Transformer Breather Operating Status Monitoring Method and System

Abstract

This application relates to the field of transformer monitoring technology, and in particular to a method and system for monitoring the operating status of transformer breather. The method collects pressure, liquid level, and vibration data related to the breather, preprocesses and extracts features from the data to obtain characteristic information representing the data fluctuation pattern, and then determines whether the breather is in a normal breathing, non-breathing, or rapid oil leakage state based on a preset mapping relationship. The system includes corresponding data acquisition, processing, and communication modules. This application achieves real-time and automated monitoring of the breather status, can promptly warn of faults, and improves the accuracy of judgment through multi-sensor data fusion. It features a simple structure, convenient installation and maintenance, and suitability for large-scale application, effectively improving the safety and reliability of transformer operation.

Description

Technical Field

[0001] This application relates to the field of transformer monitoring technology, and in particular to a method and system for monitoring the working status of transformer breathers. Background Technology

[0002] As a core power supply device in the power system, the safe and stable operation of the transformer is directly related to the reliability of the power grid. The transformer breather is a key auxiliary component to ensure the normal operation of the transformer. It is mainly used to absorb moisture entering the oil conservator to maintain the insulation strength of the transformer oil. At the same time, it dynamically balances the air pressure between the oil conservator and the outside environment through the process of inhalation and exhalation. When the transformer oil temperature changes with the ambient temperature or the heat generated during operation, the internal oil volume will expand or contract accordingly, prompting the breather to complete gas exchange and avoid abnormal air pressure inside the oil conservator, which could lead to equipment damage.

[0003] Existing monitoring methods for transformer breather status have significant limitations: some scenarios rely on regular manual inspections, which are not only labor-intensive and have long monitoring intervals, but also make it difficult to achieve continuous real-time monitoring 24 hours a day, and cannot promptly detect sudden faults such as breathing channel blockage, abnormal air pressure, and rapid oil leakage; other monitoring devices have complex structures and are cumbersome to install and operate, and their hardware deployment and subsequent maintenance require professional personnel and a large amount of resources, resulting in high monitoring costs and making it difficult to meet the routine monitoring needs of large-scale power equipment.

[0004] The shortcomings of the existing technology make it impossible to quickly identify and warn of abnormal conditions of transformer breathers, which may lead to safety hazards such as reduced transformer insulation performance, equipment failure, or even power outages. Therefore, there is an urgent need for a transformer breather condition monitoring method that is real-time, easy to operate, and has low maintenance costs to solve the core problems existing in the current technology. Summary of the Invention

[0005] This application provides a method for monitoring the working status of a transformer breather, including the following steps: Collect monitoring data related to transformer breather; The monitoring data is preprocessed and features are extracted to obtain feature information corresponding to each monitoring data. The working state of the transformer breather is determined based on the feature information and the preset mapping relationship between the working states of the transformer breather. The monitoring data includes at least one of pressure data, liquid level data, and vibration data; the feature information is used to characterize the fluctuation pattern of the corresponding monitoring data, and the fluctuation pattern includes periodic fluctuation pattern, stable no-fluctuation pattern, and instantaneous change pattern.

[0006] According to the transformer breather operating status monitoring method provided in this application, the monitoring data is preprocessed and feature extracted, including: The collected monitoring data are preprocessed by filtering, denoising, and drift compensation. The preprocessed monitoring data is subjected to time series analysis within a preset sliding time window to extract dynamic characteristic indicators that characterize its short-term fluctuation patterns.

[0007] According to the transformer breather operating status monitoring method provided in this application, the preprocessed monitoring data is subjected to time series analysis within a preset sliding time window to extract dynamic characteristic indicators that characterize its short-term fluctuation patterns, including: Within the sliding time window, the continuous change value of the pressure data is calculated; The cumulative signs of the continuously changing values ​​are accumulated to obtain the trend cumulative coefficient; The fluctuation pattern of the pressure data is determined based on the numerical characteristics of the cumulative trend coefficient and the rate of change of the pressure data within the window.

[0008] According to the transformer breather operating status monitoring method provided in this application, the fluctuation pattern of the pressure data is determined based on the numerical characteristics of the cumulative trend coefficient and the rate of change of the pressure data within a window, including: If the cumulative trend coefficient exhibits periodic alternating positive and negative changes within multiple consecutive time windows, it is determined to be a periodic fluctuation pattern. If the absolute value of the cumulative trend coefficient remains below the first preset threshold for a preset long period of time, it is determined to be a stable and unfluctuating mode. If the rate of decrease of the pressure data per unit time exceeds the second preset threshold, it is determined to be a transient change mode.

[0009] According to the transformer breather operating status monitoring method provided in this application, the preprocessed monitoring data is subjected to time series analysis within a preset sliding time window to extract dynamic characteristic indicators representing its short-term fluctuation patterns, and further includes: Analyze the temporal synchronization between the rising and falling trends of the liquid level data and the fluctuation patterns of the pressure data; The fluctuation pattern of the liquid level data is determined based on the analysis results of the synchronicity.

[0010] According to the transformer breather operating status monitoring method provided in this application, the fluctuation pattern of the liquid level data is determined based on the synchronization analysis results, including: If the rise and fall of the liquid level data is synchronized with the periodic fluctuation pattern of the pressure data in time, it is determined to be a periodic fluctuation pattern. If the change in the liquid level data remains stable within a preset long period, it is determined to be a stable and fluctuation-free mode. If the liquid level data drops sharply within a short period of time, it is determined to be a transient change mode.

[0011] According to the transformer breather operating status monitoring method provided in this application, the preprocessed monitoring data is subjected to time series analysis within a preset sliding time window to extract dynamic characteristic indicators representing its short-term fluctuation patterns, and further includes: Analyze the signal strength or spectral characteristics of the vibration data during the corresponding time periods when the pressure data exhibits a periodic fluctuation pattern; The fluctuation pattern of the vibration data is determined based on the analysis results of the signal strength or spectral characteristics.

[0012] According to the transformer breather operating status monitoring method provided in this application, the fluctuation pattern of vibration data is determined based on signal strength or spectral characteristics, including: If the vibration data synchronously exhibits periodic signal intensity peaks during the periodic fluctuation of pressure, it is determined to be a periodic fluctuation pattern. If the signal strength of the vibration data remains below the third preset threshold, it is determined to be a stable, fluctuation-free mode. If the vibration data exhibits characteristics of high frequency and large amplitude, it is determined to be a transient change mode.

