A pointer type oil level gauge for monitoring the oil level of a transformer oil storage cabinet

Abstract

The present invention relates to the technical field of oil level gauges, and discloses a pointer-type oil level gauge for monitoring the oil level of a transformer oil storage cabinet, comprising: an oil level gauge assembly and a control module, wherein the control module is electrically connected to the oil level gauge assembly and is used to control the oil level gauge assembly; the oil level gauge assembly comprises: a dial, a fixed flange, a limit rod, an angle sensor, a limit block, a rotating rod, and a float, wherein the upper end of the fixed flange is fixedly connected to the dial, the dial is electrically connected to the angle sensor, the lower end of the fixed flange is fixedly connected to the limit rod, the end of the limit rod away from the fixed flange is fixedly connected to the limit block, two limit rods are provided, a rotating rod is provided between the two limit rods, and the two ends of the rotating rod are respectively rotatably connected to the dial and the limit block, the surface of the rotating rod is axially provided with a thread, the float is sleeved on the rotating rod, and the two sides of the float are slidably connected to the two limit rods. The present application reduces deformation or fatigue damage caused by long-term stress.

Description

Technical Field

[0001] The present invention relates to the technical field of oil level gauges, in particular to a pointer type oil level gauge used for monitoring the oil level of a transformer oil storage cabinet. Background Art

[0002] Transformers are core equipment in power systems. The insulating oil within them not only provides insulation and heat dissipation but also reflects the equipment's operating status through volume changes. The oil conservator (also known as the oil pillow), a crucial component of transformers, maintains stable pressure within the tank by compensating for expansion and contraction of the insulating oil due to temperature fluctuations. Therefore, real-time monitoring of the oil in the conservator, particularly the oil level, is crucial for preventing transformer failures and ensuring the safe operation of power systems. If the oil level is too low, air or moisture can intrude, leading to insulation degradation, partial discharge, and even short circuits. If the oil level is too high, leakage or bursting can occur.

[0003] However, after long-term use, the existing oil pillow oil level gauge may experience various problems such as the float getting stuck and the telescopic rod breaking, which may lead to the "false oil level" phenomenon. However, at this time, the actual oil level may have fallen below the safety threshold, but the oil level alarm cannot be triggered, which may lead to insulation exposure and cause flashover accidents, or cause local overheating inside the transformer, accelerating the deterioration of the insulating oil. In addition, the false oil level will mask the actual oil volume change trend, causing the operation and maintenance personnel to miss the best time to refill oil.

[0004] Therefore, it is necessary to design a pointer type oil level gauge for monitoring the oil level of a transformer oil conservator to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes a pointer-type oil level gauge for monitoring the oil level of a transformer oil conservator, aiming to solve the problem of low detection accuracy of current oil level gauges.

[0006] The present invention proposes a pointer-type oil level gauge for monitoring the oil level of a transformer oil conservator, comprising:

[0007] The oil level gauge assembly and the control module are electrically connected to the oil level gauge assembly. The control module is used to control the oil level gauge assembly. The oil level gauge assembly includes:

[0008] A dial, a fixed flange, a limit rod, an angle sensor, a limit block, a rotating rod and a float, the upper end of the fixed flange is fixedly connected to the dial, the angle sensor is electrically connected to the dial, the lower end of the fixed flange is fixedly connected to the limit rod, the end of the limit rod away from the fixed flange is fixedly connected to the limit block, two limit rods are provided, the rotating rod is provided between the two limit rods, and the two ends of the rotating rod are respectively rotatably connected to the dial and the limit block, the surface of the rotating rod is axially provided with a thread, the float is sleeved on the rotating rod, and the two sides of the float are slidably connected to the two limit rods.

[0009] Furthermore, the control module includes:

[0010] an acquisition unit configured to acquire sensor data, store the sensor data at fixed time intervals, and perform denoising, the acquisition unit further configured to align timestamps of the sensor data and process missing values ​​using KNN interpolation;

[0011] a judgment unit configured to extract the sensor data and input the sensor data into a pre-trained state model for comparison, and the judgment unit is further configured to determine the current operating state of the oil level gauge assembly based on the comparison result of the state model;

[0012] The early warning unit is configured to determine an abnormal area based on the abnormal state and send corresponding early warning information based on each abnormal area when the current operating state of the oil level gauge component is abnormal.

[0013] Furthermore, when the sensor data is input into a pre-trained state model for comparison, the following steps are included:

[0014] Acquire two sets of time series of the sensor data, wherein one of the time series includes a time series waveform of the oil level value and a time series waveform of the oil temperature data, and the second time series includes an effective value of the load current, external environment temperature and humidity, and vibration intensity of the oil conservator;

[0015] In the underlying feature learning stage, the time series waveforms of the two groups of time series are spliced ​​into oil level-oil temperature joint features and auxiliary features, and enhanced time series features are formed based on the oil level-oil temperature joint features and the auxiliary features;

[0016] In the high-level feature learning stage, a dense network with a two-layer structure is used to extract high-order nonlinear features layer by layer based on the enhanced temporal features to form the state model;

[0017] Inputting the sensor data into the state model, performing waveform reconstruction and obtaining an oil level prediction value;

[0018] A reconstruction error value and a prediction deviation value of the state model are determined based on the oil level prediction value.

