determine whether a power equipment exhibits a suspected anomaly, as described in this embodiment of the invention.

[0027] Figure 3 This is a flowchart illustrating the determination of the actual mechanical state of power equipment based on the leap ratio in an embodiment of the present invention.

[0028] Figure 4 This is a flowchart illustrating the correction of vibration component weights based on the coherence coefficient ratio in an embodiment of the present invention. Detailed Implementation

[0029] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0030] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0031] This invention proposes a non-contact monitoring and early warning method for the status of power equipment based on multi-source data, comprising: Data acquisition and preprocessing: Real-time acquisition and preprocessing of vibration data of power equipment, oil tank temperature data, narrowband acoustic fingerprint data, and acoustic kurtosis data. Among them, vibration data includes cooler vibration data and oil tank vibration data. A composite anomaly index is constructed by determining the vibration component, temperature component, and acoustic penalty term based on the vibration data, the oil tank temperature data, and the narrowband acoustic signature data, respectively, and by determining the composite anomaly index based on the weighted summation of the vibration component, temperature component, and acoustic penalty term. The system monitors and issues early warnings, determines whether the equipment has a suspected abnormality based on the composite anomaly index, and detects the online status of the cooler when a suspected abnormality is detected. When the cooler is online, it determines the actual mechanical status of the equipment and the cause of the abnormality based on the ratio of the average amplitude of the oil tank vibration data within two adjacent jump detection time windows before and after the last start of the cooler by the power equipment. It also corrects the preset composite anomaly index and the composite anomaly index construction parameters based on the cause of the abnormality, or outputs an abnormality warning.

[0032] In this embodiment of the invention, the power equipment refers to power equipment that generates a large amount of heat during operation and requires a liquid medium for insulation and heat dissipation. Such power equipment needs to be equipped with a cooler and an oil tank, such as oil-immersed power transformers, oil-immersed reactors, high-voltage transformers, etc., which will not be elaborated here. When collecting vibration data from coolers in power equipment, real-time data collection should also be performed even when the cooler is not in operation to ensure data integrity. The vibration data refers to the mechanical oscillation signals generated on the surface of the mechanical structure of the power equipment under operating conditions by non-contact sensors, including time-series data of physical quantities such as displacement, velocity or acceleration changing over time. The tools for collecting vibration data include, but are not limited to, laser vibration meters (for oil tanks) and accelerometers (for coolers), which will not be elaborated here. The oil tank temperature data refers to the thermal state parameters of the power equipment oil tank collected by non-contact infrared thermal imaging equipment or contact temperature sensor. The oil tank temperature data includes, but is not limited to, winding hot spot temperature and top oil temperature. The winding hot spot temperature can be measured by infrared thermal imager to measure the surface temperature distribution of the oil tank, and the top oil temperature can be measured by infrared thermal imager to non-contactly measure the temperature of the top outer wall of the oil tank. This will not be elaborated further. Narrowband acoustic signature data refers to the frequency band characteristics and sound pressure data within the 195Hz-215Hz range radiated into the air by electrical equipment in operation, collected through a microphone array. Acoustic kurtosis data refers to the sound pressure data of electrical equipment under operating conditions, collected through a microphone array; The preprocessing steps are not limited in principle. For example, the collected data can be denoised and aligned with the time axis, followed by Fourier transform, etc. Among them, the vibration component reflects the change in mechanical stiffness. The vibration component is obtained by dividing the vibration velocity amplitude of the tank surface by the historical normal baseline value. The vibration velocity amplitude of the tank surface can be extracted from the 100Hz vibration velocity amplitude of the tank vibration amplitude spectrum obtained from the preprocessed tank vibration data, which is the vibration velocity amplitude VC of the tank surface. The temperature component reflects the load and cooling status of the power equipment. Temperature component TC = (winding hot spot temperature - top oil temperature) / 30℃. The acoustic penalty term APT is an additive correction term introduced during the construction of the composite anomaly index, based on the degree of abnormal increase in sound pressure level in the 195Hz to 215Hz frequency band of narrowband acoustic data relative to the historical baseline level. When the average sound pressure level in the 195Hz to 215Hz frequency band is detected to exceed the historical normal baseline value by more than 10dB, APT is set to 0.3; otherwise, it is set to 0. The formula for determining the composite anomaly index based on the weighted summation of vibration components, temperature components, and acoustic penalty terms is: Composite Anomaly Index MCI = VC × VCC + TC × TCC - APT, where VCC is the weighting coefficient of the vibration component, TCC is the weighting coefficient of the temperature component, VCC + TCC = 1, 0.4 ≤ VCC ≤ 0.6, and 0.4 ≤ TCC ≤ 0.6. The rise detection time window refers to a continuous time period of a preset duration used to detect the average amplitude in vibration data. When the rise detection time window is less than 10 seconds, the number of sampling points is less than 10,000, the standard error of the mean exceeds 1.0%, and the statistical fluctuation of vibration amplitude is large, which may lead to instability in the rise ratio calculation. When the rise detection time window reaches 30 seconds, the standard error of the mean drops to 0.58%, and the statistical stability meets the engineering requirements. Further increasing the window length has limited effect on improving statistical accuracy; therefore, the rise detection time window is set to a range of 10s to 30s. The average amplitude in the oil tank vibration data refers to the arithmetic mean obtained by continuously collecting and calculating the 100Hz vibration velocity amplitude of the oil tank surface within the jump detection time window. Please see Figure 1 The diagram shown is a flowchart of the non-contact monitoring and early warning method for power equipment status based on multi-source data in an embodiment of the present invention. The workflow of the non-contact monitoring and early warning method for power equipment status based on multi-source data in an embodiment of the present invention includes: S1: Real-time acquisition and preprocessing of power equipment vibration data, oil tank temperature data, narrowband acoustic signature data, and acoustic kurtosis data; S2: Determine the vibration component, temperature component, and acoustic penalty term based on the vibration data, the oil tank temperature data, and the narrowband acoustic signature data, respectively. S3: Determine the composite anomaly index based on the weighted summation of vibration components, temperature components, and acoustic penalty terms; S4: Based on the composite anomaly index, determine whether the equipment has a suspected anomaly, and when a suspected anomaly is detected, check the online status of the cooler. When the cooler is online, determine the actual mechanical status of the equipment and the cause of the anomaly based on the ratio of the average amplitude of the oil tank vibration data within two adjacent jump detection time windows before and after the last start of the cooler by the power equipment. Based on the cause of the anomaly, correct the preset composite anomaly index and the composite anomaly index construction parameters, or output an anomaly warning.

