A key feature transformer health condition detection system

Abstract

The application discloses a kind of key feature transformer health state detection systems, it is related to power equipment monitoring technical field, including: data acquisition module, data processing module, confidence calculation module, fault determination module and result output module;Data acquisition module synchronously acquires transformer oil tank surface vibration signal and neutral point direct current data;Data processing module calculates the phase difference of vibration signal and the coherence coefficient of vibration signal and direct current data;Confidence calculation module calculates mechanical fault confidence;Fault determination module judges whether mechanical fault confidence exceeds preset threshold, exceeds and then carries out winding loosening positioning processing, generates loosening coordinate;Otherwise, the multi-band resampling of vibration signal is carried out and the coherence coefficient data is recalculated;Result output module outputs loosening coordinate or interference warning signal;The application can improve the accuracy of transformer health state detection, ensure the reliability of fault positioning.

Description

Technical Field

[0001] This invention relates to the field of power equipment monitoring technology, specifically to a transformer health status detection system with key characteristic quantities. Background Technology

[0002] As a key piece of equipment in the power system, the health of transformers directly affects the stability and reliability of the power system. Transformers primarily transmit electrical energy from one circuit to another through electromagnetic induction and are widely used in various power equipment, such as power transmission, power substations, and power generation. With the continuous commissioning of power equipment, transformers, due to prolonged high-load operation, face varying degrees of aging, wear, and the risk of failure. Therefore, real-time monitoring of transformer operating status and timely detection of potential mechanical faults are essential means to ensure the safe operation of the power system.

[0003] Mechanical loosening of windings is a common fault in power transformer operation, especially during load fluctuations, where the loosening point can trigger vibrations at specific frequencies. Traditional methods use phase analysis of vibration signals to locate the loosening point, calculating phase difference abrupt changes using array sensors. However, near high-voltage direct current (HVDC) transmission systems, DC bias at the transformer neutral point is prevalent, generating 100Hz frequency vibrations highly similar to those of winding loosening. This spectral confusion leads to phase analysis mistakenly identifying the bias vibration as a mechanical fault, thus interfering with the accuracy of loosening location. In existing technologies, relying solely on vibration signal analysis is insufficient to effectively distinguish between DC bias interference and genuine mechanical faults, resulting in a high misjudgment rate.

[0004] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a transformer health status detection system with key characteristic quantities.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows:

[0007] In a first aspect, the present invention discloses a transformer health status detection system for key characteristic quantities, comprising:

[0008] The data acquisition module is used to synchronously acquire vibration signal data of the transformer tank surface and DC current data of the neutral point;

[0009] The data processing module is used to calculate the phase difference data between the sensors based on the vibration signal data, and to calculate the amplitude of the 100Hz component in the vibration signal and the coherence coefficient data of the DC current based on the vibration signal data and the DC current data.

[0010] The confidence calculation module is used to calculate the confidence level of mechanical faults based on the phase difference data and coherence coefficient data.

[0011] The fault determination module is used to determine whether the confidence level of the mechanical fault is greater than a preset confidence threshold: if yes, it performs winding loosening positioning processing based on the phase difference data and generates loosening coordinates; otherwise, it performs multi-band resampling processing on the vibration signal data and recalculates the coherence coefficient data.

[0012] The result output module is used to output the loose coordinates when the winding loosening positioning process is performed; and to output an interference warning signal when the positioning process is not triggered.

[0013] Secondly, this invention discloses a method for detecting the health status of a transformer based on key characteristic quantities, comprising the following steps:

[0014] Simultaneously acquire vibration signal data of the transformer tank surface and DC current data of the neutral point;

[0015] Calculate the phase difference data between the sensors based on the vibration signal data;

[0016] Based on the vibration signal data and DC current data, calculate the amplitude of the 100Hz component in the vibration signal and the coherence coefficient of the DC current.

[0017] Calculate the confidence level of mechanical faults based on the phase difference data and coherence coefficient data;

[0018] Determine whether the confidence level of the mechanical fault is greater than a preset confidence threshold;

[0019] If so, the winding loosening positioning process is performed based on the phase difference data to generate loosening coordinates;

[0020] Otherwise, the vibration signal data is resampled in multiple frequency bands, and the coherence coefficient data is recalculated.

[0021] When the winding loosening positioning process is executed, the loosening coordinates are output; when the positioning process is not triggered, an interference warning signal is output.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0023] 1. By introducing neutral point DC current data and calculating the coherence coefficient between vibration signal and DC current, this invention can effectively distinguish between 100Hz frequency vibration caused by DC bias and vibration caused by mechanical fault, avoiding misjudging DC bias vibration as mechanical fault and significantly reducing the misjudgment rate of fault diagnosis.