[0013] According to the transformer breather operating status monitoring method provided in this application, the preset mapping relationship is as follows: When the fluctuation pattern represented by the feature information is a periodic fluctuation pattern, it is mapped to a normal breathing state; When the fluctuation pattern represented by the feature information is a stable, non-fluctuating pattern, it is mapped to the state of disappearance of respiration; When the fluctuation pattern represented by the feature information is a transient mutation pattern, it is mapped to a rapid oil leak state.

[0014] According to the transformer breather operating status monitoring method provided in this application, the operating status of the transformer breather is determined based on the feature information and a preset mapping relationship between the transformer breather operating status, including: When making judgments based on the characteristic information of a single type of monitoring data, the corresponding working status is output directly according to the fluctuation pattern and its mapping relationship represented by the characteristic information.

[0015] According to the transformer breather operating status monitoring method provided in this application, the operating status of the transformer breather is determined based on the feature information and a preset mapping relationship between the transformer breather operating status, including: When a fusion judgment is made based on the feature information of at least two types of monitoring data, the consistency of the fluctuation pattern represented by each feature information is checked. If the check passes, the final judgment is made based on the working state mapped by the fluctuation pattern.

[0016] According to the transformer breather operating status monitoring method provided in this application, the consistency verification includes confidence fusion verification: Assign a confidence weight to the feature information of each type of monitoring data; Calculate the overall confidence level based on the aforementioned feature information and its confidence weights; If the overall confidence level exceeds the preset confidence level threshold, the verification is deemed successful.

[0017] This application also provides a transformer breather condition monitoring system for implementing the aforementioned method, characterized in that the system comprises: The data acquisition module is configured to collect monitoring data related to the transformer breather. The data processing module is communicatively connected to the data acquisition module and is configured to preprocess and extract features from the monitoring data to obtain feature information, and determine the working status according to a preset mapping relationship. A data communication module, connected to the data processing module, is configured to upload the judgment result and / or the collected monitoring data to the remote monitoring platform.

[0018] According to the transformer breather condition monitoring system provided in this application, the data acquisition module includes at least one of a pressure sensor, a liquid level sensor, and a vibration sensor; The pressure sensor is installed at the connection flange of the transformer breather. The liquid level sensor is installed inside the oil cup of the respirator; The vibration sensor is located inside the oil cup or on the respirator body.

[0019] The above-mentioned one or more technical solutions provided in this application include at least the following technical effects: The beneficial effects of this technical solution are primarily reflected in its realization of 24 / 7 automated intelligent monitoring of the transformer breather's operating status. It completely changes the outdated model that relies on periodic manual inspections. By deploying pressure, level, and vibration sensors and establishing a real-time data acquisition and analysis chain, it can continuously capture subtle dynamic changes in the breather. This means that critical potential faults such as blocked breathing channels, pressure imbalance failures, or sudden rapid oil leaks can be accurately identified by the system in their early stages or even in their nascent phase. This provides valuable early warning time for maintenance personnel, greatly enhancing the initiative and preventative capabilities of power equipment safety management, and effectively avoiding transformer insulation degradation or even power outages that may be caused by breather malfunction.

[0020] Secondly, this solution defines and identifies three core fluctuation patterns: periodic fluctuations, stable fluctuations, and transient abrupt changes, constructing a language system that accurately maps the health status of respirators. Its advantage lies not only in analyzing single data points but also in its multi-source information fusion and verification mechanism. When the system uses pressure, level, and vibration data for comprehensive judgment, it performs rigorous logical consistency or confidence level fusion verification, effectively providing multiple layers of insurance for the status diagnosis results. This design significantly enhances the system's anti-interference capability and the reliability of diagnostic results in complex industrial environments, effectively filtering out false alarms caused by occasional sensor errors or environmental noise. It ensures that every reported "disappearance of respiration" or "rapid oil leakage" alarm has extremely high credibility, greatly improving the accuracy of operational and maintenance decisions.

[0021] Finally, from the perspective of engineering implementation and economic benefits, this solution demonstrates good practicality and scalability. The entire monitoring system architecture is clear, the sensor selection and installation scheme are specific and well-defined, and the data processing core is deployed on edge-side embedded devices, reducing reliance on the continuous bandwidth of the central system. Simultaneously, the system supports dual-channel data communication via RS485 and ZigBee, balancing reliability and deployment flexibility. This integrated design simplifies system installation and commissioning, controls subsequent maintenance costs, and overcomes the shortcomings of some previous monitoring devices, such as complex structures, high professional requirements, and high costs. This enables the technical solution to be adapted for large-scale, routine deployment in a large number of transformer clusters, providing a practical and feasible technical path for power grid companies to achieve intelligent, low-cost, and universal monitoring of key auxiliary equipment. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in this application 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart of a transformer breather operating status monitoring method provided in some embodiments of this application; Figure 2 This is a schematic diagram of the structure of a transformer breather operating status monitoring system provided in some embodiments of this application; Figure 3 This is a pressure data change graph in a transformer breather operating status monitoring method provided in some embodiments of this application; Figure 4 This is another pressure data change graph in the transformer breather operating status monitoring method provided in some embodiments of this application; Figure 5 This is yet another pressure data change graph in the transformer breather working status monitoring method provided in some embodiments of this application. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0025] The terms "first" and "second" in the specification and claims of this application may explicitly or implicitly include one or more of the features. In the description of this application, unless otherwise stated, "multiple" means two or more. Furthermore, "and / or" in the specification and claims indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0026] In the description of this application, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc., indicating the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0027] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0028] See Figure 1 This application provides a method for monitoring the working status of a transformer breather, comprising the following steps: S1. Collect monitoring data related to transformer breather; S2. Preprocess and extract features from the monitoring data to obtain feature information corresponding to each monitoring data; S3. Determine the working state of the transformer breather based on the feature information and the preset mapping relationship between the transformer breather working states; The monitoring data includes at least one of pressure data, liquid level data, and vibration data; the feature information is used to characterize the fluctuation pattern of the corresponding monitoring data, and the fluctuation pattern includes periodic fluctuation pattern, stable no-fluctuation pattern, and instantaneous change pattern.