[0033] Specifically, the process of determining whether a device has a suspected anomaly based on the composite anomaly index includes, If the composite anomaly index is less than or equal to the preset composite anomaly index, the power equipment is determined to be operating normally. If the composite anomaly index is greater than the preset composite anomaly index, it is determined that the power equipment has a suspected anomaly, and the cooler is checked to see if it is online. If the cooler is online, the actual mechanical state of the equipment is determined based on the ratio of the average amplitude of the vibration data in the two adjacent jump detection time windows before and after the most recent start of the cooler. If the cooler is offline, the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic signature within the superimposed detection time window is used to determine whether a suspected fault superposition has occurred.

[0034] Among them, suspected anomaly refers to an intermediate state in which the system determines that the power equipment may have an abnormal mechanical condition when the composite anomaly index exceeds the preset threshold, but the actual source of the fault has not yet been confirmed by the branch verification process.

[0035] The vibration component reflects the vibration intensity of the equipment's mechanical structure, the temperature component reflects the load and cooling status, and the acoustic penalty term reflects high-frequency impact characteristics. These three physical quantities describe the equipment's operating status from different dimensions and are relatively independent of each other. When a physical quantity exhibits an abnormal change, the composite anomaly index will increase accordingly, indicating a possible abnormal situation. Compared to single-parameter monitoring, the composite anomaly index integrates multi-dimensional information and can, to some extent, eliminate occasional fluctuations of a single physical quantity, thus providing higher accuracy and reliability in anomaly detection. When the composite anomaly index is less than or equal to the preset composite anomaly index, it indicates that all physical quantities are currently within the normal fluctuation range, the equipment is operating well, no abnormal signs requiring attention are detected, and the system continues to maintain monitoring. However, when the composite anomaly index exceeds the preset composite anomaly index, it is first determined whether the cooler is online. If it is offline, it indicates that cooler vibration has been ruled out. The possibility of coupling interference means that an increase in the composite anomaly index must originate from a fault in the equipment itself, thus the current anomaly can be directly determined as a real anomaly. However, when the cooler is online, the cooler itself will vibrate during operation, and when its vibration frequency is close to the mechanical modal frequency of the oil tank, a vibration coupling phenomenon will occur, resulting in a falsely high vibration amplitude on the surface of the oil tank, thereby passively increasing the composite anomaly index. At this time, the composite anomaly index itself cannot distinguish whether the vibration increase originates from cooler coupling or a fault in the equipment itself. Therefore, this application needs to combine the online status of the cooler with the actual vibration situation for further judgment to determine whether the current anomaly is a pseudo-anomaly caused by cooler coupling. Through the composite anomaly index, this invention can maintain high detection sensitivity to real faults and eliminate pseudo-anomalies through subsequent hierarchical verification mechanisms, thereby achieving accurate monitoring and early warning of the mechanical status of power equipment, effectively reducing the false alarm rate, and avoiding unnecessary maintenance due to misjudgment of a single parameter.

[0036] Please see Figure 2 As shown, this is a flowchart illustrating the process of determining whether power equipment exhibits a suspected anomaly based on a composite anomaly index in an embodiment of the present invention. The process of determining whether equipment exhibits a suspected anomaly based on the composite anomaly index in an embodiment of the present invention includes: The composite anomaly index (MCI) is obtained and compared with the preset composite anomaly index (MCI1). The composite anomaly index is composed of a weighted sum of vibration component, temperature component, and acoustic penalty term. Under normal operating conditions with no mechanical faults confirmed, at night with low load and the cooler completely stopped, the 100Hz vibration amplitude is close to the historical baseline, with the vibration component being approximately 1.0, the temperature component being approximately 0.1 to 0.3, and the acoustic penalty term being 0 when there is no high-frequency impact. The composite anomaly index is approximately 0.5 to 0.6. During the day with high load and the cooler running, the 100Hz vibration amplitude can rise to 2.0 to 2.5 times the baseline due to the increased electromagnetic force, with the vibration component reaching 2.0 to 2.5. The temperature difference between the winding hot spot temperature and the oil temperature can reach 15℃ to 25℃, with the temperature component being 0.5 to 0.8. At this time, the theoretical calculated value of the composite anomaly index can reach 1.15-1.49. To avoid the preset composite anomaly index (MCI1) being too sensitive, the preset composite anomaly index (MCI1) is set to [1.2, 1.5]. If the composite anomaly index MCI is less than or equal to the preset composite anomaly index MCI1, the power equipment is determined to be operating normally. If the composite anomaly index MCI is greater than the preset composite anomaly index MCI1, it is determined that the power equipment has a suspected anomaly, and the cooler is checked to see if it is online. If the cooler is online, the actual mechanical state of the equipment is determined based on the ratio of the average amplitude of the vibration data in the two adjacent jump detection time windows before and after the most recent start of the cooler. If the cooler is offline, the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic data within the superimposed detection time window is used to determine whether a suspected fault superposition has occurred.

[0037] Specifically, the process of determining the actual mechanical state of the equipment based on the ratio of the average amplitude of the vibration data within two adjacent rise detection time windows before and after the most recent start-up of the cooler includes the following: The jump ratio is determined based on the ratio of the average amplitude of the vibration data within two adjacent jump detection time windows before and after the most recent start-up of the cooler by the power equipment. If the jump ratio is less than or equal to the preset jump ratio, the suspected abnormality is determined to be a true abnormality. The number of local maxima in the 195Hz-215Hz frequency band of the narrowband voiceprint data within the superimposed detection time window is used to determine whether a suspected fault superposition occurs. If the jump ratio is greater than the preset jump ratio, the actual mechanical state of the power equipment is determined based on the cooler vibration data and the oil tank vibration data in the superimposed detection time window. The superimposed detection time window refers to a continuous time period of preset duration that is taken forward from the moment when a suspected abnormality is determined. The value of the superimposed detection time window is not limited in principle, and technicians can set it according to their needs. For example, the preset duration of the superimposed detection time window can be between 2 seconds and 10 seconds. The jump ratio reflects the degree of influence of the cooler startup event on the oil tank vibration amplitude. When the cooler starts, if the oil tank vibration amplitude shows a significant jump, it indicates a strong coupling relationship between the oil tank vibration and the cooler operation. Conversely, if the oil tank vibration amplitude does not change significantly, it indicates that the oil tank vibration mainly originates from the equipment body rather than the cooler. When the jump ratio is greater than the preset jump ratio, it indicates that the cooler startup caused a significant jump in the oil tank vibration. At this time, there are two possibilities: one is that the anomaly is entirely caused by cooler coupling, and the other is that cooler coupling and equipment body failure coexist. Therefore, when the jump ratio is large, the anomaly type cannot be directly determined. It is necessary to further distinguish between the coherence coefficients of cooler vibration and oil tank vibration. Through the jump ratio, this invention can quickly screen out operating conditions with a high probability of cooler coupling, and effectively eliminate false anomalies.