[0024] 2. By constructing a vibration wave propagation vector field based on phase difference data and scanning the abrupt change region of phase difference, the accurate location of winding loosening faults can be achieved. This invention uses a multi-point array sensor, which makes fault location more accurate and can effectively avoid erroneous location caused by local noise.

[0025] 3. By introducing a mechanism for vibration energy concentration and dynamic adjustment of confidence threshold, this invention can adaptively adjust the fault judgment criteria according to different operating environments. When the vibration energy distribution is not concentrated, the system will automatically improve the sensitivity of fault judgment, effectively avoid misjudgment caused by local vibration energy dispersion, and ensure high accuracy under complex working conditions. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is an overall block diagram of the system according to Embodiment 1 of the present invention;

[0028] Figure 2 This is a flowchart illustrating the overall execution process of the system according to Embodiment 1 of the present invention.

[0029] Figure 3 This is an overall block diagram of the method in Embodiment 2 of the present invention. Detailed Implementation

[0030] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] Application Overview: In power transformer operation monitoring, the vibration signal caused by mechanical loosening of the windings overlaps with the 100Hz vibration generated by DC bias magnetization, making traditional fault location methods based on abrupt phase difference ineffective in distinguishing the fault source. When the transformer is in a high-voltage DC transmission environment, fluctuations in the neutral point DC current excite the magnetostrictive effect of the core, generating vibration components highly similar to those of mechanical loosening, causing coupling interference between phase difference data and coherence coefficient data. This interference directly leads to deviations in the calculation of mechanical fault confidence, preventing the system from accurately identifying the true fault type during the fault determination phase.

[0032] For example, in the operating scenario of a supporting transformer in a ±800 kV UHV converter station, the DC current at the neutral point fluctuates randomly within the range of 0.5 - 5 A, generating a 100 Hz vibration component with an amplitude of 0.8 m / s² on the surface of the oil tank through the magnetostrictive effect. At this time, for the vibration signal collected by the six-channel acceleration sensor array arranged on the surface of the oil tank, the spatial distribution gradient of the phase difference obtained by cross-correlation calculation reaches 12° / m, completely overlapping with the phase difference characteristics of typical winding looseness faults. At the same time, the Pearson correlation coefficient between the amplitude of the 100 Hz component of the vibration signal and the DC current fluctuates within the range of 0.65 - 0.85, and effective discrimination cannot be achieved through a fixed threshold. The system will misjudge as a high-confidence fault state during the confidence calculation stage, triggering an incorrect winding looseness localization process.

[0033] When facing the above problems, this application first analyzes the overlapping mechanism of DC bias and mechanical looseness vibration characteristics, and finds that the traditional method relying solely on the sudden change of phase difference for fault judgment has inherent defects. In response, this application considers introducing the DC current at the neutral point as an auxiliary criterion, and distinguishing the interference source through the correlation analysis between the vibration signal and the DC current. Further research finds that there is a strong correlation between the amplitude of the 100 Hz vibration component generated by DC bias and the DC current at the neutral point, while the vibration amplitude caused by mechanical looseness has nothing to do with the DC current. Based on this, this application proposes to jointly model the phase difference mutation intensity and the vibration-current coherence coefficient, and construct a dynamic confidence calculation mechanism. When the confidence exceeds the threshold, positioning is executed, otherwise a resampling process is triggered to exclude instantaneous interference, so as to achieve fault feature decoupling in the decision logic.

[0034] Embodiment 1:

[0035] As Figure 1-2 shown, a key feature quantity transformer health state detection system includes the following steps: a data acquisition module for synchronously acquiring the vibration signal data on the surface of the transformer oil tank and the DC current data at the neutral point; a data processing module for calculating the phase difference data between sensors according to the vibration signal data, and calculating the coherence coefficient data between the amplitude of the 100 Hz component in the vibration signal and the DC current according to the vibration signal data and the DC current data; a confidence calculation module for calculating the mechanical fault confidence according to the phase difference data and the coherence coefficient data; a fault determination module for judging whether the mechanical fault confidence is greater than a preset confidence threshold: if yes, performing winding looseness localization processing according to the phase difference data to generate a looseness coordinate; otherwise, performing multi-band resampling processing on the vibration signal data and recalculating the coherence coefficient data; a result output module for outputting the looseness coordinate when performing winding looseness localization processing; and outputting an interference warning signal when no localization processing is triggered.

[0036] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0037] The data acquisition module consists of multiple accelerometers and one DC current sensor. The accelerometers are arranged in an array on the surface of the transformer tank, with a sampling frequency of 1 kHz. The DC current sensor is installed on the transformer neutral grounding wire, with a sampling frequency of 10 kHz. Both types of sensors are synchronized using the same clock source.