[0019] Furthermore, when determining the reconstruction error value and the prediction deviation value of the state model based on the oil level prediction value, the method includes:

[0020] determining the reconstruction error value based on a time series waveform of the oil level value, and performing temperature coefficient compensation on the reconstruction error value when the external ambient temperature fluctuates;

[0021] determining a reconstruction error threshold based on historical normal oil level values;

[0022] Acquiring actual data of the oil level value, and determining the prediction deviation value based on the oil level prediction value;

[0023] The prediction deviation values ​​are arranged in ascending order, and the 95% quantile is taken as the prediction deviation threshold.

[0024] Furthermore, when determining the current operating state of the oil level gauge assembly based on the comparison result of the state model, the method includes:

[0025] Determine anomaly scores based on prediction deviation threshold and reconstruction error threshold:

[0026] ;

[0027] in, For abnormality score, and is the weight coefficient, and + =1, is the reconstruction error value, is the reconstruction error threshold, is the prediction deviation value, is the threshold of the prediction deviation value.

[0028] Furthermore, when the current operating state of the oil level gauge component is abnormal, the method includes:

[0029] When the abnormality score is less than or equal to 1, it is determined that the oil level gauge assembly is normal;

[0030] When the abnormality score is greater than 1, the oil level gauge component is determined to be abnormal, and a fault feature analysis is triggered;

[0031] The fault feature analysis is divided into false oil level, oil leakage and sensor failure;

[0032] when When the value is greater than 0.7, the reconstruction error value is greater than twice the reconstruction error threshold, and the prediction deviation value is less than the prediction deviation threshold, it is determined to be a false oil level fault;

[0033] when When the value is greater than 0.7, the reconstruction error value is less than twice the reconstruction error threshold, and the prediction deviation value is greater than twice the prediction deviation threshold, it is determined that the sensor is faulty;

[0034] when When the value of the reconstruction error is greater than 0.7, the reconstruction error value is less than twice the reconstruction error threshold, and the prediction deviation value is greater than twice the prediction deviation threshold, it is determined to be an oil leakage fault.

[0035] Furthermore, when the current operating state of the oil level gauge assembly is abnormal, the method further includes:

[0036] When the fault feature analysis determines that it is a false oil level fault, but the oil level value changes suddenly, it is preferentially determined to be a sensor fault;

[0037] When the fault characteristic analysis determines that it is a false oil level fault, but the effective value of the load current increases suddenly and the vibration intensity of the oil conservator changes synchronously, it is preferentially determined to be vibration interference.

[0038] Furthermore, when determining abnormal areas based on the abnormal state and sending corresponding warning information based on each abnormal area, it includes:

[0039] When the warning unit determines that the oil level is false, a first-level warning message is generated;

[0040] When the early warning unit determines that there is an oil leak, a second-level early warning message is generated;

[0041] When the warning unit determines that the sensor is faulty, it generates a third-level warning message.

[0042] Furthermore, when the early warning unit determines that the current operating state is abnormal, the abnormal information is stored in the history database and the fault type is marked, and the state model is trained and the threshold is updated at fixed time intervals.

[0043] Furthermore, it also includes: a remote monitoring unit, which is electrically connected to the early warning unit, and the remote monitoring unit is configured to monitor the operating status of the early warning unit and send operating status monitoring information to the Internet and a cloud data platform.

[0044] Compared with the prior art, the beneficial effects of the present invention are: through the rigid connection between the fixed flange and the limit rod, the mechanical strength of the overall structure is enhanced, the impact of external vibration or pressure on key components can be dispersed, and deformation or fatigue damage caused by long-term stress can be reduced. The sliding setting of the float along the limit rod replaces the free floating mode of the traditional float, and the movement trajectory of the float is constrained by physical guidance, which avoids the jamming problem caused by mechanical loosening or offset. The double-sided support of the limit rod further improves the stability of the float movement, ensuring that it can still maintain a linear response when the oil density changes or the temperature fluctuates, thereby improving the reliability of long-term operation. In addition, the cooperation between the rotating rod and the limit rod improves the force transmission path and reduces the risk of wear due to local stress concentration. The direct linkage between the angle sensor and the rotating rod eliminates the indirect transmission in traditional indirect transmission (such as gears and chains). The system eliminates gap errors and can capture minute oil level changes in real time. The linear motion of the threaded rotating rod converts the vertical displacement of the float into a rotational angle signal, avoiding nonlinear errors caused by gear meshing or lever swing. This not only improves oil level detection sensitivity but also shortens signal transmission response time, enabling the control module to quickly reflect dynamic oil level changes. Furthermore, the axial threads on the rotating rod surface increase the contact area between the float and the rod, reducing motion resistance caused by oil viscosity or impurity accumulation, further improving measurement consistency. The physical constraint of the limit rod on the float's motion suppresses false operation caused by severe oil fluctuations or mechanical vibration, and filters high-frequency interference, ensuring that the float only moves when the oil level actually changes. The electrical connection between the control module and the angle sensor enables real-time acquisition and digital conversion of oil level data, providing a foundation for remote monitoring and automated analysis. The integrated control module can identify oil level trend characteristics, provide early warning of potential faults (such as leaks and jams), and reduce reliance on manual inspections. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present invention. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0046] Figure 1 A schematic structural diagram of a pointer-type oil level gauge for monitoring the oil level of a transformer oil conservator provided by an embodiment of the present invention.