[0038] Please see Figure 3 As shown, this is a flowchart illustrating the process of determining the actual mechanical state of power equipment based on the rise ratio in an embodiment of the present invention. The process of determining the actual mechanical state of power equipment based on the rise ratio in an embodiment of the present invention includes: The jump ratio LAB is determined based on the ratio of the average amplitude of the vibration data within two adjacent jump detection time windows before and after the most recent start-up of the cooler by the power equipment. The jump ratio LAB is compared with the preset jump ratio LAB1. Under the condition that the cooler is installed in a good condition, the vibration damping pads are not aged, and the oil tank structure is not abnormal, the change in the 100Hz vibration amplitude of the oil tank caused by the start-up of the cooler is usually between 10% and 30%, that is, the jump ratio is in the range of 1.1 to 1.3. However, when cooler coupling occurs, the jump ratio may increase significantly. At the same time, in order to avoid being too sensitive, the preset jump ratio LAB1 is set to [1.1, 1.4]. If the rise ratio LAB is less than or equal to the preset rise ratio LAB1, the suspected abnormality is determined to be a true abnormality. The number of local maxima in the 195Hz-215Hz frequency band of the narrowband voiceprint data within the superimposed detection time window is used to determine whether a suspected fault superposition occurs. If the jump ratio LAB is greater than the preset jump ratio LAB1, the actual mechanical state of the power equipment is determined based on the cooler vibration data and the oil tank vibration data in the superimposed detection time window.

[0039] Specifically, the process of determining the actual mechanical state of the power equipment based on the cooler vibration data and the oil tank vibration data within the superimposed detection time window includes: Based on the cooler vibration data and the oil tank vibration data in the superimposed detection time window, the cross power spectrum, cooler power spectrum and oil tank power spectrum are determined. The coherence coefficient is determined based on the cross power spectrum, the cooler power spectrum, and the oil tank power spectrum; The actual mechanical state of the power equipment is determined based on the coherence coefficient. The power spectrum is obtained by performing a Fourier transform on the original vibration data to obtain a 100Hz vibration signal, and then calculating it using the single-sided power spectrum. The calculation method of the cross power spectrum is not limited. For example, the oil tank vibration data (100Hz) can be complex conjugate, and then the cross power spectrum can be calculated using the cross power spectrum formula.

[0040] Cross power spectrum

[0041] Where X(f) represents the cooler vibration signal;

[0042] This indicates a vibration signal from the fuel tank.

[0043] N represents the number of sampling points. Indicates the sampling frequency;

[0044] Coherence coefficient =

[0045] in, This represents the autopower spectrum of the cooler vibration signal. The power spectrum representing the vibration signal of the fuel tank;

[0046] If the coherence coefficient is less than or equal to the preset coherence coefficient, the suspected anomaly is determined to be a superposition of a true anomaly and a false anomaly coupled to the cooler. Based on the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic data within the superposition detection time window, it is determined whether a suspected fault superposition occurs. If the coherence coefficient is greater than the preset coherence coefficient, the suspected anomaly is determined to be a false anomaly of the cooler coupling, and the weighting coefficient of the vibration component is adjusted when constructing the composite anomaly index based on the ratio of the preset coherence coefficient to the coherence coefficient.

[0047] The coherence coefficient reflects the strength of the linear causal relationship between cooler vibration and tank vibration in the frequency domain. Its value is equal to the square of the cross-power spectral modulus of the cooler vibration and tank vibration divided by the product of their respective power spectra. The value ranges from 0 to 1. When the coherence coefficient is greater than the preset coherence coefficient, it indicates that the tank vibration and cooler vibration are highly coherent, and almost all the vibration energy on the tank surface originates from the transmission of cooler vibration. There is no real mechanical fault in the equipment itself; therefore, the current anomaly is determined to be a pure cooler-coupled spurious anomaly. When the coherence coefficient is less than or equal to the preset coherence coefficient, it indicates that only a portion of the energy in the tank vibration is related to cooler vibration, while the remaining vibration... The energy originates from the actual mechanical defects of the equipment itself, thus determining that the current anomaly is a superposition of a true anomaly and a spurious anomaly coupled with the cooler. The coherence coefficient, based on the physical nature of energy transfer, accurately quantifies the contribution ratio of cooler vibration to oil tank vibration, enabling quantitative differentiation between two states: pure cooler-coupled spurious anomaly and coupled superimposed real fault. This provides a reliable basis for the formulation of subsequent differentiated processing strategies, avoiding both ineffective maintenance caused by misjudging pure spurious anomalies as real faults and underreporting of real faults caused by simply classifying superimposed anomalies as spurious anomalies. This significantly improves the diagnostic accuracy and engineering practicality of the monitoring system.

[0048] Specifically, the process of determining the actual mechanical state of the power equipment based on the cooler vibration data and the oil tank vibration data includes, Based on the cooler vibration data and the oil tank vibration data, the cross power spectrum, cooler power spectrum and oil tank power spectrum are determined. The coherence coefficient CC is determined based on the cross power spectrum, the cooler power spectrum, and the oil tank power spectrum. The coherence coefficient CC is compared with the preset coherence coefficient CC1. Theoretical analysis shows that when the tank vibration is caused entirely by the cooler vibration transmitted through the structure, the two should have a high degree of consistency under the action of the transfer function, and the theoretical value of the coherence coefficient should be greater than 0.95. However, in actual measurement, there are sensor noise, nonlinear transmission paths and environmental vibration interference. The measured coherence coefficient under pure coupling conditions is usually between 0.85 and 0.98. When there are real faults such as loose windings in the equipment body, the component of the tank vibration from electromagnetic force is unrelated to the cooler vibration and will dilute the coherence coefficient. Therefore, the preset coherence coefficient CC1 is set to a value range of 0.8-0.85. If the coherence coefficient CC is less than or equal to the preset coherence coefficient CC1, the suspected abnormality is determined to be a superposition of a true abnormality and a false abnormality coupled with the cooler. Based on the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic data within the superposition detection time window, it is determined whether a suspected fault superposition occurs. If the coherence coefficient CC is greater than the preset coherence coefficient CC1, the suspected anomaly is determined to be a false anomaly of the cooler coupling. The weighting coefficient of the vibration component is adjusted when constructing the composite anomaly index based on the ratio of the preset coherence coefficient to the coherence coefficient.

[0049] Specifically, the process of constructing the weighting coefficients of the vibration components when modifying the composite anomaly index based on the ratio of the coherence coefficient to the preset coherence coefficient includes the following steps: The coherence coefficient ratio is determined based on the ratio of the coherence coefficient to the preset coherence coefficient, and the vibration component weighting coefficient is reduced based on the coherence coefficient ratio, wherein the reduction in the vibration component weighting coefficient is proportional to the coherence coefficient ratio.