[0038] The data processing module first performs cross-correlation analysis on the vibration signal to calculate the phase difference between each pair of sensors. It then statistically analyzes the spatial distribution gradient of the phase difference to generate phase difference abrupt change intensity data. Simultaneously, it extracts the time-varying sequence of the 100Hz component amplitude of the vibration signal and calculates the Pearson correlation coefficient with the DC current data to obtain the coherence coefficient data.

[0039] The confidence calculation module uses the formula: Mechanical fault confidence = (1 - coherence coefficient) × phase difference abrupt change intensity, where the phase difference abrupt change intensity is taken as the maximum value of the spatial distribution gradient.

[0040] The fault determination module has a preset confidence threshold of 0.6. When the mechanical fault confidence level is greater than 0.6, a winding loosening location process is triggered. The location process is achieved by constructing a vibration wave propagation vector field and scanning the phase difference abrupt change region. When the confidence level is less than 0.6, the vibration signal is resampled in multiple frequency bands from 50Hz to 500Hz, and the coherence coefficient and mechanical fault confidence level are recalculated.

[0041] The output module outputs loose coordinates, including coordinate values ​​and confidence levels, during the positioning process. When positioning is not triggered, it outputs an interference warning signal, including the interference type and intensity estimate.

[0042] Through the above-described scheme, this application achieves an effective distinction between transformer mechanical faults and DC bias interference. By introducing neutral point DC current data and establishing a vibration-current correlation analysis model, the limitations of traditional methods that rely solely on phase difference abrupt changes are overcome. The dynamic confidence calculation mechanism improves the accuracy of fault diagnosis, and multi-band resampling processing enhances the system's anti-interference capability. This scheme can accurately identify transformer winding loosening faults in high-voltage DC transmission environments, reducing false positive and false negative rates and improving the reliability of transformer health status monitoring.

[0043] This application further proposes the following method for synchronously acquiring vibration signal data of the transformer tank surface and DC current data of the neutral point: arranging several accelerometers on the surface of the transformer tank to form a monitoring array and acquiring vibration waveform data; installing a DC current sensor on the neutral point grounding wire of the transformer and acquiring current waveform data; and using a unified clock source to timestamp the vibration waveform data and current waveform data to generate time-aligned vibration signal data and DC current data.

[0044] Among them, the accelerometer is arranged in an array to cover the surface of the fuel tank, capturing vibration waveforms at different locations through spatial distribution; the DC current sensor is directly installed on the neutral point grounding wire to avoid interference introduced by the signal transmission path; a unified clock source provides a synchronous time reference for all sensors, ensuring that the timestamp marking accuracy of vibration and current signals reaches the microsecond level, and eliminating phase shift caused by asynchronous sampling.

[0045] Specifically, the accelerometers in the monitoring array are evenly arranged at preset intervals, covering key vibration-sensitive areas on the tank surface. The spatial distribution characteristics of the vibration waveform are acquired through multi-channel synchronous acquisition. The DC current sensor uses a non-invasive installation, measuring the DC current waveform on the neutral grounding wire through the Hall effect or fluxgate principle. A unified clock source further synchronizes the sampling clocks of all sensors via hardware trigger signals, ensuring strict alignment of the timestamps of the vibration waveform data and the current waveform data. The resulting vibration signal data and DC current data maintain precise correspondence in the time dimension, providing highly consistent input for subsequent phase difference calculations and coherence coefficient analysis, effectively reducing the risk of misjudgment due to time asynchrony. For example, when the 100Hz vibration component caused by DC bias overlaps with mechanical loosening vibration in the time domain, the time-aligned data can accurately distinguish the correlation between the two, avoiding interference from spectral confusion in confidence calculations.

[0046] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0047] An acceleration sensor array is formed by arranging acceleration sensors on the surface of the transformer tank. Specifically, 16 acceleration sensors can be evenly arranged around the tank, each with a sensitivity of 100mV / g and a frequency response range of 0.5Hz-10kHz. These sensors are connected to a data acquisition device via signal lines to collect vibration waveform data.

[0048] Install a DC current sensor on the neutral point grounding wire of the transformer. For example, a Hall effect current sensor with a range of 0-100A and an accuracy of ±0.5%FS can be used. This sensor is also connected to a data acquisition device to collect current waveform data.

[0049] A GPS timing module provides a unified clock source for the data acquisition equipment. The equipment synchronously acquires vibration and current signals at a sampling rate of 10kHz, adding a timestamp accurate to the microsecond level to each data point. This generates time-aligned vibration signal data and DC current data.

[0050] Through the above technical solution, this application achieves high-precision synchronous acquisition of vibration signals from the transformer tank surface and DC current at the neutral point. This allows for accurate capture of the correlation between vibration and current, providing a reliable data foundation for subsequent analysis. Furthermore, the use of a high sampling rate and precise timestamps ensures data temporal consistency, contributing to improved fault location accuracy.