[0029] The core of step S1 is to acquire relevant monitoring data that accurately reflects the working status of the transformer breather, providing reliable raw data support for subsequent preprocessing, feature extraction, and status judgment. The collected monitoring data needs to be accurately correlated with the core working characteristics of the breather. Among them, pressure data is used to capture the air pressure fluctuations inside the breather caused by changes in the transformer oil volume; liquid level data is used to reflect the dynamic adjustment of the liquid level in the breather's oil cup during the breathing process; and vibration data is used to record the mechanical vibrations caused by the generation of air bubbles or the release of pressure in the breather's oil cup during the breathing process. These three types of data can be collected individually or in combination to adapt to the needs of different monitoring scenarios.

[0030] In some embodiments, monitoring data is acquired by deploying corresponding types of sensors. The installation position and fixing method of each sensor must match the structural characteristics of the respirator to ensure the accuracy of data acquisition. The pressure sensor is embedded in the flange connecting the transformer respirator and the pipeline and is fixed with screws to ensure that the sensor can directly sense the pressure changes inside the respirator. The level sensor is fixed to the inner wall of the respirator's oil cup, while the vibration sensor is placed in the transformer oil inside the oil cup. The data cables of both are fixed with a combination of straps and tape to prevent the cables from loosening or shifting due to vibration during equipment operation. All sensors are connected to the monitoring system's host computer via RS485 data cables, and the host computer is connected to a 220V power supply for stable power.

[0031] In some embodiments, the acquisition of different types of monitoring data has clear physical meaning and parameter standards. The sampling accuracy of pressure data reaches ±0.5%, which is used to reflect the dynamic changes of air pressure in the transformer oil conservator in real time; the unit of liquid level data is millimeters, which is used to accurately record the fluctuation of oil level height in the oil cup, with a sampling accuracy of ±1 millimeter; the unit of vibration data is also millimeters, which mainly captures the liquid surface vibration signal when the oil cup expels air bubbles during the breather's exhalation, with a sampling accuracy of ±2%FS. The acquisition accuracy design of each type of data lays the foundation for subsequent accurate identification of fluctuation patterns.

[0032] In some embodiments, data acquisition employs an adaptive frequency adjustment strategy to balance real-time monitoring and data validity. Under normal monitoring conditions, the system samples various monitoring data at a frequency of 1-2 seconds per sample, preferably 1 second per sample. The sampled data is first stored locally, and then packaged and organized every 10-20 minutes per sample, preferably every 15 minutes per sample. When any of the collected pressure, level, or vibration data exceeds its corresponding preset threshold, the system immediately triggers a high-frequency acquisition mode, adjusting to a sampling frequency of 0.2-1 seconds per sample, preferably 0.5 seconds per sample, until all monitoring data recovers to within the preset threshold range. Specifically, the preset threshold for pressure data is ±1 kPa, for level data it is 1-10 cm, and for vibration data it is ±5 mm. This strategy can reasonably control the amount of data under normal conditions and accurately capture key changes under abnormal conditions.

[0033] In some embodiments, time-series alignment processing needs to be performed synchronously during data acquisition to ensure temporal consistency among multiple data types. By adding a standardized timestamp to each set of acquired pressure, level, and vibration data, time-series matching of data from different sensors is achieved, avoiding data misalignment due to acquisition delays or transmission differences. For data gaps that may occur during acquisition, interpolation is used to supplement the data, ensuring the integrity of the original data and providing a time-consistent, complete, and continuous data source for subsequent multi-data fusion analysis and fluctuation pattern determination.

[0034] Step S2 involves systematically processing and extracting key information from the collected monitoring data. Preprocessing eliminates data interference and corrects data deviations. Then, feature extraction uncovers patterns in the monitoring data related to the transformer breather's operating status, ultimately forming feature information that accurately characterizes the fluctuation patterns of each monitoring data point. This provides a reliable analytical basis for subsequent judgment of the breather's operating status based on a pre-defined mapping relationship. This step requires adopting appropriate data processing strategies for the different physical characteristics of pressure, level, and vibration data to ensure that the preprocessed data quality meets analytical requirements and that the extracted feature information effectively correlates the breather's normal breathing, loss of breathing, and rapid oil leakage states.

[0035] In some embodiments, step S2 specifically includes two subordinate steps, S21 and S22: S21. Perform preprocessing operations such as filtering, denoising, and drift compensation on the collected monitoring data. S22. Perform time series analysis on the preprocessed monitoring data within a preset sliding time window to extract dynamic characteristic indicators that characterize its short-term fluctuation patterns.

[0036] The S21 preprocessing operation is performed first to optimize the quality of the original monitoring data, and the S22 feature extraction operation is performed later to extract key features from the optimized data. The two are connected to form a complete S2 processing flow.

[0037] Step S21, the preprocessing operation, mainly includes filtering and denoising, drift compensation, and time synchronization to eliminate interference signals in the original data, correct data offsets over long-term operation, and ensure the temporal consistency of multiple types of monitoring data. For high-frequency noise and random spikes that may exist in the original monitoring data, a combination of digital filtering and moving average is used to smooth data fluctuations and filter out irrelevant high-frequency interference. For long-term drift in pressure data that may be caused by environmental factors or equipment characteristics, a pressure baseline is calculated by referring to the expiratory peak and inspiratory trough values ​​of pressure data within each day or work cycle. Zero-point adjustment is then performed based on this baseline to achieve drift compensation and avoid misjudgment of features caused by drift. For potential time differences in the acquisition of multiple types of monitoring data, pressure data, liquid level data, and vibration data are time-aligned based on the timestamp at the time of acquisition. If data is missing, it is supplemented by interpolation to ensure that different types of data correspond one-to-one in the time dimension, laying the foundation for subsequent multi-data fusion analysis.

[0038] Step S22, feature extraction, needs to be carried out within a preset sliding time window. Dynamic feature indicators that can characterize the short-term fluctuation pattern of monitoring data are extracted through time series analysis. Different extraction strategies are adopted for different physical meanings of pressure data, liquid level data, and vibration data.