[0050] When a suspected anomaly is determined to be a false anomaly due to cooler coupling, it indicates that the current increase in the composite anomaly index is mainly caused by cooler vibration coupling, rather than a real fault in the equipment itself. In this case, the contribution of the vibration component to the composite anomaly index is artificially amplified by the coupling component. If it is not corrected, the composite anomaly index will continue to exceed the threshold during normal cooler operation in subsequent monitoring, causing the system to repeatedly trigger false alarms. The coherence coefficient ratio reflects the strength of cooler coupling. By reducing the vibration component weight coefficient based on the coherence coefficient ratio, and the reduction of the vibration component weight coefficient is proportional to the coherence coefficient ratio, the correction magnitude and coupling strength are adaptively matched, avoiding the problems of insufficient or excessive correction that may be caused by fixed magnitude correction. Secondly, the corrected composite anomaly index can truly reflect the mechanical state of the equipment itself. When the cooler is operating normally but the equipment itself is fault-free, the composite anomaly index will no longer exceed the threshold due to coupling, thereby effectively eliminating false alarms. By reducing the vibration component weight coefficient based on the coherence coefficient ratio, this application can enable the composite anomaly index to maintain sensitivity to real faults while effectively suppressing interference caused by cooler coupling, thereby improving the reliability and accuracy of the monitoring system.

[0051] Please see Figure 4 The diagram shows a flowchart illustrating the process of correcting vibration component weights based on the coherence coefficient ratio in an embodiment of the present invention. The process of correcting vibration component weights based on the coherence coefficient ratio in this embodiment includes: The coherence coefficient ratio CCB is determined based on the ratio of the coherence coefficient to the preset coherence coefficient. The coherence coefficient ratio CCB is compared with the set first preset coherence coefficient ratio CCB1 and second preset coherence coefficient ratio CCB2. According to engineering experience, when the actual coherence coefficient exceeds the preset value by less than 20%, it is still considered a slightly enhanced coupling scenario. When it exceeds 1.5, it is considered that the current coupling strength is high. When it is greater than 1.5, it is considered that the current scenario is a strong coupling scenario. Therefore, the first preset coherence coefficient ratio CCB1 is set to (1, 1.2] and the second preset coherence coefficient ratio CCB2 is set to (1.2, 1.5]. If the coherence coefficient ratio CCB is less than or equal to the first preset coherence coefficient ratio CCB1, then the vibration component weight coefficient is corrected by the first vibration weight correction coefficient α1. The corrected vibration component weight coefficient VCC' = VCC × α1, where the first vibration weight correction coefficient α1 is set to 0.98. If the coherence coefficient ratio CCB is greater than the first preset coherence coefficient ratio CCB1 and less than or equal to the second preset coherence coefficient ratio CCB2, then the vibration component weight coefficient is corrected by the second vibration weight correction coefficient α2. The corrected vibration component weight coefficient VCC' = VCC × α2, where the second vibration weight correction coefficient α2 is set to 0.95. If the coherence coefficient ratio CCB is greater than the second preset coherence coefficient ratio CCB2, then the vibration component weight coefficient is corrected by the third vibration weight correction coefficient α3. The corrected vibration component weight coefficient VCC' = VCC × α3, where the third vibration weight correction coefficient α3 is set to 0.9.

[0052] Specifically, the process of determining whether a suspected fault superposition occurs based on the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic data within the superposition detection time window includes: The process of determining whether a suspected fault superposition occurs based on the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic data within the superposition detection time window includes: The number of maxima is determined based on the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic text data within the superimposed detection time window. If the maximum number is 0, the current anomaly is determined to be a normal anomaly, and the cause of the anomaly is determined based on preset rules; If the maximum value is greater than 0 and less than or equal to the preset maximum value, the speed of the suspected fan is reduced based on the speed test coefficient, and the uncorrected kurtosis and corrected kurtosis are determined based on the acoustic kurtosis data in the last verification cycle before correction and the second bearing fault verification window after correction, respectively. The anomaly type is determined based on the difference between the corrected kurtosis and the uncorrected kurtosis. If the maximum value is greater than the preset maximum value, the current anomaly is determined to be a fault superposition, and an early warning of fan bearing fault and winding loosening fault is output. The suspected fan refers to the cooling fan identified as generating an abnormal sound source based on the narrowband acoustic data. The method for locating the suspected fan is not limited; for example, a microphone array can be used to collect narrowband acoustic data, and a beamforming algorithm can be used to calculate the directional angle of the sound source within the 195Hz-215Hz frequency band. The beamforming algorithm calculates the energy distribution in different directions by summing the signals received by each microphone in the array after a delay. The direction with the highest energy is the direction of the sound source. The calculated directional angle of the sound source is compared with the installation coordinates of each cooling fan to determine that the fan group containing the sound source is the suspected fan. The speed test coefficient is a preset scaling factor used to control the rate of decrease in the speed of the suspected fan. The value of the speed test coefficient can be taken from 0.9 to 0.7. Among them, when determining whether the current anomaly is a normal anomaly, the cause of the anomaly is determined based on preset rules. The preset rules are not limited, and technical personnel can directly judge and distinguish them through thresholds, which will not be elaborated here.

[0053] The maximum value count refers to the number of local maxima found in the 195Hz to 215Hz frequency band after Fourier transform of the narrowband acoustic data within the superimposed detection time window. This parameter reflects the number of independent frequency components within this frequency band. Both early fan bearing faults and slight winding loosening will generate independent frequency components within this frequency band. When one of these faults exists, a local maximum is present in the frequency band; when both faults exist simultaneously, two distinguishable local maxima are present; when neither fault exists, there are no local maxima in the frequency band. Therefore, when the maximum value count is 0, it indicates that there are no independent frequency components within the 195Hz to 215Hz frequency band. The current anomaly does not belong to early fan bearing faults or slight winding loosening, and is therefore judged as a common anomaly, which is another mechanical fault that can be directly identified by preset rules (such as iron core magnetostriction anomalies, oil pump bearing faults, etc.), without needing to enter the special verification flow for feature overlap type faults. When the maximum value is greater than 0 and less than or equal to the preset maximum value (usually 1), it indicates that there is an independent frequency component in the frequency band. The current anomaly may be either an early failure of the fan bearing or a slight loosening of the winding, but it is impossible to directly determine which one it is. Therefore, it is necessary to further verify it by reducing the speed of the suspected fan. The essence of an early failure of the fan bearing is the periodic mechanical impact between the rolling element and the raceway. Kurtosis reflects the impact of the acoustic signal. If it is a bearing failure, the reduction of the speed means that the impact energy is weakened, so there will be a decrease in kurtosis. When the maximum value is greater than the preset maximum value, it indicates that there are two or more independent frequency components in the frequency band, that is, an early failure of the fan bearing and a slight loosening of the winding coexist. At this time, it can be directly determined as a superposition of faults, without the need for further individual fault differentiation verification. Through the maximum value, this invention realizes rapid diversion of fault types, effectively optimizing the allocation of computing resources while ensuring diagnostic accuracy.