[0051] This application further proposes a phase difference data calculation process including: performing cross-correlation calculation on vibration signal data to obtain the phase difference between each pair of sensors; and statistically analyzing the spatial distribution gradient of all phase differences to generate phase difference abrupt change intensity data.

[0052] Among them, cross-correlation calculation determines the phase offset of vibration signals from two sensors through time-domain or frequency-domain analysis methods, such as using fast Fourier transform to extract the phase angle of the signal and then calculating the difference; spatial distribution gradient statistics identifies abrupt change regions by calculating the rate of change of phase difference between adjacent sensors, such as using a two-dimensional gradient operator to perform convolution operation on the phase difference distribution matrix of the sensor array and extracting the maximum gradient amplitude as the abrupt change intensity.

[0053] Specifically, after cross-correlation calculation of the vibration signal data, the phase difference between each pair of sensors is obtained, forming a phase difference distribution matrix. Subsequently, spatial gradient statistics are performed on the matrix to calculate the rate of change of the phase difference between adjacent sensors at each location, generating a gradient distribution map. By scanning the gradient distribution map, the maximum gradient value is extracted as the phase difference abrupt change intensity data. This process can effectively distinguish between phase abrupt changes caused by winding loosening and local fluctuations caused by noise. For example, when the maximum gradient value exceeds a preset threshold, a phase abrupt change region is determined to exist, thereby improving the accuracy of loosening location.

[0054] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0055] The calculation of phase difference data involves two steps. First, a cross-correlation calculation is performed on the vibration signal data to obtain the phase difference between each pair of sensors. Specifically, the vibration signals of two sensors are selected as inputs, and the phase difference between the two signals is calculated using a cross-correlation function. This step is repeated for all sensor pairs to obtain a complete phase difference matrix.

[0056] Secondly, the spatial distribution gradients of all phase differences are statistically analyzed to generate phase difference abrupt change intensity data. In practice, the central difference method can be used to calculate the phase difference gradient between adjacent sensors. Then, all gradient values ​​are statistically analyzed, such as calculating the mean and standard deviation of the gradients, or using histograms to statistically analyze the gradient distribution. Finally, a threshold is determined based on the statistical results, and gradients exceeding this threshold are marked as abrupt change points. The number and intensity of these abrupt change points are then calculated to form phase difference abrupt change intensity data.

[0057] Through the above technical solution, this application can effectively detect the spatial discontinuity of vibration signals, thereby identifying the location of potential mechanical faults. Due to the use of cross-correlation calculation and spatial gradient analysis, this method has strong anti-interference capabilities against noise and can accurately capture the abrupt changes in phase difference. Furthermore, by generating phase difference abrupt change intensity data, it provides important basis for subsequent fault location and confidence calculation, improving the detection accuracy and reliability of the entire system.

[0058] This application further proposes a process for calculating coherence coefficient data, which includes: extracting the amplitude sequence of the 100Hz component in the vibration signal as a function of time; and calculating the Pearson correlation coefficient between the amplitude sequence and the DC current data as the coherence coefficient.

[0059] The extraction of the 100Hz component's amplitude variation over time is achieved using a bandpass filter or Fast Fourier Transform. Amplitude sampling points are acquired within a preset time window, forming a discrete data sequence in the time dimension. The Pearson correlation coefficient is calculated using the covariance to standard deviation ratio formula. The amplitude sequence and DC current data are standardized to eliminate the influence of dimensional differences. The time alignment accuracy between the amplitude sequence and current data is controlled at the millisecond level to ensure a one-to-one correspondence between data points. As a preferred implementation, the sampling interval of the amplitude sequence is set to an integer multiple of the DC current fluctuation period, for example, 30 consecutive data points are collected at 10-millisecond intervals.

[0060] Specifically, after the vibration signal undergoes spectral analysis, a sliding time window mechanism is used to extract the amplitude data of the 100Hz component, with a window length of 1 second and a step size of 0.1 seconds. Within each window, the average amplitude is calculated through integration to form time-series data. The DC current data, after low-pass filtering, is then mean-processed within the same time window to generate a corresponding sequence. The two sequences are input to the Pearson correlation coefficient calculation module, and the linear correlation index is obtained through covariance matrix operations. When the correlation coefficient approaches 1, it indicates that the vibration and current changes are highly synchronized, which can be identified as DC bias interference; when the correlation coefficient is below 0.5, it indicates that the vibration source is independent, supporting mechanical fault diagnosis. This calculation process effectively improves the accuracy of vibration source identification by eliminating phase shift errors in the time dimension.