[0039] In some embodiments, when performing time-series analysis and dynamic feature extraction on the preprocessed pressure data in step S22, it is first necessary to determine the parameters of the preset sliding time window. The length of this window can be flexibly configured according to actual monitoring needs and sampling frequency. Its setting must take into account both the timeliness and accuracy of feature extraction—a window that is too short may result in incomplete feature extraction and fail to reflect the fluctuation pattern of the data, while a window that is too long may delay feature recognition and affect the real-time performance of state judgment. Considering that the sampling frequency of pressure data in a normal monitoring scenario is once every second, the number of sampling points N in the sliding time window can be set to 10-20, preferably N=10. In this case, the window duration is 10 seconds. This duration can cover the pressure changes during a single exhalation or inhalation of the respirator, and can also avoid feature delay due to an excessively long window, while adapting to the subsequent second-level response state judgment requirements. If in an abnormally high-frequency acquisition mode (sampling frequency of once per second), the number of sampling points N in the window can be adjusted to 20, keeping the window duration stable at 10 seconds to ensure the consistency of feature extraction under different acquisition modes.

[0040] After determining the sliding time window, the continuous change in pressure data within the window is first calculated, specifically using the following formula:

[0041] in This represents the pressure value (in Pascals) of the i-th sample within the window. This is the difference between the pressure values ​​sampled in the i-th and i+1-th sampling.

[0042] The core function of this calculation step is to capture the instantaneous trend of pressure data changes. A positive value indicates that the pressure is increasing within that sampling interval; when... When the value is negative, it indicates that the pressure is decreasing; when A value of zero indicates that the pressure remains unchanged within that interval. This is achieved through continuous calculations. This process can transform discrete pressure samples within a window into a sequence reflecting dynamic changes, providing foundational data for subsequent trend quantification analysis. Simultaneously, this calculation can initially filter out outliers caused by single-sample errors—if a certain... The absolute value far exceeds the normal fluctuation range, which can be combined with adjacent values. The changing trend can be used to determine whether it is a sampling error, thus providing a basis for verifying the effectiveness of filtering and denoising.

[0043] Then, all continuously changing values ​​within the window were analyzed. The cumulative signs are accumulated to obtain the trend cumulative coefficient, which is calculated using the following formula:

[0044]

[0045] in, This is the cumulative trend coefficient. The number of sampling points in the sliding time window. It is a symbolic function.

[0046] The physical meaning of the cumulative trend coefficient k lies in quantifying the overall trend of pressure data within the window: if the value of k is large and positive (e.g., k≥+90, this threshold is based on a large amount of normal operating data of respirators and can effectively distinguish between obvious upward trends and random fluctuations), it indicates that the pressure within the window is generally on an upward trend, corresponding to the respirator possibly being in the process of inhalation, at which time the pressure in the oil reservoir gradually increases due to the contraction of oil volume; if the value of k is small and negative (e.g., k≤-90), it indicates that the pressure within the window is generally on a downward trend, corresponding to the respirator possibly being in the process of exhalation, at which time the pressure in the oil reservoir gradually decreases due to the expansion of oil volume; if the value of k is between +90 and -90, it is necessary to further judge in conjunction with the cumulative trend coefficient of the previous sliding time window. For example, if k≥+90 (upward trend) in the previous window, and k is in the middle range in the current window, it indicates that the pressure has transitioned from rising to stable, corresponding to the stage of switching from inhalation to exhalation, and vice versa. This cumulative calculation process can effectively filter out the interference of short-term random fluctuations on trend judgment, strengthen the overall trend characteristics through symbol accumulation, and make the direction of pressure changes clearer.

[0047] Finally, by combining the numerical characteristics of the cumulative trend coefficient k with the rate of change of the pressure data within the window, the fluctuation pattern of the pressure data is comprehensively determined.

[0048] When the cumulative trend coefficient k exhibits periodic alternating positive and negative changes within multiple consecutive sliding time windows (e.g., k ≥ +90 initially, then k falls within the middle range, then k ≤ -90, and returns to the middle range, repeating this cycle), and the pressure change rate remains stable within a preset normal range (this range is set based on pressure fluctuation data during normal breathing of the respirator, such as pressure changes not exceeding ±500 Pascals per 10 seconds), the pressure data fluctuation pattern is determined to be a periodic fluctuation pattern. This pattern corresponds to the normal breathing state of the respirator, indicating that the pressure inside the oil reservoir achieves dynamic equilibrium with changes in oil temperature. The pressure data change graph for one state is shown below. Figure 3 As shown.

[0049] When the cumulative trend coefficient k remains below +90 and above -90 for a preset long period (e.g., 24 hours, a complete work cycle), and the pressure change rate approaches zero (e.g., pressure change less than ±50 Pascals per 10 seconds), it is determined to be a stable, non-fluctuating mode. This mode corresponds to the absence of breathing, which may be caused by desiccant blockage, air passage obstruction, or valve jamming, preventing normal pressure fluctuations. A pressure data change graph for one of these states is shown below. Figure 4 As shown.

[0050] When the rate of change of pressure data within a single sliding time window far exceeds the normal range (e.g., a sudden pressure drop rate > 10 Pa / s during normal breathing, which is around 3 Pa / s, indicates oil leakage), or when the lowest pressure value within the window is less than 0.5 Pa / s during inspiration... (This threshold is set based on the pressure mutation characteristics during oil leakage, effectively distinguishing between normal fluctuations and rapid pressure drops caused by leakage.) When this threshold is reached, it is determined to be a transient mutation mode. This mode corresponds to a rapid oil leakage state, indicating that a tank or pipeline rupture causes a rapid drop in oil volume, which in turn leads to a rapid decrease in pressure within a short period of time, stabilizing in a negative pressure zone. The pressure data change graph for one of these states is shown below. Figure 5 As shown.

[0051] The calculation of the pressure change rate requires considering the difference between the maximum and minimum pressure values ​​within the window, combined with the window duration, to determine the rate of pressure change per unit time. Specifically, if the rate of pressure decrease per unit time exceeds a second preset threshold, it is classified as a transient change pattern. The second preset threshold is obtained by multiplying the maximum rate of pressure change from routine monitoring over the most recent day by 3.

[0052] in, The respiratory extreme (average pressure value within the window) is calculated using the following formula:

[0053] By combining the cumulative trend coefficient and the rate of change, we can ensure the accuracy of fluctuation pattern determination and cover the pressure characteristics under different working conditions of the respirator, providing a reliable basis for pressure characteristics for subsequent multi-source fusion judgment based on liquid level and vibration data.