[0054] Specifically, the process of determining whether a suspected fault superposition occurs based on the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic data within the superposition detection time window includes: The number of maxima MN is determined based on the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic text data within the superimposed detection time window. The maximum number MN is compared with the preset maximum number MN1, wherein the preset maximum number MN1 is set to 1; If the maximum number MN is 0, the current anomaly is determined to be a normal anomaly, and the cause of the anomaly is determined based on preset rules; If the maximum value MN is greater than 0 and less than or equal to the preset maximum value MN1, then the suspected fan speed = speed test coefficient multiplied by the current suspected fan speed, and the uncorrected kurtosis and corrected kurtosis are determined based on the acoustic kurtosis data in the last verification cycle before correction and the second bearing fault verification window after correction, respectively, and the anomaly type is determined based on the difference between the corrected kurtosis and the uncorrected kurtosis. If the maximum value MN is greater than the preset maximum value MN1, the current abnormality is determined to be a fault superposition, and an early warning of fan bearing fault and winding loosening fault is output. Specifically, the corrected kurtosis is determined based on the acoustic kurtosis data in the second bearing fault verification window after correction, the uncorrected kurtosis is determined based on the acoustic kurtosis data in the last verification cycle before correction, and the kurtosis difference is determined based on the absolute value of the difference between the corrected kurtosis and the uncorrected kurtosis. The anomaly type is determined based on the kurtosis difference value; If the kurtosis difference value is greater than the preset kurtosis difference value, the current anomaly is determined to be an early failure of the fan bearing, a fault warning is output and the suspected fan speed is restored; If the kurtosis difference is less than or equal to the preset kurtosis difference, the load of the power equipment is corrected based on the ratio of the composite anomaly index to the preset composite anomaly index. The bearing fault verification window refers to a preset continuous time period used to determine the kurtosis of the acoustic kurtosis data, and the preset duration of the bearing fault verification window ranges from 2 seconds to 5 seconds.

[0055] The formula for calculating kurtosis is: ;

[0056] in, This represents the signal value at the i-th sampling point;

[0057] N represents the total number of sampling points;

[0058] μ represents the average amplitude of the acoustic data;

[0059] σ represents the average amplitude of the acoustic data;

[0060] Kurtosis difference refers to the difference between corrected kurtosis and uncorrected kurtosis. Uncorrected kurtosis is determined based on acoustic kurtosis data from the last verification period before correction, representing the impact level before fan speed adjustment. Corrected kurtosis is determined based on acoustic kurtosis data from the second bearing fault verification window after correction, representing the impact level after fan speed adjustment and stabilization. The reason for choosing the second verification window after correction instead of the first is that after fan speed adjustment, the mechanical motion state of the bearing needs time to reach a new stable equilibrium. The first verification window may still be in the speed transition period, at which time the collected acoustic kurtosis data contains transient interference, and the kurtosis value is unstable. The second verification window is located after the speed stabilizes, and can truly reflect the steady-state impact characteristics after correction. Choosing the last verification period before correction as the benchmark for uncorrected kurtosis ensures that the two comparison windows are closest in time, thereby minimizing the influence of other operating condition changes on kurtosis. When the kurtosis difference is greater than the preset kurtosis... When the kurtosis difference is less than or equal to the preset kurtosis difference, it indicates that the kurtosis decreases significantly after reducing the speed, meaning that the impact characteristics of the bearing failure weaken as the speed decreases. This aligns with the physical law that the characteristic frequency of the bearing failure is proportional to the speed, thus the current anomaly is determined to be an early-stage fan bearing failure. When the kurtosis difference is less than or equal to the preset kurtosis difference, it indicates that the kurtosis does not change significantly after reducing the speed, meaning that the impact characteristics do not change with the speed, which does not conform to the law of bearing failure. Therefore, the current anomaly is determined not to be a bearing failure, and further verification is conducted by adjusting the load to determine if it is a loose winding. This design establishes a clear causal relationship between the fault type and the response characteristics by actively changing the speed and observing the changes in kurtosis. Secondly, kurtosis calculation is simple, sensitive to impact signals, and does not require precise extraction of the fault characteristic frequency, resulting in high computational efficiency. Thirdly, selecting a second verification window for corrected kurtosis calculation effectively avoids transient interference during the speed adjustment transition period, improving the accuracy of the judgment. Finally, binary judgment based on kurtosis difference is simple, clear, and easy to implement in engineering.

[0061] Specifically, the corrected kurtosis is determined based on the acoustic kurtosis data in the second bearing fault verification window after correction, the uncorrected kurtosis is determined based on the acoustic kurtosis data in the last verification cycle before correction, and the kurtosis difference KD is determined based on the difference between the corrected kurtosis and the uncorrected kurtosis. The kurtosis difference KD is compared with the preset kurtosis difference KD1. Kurtosis is a fourth-order statistic. The kurtosis value of a normal Gaussian distribution signal is 3. The impact of bearing failure will significantly increase the kurtosis to more than 5. Generally speaking, a kurtosis decrease of more than 0.5 is considered a significant change and is judged as a bearing failure. It is suitable for sensitive detection of early weak faults. The upper limit of 2.0 requires that the kurtosis must change significantly to be judged as a bearing failure. It is suitable for avoiding misjudgment caused by other interference factors. Therefore, the preset kurtosis difference KD1∈[0.5,2] is set. If the kurtosis difference KD is less than or equal to the preset kurtosis difference KD1, the load of the power equipment is corrected based on the ratio of the composite anomaly index to the preset composite anomaly index. If the kurtosis difference KD is greater than the preset kurtosis difference KD1, the current anomaly is determined to be an early failure of the fan bearing, a fault warning is output, and the suspected fan speed is restored.

[0062] Specifically, the process of correcting the load of the power equipment based on the ratio of the composite anomaly index to the preset composite anomaly index includes: The composite anomaly index ratio is determined based on the ratio of the composite anomaly index to the preset composite anomaly index, and the load of the power equipment is reduced based on the composite anomaly index ratio, wherein the reduction in the load of the power equipment is proportional to the composite anomaly index.