[0061] Through the above technical solution, this application can effectively distinguish between vibrations caused by transformer winding loosening faults and vibrations caused by DC bias at the neutral point. Since the vibrations caused by winding loosening faults are unrelated to the neutral point DC current, while the vibrations caused by DC bias are highly correlated with the neutral point DC current, the source of the vibration can be determined by calculating the coherence coefficient between the 100Hz component of the vibration signal and the DC current. This method improves the accuracy of transformer mechanical fault diagnosis, avoids misjudging DC bias vibrations as mechanical faults, and provides a reliable basis for transformer health status assessment.

[0062] In some of the solutions described above in this application, the calculation method for mechanical fault confidence fails to effectively distinguish between vibration interference caused by DC bias and actual winding mechanical loosening faults, resulting in a high misjudgment rate. Traditional methods rely solely on a single indicator such as phase difference or coherence coefficient for judgment, which cannot accurately identify the fault type when the two fault characteristics overlap, affecting the reliability of the detection system.

[0063] This application further proposes that the confidence level of mechanical faults satisfies: Confidence level of mechanical faults = (1 - coherence coefficient) × phase difference abrupt change intensity; where the phase difference abrupt change intensity is the maximum value of the phase difference spatial distribution gradient.

[0064] The coherence coefficient reflects the correlation between the amplitude of the 100Hz component in the vibration signal and the neutral point DC current data. When DC bias interference is present, this coefficient approaches 1, causing (1 - coherence coefficient) to approach 0, thus reducing the confidence level of mechanical faults. The phase difference abrupt change intensity is characterized by the severity of the phase anomaly abrupt change in the vibration wave propagation path by calculating the maximum value of the phase difference spatial distribution gradient in the sensor array. The multiplication logic of these two parameters combines the suppression mechanism of DC bias interference with the enhancement mechanism of mechanical fault characteristics to form a composite criterion.

[0065] Specifically, the calculation process first extracts the phase difference between each sensor from the vibration signal data, constructs a spatial distribution map of the phase difference, and calculates its gradient field, determining the maximum gradient value as the phase difference abrupt change intensity. Simultaneously, the Pearson correlation coefficient is calculated between the 100Hz component amplitude sequence of the vibration signal and the DC current data to obtain the coherence coefficient. By multiplying (1 - coherence coefficient) by the phase difference abrupt change intensity, when DC bias interference dominates, the coherence coefficient approaches 1, the calculation result approaches 0, and the system determines it as interference; when winding loosening actually exists, the coherence coefficient is low while the phase difference abrupt change intensity is high, the calculation result exceeds the threshold, triggering location processing. For example, when the coherence coefficient is 0.2 and the phase difference abrupt change intensity is 8, the confidence level calculation result is 6.4, higher than the typical preset threshold of 5, and the system will perform winding loosening location. This calculation method transforms the physical meaning of two types of characteristic quantities into quantifiable decision indicators through mathematical correlation, effectively improving the anti-interference capability of fault diagnosis.

[0066] Through the above technical solution, this application can accurately calculate the confidence level of mechanical faults and effectively reduce the impact of neutral point magnetic interference on winding loosening and positioning. This method fully utilizes the correlation between vibration signals and DC current, as well as the spatial distribution characteristics of vibration phase differences, improving the accuracy and reliability of fault diagnosis. Furthermore, this calculation method is simple and intuitive, easy to implement in engineering, and can achieve real-time monitoring and early warning of transformer health status.

[0067] This application further proposes a process for performing winding loosening location processing, which includes: constructing a vibration wave propagation vector field based on phase difference data; scanning the abrupt change region of phase difference in the vector field; and generating the loosening coordinates of the region if there is a region where the phase difference change exceeds a preset angle threshold.

[0068] The vibration wave propagation vector field maps the phase difference data of each sensor node into a spatial vector, forming a spatial distribution model of the vibration propagation direction. The scanning of abrupt phase difference regions employs a gradient detection algorithm, with a preset angle threshold set based on historical fault data statistics. For example, regions with a phase difference exceeding 30 degrees between adjacent sensors are considered valid abrupt changes. The loosening coordinates are generated using a spatial interpolation algorithm to determine the geometric center of the abrupt change region.

[0069] Specifically, the construction of the vibration wave propagation vector field converts the phase difference of each sensor node into the vibration wave arrival time difference, and calculates the vibration wave propagation direction vector by combining it with the spatial coordinates of the sensor array. During the scanning process, the vector field is divided into regional grids, and the phase difference gradient value is calculated grid by grid. When the phase difference gradient in a certain grid exceeds a preset angle threshold, it is identified as a potential loosening area. For example, in a 5×5 grid, if the direction change of adjacent vectors in a certain grid exceeds 30 degrees, coordinate generation is triggered. This process effectively distinguishes between mechanical loosening and phase disturbances caused by DC bias magnetization by eliminating interference signals with low gradient changes and retaining only areas with significant phase abrupt changes. After coordinate generation, the system maps the loosening location to the transformer's 3D model and outputs the corresponding tank surface coordinate values, providing accurate positioning information for maintenance.