[0054] In some embodiments, when the pressure data within the current sliding time window corresponds to the ventilator's exhalation process through a synergistic analysis of the trend cumulative coefficient and the pressure change rate, the system will calculate the average pressure value within that window. This is directly used as the reference for the peak expiratory volume during this expiratory process and recorded as... .

[0055] The core logic of this design is that during exhalation, pressure data undergoes a gradual decline after reaching a peak. The average pressure value within the window effectively represents the typical peak pressure level of that exhalation, providing a reliable basis for capturing the pressure extremes of each subsequent exhalation. Similarly, when the pressure data within the current window is determined to correspond to the inhalation process, the system uses the average pressure value of that window. As a reference for the inhalation trough value during this inhalation process, it is denoted as... Because pressure data gradually rises after dropping to a trough during inspiration, the average value within a window accurately reflects the typical pressure trough characteristics of that inspiration. By recording the peak and trough values ​​of each expiration and inspiration cycle in real time, the system can continuously accumulate key pressure extreme value data within the ventilator's respiratory cycle, laying the foundation for subsequent pressure baseline calculations.

[0056] At the end of each day, or when the transformer completes a full work cycle (such as the complete process from startup to shutdown maintenance), the system will retrieve all records from that day or work cycle. The highest value was selected from all values ​​as the highest air pressure (i.e., peak expiratory pressure) for the day or cycle, while the highest value was selected from all values. The lowest value is selected from the data and used as the lowest air pressure for that day or cycle (i.e., the minimum inhalation valley value). Then, based on the selected highest and lowest air pressures, the system uses the formula:

[0057] The pressure baseline for this period was calculated. The pressure baseline is essentially the central reference value of the pressure fluctuation range under normal operating conditions of the respirator. Its core function is to achieve automatic zero-point adjustment of the system. Because sensors may experience pressure data drift due to factors such as changes in ambient temperature and equipment aging during long-term operation, the original monitoring data may deviate from the actual pressure level. By periodically calculating and updating the pressure baseline, the system can compare subsequently collected pressure data with this baseline, automatically calibrating the zero-point deviation and eliminating measurement errors caused by drift. For example, if the subsequently collected pressure data is generally too high or too low, the system will correct the data based on the latest pressure baseline, ensuring that the pressure data always remains consistent with the actual pressure state under operating conditions. This effectively avoids misjudgments of breathing status caused by data drift, improving the long-term reliability and accuracy of the entire monitoring system.

[0058] In some embodiments, when performing time-series analysis and dynamic feature extraction on the preprocessed liquid level data in step S22, the core logic is based on the common physical mechanism of liquid level change and pressure fluctuation—both are driven by the breathing action of the breather caused by the change in transformer oil volume with temperature. Therefore, the signal characteristics collected by the liquid level sensor are completely consistent with those of the pressure sensor, and the determination of its fluctuation mode needs to be based on the time-series synchronization and dynamic feature consistency verification with the pressure data.

[0059] Specifically, time-series synchronization analysis requires timestamp alignment, matching preprocessed liquid level data and pressure data at the same time granularity (e.g., 1 second per data point), comparing their changing trends at corresponding time nodes. Dynamic feature consistency focuses on the synergy of fluctuation period, rate of change, and amplitude range, ensuring that the characteristics driven by the same respiratory action overlap.

[0060] When the rising and falling trend of liquid level data corresponds perfectly with the periodic fluctuation of pressure data at the same time point, that is, when the pressure rises, the liquid level rises synchronously and when the pressure falls, the liquid level falls synchronously, and the period of liquid level fluctuation is completely consistent with the period of pressure periodic fluctuation (such as 3 minutes / cycle under normal breathing conditions), and the rate of liquid level change is positively correlated with the rate of pressure change, the fluctuation pattern of liquid level data is determined to be a periodic fluctuation pattern.

[0061] When the change in liquid level data remains highly stable within a preset long period (e.g., 24 hours, covering the complete working cycle of the transformer), and this stable state is completely synchronized with the stable and unfluctuating mode of pressure data in terms of time sequence, that is, when there is no periodic fluctuation in pressure, there is no dynamic change in liquid level, it is determined to be a stable and unfluctuating mode. This state corresponds to the cessation of breathing action caused by blockage of the breathing channel of the respirator or valve jamming.

[0062] When the liquid level data drops sharply in a short period of time (e.g., within 10 seconds), that is, the rate of drop exceeds 5 times the rate of change under the normal fluctuation cycle, and this drop process completely coincides with the instantaneous change pattern of the pressure data (short-term sharp drop in pressure), and the downward trend does not decay and continues until the liquid level stabilizes in the low liquid level range, it is determined to be an instantaneous change pattern, corresponding to a rapid oil leakage scenario caused by the rupture of the respirator tank or pipeline.

[0063] The breathing cycle of a transformer breather is a cycle of exhalation and inhalation that takes several hours or more to complete. The oscillation cycle of the breather is the period of vibration caused by one exhalation or one inhalation within the breathing cycle.

[0064] In some embodiments, when performing time-series analysis and dynamic characteristic index extraction on the preprocessed vibration data in step S22, the basis is the physical relationship between the vibration signal and the breathing action of the respirator. The vibration is caused by the inhalation and exhalation of bubbles on the oil surface during the respirator's exhalation or inhalation process. Its signal characteristics (amplitude, period, attenuation law) are directly related to the intensity and frequency of the breathing action. Therefore, it is necessary to combine the fluctuation pattern of the pressure data and focus on analyzing the signal amplitude, duration, periodic characteristics and attenuation law of the vibration data.

[0065] For periodic fluctuation patterns, the determination must meet the requirements of temporal correspondence with the periodic fluctuations of pressure data and the inherent characteristics of the vibration signal. During the period when the pressure data exhibits periodic fluctuations, the signal collected by the vibration sensor shows pulse-like vibrations with a fixed period: the typical fluctuation period is set to 3 minutes, and a set of vibration signals is generated only when the ventilator expels bubbles within each period. The vibration amplitude is within ±5 mm, and the signal shows obvious attenuation characteristics—the vibration amplitude reaches its peak at the moment of expiring bubbles, and then rapidly decays within 20 seconds until the amplitude is below the detection limit. The vibration data remains stable without significant fluctuations in the remaining period until the expiring bubble action of the next fluctuation period triggers a new vibration signal. This pattern is completely synchronized with the exhalation and inhalation alternation process of the periodic pressure fluctuations. When the above characteristics are met, it is determined to be a periodic fluctuation pattern.