[0063] After verification through speed correction, the kurtosis difference is less than the preset kurtosis difference, indicating that reducing the fan speed did not cause a significant decrease in kurtosis. This means the impact characteristics of the acoustic signal do not change with the speed, which does not conform to the physical laws of early fan bearing failure. Therefore, the current anomaly is determined not to be a bearing failure, but possibly a slight loosening of the winding. The essence of slight winding loosening is a decrease in structural stiffness caused by a decrease in winding clamping force. Its vibration characteristics change with the load current; the electromagnetic force is proportional to the square of the current. The greater the load, the greater the electromagnetic force, and the stronger the winding vibration. The composite anomaly index ratio refers to the ratio of the measured composite anomaly index to the preset composite anomaly index. This parameter reflects the severity of the current anomaly relative to the anomaly threshold. The larger the composite anomaly index, the more significant the vibration, temperature, or acoustic anomaly. In this scheme, the reduction in the load of the power equipment is designed to be proportional to the composite anomaly index ratio. A high composite anomaly index ratio indicates a more severe anomaly, suggesting a higher probability or degree of winding loosening. In this case, a significant load reduction is required to substantially decrease the electromagnetic force, thereby observing a significant drop in vibration amplitude. Conversely, a low composite anomaly index ratio indicates a milder anomaly. In this case, a slight load reduction is sufficient to observe vibration changes, avoiding unnecessary disruption to normal equipment operation. This design directly verifies the winding loosening fault mode by actively changing the load and observing the vibration response. Secondly, the load reduction magnitude is adaptively matched to the severity of the anomaly, ensuring thorough verification during severe anomalies and minimizing intervention during minor anomalies, achieving a balance between verification effectiveness and operational impact. Thirdly, the load adjustment is a proactive and reversible operation; the original load can be immediately restored after verification, minimizing the impact on the normal operation of the power system.

[0064] Specifically, the process of correcting the load of the power equipment based on the ratio of the composite anomaly index to the preset composite anomaly index includes: The composite anomaly index ratio (MCID) is determined based on the ratio of the composite anomaly index to the preset composite anomaly index. The composite anomaly index ratio MCID is compared with the first preset composite anomaly index ratio MCID1 and the second preset composite anomaly index ratio MCID2. The first preset composite anomaly index ratio is used to divide the boundary between slight load reduction and moderate load reduction. Based on engineering experience, when the composite anomaly index ratio is less than 1.3, it indicates that the degree of anomaly is relatively slight. When it exceeds 1.3 and reaches 1.5, it means that a larger load reduction strategy needs to be adopted for verification. When it is greater than 1.5, it indicates that the degree of anomaly is severe and deep load reduction is required for full verification. At the same time, it also provides protective load reduction for power equipment. Therefore, the first preset composite anomaly index ratio MCID1∈(1,1.3] and the second preset composite anomaly index ratio MCID2∈(1.3,1.5] are set. If the composite anomaly index ratio MCID is less than or equal to the first preset composite anomaly index ratio MCID1, then the load of the power equipment is corrected using the first load correction threshold θ1, and the corrected load of the power equipment EE' = EE × θ1, wherein the first load correction threshold θ1 is set to 0.96. If the composite anomaly index ratio MCID is greater than the first preset composite anomaly index ratio MCID1 and less than or equal to the second preset composite anomaly index ratio MCID2, then the power equipment load is corrected using the second load correction threshold θ2, and the corrected power equipment load EE' = EE × θ2, wherein the second load correction threshold θ2 is set to 0.9. If the composite anomaly index ratio MCID is greater than the second preset composite anomaly index ratio MCID2, then the load of the power equipment is corrected using the third load correction threshold θ3. The corrected load of the power equipment is EE'=EE×θ3, where the third load correction threshold θ3 is set to 0.82.

[0065] Specifically, the original amplitude ratio is determined based on the ratio of the amplitude of the 200Hz frequency component to the amplitude of the 100Hz frequency component in the tank vibration data in the first winding fault verification window before correction, and the corrected component amplitude ratio is determined based on the ratio of the amplitude of the 200Hz frequency component to the amplitude of the 100Hz frequency component in the tank vibration data in the second winding fault verification window. The absolute value of the amplitude difference is determined based on the absolute value of the difference between the original amplitude ratio and the amplitude ratio of the modified component; Based on determining the cause of the abnormal absolute value of the amplitude difference; If the absolute value of the amplitude difference is less than or equal to the set absolute value of the preset amplitude difference, it is determined that the preset composite anomaly index is set unreasonably, the load of the power equipment is restored, and the 100Hz vibration amplitude is determined based on the oil tank vibration data in the second winding fault verification window after restoration. The preset composite anomaly index is corrected based on the ratio of the 100Hz vibration amplitude to the set preset 100Hz vibration amplitude. If the absolute value of the amplitude difference is greater than the preset absolute value of the amplitude difference, the current anomaly is determined to be a winding loosening fault, a fault warning is output, and the load of the power equipment is restored. The winding fault verification window refers to a preset continuous time period used to determine the corrected 100Hz vibration amplitude, and the value range of the winding fault verification window is set to 30s-60s.

[0066] The component amplitude ratio refers to the ratio of the amplitude of the 200Hz frequency component to the amplitude of the 100Hz frequency component in the tank vibration data, i.e., the harmonic ratio. This parameter reflects the degree of nonlinear distortion of the transformer's magnetic circuit. The absolute value of the amplitude difference refers to the change in the harmonic ratio before and after reducing the load, i.e., the absolute value of the difference between the original amplitude ratio and the corrected component amplitude ratio. This parameter reflects the sensitivity of the harmonic ratio to load changes. When there is looseness in the winding, reducing the load will reduce the electromagnetic force, and the nonlinear distortion will weaken accordingly. The harmonic ratio will drop significantly from a higher value, so the absolute value of the amplitude difference is large. When the absolute value of the amplitude difference is small, it indicates that the harmonic ratio is not sensitive to load changes, which does not conform to the physical characteristics of loose winding. If the cooler coupling false anomaly and early fan bearing failure have been ruled out through the aforementioned steps, it indicates that there is a problem with the setting of the composite anomaly index itself, i.e., the preset composite anomaly index setting is unreasonable. Conversely, when the absolute value of the amplitude difference is large, it indicates that the harmonic ratio has changed significantly before and after the load is reduced, which is consistent with the physical law of winding loosening. Therefore, it is judged as a winding loosening fault. Using the change in harmonic ratio before and after the load reduction, rather than the absolute value, as the basis for judgment eliminates the influence of individual equipment differences and operating condition differences on the absolute value of the harmonic ratio, making the judgment rule more universal and robust.