[0047] Figure 2 This is a functional block diagram of a pointer-type oil level gauge for monitoring the oil level in a transformer oil conservator provided by an embodiment of the present invention.

[0048] Among them, 1. Oil level gauge assembly; 101. Dial; 102. Fixed flange; 103. Limit rod; 104. Angle sensor; 105. Rotating rod; 106. Float; 107. Limit block; 108. Thread. DETAILED DESCRIPTION

[0049] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that, unless there is a conflict, the embodiments of the present disclosure and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0050] In some embodiments of the present application, see Figure 1-2 As shown, a pointer type oil level gauge for monitoring the oil level of a transformer oil conservator includes:

[0051] The oil level gauge assembly 1 and the control module are electrically connected to the oil level gauge assembly and are used to control the oil level gauge assembly. The oil level gauge assembly includes:

[0052] The dial 101, the fixed flange 102, the limit rod 103, the angle sensor 104, the limit block 107, the rotating rod 105 and the float 106. The upper end of the fixed flange 102 is fixedly connected to the dial 101, and the angle sensor 104 is electrically connected to the dial 101. The lower end of the fixed flange 102 is fixedly connected to the limit rod 103. The end of the limit rod 103 away from the fixed flange 102 is fixedly connected to the limit block 107. There are two limit rods 103. A rotating rod 105 is provided between the two limit rods 103, and the two ends of the rotating rod 105 are respectively rotatably connected to the dial 101 and the limit block 107. The surface of the rotating rod 105 is axially provided with a thread 108. The float 106 is sleeved on the rotating rod 105, and the two sides of the float 106 are slidably connected to the two limit rods 103.

[0053] Specifically, the dial 101 is arranged above the fixed flange 102, and two limit rods 103 are arranged below the fixed flange 102. A rotating rod 105 is arranged between the two limit rods 103, and the rotating rod 105 is rotatably connected to the dial 101, and the dial 101 is electrically connected to the angle sensor 104. When the oil level gauge assembly 1 is in operation, the float 106 is located on the upper surface of the cooling oil, and the float 106 floats, driving the rotating rod 105 to rotate, so that the dial 101 can display the oil level data at the current moment. At the same time, the angle sensor 104 can obtain the oil level data more accurately, providing a data basis for the control module. The control module obtains the oil temperature, oil level and other data of the oil level gauge to predict the future oil level change trend in advance. When it is predicted that the change exceeds the threshold, an early warning information can be generated to facilitate the staff to determine the abnormal area or determine the cause of the abnormal event, so that abnormal events can be dealt with more conveniently.

[0054] As can be appreciated, the rigid connection between the fixed flange 102 and the stop rod 103 enhances the structural stability of the oil level gauge assembly 1, dissipates stress caused by external mechanical vibration or oil fluctuations, and reduces deformation or wear of components due to long-term uneven stress. The guide arrangement of the float 106 sliding along the stop rod 103 replaces the traditional free-floating structure, eliminating the problem of float 106 jamming due to mechanical loosening or offset, thereby improving long-term operational reliability. The coordinated cooperation between the rotating rod 105 and the stop rod 103 further optimizes the motion trajectory of the float 106, ensuring a linear response even with changes in oil density or temperature fluctuations, and reducing measurement deviations caused by environmental factors. The direct linkage between the angle sensor 104 and the rotating rod 105 eliminates the backlash errors associated with traditional indirect transmissions (such as gears and chains), enabling real-time capture of subtle changes in the oil level. The linear motion mechanism of the threaded rotating rod 105 108 converts the vertical displacement of the float 106 into a precise rotational angle signal, avoiding nonlinear errors caused by gear meshing or lever swing. This not only improves the sensitivity of oil level detection, but also shortens the response time of signal transmission, allowing the control module to quickly reflect dynamic changes in the oil level. In addition, the contact between the float 106 and the axial thread 108 of the rotating rod 105 also increases the moving contact area, reduces the movement resistance caused by oil viscosity or impurity accumulation, and further improves measurement consistency. The physical constraint of the limit rod 103 on the movement of the float 106 suppresses malfunctions caused by violent oil fluctuations or mechanical vibrations. In scenarios with frequent external vibrations, this rigid guide structure can filter out high-frequency interference and ensure that the float 106 only moves when the oil level actually changes. The dial 101 is a sealed dial that can isolate external dust, oil and moisture, reduce the risk of contamination of the angle sensor 104, and extend the service life of key components. The control module can dynamically analyze the trend characteristics of oil level changes by collecting oil level data from the angle sensor 104 in real time and combining it with parameters such as oil temperature. Based on the predictive model, potential anomalies (such as leakage and jamming) can be identified in advance and early warning information can be generated, shortening the fault response time and reducing reliance on manual inspections. At the same time, by locating the cause of the anomaly (such as distinguishing between mechanical failure and sensor failure), the work complexity of operation and maintenance personnel is reduced.

[0055] In some embodiments of the present application, the control module includes:

[0056] The control module includes:

[0057] The acquisition unit is configured to acquire sensor data, store the sensor data at fixed time intervals, and perform denoising. The acquisition unit is also configured to align timestamps of the sensor data and use KNN interpolation to handle missing values.