[0066] For a stable, unfluctuated mode, the key to judgment lies in the intensity threshold and continuous stability verification of the vibration signal. The third preset threshold is calibrated based on the ambient background noise level and is usually set to ±0.5mm (this value is lower than the minimum amplitude of normal bubbling vibration and covers the inherent noise range of the sensor). If, within the fluctuation period corresponding to the pressure data, the signal intensity of the vibration data remains below this threshold, with no obvious pulsed vibration peaks, and remains stable throughout a preset long period (e.g., 24 hours), without any signals exhibiting periodic fluctuation characteristics, and is synchronized with the stable, unfluctuated mode of the pressure data, then it is determined to be a stable, unfluctuated mode, corresponding to the state where the respirator's breathing action has stopped.

[0067] For transient change patterns, the criteria for determination are sudden changes in the frequency and amplitude of the vibration signal, and a temporal correlation with the transient change pattern of the pressure data. When the breathing rate increases significantly due to rapid oil leakage from the respirator, the vibration signal exhibits significant high-frequency and high-amplitude characteristics: the fluctuation period shortens from the normal 3 minutes to less than 30 seconds, while the vibration amplitude exceeds ±10mm, far exceeding the amplitude range of normal periodic fluctuations. The signal frequency perfectly matches the rapid fluctuation frequency of the pressure data (such as multiple pressure changes per second). When the above characteristics are met, it is determined to be a transient change pattern, corresponding to an emergency leakage scenario caused by the rupture of the respirator tank or pipeline.

[0068] Step S3 is the decision-making stage of the transformer breather operating status monitoring method. Its core logic lies in matching and analyzing the characteristic information corresponding to the various monitoring data (pressure data, liquid level data, vibration data) extracted in step S2 with the preset operating status mapping relationship, and finally outputting the specific operating status of the breather. The characteristic information carries the fluctuation pattern of each monitoring data (periodic fluctuation pattern, stable no-fluctuation pattern, instantaneous change pattern), while the preset mapping relationship is a "fluctuation pattern combination - operating status" correspondence rule established based on the working principle of the transformer breather and a large amount of typical operating condition test data. This includes both the basic correspondence between the fluctuation pattern and status of a single monitoring data point, and the fusion correspondence between the fluctuation patterns and status of multiple monitoring data points. Through this step, accurate identification of three core states of the breather—normal breathing, disappearance of breathing, and rapid oil leakage—can be achieved, while providing a decision-making basis for subsequent abnormal early warning and remote management.

[0069] In some embodiments, the preset mapping relationship of transformer breather operating states needs to be established through prior operating condition calibration and dynamic calibration to ensure the scientific validity and adaptability of the mapping rules. Specifically, by simulating three typical states—normal breathing, no breathing, and rapid oil leakage—characteristic information of pressure, liquid level, and vibration data is collected under each state. The fluctuation pattern (periodic fluctuation, stable without fluctuation, instantaneous change) corresponding to each characteristic information is labeled, and then the mapping rules are determined based on statistical analysis: for example, the periodic fluctuation pattern corresponds to the normal breathing state, the stable without fluctuation pattern corresponds to the no breathing state, and the instantaneous change pattern corresponds to the rapid oil leakage state. At the same time, this mapping relationship needs to be dynamically calibrated in conjunction with sensor characteristics. For example, considering the sampling accuracy of ±0.5% for pressure sensors and ±1mm for liquid level sensors, a small fluctuation tolerance range is set in the mapping rules to avoid misjudgments caused by inherent sensor errors and ensure the reliability and adaptability of the mapping relationship.

[0070] In some embodiments, when determining the working status based on feature information, the determination can be made initially based on the feature information of a single type of monitoring data. In this case, the physical correlation between the data and the respirator's status must be fully considered to enhance the rationality of the determination. For example, when determining based solely on pressure data feature information, if the fluctuation pattern is periodic, it is directly mapped to a normal breathing state; if the pressure data fluctuation pattern is stable with no fluctuation, it is mapped to a state where breathing has ceased; if the pressure data fluctuation pattern is a sudden change, it is mapped to a state of rapid oil leakage. This single-data determination method is suitable for simple scenarios where the sensor portion is deployed, while ensuring determination accuracy through physical correlation and formula calculation.

[0071] In some embodiments, when determining the operating status of a transformer breather based on feature information, if a fusion judgment is performed based on feature information from at least two types of monitoring data (i.e., any two or three of pressure data, liquid level data, and vibration data), the core is to first verify the consistency of the fluctuation patterns represented by each of these feature information types. After the verification passes, the final determination of the breather's operating status is made based on the mapping relationship between the fluctuation patterns and preset operating states. The core purpose of this fusion judgment method is to avoid misjudgments caused by environmental interference, sensor bias, and other factors due to the collaborative verification of multi-dimensional data, thereby improving the reliability of status judgment.

[0072] Consistency verification can specifically employ a confidence-based fusion verification strategy. This strategy not only considers whether the judgment results from each data source are consistent, but also introduces a quantitative assessment and synthesis of the reliability of each judgment result, thereby achieving a higher level of intelligent decision-making.

[0073] The core of confidence fusion verification lies in assigning a dynamic or preset confidence weight to the feature information generated by each type of monitoring data, and calculating a comprehensive confidence level that reflects the overall reliability of the judgment.

[0074] Specifically, the system first calculates a basic confidence score for the fluctuation pattern determined by each type of data based on the time series analysis results. The basic confidence scores of each type of data are calculated by "weighted synthesis of multi-dimensional indicators" to ensure the comprehensiveness and accuracy of the evaluation results.

[0075] For stress data, its baseline confidence score The calculation assesses the degree of match between the current pressure fluctuation pattern and the ideal / historical breathing pattern. This is achieved through a weighted synthesis of multi-dimensional indicators, using the following formula:

[0076] in , , The sub-weights for each indicator can be adjusted according to actual monitoring needs. In a specific example, they can all be set to 1 / 3.