[0067] Specifically, the process of determining the cause of the anomaly based on the component amplitude ratio includes: The original amplitude ratio is determined based on the ratio of the amplitude of the 200Hz frequency component to the amplitude of the 100Hz frequency component in the tank vibration data in the first winding fault verification window before correction, and the corrected component amplitude ratio is determined based on the ratio of the amplitude of the 200Hz frequency component to the amplitude of the 100Hz frequency component in the tank vibration data in the second winding fault verification window. The absolute value of the amplitude difference MR is determined based on the absolute value of the difference between the original amplitude ratio and the amplitude ratio of the modified component; The absolute value of the amplitude difference MR is compared with the preset absolute value of the amplitude difference MR1. When the winding is slightly loose, the structural stiffness decreases, which leads to an increase in magnetic circuit distortion. The 200Hz component can increase to 50% to 80% of the 100Hz component. When the load is reduced, for example, from 100% to 80%, the 200Hz component will decrease by 30%-50%. Therefore, the preset absolute value of the amplitude difference MR1 is set to [0.1, 0.3]. If the absolute value of the amplitude difference MR is less than or equal to the absolute value of the preset amplitude difference MR1, then the preset composite anomaly index is determined to be unreasonable, the load of the power equipment is restored, and the 100Hz vibration amplitude is determined based on the oil tank vibration data in the second winding fault verification window after restoration. The preset composite anomaly index is corrected based on the ratio of the 100Hz vibration amplitude to the preset 100Hz vibration amplitude. If the absolute value of the amplitude difference MR is greater than the preset absolute value of the amplitude difference MR1, then the current abnormality is determined to be a winding loosening fault, a fault warning is output and the load of the power equipment is restored; Specifically, the process of correcting the preset composite anomaly index based on the ratio of the 100Hz vibration amplitude to a preset 100Hz vibration amplitude includes, as follows: The vibration amplitude of 100Hz is determined based on the oil tank vibration data in the second winding fault verification window after recovery. The vibration amplitude ratio is determined based on the ratio of the 100Hz vibration amplitude to the preset 100Hz vibration amplitude. The preset composite anomaly index is increased based on the vibration amplitude ratio, and the increase of the preset composite anomaly index is proportional to the vibration amplitude ratio.

[0068] The vibration amplitude ratio refers to the ratio of the 100Hz vibration amplitude in the second winding fault verification window after load recovery to the preset 100Hz vibration amplitude. This parameter reflects the degree of deviation of the current vibration level from the normal baseline after excluding all specific faults such as cooler coupling, bearing failure, and winding looseness. The preset composite anomaly index is increased based on the vibration amplitude ratio, and the increase of the preset composite anomaly index is proportional to the vibration amplitude ratio. This helps to ensure that the correction magnitude and the degree of deviation are adaptively matched, thereby effectively improving the stability and reliability of the system.

[0069] Specifically, the process of correcting the preset composite anomaly index based on the ratio of the 100Hz vibration amplitude to the preset 100Hz vibration amplitude includes: The 100Hz vibration amplitude is determined based on the oil tank vibration data in the second winding fault verification window after recovery. The vibration amplitude ratio OVAD is determined based on the ratio of the 100Hz vibration amplitude to the preset 100Hz vibration amplitude. The value of the preset 100Hz vibration amplitude is not limited and can be determined by technicians based on historical data baselines. The vibration amplitude ratio OVAD is compared with the first preset vibration amplitude ratio OVAD1 and the second preset vibration amplitude ratio OVAD2. The specific values ​​of the first preset vibration amplitude ratio OVAD1 and the second preset vibration amplitude ratio OVAD2 are not limited in principle. Technicians can set them according to the historical data of different power equipment, which will not be elaborated here. If the vibration amplitude ratio OVAD is less than or equal to the first preset vibration amplitude ratio OVAD1, then the preset composite anomaly index is corrected by the first composite correction coefficient λ1. The corrected preset composite anomaly index MCI1' = MCI1 × λ1, where the first composite correction coefficient λ1 is set to 1.02. If the vibration amplitude ratio OVAD is greater than the first preset vibration amplitude ratio OVAD1 and less than or equal to the second preset vibration amplitude ratio OVAD2, then the preset composite anomaly index is corrected using the second composite correction coefficient λ2. The corrected preset composite anomaly index MCI1' = MCI1 × λ2, where the second composite correction coefficient λ2 is set to 1.05. If the vibration amplitude ratio OVAD is greater than the second preset vibration amplitude ratio OVAD2, then the preset composite anomaly index is corrected by the third composite correction coefficient λ3. The corrected preset composite anomaly index MCI1' = MCI1 × λ3, where the third composite correction coefficient λ3 is set to 1.1.

[0070] In summary, this invention addresses the challenges of distinguishing "pseudo-anomalies" caused by cooler vibration coupling and the difficulty in tracing fault sources due to highly overlapping characteristics of multi-source mechanical faults in existing non-contact condition monitoring technologies for power equipment. It proposes a non-contact monitoring and early warning method for power equipment condition based on multi-source data fusion and hierarchical progressive verification.

[0071] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A non-contact monitoring and early warning method for the status of power equipment based on multi-source data, characterized in that, include, Data acquisition and preprocessing: Real-time acquisition and preprocessing of vibration data of power equipment, oil tank temperature data, narrowband acoustic fingerprint data and acoustic kurtosis data, including vibration data of cooler and oil tank. A composite anomaly index is constructed by determining the vibration component, temperature component, and acoustic penalty term based on the vibration data, the oil tank temperature data, and the narrowband acoustic signature data, respectively, and by determining the composite anomaly index based on the weighted summation of the vibration component, temperature component, and acoustic penalty term. The system monitors and issues early warnings, determining whether a suspected anomaly has occurred in the equipment based on the composite anomaly index. When a suspected anomaly is detected, the system checks the online status of the cooler. When the cooler is online, it determines the jump ratio based on the ratio of the average amplitude of the vibration data in the oil tank within two adjacent jump detection time windows before and after the last start-up of the cooler. Based on the jump ratio, it determines the actual mechanical state of the equipment and the cause of the anomaly. Based on the cause of the anomaly, it corrects the preset composite anomaly index and its construction parameters, or outputs an anomaly warning. The jump detection time window refers to a continuous time period of a preset duration extracted when detecting the average amplitude of the vibration data. The acoustic penalty term refers to the additive correction term introduced during the construction of the composite anomaly index, based on the degree of abnormal increase in sound pressure level in the 195Hz to 215Hz frequency band relative to the historical baseline level in narrowband acoustic data and acoustic kurtosis data. The process of determining the actual mechanical state of power equipment based on the lift ratio includes: If the jump ratio is less than or equal to the preset jump ratio, the suspected abnormality is determined to be a true abnormality. The number of local maxima in the 195Hz-215Hz frequency band of the narrowband voiceprint data within the superimposed detection time window is used to determine whether a suspected fault superposition occurs. If the jump ratio is greater than the preset jump ratio, the actual mechanical state of the power equipment is determined based on the cooler vibration data and the oil tank vibration data in the superimposed detection time window. The superimposed detection time window refers to a continuous time period of preset duration that is taken forward from the moment when a suspected anomaly is determined to occur. The process of determining the actual mechanical state of the power equipment based on the vibration data of the cooler and the vibration data of the oil tank within the superimposed detection time window includes the following steps: Based on the cooler vibration data and the oil tank vibration data in the superimposed detection time window, the cross power spectrum, cooler power spectrum and oil tank power spectrum are determined. The coherence coefficient is determined based on the cross power spectrum, the cooler power spectrum, and the oil tank power spectrum; The actual mechanical state of the power equipment is determined based on the coherence coefficient. If the coherence coefficient is less than or equal to the preset coherence coefficient, the suspected anomaly is determined to be a superposition of a true anomaly and a false anomaly coupled to the cooler. Based on the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic data within the superposition detection time window, it is determined whether a suspected fault superposition occurs. If the coherence coefficient is greater than the preset coherence coefficient, the suspected anomaly is determined to be a false anomaly of the cooler coupling, and the weighting coefficient of the vibration component is adjusted when constructing the composite anomaly index based on the ratio of the preset coherence coefficient to the coherence coefficient.