[0070] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0071] When performing winding loosening location processing, the vibration wave propagation vector field is first constructed based on the phase difference data. Specifically, the surface of the transformer tank is divided into a grid, with each grid point representing a position coordinate. For each grid point, its phase difference with surrounding sensors is calculated, forming a direction vector. These vectors together constitute the vibration wave propagation vector field of the entire tank surface.

[0072] Next, regions of abrupt phase difference changes in the scan vector field are identified. Specifically, a scan window is defined and moved across the entire vector field. For each scan position, the angular difference between adjacent vectors within the window is calculated. If a region exists where the phase difference change exceeds a preset angular threshold, it is considered that the region may have a loose winding.

[0073] For example, a preset angle threshold can be set to 30 degrees. When the angle difference between adjacent vectors within the scanning window exceeds 30 degrees, the region is marked as a potential loosening region. Furthermore, for each potential loosening region, the coordinates of its center point are calculated and output as loosening coordinates.

[0074] Therefore, by analyzing the spatial characteristics of vibration wave propagation, the location of the loose winding can be accurately located, providing precise spatial information for subsequent maintenance.

[0075] Through the above technical solution, this application can effectively identify and locate loose faults in transformer windings. By constructing a vibration wave propagation vector field, the spatial characteristics of vibration propagation can be intuitively displayed. Furthermore, by scanning the phase difference abrupt change region in the vector field, local vibration anomalies caused by winding looseness can be accurately captured. This method can not only detect the existence of loose faults but also accurately locate the specific location of the fault, providing important spatial information guidance for transformer maintenance and repair. Compared with traditional single-point measurement methods, this solution utilizes the spatial distribution advantages of multi-point array sensors, improving the accuracy and reliability of fault location. In addition, by setting a preset angle threshold, minute phase fluctuations can be effectively filtered out, reducing the possibility of false alarms. In summary, this fault location method based on vector field analysis provides an efficient and accurate technical means for transformer health status monitoring.

[0076] This application further proposes that before outputting the loosening coordinate data, the following steps are included: calculating the vibration energy concentration of the area corresponding to the loosening coordinate; determining whether the vibration energy concentration is greater than a preset energy threshold: if yes, outputting the loosening coordinate data and a high confidence flag; otherwise, outputting a flag to be re-inspected and updating the confidence threshold; the process of calculating the vibration energy concentration of the area corresponding to the loosening coordinate includes: determining the target position corresponding to the loosening coordinate, and selecting all acceleration sensors within a preset radius centered on that position as the target sensor group; extracting the vibration signal data of the target sensor group, calculating the effective value of the vibration signal of each sensor within a preset time window, and generating energy value data of each sensor; calculating the variance of the energy value data of all target sensors as the vibration energy concentration.

[0077] The selection of the target sensor group is limited to a spatial area by a preset radius range, which is set according to the size of the transformer tank and the vibration propagation attenuation characteristics, for example, a value of 0.5 meters to 1.5 meters; the effective value is calculated using the root mean square algorithm, and the preset time window length is set to the integer period of the 100Hz component in the vibration signal, for example, 0.1 seconds; the variance calculation is used to quantify the dispersion of energy distribution, and the lower the variance value, the more concentrated the energy is at the target location.

[0078] Specifically, once a region of abrupt phase difference change is identified as a loose coordinate, it is necessary to verify whether this region is accompanied by concentrated vibration energy. After the effective value of the vibration signal from the target sensor group is extracted, if the variance is lower than a preset energy threshold, it indicates concentrated energy distribution, confirming the reliability of the loose coordinate and outputting a high-confidence flag. If the variance is higher than the threshold, it is judged as interference or local noise, triggering a re-inspection flag and dynamically adjusting the confidence threshold. For example, when the preset energy threshold is 0.8, if the calculated variance is 0.5, a high-confidence flag is output; if the variance is 1.2, the confidence threshold is updated to reduce the probability of subsequent misjudgments. Through dual verification of energy concentration and variance, the system effectively distinguishes between real mechanical faults and noise interference, improving positioning accuracy.

[0079] Through the above technical solution, this application can effectively verify the reliability of loosening location results and avoid misjudgments caused by the uneven distribution of vibration energy. By introducing a vibration energy concentration index, it is possible to distinguish between genuine mechanical loosening and vibrations caused by environmental disturbances. Simultaneously, the mechanism for dynamically adjusting the confidence threshold can adapt to different operating environments and improve the robustness of the system. Therefore, this solution significantly improves the accuracy and reliability of transformer winding loosening fault diagnosis, providing a strong guarantee for the safe and stable operation of the power system.