[0077] The regularity score is reflected by the cumulative trend coefficient k: a complete fluctuation should produce a sequence of k values ​​that are first positive and then negative. The regularity is reflected in the stability of the period and amplitude. The calculation method is to first calculate the standard deviation of k within a time window of a complete fluctuation period T. Then calculate using the following formula:

[0078] in, The smaller the value, the more stable the k-value sequence, and the closer the score is to 1.

[0079] The matching score based on the historical model is based on the idea of ​​comparing the current window's pressure-time curve with the standard template curve (or mean curve) of normal breathing in the past. The calculation method is to calculate the cross-correlation coefficient of the two curves or to use the dynamic time warping (DTW) distance.

[0080] If the cross-correlation coefficient is used, this coefficient is directly used as the model matching score; If DTW distance is used, the model matching score is calculated using the following formula:

[0081] The stability score reflects the stability of the rate of change of data. The core idea is the rate of change of pressure under normal breathing (i.e., ΔP). i The sequence should be relatively stationary without drastic jumps. Its calculation method is as follows: Calculate ΔP i Standard deviation of the sequence And the mean, then calculate the coefficient of variation:

[0082] The stability score is then calculated using the following formula:

[0083] The more stable the rate of change, the higher the score.

[0084] Basic confidence score of liquid level data Baseline confidence score of vibration data All data were calculated using the "multi-dimensional index weighted synthesis" method for the aforementioned stress data, which involves obtaining the basic confidence level through a weighted synthesis of "regularity score + model matching score + stability score".

[0085] Among them, the sub-weights can be flexibly configured according to the sensor characteristics and monitoring scenarios to ensure that the evaluation logic is consistent with the pressure data and is adapted to its own signal characteristics.

[0086] Subsequently, the system synthesizes these basic confidence scores according to a preset weighted fusion model to calculate the final comprehensive confidence score. :

[0087] in , , These represent the confidence weights assigned to pressure, level, and vibration data, respectively. These weights can be set and adjusted based on the historical reliability of different sensors in specific condition diagnostics, the importance of the installation environment, or expert experience.

[0088] Based on the actual performance of various sensors, experimental verification determined that the pressure sensor provides the most accurate, stable, and intuitive data, followed by the level sensor, while the vibration sensor performs the worst. Therefore, the reliability weights are assigned as follows: Pressure Data Weight =0.5, weight of liquid level data =0.3, vibration data weight =0.2.

[0089] After calculating the overall confidence level, the system compares it with a preset confidence threshold (e.g., 0.9).

[0090] If the overall confidence level exceeds the threshold, it indicates that the result of the multi-sensor fusion judgment has a high degree of credibility and the consistency verification is passed. The system will map and output the final working state (such as normal breathing state) based on the fluctuation pattern corresponding to the fusion result (such as all being periodic fluctuation patterns). If the overall confidence level does not reach the threshold, further investigation is required by combining the number of sensors deployed and the combination of modes to ensure the reliability of the judgment result.

[0091] The advantage of this confidence fusion verification method lies in its ability to more scientifically handle situations where data quality varies across different sensors or where some data is ambiguous. Through quantified weights and confidence calculations, it achieves soft decision fusion of multi-source information. Compared to simple, rigid consistency rules, its decision-making process is more flexible and adaptable, better suppressing random errors from individual sensors, thereby further improving the overall accuracy of state judgment and system robustness under complex operating conditions. This method also facilitates integration with potential subsequent machine learning modules, allowing for adaptive optimization of weights using historical data, enabling the system to continuously evolve.

[0092] This application also provides a transformer breather condition monitoring system for implementing the above-mentioned method, so as to realize real-time and accurate monitoring of the working status of the transformer breather, solve the problems of the inability to monitor in real time, complex structure, and high operation and maintenance costs in the prior art, and provide a guarantee for the safe and stable operation of the transformer.

[0093] See Figure 2 The transformer breather condition monitoring system provided in this application is an embedded hardware device based on a modular design, whose internal structure and connection relationships clearly correspond to the various steps of the executed method. The system includes a data acquisition module, a data processing module, and a data communication module that work together physically.

[0094] The data acquisition module consists of a set of dedicated sensors, including a pressure sensor installed at the transformer breather connection flange, a liquid level sensor located inside the breather's oil cup, and a vibration sensor placed inside the oil cup or attached to the breather body. These sensors are configured to synchronously transmit the real-time sensed pressure, liquid level, and vibration analog or digital signals to the system's data processing module via a standard RS485 bus.

[0095] The data processing module is implemented with an embedded data processing chip at its core. This chip receives all raw monitoring data from the data acquisition module via an RS485 bus interface. Subsequently, the firmware running within the chip performs preprocessing on the raw data, including filtering to eliminate noise, removing signal glitches, and compensating for baseline drift. After preprocessing, the chip performs time-series analysis on the data within a preset sliding time window. By calculating the cumulative trend coefficient of pressure data, analyzing the synchronicity of liquid level data, and extracting characteristics of vibration data, it determines the fluctuation pattern of each data point, i.e., generates corresponding feature information. Finally, based on pre-existing internal rules that map fluctuation patterns to specific operating states, the chip performs logical or confidence fusion judgments and outputs a diagnostic conclusion regarding the operating status of the transformer breather.

[0096] The data communication module is a composite communication unit integrated within the system host, responsible for establishing a connection with the remote monitoring platform. After receiving status judgment results and related data packets from the data processing module, this module forwards them through a dual-communication channel mechanism to ensure reliability. One channel involves the RS485 module within the host, which packages the data into standard RS485 protocol frames for access to wired industrial networks. The other channel uses a ZigBee chip integrated within the host to wirelessly transmit data to a nearby respirator gateway device. This gateway, acting as a network aggregation node, uploads the received data to the IoT cloud platform via the Internet through the TCP / IP protocol, thereby completing remote data storage, visualization, and management. This module also supports configurable transmission strategies, such as periodically reporting routine data and instantly reporting abnormal alarms.