[0058] The judgment unit is configured to extract sensor data and input the sensor data into a pre-trained state model for comparison. The judgment unit is also configured to determine the current operating state of the oil level gauge component based on the comparison result of the state model.

[0059] The early warning unit is configured to determine an abnormal area based on the abnormal state and send corresponding early warning information based on each abnormal area when the current operating state of the oil level gauge component is abnormal.

[0060] Specifically, the acquisition unit collects sensor data, including but not limited to oil level value data, oil temperature data, transformer load current effective value data, external environment temperature and humidity data, oil pillow (oil storage cabinet) vibration intensity, etc., and the state model is the LSTM-Autoencoder model, which can predict the short-term change trend of oil level. At the same time, it is more sensitive to latent faults such as slow leakage and false oil level, and can identify these latent faults. Through state model prediction, it can capture the complex time series relationship between oil level and oil temperature, adapt to nonlinear changes, and then improve detection reliability and reduce false alarm rate through the dual indicators of reconstruction error and prediction deviation. In actual use, the oil level will suddenly change, so KNN interpolation is used to process missing values ​​of sensor signals. At the same time, Kalman filtering combined with wavelet denoising can be used to adjust the filter parameters to suppress high-frequency noise and low-frequency drift, and for low-frequency parameters (such as external The sensor data is upsampled and aligned with the high-frequency oil level data, and the sensor data is stored in the historical database. The parameters of the model are regularly fine-tuned with new historical data to adapt to changes in data distribution. The judgment unit can identify abnormal situations by comparing the reconstructed oil level data. Since all the data used in state model training are operating under normal conditions, when an abnormal situation occurs, the data cannot match the state model, and reconstruction errors will occur, which can predict future trends and identify abnormal situations. Combined with the actual measured data, such as the actual oil level data, the actual oil temperature data, etc., the difference between the predicted situation and the actual situation can be seen. Combined with the output prediction value of the state model, the abnormal area or the cause of the abnormal situation can be determined more accurately, and then the warning information is sent through the warning unit to facilitate the staff to quickly find the abnormal situation and solve it.

[0061] As can be seen, the integration of multi-source sensor data acquisition and intelligent preprocessing improves data integrity and reliability. Taking into account the dynamic characteristics of parameters such as oil level, oil temperature, and load current, time alignment and interpolation techniques are used to address data loss caused by sensor communication delays or brief failures. Combined with an adaptive filtering algorithm, this algorithm suppresses interference from high-frequency noise and low-frequency drift on the original signal, reducing the risk of measurement distortion caused by environmental factors. Furthermore, upsampling of low-frequency parameters enables simultaneous analysis of multi-frequency data, enhancing the accuracy of cross-parameter correlation analysis and providing a consistent input foundation for subsequent model inference. Based on deep time series modeling technology, the model demonstrates the ability to characterize the dynamic changes in oil level and the coupled relationships between multiple parameters. By capturing the complex nonlinear relationships between oil level, oil temperature, load current, and other parameters, the model can identify hidden fault modes (such as slow leakage and float sticking) that are difficult to detect using traditional threshold methods. The dual-criteria design of reconstruction error and prediction bias reduces the probability of misjudgment caused by a single metric while enhancing the ability to distinguish between sudden anomalies and progressive faults. During model training, the system focuses on data from normal operating conditions, making it easier to detect pattern deviations under abnormal conditions, thereby increasing sensitivity to early fault detection. The continuous update mechanism of the historical database and the fine-tuning strategy for model parameters ensure the system's adaptability to long-term drift due to equipment aging, environmental changes, and other factors. Regularly incorporating new data to re-optimize the model avoids performance degradation caused by data distribution shifts. The threshold adjustment function further balances false positive and false negative rates, ensuring the system maintains stable detection performance across different seasons and operating phases, enhancing its sustainable service capabilities in complex industrial scenarios. The combination of anomaly detection results and multi-dimensional parameter correlation analysis supports root cause identification. By comparing the deviation characteristics of predicted trends with actual measurements, it can distinguish between different anomaly types, such as mechanical failure, sensor failure, or external interference, and generate targeted alerts. This targeted diagnostic capability shortens the troubleshooting process for maintenance personnel and reduces the secondary risks associated with blind repairs. Furthermore, the structured storage of historical data provides a data foundation for fault mode tracing and experience accumulation, helping to develop preventive maintenance strategies. To address common challenges in industrial environments, such as electromagnetic interference and mechanical vibration, the signal processing stage employs multi-stage filtering and redundant checking to minimize the impact of transient interference on detection results. The model design incorporates temporal dependency modeling to distinguish between actual oil level changes and short-term fluctuations caused by noise, preventing ineffective alarms triggered by occasional anomalies.

[0062] In some embodiments of the present application, when sensor data is input into a pre-trained state model for comparison, the process includes:

[0063] Two sets of time series of sensor data are obtained, one of which includes the time series waveform of the oil level value and the time series waveform of the oil temperature data, and the other time series includes the effective value of the load current, the external environment temperature and humidity, and the vibration intensity of the oil storage cabinet.

[0064] In the underlying feature learning stage, the time series waveforms of the two groups of time series are spliced ​​into oil level-oil temperature joint features and auxiliary features, and enhanced time series features are formed based on the oil level-oil temperature joint features and auxiliary features.