[0080] This application further proposes a process for updating the confidence threshold, which includes: new confidence threshold = original confidence threshold × (1 + learning rate × (1 - vibration energy concentration / preset energy threshold)), where the learning rate is a preset correction coefficient.

[0081] The original confidence threshold serves as the baseline value for the current system to determine mechanical faults. The learning rate, as a preset correction coefficient, controls the threshold adjustment range. Vibration energy concentration is a quantified value of the dispersion of vibration energy distribution in the area corresponding to the loosening coordinate. The preset energy threshold is a pre-set standard for acceptable energy concentration. The calculation of the new confidence threshold establishes a linear relationship between the degree of energy concentration deviation and the threshold adjustment amount by introducing the normalized difference of (1 - vibration energy concentration / preset energy threshold). The learning rate, as a coefficient, controls the adjustment range, and its value range is determined through experimental data, with a typical value of 0.1-0.3.

[0082] Specifically, when the vibration energy concentration is lower than the preset energy threshold, (1 - vibration energy concentration / preset energy threshold) generates a positive value, and the new confidence threshold is increased proportionally. The increase is determined by the product of the learning rate and the energy deviation, and the learning rate, as a preset parameter, can be configured according to the operating environment of different transformers. For example, when the vibration energy concentration is 80% of the preset energy threshold, (1 - 0.8) = 0.2 is calculated. If the learning rate is set to 0.2, the increase in the new confidence threshold is 0.2 × 0.2 = 4%. This calculation process dynamically adjusts the confidence threshold, enabling the system to automatically raise the fault judgment standard when the energy concentration is insufficient, avoiding mislocation problems caused by local vibration energy dispersion. At the same time, the introduction of the learning rate parameter enables controllability of the adjustment range, preventing threshold abrupt changes caused by single energy detection anomalies and ensuring the stability of the system's judgment.

[0083] Through the above technical solution, this application achieves adaptive adjustment of the fault judgment threshold based on the vibration energy distribution characteristics, effectively solving the problem of high false judgment rate of mechanical faults under DC bias interference. When the vibration energy concentration is insufficient, the detection sensitivity is improved by lowering the confidence threshold to avoid missing potential winding loosening faults; when the energy concentration is too high, the false judgment is suppressed by raising the threshold, thereby maintaining the accuracy of fault detection under complex working conditions.

[0084] This application further proposes a multi-band resampling process, which includes: extending the sampling frequency band of vibration signal data to 50Hz-500Hz; and recalculating the coherence coefficient data and mechanical fault confidence based on the resampled data.

[0085] As a preferred embodiment, the solution of this application is implemented as follows: When the confidence level of the detected mechanical fault does not exceed a preset confidence threshold, the vibration signal data is input to a digital signal processing unit for multi-band resampling processing. Specifically, firstly, the sampling frequency band of the original vibration signal is extended from 0-100Hz to 50Hz-500Hz using an FIR bandpass filter, covering the fundamental and higher harmonic components. Subsequently, a variable rate sampling algorithm is used to resample the signal within the extended frequency band, generating time-domain waveform data containing multi-band information. Based on the resampled data, the amplitude sequence of the 100Hz component in the vibration signal is re-extracted, and a sliding window Pearson correlation coefficient is calculated with the DC current data to update the coherence coefficient data. Finally, the updated coherence coefficient is substituted into the mechanical fault confidence calculation model to generate a new confidence assessment result.

[0086] Through the above technical solution, this application effectively solves the technical problem that it is difficult to distinguish between DC bias interference and real mechanical vibration in a single frequency band. By expanding the sampling frequency band and recalculating the coherence coefficient, the characteristic harmonic components of real mechanical faults in the high frequency band can be separated, reducing the coupling effect of 100Hz vibration interference caused by neutral point DC bias on confidence calculation, thereby improving the anti-interference ability and accuracy of winding loosening fault identification.

[0087] Example 2:

[0088] like Figure 3 As shown, a method for detecting the health status of a transformer based on key characteristic quantities includes the following steps:

[0089] Simultaneously acquire vibration signal data of the transformer tank surface and DC current data of the neutral point;

[0090] Calculate the phase difference data between the sensors based on the vibration signal data;

[0091] Based on the vibration signal data and DC current data, calculate the amplitude of the 100Hz component in the vibration signal and the coherence coefficient of the DC current.

[0092] Calculate the confidence level of mechanical faults based on the phase difference data and coherence coefficient data;

[0093] Determine whether the confidence level of the mechanical fault is greater than a preset confidence threshold;

[0094] If so, the winding loosening positioning process is performed based on the phase difference data to generate loosening coordinates;

[0095] Otherwise, the vibration signal data is resampled in multiple frequency bands, and the coherence coefficient data is recalculated.