[0097] The system's technical effectiveness lies in its clear division of labor and close collaboration among the three modules at the hardware and protocol levels, materializing the monitoring method into a stand-alone industrial product. The data acquisition module provides precise physical signal sensing, the data processing module performs intelligent analysis and diagnosis at the edge, and the data communication module ensures that diagnostic results are reliably delivered to remote maintenance personnel. This integrated design not only realizes all the method's functions but also enhances the system's overall reliability and practicality in complex industrial environments through engineering optimizations such as dual communication redundancy.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for monitoring the working status of a transformer breather, characterized in that, Includes the following steps: Collect monitoring data related to transformer breather; The monitoring data is preprocessed and features are extracted to obtain feature information corresponding to each monitoring data. The working state of the transformer breather is determined based on the feature information and the preset mapping relationship between the working states of the transformer breather. The monitoring data includes at least one of pressure data, liquid level data, and vibration data; The feature information is used to characterize the fluctuation pattern of the corresponding monitoring data, including periodic fluctuation pattern, stable no-fluctuation pattern and instantaneous change pattern.

2. The method according to claim 1, characterized in that, The monitoring data is preprocessed and its features are extracted, including: The collected monitoring data are preprocessed by filtering, denoising, and drift compensation. The preprocessed monitoring data is subjected to time series analysis within a preset sliding time window to extract dynamic characteristic indicators that characterize its short-term fluctuation patterns.

3. The method according to claim 2, characterized in that, For the preprocessed monitoring data, time-series analysis is performed within a preset sliding time window to extract dynamic characteristic indicators that characterize its short-term fluctuation patterns, including: Within the sliding time window, the continuous change value of the pressure data is calculated; The cumulative signs of the continuously changing values ​​are accumulated to obtain the trend cumulative coefficient; The fluctuation pattern of the pressure data is determined based on the numerical characteristics of the cumulative trend coefficient and the rate of change of the pressure data within the window.

4. The method according to claim 3, characterized in that, Based on the numerical characteristics of the cumulative trend coefficient and the rate of change of the pressure data within the window, the fluctuation pattern of the pressure data is determined, including: If the cumulative trend coefficient exhibits periodic alternating positive and negative changes within multiple consecutive time windows, it is determined to be a periodic fluctuation pattern. If the absolute value of the cumulative trend coefficient remains below the first preset threshold for a preset long period of time, it is determined to be a stable and unfluctuating mode. If the rate of decrease of the pressure data per unit time exceeds the second preset threshold, it is determined to be a transient change mode.

5. The method according to claim 2, characterized in that, For the preprocessed monitoring data, time-series analysis is performed within a preset sliding time window to extract dynamic characteristic indicators that characterize its short-term fluctuation patterns, including: Analyze the temporal synchronization between the rising and falling trends of the liquid level data and the fluctuation patterns of the pressure data; The fluctuation pattern of the liquid level data is determined based on the analysis results of the synchronicity.

6. The method according to claim 5, characterized in that, The fluctuation pattern of the liquid level data is determined based on the synchronization analysis results, including: If the rise and fall of the liquid level data is synchronized with the periodic fluctuation pattern of the pressure data in time, it is determined to be a periodic fluctuation pattern. If the change in the liquid level data remains stable within a preset long period, it is determined to be a stable and fluctuation-free mode. If the liquid level data drops sharply within a short period of time, it is determined to be a transient change mode.

7. The method according to claim 2, characterized in that, For the preprocessed monitoring data, time-series analysis is performed within a preset sliding time window to extract dynamic characteristic indicators that characterize its short-term fluctuation patterns, including: Analyze the signal strength or spectral characteristics of the vibration data during the corresponding time periods when the pressure data exhibits a periodic fluctuation pattern; The fluctuation pattern of the vibration data is determined based on the analysis results of the signal strength or spectral characteristics.

8. The method according to claim 7, characterized in that, The fluctuation pattern of vibration data is determined based on signal strength or spectral characteristics, including: If the vibration data synchronously exhibits periodic signal intensity peaks during the periodic fluctuation of pressure, it is determined to be a periodic fluctuation pattern. If the signal strength of the vibration data remains below the third preset threshold, it is determined to be a stable, fluctuation-free mode. If the vibration data exhibits characteristics of high frequency and large amplitude, it is determined to be a transient change mode.

9. The method according to claim 1, characterized in that, The preset mapping relationship is as follows: When the fluctuation pattern represented by the feature information is a periodic fluctuation pattern, it is mapped to a normal breathing state; When the fluctuation pattern represented by the feature information is a stable, non-fluctuating pattern, it is mapped to the state of disappearance of respiration; When the fluctuation pattern represented by the feature information is a transient mutation pattern, it is mapped to a rapid oil leak state.

10. The method according to claim 9, characterized in that, Based on the feature information and the preset mapping relationship between transformer breather operating states, the operating state of the transformer breather is determined, including: When making judgments based on the characteristic information of a single type of monitoring data, the corresponding working status is output directly according to the fluctuation pattern and its mapping relationship represented by the characteristic information.

11. The method according to claim 9, characterized in that, Based on the feature information and the preset mapping relationship between transformer breather operating states, the operating state of the transformer breather is determined, including: When a fusion judgment is made based on the feature information of at least two types of monitoring data, the consistency of the fluctuation pattern represented by each feature information is checked. If the check passes, the final judgment is made based on the working state mapped by the fluctuation pattern.

12. The method according to claim 11, characterized in that, The consistency check includes a confidence fusion check: Assign a confidence weight to the feature information of each type of monitoring data; Calculate the overall confidence level based on the aforementioned feature information and its confidence weights; If the overall confidence level exceeds the preset confidence level threshold, the verification is deemed successful.

13. A transformer breather condition monitoring system, used to implement the method according to any one of claims 1 to 12, characterized in that, The system includes: The data acquisition module is configured to collect monitoring data related to the transformer breather. The data processing module is communicatively connected to the data acquisition module and is configured to preprocess and extract features from the monitoring data to obtain feature information, and determine the working status according to a preset mapping relationship. A data communication module, connected to the data processing module, is configured to upload the judgment result and / or the collected monitoring data to the remote monitoring platform.

14. The system according to claim 13, characterized in that: The data acquisition module includes at least one of a pressure sensor, a liquid level sensor, and a vibration sensor; The pressure sensor is installed at the connection flange of the transformer breather. The liquid level sensor is installed inside the oil cup of the respirator; The vibration sensor is located inside the oil cup or on the respirator body.

Images