2. The non-contact monitoring and early warning method for power equipment status based on multi-source data according to claim 1, characterized in that, The process of determining whether a device has a suspected anomaly based on the composite anomaly index includes: If the composite anomaly index is less than or equal to the preset composite anomaly index, the power equipment is determined to be operating normally. If the composite anomaly index is greater than the preset composite anomaly index, it is determined that the power equipment has a suspected anomaly, and the cooler is checked to see if it is online. If the cooler is online, the actual mechanical state of the equipment is determined based on the ratio of the average amplitude of the vibration data in the two adjacent jump detection time windows before and after the most recent start of the cooler. If the cooler is offline, the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic data within the superimposed detection time window is used to determine whether a suspected fault superposition has occurred.

3. The non-contact monitoring and early warning method for power equipment status based on multi-source data according to claim 2, characterized in that, The process of constructing the weighting coefficients of the vibration components based on the ratio of the coherence coefficient to the preset coherence coefficient includes the following steps: The coherence coefficient ratio is determined based on the ratio of the coherence coefficient to the preset coherence coefficient, and the vibration component weighting coefficient is reduced based on the coherence coefficient ratio, wherein the reduction in the vibration component weighting coefficient is proportional to the coherence coefficient ratio.

4. The non-contact monitoring and early warning method for power equipment status based on multi-source data according to claim 3, characterized in that, The process of determining whether a suspected fault superposition occurs based on the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic data within the superposition detection time window includes: The number of maxima is determined based on the number of local maxima in the 195Hz-215Hz frequency band of the narrowband acoustic text data within the superimposed detection time window. If the maximum number is 0, the current anomaly is determined to be a normal anomaly, and the cause of the anomaly is determined based on preset rules; If the maximum value is greater than 0 and less than or equal to the preset maximum value, the speed of the suspected fan is reduced based on the speed test coefficient, and the uncorrected kurtosis and corrected kurtosis are determined based on the acoustic kurtosis data in the last verification cycle before correction and the second bearing fault verification window after correction, respectively. The anomaly type is determined based on the difference between the corrected kurtosis and the uncorrected kurtosis. If the maximum value is greater than the preset maximum value, the current anomaly is determined to be a fault superposition, and an early warning of fan bearing fault and winding loosening fault is output. The suspected fan refers to the cooling fan that generates abnormal sound sources, as determined based on the narrowband acoustic data.

5. The non-contact monitoring and early warning method for power equipment status based on multi-source data according to claim 4, characterized in that, The corrected kurtosis is determined based on the acoustic kurtosis data in the second bearing fault verification window after correction, the uncorrected kurtosis is determined based on the acoustic kurtosis data in the last verification cycle before correction, and the kurtosis difference is determined based on the absolute value of the difference between the corrected kurtosis and the uncorrected kurtosis. The anomaly type is determined based on the kurtosis difference value; If the kurtosis difference is less than or equal to the preset kurtosis difference, the load of the power equipment is corrected based on the ratio of the composite anomaly index to the preset composite anomaly index. If the kurtosis difference value is greater than the preset kurtosis difference value, the current anomaly is determined to be an early failure of the fan bearing, a fault warning is output and the suspected fan speed is restored; The bearing fault verification window refers to a preset continuous time period used to determine the kurtosis of the acoustic kurtosis data.

6. The non-contact monitoring and early warning method for power equipment status based on multi-source data according to claim 5, characterized in that, The process of correcting the load of the power equipment based on the ratio of the composite anomaly index to the preset composite anomaly index includes, The composite anomaly index ratio is determined based on the ratio of the composite anomaly index to the preset composite anomaly index, and the load of the power equipment is reduced based on the composite anomaly index ratio, wherein the reduction in the load of the power equipment is proportional to the composite anomaly index.

7. The non-contact monitoring and early warning method for power equipment status based on multi-source data according to claim 6, characterized in that, The original amplitude ratio is determined based on the ratio of the amplitude of the 200Hz frequency component to the amplitude of the 100Hz frequency component in the tank vibration data in the first winding fault verification window before correction, and the corrected component amplitude ratio is determined based on the ratio of the amplitude of the 200Hz frequency component to the amplitude of the 100Hz frequency component in the tank vibration data in the second winding fault verification window. The absolute value of the amplitude difference is determined based on the absolute value of the difference between the original amplitude ratio and the amplitude ratio of the modified component; The cause of the anomaly is determined based on the absolute value of the amplitude difference. If the absolute value of the amplitude difference is less than or equal to the set absolute value of the preset amplitude difference, it is determined that the preset composite anomaly index is set unreasonably, the load of the power equipment is restored, and the 100Hz vibration amplitude is determined based on the oil tank vibration data in the second winding fault verification window after restoration. The preset composite anomaly index is corrected based on the ratio of the 100Hz vibration amplitude to the set preset 100Hz vibration amplitude. If the absolute value of the amplitude difference is greater than the preset absolute value of the amplitude difference, the current anomaly is determined to be a winding loosening fault, a fault warning is output, and the load of the power equipment is restored. The winding fault verification window refers to a preset continuous time period used to determine the corrected 100Hz vibration amplitude.

8. The non-contact monitoring and early warning method for power equipment status based on multi-source data according to claim 7, characterized in that, The process of correcting the preset composite anomaly index based on the ratio of the 100Hz vibration amplitude to the preset 100Hz vibration amplitude includes the following steps: The vibration amplitude of 100Hz is determined based on the oil tank vibration data in the second winding fault verification window after recovery. The vibration amplitude ratio is determined based on the ratio of the 100Hz vibration amplitude to the preset 100Hz vibration amplitude. The preset composite anomaly index is increased based on the vibration amplitude ratio, and the increase of the preset composite anomaly index is proportional to the vibration amplitude ratio.

Images