[0096] When the winding loosening positioning process is executed, the loosening coordinates are output; when the positioning process is not triggered, an interference warning signal is output.

[0097] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

[0098] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0099] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A transformer health status detection system with key characteristic quantities, characterized in that, include: The data acquisition module is used to synchronously acquire vibration signal data of the transformer tank surface and DC current data of the neutral point; The synchronous acquisition includes: arranging several accelerometers on the surface of the transformer tank to form a monitoring array and acquiring vibration waveform data; installing a DC current sensor on the neutral point grounding wire of the transformer and acquiring current waveform data; and using a unified clock source to timestamp the vibration waveform data and current waveform data to generate time-aligned vibration signal data and DC current data. The data processing module is used to calculate the phase difference data between each pair of acceleration sensors based on the vibration signal data, and to calculate the amplitude of the 100Hz component in the vibration signal and the coherence coefficient data of the DC current based on the vibration signal data and the DC current data. The calculation process of the phase difference data includes: performing cross-correlation calculation on the vibration signal data to obtain the phase difference between each pair of acceleration sensors; statistically analyzing the spatial distribution gradient of all phase differences to generate phase difference abrupt change intensity data; wherein, the phase difference abrupt change intensity is the maximum value of the spatial distribution gradient of the phase difference; The confidence calculation module is used to calculate the confidence level of mechanical faults based on the phase difference data and coherence coefficient data; the confidence level of mechanical faults = (1 - coherence coefficient) × phase difference abrupt change intensity; The fault determination module is used to determine whether the confidence level of the mechanical fault is greater than a preset confidence threshold: if yes, it performs winding loosening positioning processing based on the phase difference data and generates loosening coordinates; otherwise, it performs multi-band resampling processing on the vibration signal data and recalculates the coherence coefficient data. The result output module is used to output the loose coordinates when the winding loosening positioning process is performed; and to output an interference warning signal when the positioning process is not triggered.

2. The key characteristic quantity transformer health status detection system according to claim 1, characterized in that: The calculation process for the coherence coefficient data includes: Extract the time sequence of the amplitude of the 100Hz component in the vibration signal; Calculate the Pearson correlation coefficient between the amplitude sequence and the DC current data, and use it as the coherence coefficient.

3. The key characteristic quantity transformer health status detection system according to claim 1, characterized in that: The process of performing the winding loosening positioning treatment includes: Construct a vibration wave propagation vector field based on phase difference data; If there is a region in the scanning vector field where the phase difference changes abruptly, and the phase difference change exceeds a preset angle threshold, then the loose coordinates of that region are generated.

4. The key characteristic quantity transformer health status detection system according to claim 3, characterized in that: Before outputting the loosening coordinate data, the following is also included: Calculate the vibration energy concentration in the region corresponding to the loosening coordinates; Determine if the vibration energy concentration is greater than the preset energy threshold: if yes, output the loosening coordinate data and high confidence flag; otherwise, output the pending re-inspection flag and update the confidence threshold. The process of calculating the vibration energy concentration in the region corresponding to the loosening coordinates includes: Determine the target location corresponding to the loosening coordinates, and select all acceleration sensors within a preset radius centered on that location as the target sensor group; Extract vibration signal data from the target sensor group, calculate the effective value of each sensor's vibration signal within a preset time window, and generate energy value data for each sensor. The variance of the energy values ​​from all target sensors is calculated as the vibration energy concentration.

5. A key characteristic quantity transformer health status detection system according to claim 4, characterized in that: The process of updating the confidence threshold includes: New confidence threshold = original confidence threshold × (1 + learning rate × (1 - vibration energy concentration / preset energy threshold)). The learning rate is a preset correction coefficient.

6. The key characteristic quantity transformer health status detection system according to claim 1, characterized in that: The multi-band resampling process includes: Extend the sampling frequency band of vibration signal data to 50Hz-500Hz; The coherence coefficient data and mechanical fault confidence were recalculated based on the resampled data.

7. A method for detecting the health status of a transformer based on the system described in any one of claims 1-6, comprising the following steps: Simultaneously acquire vibration signal data of the transformer tank surface and DC current data of the neutral point; Calculate the phase difference data between the sensors based on the vibration signal data; Based on the vibration signal data and DC current data, calculate the amplitude of the 100Hz component in the vibration signal and the coherence coefficient of the DC current. Calculate the confidence level of mechanical faults based on the phase difference data and coherence coefficient data; Determine whether the confidence level of the mechanical fault is greater than a preset confidence threshold; If so, the winding loosening positioning process is performed based on the phase difference data to generate loosening coordinates; Otherwise, the vibration signal data is resampled in multiple frequency bands, and the coherence coefficient data is recalculated. When the winding loosening positioning process is executed, the loosening coordinates are output; when the positioning process is not triggered, an interference warning signal is output.

Images