[0065] In the high-level feature learning stage, a dense network with a two-layer structure is used to extract high-order nonlinear features layer by layer based on enhanced temporal features to form a state model.

[0066] The sensor data is input into the state model to reconstruct the waveform and obtain the oil level prediction value.

[0067] The reconstruction error value and the prediction deviation value of the state model are determined based on the oil level prediction value.

[0068] Specifically, by obtaining oil level data, oil temperature data and auxiliary time series data, the auxiliary time series data includes but is not limited to the load current effective value: reflecting the load condition, used to analyze the operating status and load changes of the equipment, ambient temperature: monitoring the ambient temperature around the oil storage cabinet, used to evaluate the impact of temperature on the operation of the equipment, ambient humidity: monitoring the ambient humidity around the oil storage cabinet, used to evaluate the impact of humidity on the insulation performance of the equipment, power factor: reflecting the power utilization efficiency of the equipment, used to analyze the operating efficiency and energy consumption of the equipment, active power and reactive power: used to evaluate the actual power output and energy consumption of the equipment, equipment temperature: monitoring the oil temperature, used to evaluate the operating status and overheating risk of the equipment, vibration data: monitoring the vibration of the equipment, used to judge the mechanical state and potential failure of the equipment. The state model is an LSTM-Autoencoder model. The encoder extracts time series features by inputting multivariate time series data (such as oil level, oil temperature, load current, etc.) within the time window, compresses the potential space representation, and then predicts the oil level value of the future time step through the decoder and calculates the reconstruction error. In the underlying feature learning stage, the original waveform is input to obtain a low-dimensional time series feature vector, and the auxiliary data encoding is processed by the fully connected encoder by inputting load current, temperature, humidity and other data. Finally, the oil level-oil temperature features and auxiliary features are spliced ​​together to obtain enhanced time series features. In the high-level feature learning, dense network training is performed. Finally, by inputting new data that has not participated in the training, the reconstruction error value and prediction deviation value can be obtained, that is, the reconstruction error value:

[0069] ,in, is the reconstruction error value, is the actual oil level value at time t (unit %), is the oil level value reconstructed by the model (unit %), T is the time series window length, and the prediction deviation value: ,in, is the prediction deviation value, is the number of prediction time steps, is the actual oil level value at the kth time point in the future, The oil level value at the kth time point in the future predicted by the model.

[0070] In some embodiments of the present application, determining a reconstruction error value and a prediction deviation value of a state model based on an oil level prediction value includes:

[0071] A reconstruction error value is determined based on a time series waveform of the oil level value, and when the external ambient temperature fluctuates, a temperature coefficient compensation is performed on the reconstruction error value.

[0072] A reconstruction error threshold is determined based on historical normal oil level values.

[0073] Actual data of the oil level value is obtained, and a prediction deviation value is determined based on the predicted oil level value.

[0074] The prediction deviation values ​​are sorted in ascending order, and the 95% quantile is taken as the prediction deviation threshold.

[0075] Specifically, temperature coefficient compensation: ,in The ambient temperature changes. If the ambient temperature fluctuates (e.g., the temperature change is greater than 15°C within 24 hours), the reconstruction error is multiplied by the temperature compensation coefficient. , the reconstruction error threshold is determined by using the data of the trained state model: ,in, is the quantile of the dataset used to train the state model, is the set of reconstruction errors of the dataset used to train the state model, is the reconstruction error threshold, for all prediction errors Arrange them in ascending order and take the 95% quantile as the prediction deviation threshold.

[0076] In some embodiments of the present application, determining the current operating state of the oil level gauge assembly based on the comparison result of the state model includes:

[0077] Determine anomaly scores based on prediction deviation threshold and reconstruction error threshold:

[0078] .

[0079] in, For abnormality score, and is the weight coefficient, and + =1, is the reconstruction error value, is the reconstruction error threshold, is the prediction deviation value, is the threshold of the prediction deviation value.

[0080] In some embodiments of the present application, when the current operating state of the oil level gauge assembly is abnormal, it includes:

[0081] When the abnormality score is less than or equal to 1, the oil level gauge component is determined to be normal.

[0082] When the abnormality score is greater than 1, the oil level gauge component is determined to be abnormal and the fault feature analysis is triggered.

[0083] Fault signature analysis is divided into false oil level, oil leakage and sensor failure.

[0084] when When the value is greater than 0.7, the reconstruction error value is greater than twice the reconstruction error threshold, and the prediction deviation value is less than the prediction deviation threshold, it is determined to be a false oil level fault.

[0085] when When the value is greater than 0.7, the reconstruction error value is less than twice the reconstruction error threshold, and the prediction deviation value is greater than twice the prediction deviation threshold, it is determined to be a sensor failure.

[0086] when When the value is greater than 0.7, the reconstruction error value is less than twice the reconstruction error threshold, and the prediction deviation value is greater than twice the prediction deviation threshold, it is determined to be an oil leakage fault.

[0087] It's understandable that integrating the dual criteria of reconstruction error and prediction bias allows for more comprehensive detection of abnormal signals in oil level monitoring. Reconstruction error emphasizes sensitivity to deviations from current data patterns, while prediction bias enhances early warning capabilities for future trend anomalies. The weighted combination of the two reduces the risk of misjudgment that could result from a single metric. For example, relying solely on reconstruction error can miss early signs of slow leaks, while relying solely on prediction bias can overreact to transient noise. This synergistic effect improves the coverage and reliability of anomaly detection, ensuring accurate capture of complex fault modes such as intermittent jams or gradual oil leaks. The decision logic, based on a combination of anomaly score component proportions and thresholds, enhances the accuracy of root cause diagnosis. By setting differentiated thresholds for the weight coefficients (e.g., false oil levels prioritize reconstruction error over prediction bias, while leaks prioritize prediction bias over mechanical faults, sensor failures, and true oil level anomalies), it is possible to distinguish between mechanical faults, sensor failures, and true oil level anomalies. For example, the combination of excessive reconstruction error and relatively stable prediction bias in false oil level determination can eliminate interference from sensor noise or environmental interference. In oil leak detection, persistently exceeding prediction error limits and maintaining a controllable reconstruction error range reflect the real risk of a persistent oil level drop. This refined classification reduces the possibility of fault confusion and provides a clear basis for targeted maintenance. Multi-level thresholds (such as twice the error threshold) are employed to adapt to fault scenarios of varying severity. Significantly exceeding the limits for reconstruction error or prediction error (e.g., exceeding twice the threshold) allows for rapid identification of high-risk faults (such as sudden sensor failure or severe oil leaks), shortening response time to critical faults. For minor exceeding limits, a comprehensive scoring mechanism, combined with other parameters, provides further verification to avoid over-response to occasional fluctuations, thereby balancing detection sensitivity and specificity and improving stability under complex operating conditions. The direct mapping between anomaly scores and fault types simplifies the diagnostic process for maintenance personnel. By clearly distinguishing between false oil levels, oil leaks, and sensor failures, targeted alerts can be generated, guiding maintenance personnel to prioritize high-probability fault points. For example, a false oil level alert can directly prompt inspection of the float mechanism or transmission components, while a sensor failure alert prioritizes verification of the sensor circuitry or communication link. Targeted guidance reduces the time lost in blind troubleshooting and reduces unnecessary equipment downtime caused by misjudgment.

[0088] In some embodiments of the present application, when the current operating state of the oil level gauge assembly is abnormal, the method further includes:

[0089] When the fault feature analysis determines that it is a false oil level fault, but the oil level value changes suddenly, it is preferentially determined to be a sensor fault.

[0090] When the fault feature analysis determines that it is a false oil level fault, but the load current effective value suddenly increases and the vibration intensity of the oil conservator changes synchronously, it is preferentially determined to be vibration interference.

[0091] In some embodiments of the present application, determining abnormal areas based on abnormal conditions and sending corresponding warning information based on each abnormal area includes:

[0092] When the early warning unit determines that the oil level is false, a first-level early warning message is generated.

[0093] When the early warning unit determines that there is an oil leak, a second-level early warning message is generated.

[0094] When the warning unit determines that the sensor is faulty, it generates a level 3 warning message.

[0095] As can be seen, the introduction of a dynamic priority determination mechanism enables more accurate differentiation of similar fault scenarios, avoiding diagnostic bias caused by misjudgment based on a single characteristic. For example, when a false oil level fault and a sudden oil level change occur simultaneously, the sensor fault is prioritized, combining the transient nature of the sudden oil level change signal with the persistent nature of the mechanical jam, thereby eliminating the influence of the mechanical fault. This multi-parameter cross-validation mechanism enhances fault location reliability, reduces reliance on a single data source, and reduces the risk of misjudgment due to localized anomalies. In scenarios where a sudden increase in load current and simultaneous changes in vibration occur, mechanical vibration interference is prioritized over false oil level, improving the ability to distinguish complex anomalies. By correlating the coordinated changes in electrical parameters (load current) and mechanical parameters (vibration intensity), indirect effects of external environmental factors (such as transformer overload) on oil level data can be identified, avoiding the misattribution of oil level fluctuations caused by mechanical vibration to a stuck float or sensor failure. Multimodal data analysis improves robustness in complex industrial environments. A graded warning mechanism (levels 1 to 3) enables differentiated alert notification based on fault type and severity. For example, a level one warning (false oil level) prompts a mechanical component inspection, a level two warning (oil leak) triggers an emergency shutdown, and a level three warning (sensor failure) recommends circuit maintenance. If the warning unit identifies vibration interference, it re-compares the current data and re-evaluates based on the new state model. If vibration interference is still detected, no alarm is issued. If another anomaly is detected, an alarm is issued based on the other anomaly information. This hierarchical strategy helps maintenance personnel quickly prioritize and shorten response times for critical faults (such as oil leaks), while avoiding overreaction to low-risk events and optimizing human resource allocation and maintenance costs. Prioritization rules mitigate the limitations of traditional static threshold methods. For example, in the case of sudden oil level changes, dynamic switching of the judgment logic prioritizes transient sensor failures rather than directly assigning them to mechanical anomalies. This reduces over-reliance on preset thresholds, adapts to complex scenarios with sudden interference or transient signal anomalies, and reduces false alarms caused by sporadic events.

[0096] In some embodiments of the present application, when the early warning unit determines that the current operating state is abnormal, the abnormal information is stored in the historical database and the fault type is marked, and the state model is trained and the threshold is updated at fixed time intervals.

[0097] As can be understood, by storing anomaly information in a historical database and annotating it with fault types, we can continuously accumulate multi-dimensional operational data and fault cases. This transforms historical data from isolated event records into a traceable and analyzable knowledge base. By annotating fault types, the data possesses clear semantic information, providing a structured foundation for subsequent root cause analysis and pattern mining, enhancing the reuse of fault cases and reducing the risk of analytical gaps caused by reliance on experience or personnel turnover. Regularly training the state model at fixed intervals ensures that the model can dynamically update with changes in equipment status or the emergence of new fault modes. By incorporating the latest anomaly data, the model continuously learns new feature distributions and fault patterns, avoiding performance degradation caused by equipment aging, environmental changes, or operating condition adjustments. This improves the model's generalization to new scenarios and reduces the probability of misjudgment caused by data distribution shifts in traditional static models. Furthermore, the incremental training strategy reduces reliance on the entire historical data set and optimizes computing resource utilization. Based on the updated model and historical anomaly data, the thresholds for reconstruction error and prediction deviation can be dynamically adjusted, overcoming the potential for insensitivity or overreaction with traditional fixed thresholds over long-term operation. For example, after a device enters a stable aging phase, the threshold can be moderately relaxed to avoid excessive alerts. In high-precision monitoring scenarios, the threshold can be tightened to catch early signs of anomalies. This flexibility improves adaptability to different lifecycle stages or external environmental changes, reducing the burden of frequent manual calibration. Instead, thresholds are updated based on new data: ,in, is the forgetting factor (usually 0.9), The threshold is calculated for the most recent data, is the old threshold, is the updated threshold.

[0098] In some embodiments of the present application, it also includes: a remote monitoring unit, which is electrically connected to the early warning unit, and the remote monitoring unit is configured to monitor the operating status of the early warning unit and send the operating status monitoring information to the Internet and the cloud data platform.

[0099] It is understandable that by connecting the remote monitoring unit with the early warning unit, real-time tracking and cloud synchronization of oil level monitoring status are achieved. This allows operation and maintenance personnel to obtain equipment operating status without geographical restrictions and promptly grasp the occurrence of abnormal situations. The instant transmission of monitoring information ensures the timeliness of fault response, shortens the time window from abnormality occurrence to processing intervention, and reduces the risk of fault expansion due to information delays. At the same time, access to the cloud platform facilitates multi-terminal access, supports multi-channel monitoring such as mobile terminals and PC terminals, and enhances the flexibility of emergency response. The introduction of the cloud data platform enables the centralized storage and management of dispersed monitoring data. The centralized processing model breaks the space limitations of traditional local storage and provides a foundation for the long-term preservation and rapid retrieval of massive historical data. Through cloud-based data aggregation, horizontal comparative analysis across devices and regions can be achieved to identify potential regional risks or common failure modes.

[0100] In summary, the beneficial effects of the present invention are: through the rigid connection between the fixed flange and the limit rod, the mechanical strength of the overall structure is enhanced, the impact of external vibration or pressure on key components can be dispersed, and deformation or fatigue damage caused by long-term stress can be reduced. The sliding setting of the float along the limit rod replaces the free floating mode of the traditional float, and the movement trajectory of the float is constrained by physical guidance, which avoids the jamming problem caused by mechanical looseness or offset. The double-sided support of the limit rod further improves the stability of the float movement, ensuring that it can still maintain a linear response when the oil density changes or the temperature fluctuates, thereby improving the reliability of long-term operation. In addition, the cooperation between the rotating rod and the limit rod improves the force transmission path and reduces the risk of wear due to local stress concentration. The direct linkage between the angle sensor and the rotating rod eliminates the gap error in traditional indirect transmission (such as gears and chains). The linear motion of the threaded rotating rod converts the vertical displacement of the float into a rotational angle signal, avoiding nonlinear errors caused by gear meshing or lever swing. This not only improves the sensitivity of oil level detection but also shortens the response time of signal transmission, enabling the control module to quickly reflect dynamic oil level changes. Furthermore, the axial threads on the rotating rod surface increase the contact area between the float and the rod, reducing the motion resistance caused by oil viscosity or impurity accumulation, further improving measurement consistency. The physical constraint of the limit rod on the float's motion suppresses false operation caused by severe oil fluctuations or mechanical vibration, and filters high-frequency interference, ensuring that the float only moves when the oil level actually changes. The electrical connection between the control module and the angle sensor enables real-time acquisition and digital conversion of oil level data, providing a foundation for remote monitoring and automated analysis. The integrated control module can identify trend characteristics of oil level changes, provide early warning of potential faults (such as leaks and jams), and reduce reliance on manual inspections.

[0101] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0102] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0103] These computer program instructions may also be stored in a computer-readable storage device that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable storage device produce an article of manufacture comprising an instruction device that implements the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0104] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A pointer type oil level gauge for monitoring the oil level of a transformer oil conservator, characterized in that: include: An oil level gauge assembly and a control module, wherein the control module is electrically connected to the oil level gauge assembly and is used to control the oil level gauge assembly; The oil level gauge assembly includes: A dial, a fixed flange, a limit rod, an angle sensor, a limit block, a rotating rod and a float, the upper end of the fixed flange is fixedly connected to the dial, the angle sensor is electrically connected to the dial, the lower end of the fixed flange is fixedly connected to the limit rod, the end of the limit rod away from the fixed flange is fixedly connected to the limit block, two limit rods are provided, the rotating rod is provided between the two limit rods, and the two ends of the rotating rod are respectively rotatably connected to the dial and the limit block, the surface of the rotating rod is axially provided with a thread, the float is sleeved on the rotating rod, and the two sides of the float are slidably connected to the two limit rods; The control module includes: an acquisition unit configured to acquire sensor data, store the sensor data at fixed time intervals, and perform denoising, the acquisition unit further configured to align timestamps of the sensor data and process missing values ​​using KNN interpolation; a judgment unit configured to extract the sensor data and input the sensor data into a pre-trained state model for comparison, and the judgment unit is further configured to determine the current operating state of the oil level gauge assembly based on the comparison result of the state model; an early warning unit configured to, when the current operating state of the oil level gauge assembly is abnormal, determine an abnormal area based on the abnormal state and send corresponding early warning information based on each abnormal area; When the sensor data is input into the pre-trained state model for comparison, it includes: Acquire two sets of time series of the sensor data, wherein one of the time series includes a time series waveform of the oil level value and a time series waveform of the oil temperature data, and the second time series includes an effective value of the load current, external environment temperature and humidity, and vibration intensity of the oil conservator; In the underlying feature learning stage, the time series waveforms of the two groups of time series are spliced ​​into oil level-oil temperature joint features and auxiliary features, and enhanced time series features are formed based on the oil level-oil temperature joint features and the auxiliary features; In the high-level feature learning stage, a dense network with a two-layer structure is used to extract high-order nonlinear features layer by layer based on the enhanced temporal features to form the state model; Inputting the sensor data into the state model, performing waveform reconstruction and obtaining an oil level prediction value; determining a reconstruction error value and a prediction deviation value of the state model based on the oil level prediction value; Determining the reconstruction error value and the prediction deviation value of the state model based on the oil level prediction value includes: determining the reconstruction error value based on a time series waveform of the oil level value, and performing temperature coefficient compensation on the reconstruction error value when the external ambient temperature fluctuates; determining a reconstruction error threshold based on historical normal oil level values; Acquiring actual data of the oil level value, and determining the prediction deviation value based on the oil level prediction value; The prediction deviation values ​​are arranged in ascending order, and the 95% quantile is taken as the prediction deviation threshold.

2. The pointer type oil level gauge for monitoring the oil level of a transformer oil conservator according to claim 1 is characterized in that: Determining the current operating state of the oil level gauge assembly based on the comparison result of the state model includes: Determine anomaly scores based on prediction deviation threshold and reconstruction error threshold: ； in, For abnormality score, and is the weight coefficient, and + =1, is the reconstruction error value, is the reconstruction error threshold, is the prediction deviation value, is the threshold of the prediction deviation value.

3. The pointer type oil level gauge for monitoring the oil level of a transformer oil conservator according to claim 2 is characterized in that: When the current operating state of the oil level gauge component is abnormal, it includes: When the abnormality score is less than or equal to 1, it is determined that the oil level gauge assembly is normal; When the abnormality score is greater than 1, the oil level gauge component is determined to be abnormal, and a fault feature analysis is triggered; The fault feature analysis is divided into false oil level, oil leakage and sensor failure; when When the value is greater than 0.7, the reconstruction error value is greater than twice the reconstruction error threshold, and the prediction deviation value is less than the prediction deviation threshold, it is determined to be a false oil level fault; when When the value is greater than 0.7, the reconstruction error value is less than twice the reconstruction error threshold, and the prediction deviation value is greater than twice the prediction deviation threshold, it is determined that the sensor is faulty; when When the value of the reconstruction error is greater than 0.7, the reconstruction error value is less than twice the reconstruction error threshold, and the prediction deviation value is greater than twice the prediction deviation threshold, it is determined to be an oil leakage fault.

4. The pointer type oil level gauge for monitoring the oil level of a transformer oil conservator according to claim 3 is characterized in that: When the current operating state of the oil level gauge component is abnormal, the method further includes: When the fault feature analysis determines that it is a false oil level fault, but the oil level value changes suddenly, it is preferentially determined to be a sensor fault; When the fault characteristic analysis determines that it is a false oil level fault, but the effective value of the load current increases suddenly and the vibration intensity of the oil conservator changes synchronously, it is preferentially determined to be vibration interference.

5. The pointer type oil level gauge for monitoring the oil level of a transformer oil conservator according to claim 4, characterized in that: When determining abnormal areas based on abnormal conditions and sending corresponding warning information based on each abnormal area, it includes: When the warning unit determines that the oil level is false, a first-level warning message is generated; When the early warning unit determines that there is an oil leak, a second-level early warning message is generated; When the warning unit determines that the sensor is faulty, it generates a third-level warning message.

6. The pointer type oil level gauge for monitoring the oil level of a transformer oil conservator according to claim 5, characterized in that: When the early warning unit determines that the current operating state is abnormal, the abnormal information is stored in the history database and the fault type is marked, and the state model is trained and the threshold is updated at fixed time intervals.

7. The pointer type oil level gauge for monitoring the oil level of a transformer oil conservator according to claim 6, characterized in that: Also includes: A remote monitoring unit is electrically connected to the early warning unit, and is configured to monitor the operating status of the early warning unit and send operating status monitoring information to the Internet and a cloud data platform.

Images