Fluorescent optical fiber temperature measurement method and system for transformer

Abstract

The invention discloses a fluorescent optical fiber temperature measurement method and system for a transformer, and relates to the technical field of optical fiber temperature measurement, and the method comprises the steps: calibrating fluorescent optical fibers at different preset temperatures, and arranging the calibrated fluorescent optical fibers at preset temperature measurement points of the transformer; acquiring a fluorescence signal and a fluorescence time sequence signal of the preset temperature measurement point by using the fluorescence optical fiber sensor, and performing attenuation characteristic analysis on the fluorescence time sequence signal to obtain an attenuation signal; performing signal interaction on the fluorescence signal, the fluorescence time sequence signal and the attenuation signal to obtain different types of fluorescence interaction signals; fusing the fluorescence interaction signals to obtain a fused attention signal; monitoring an abnormal temperature point in the temperature curve and an abnormal coordinate interval where the abnormal temperature point is located; and generating a temperature gradient distribution map based on the temperature signal of each preset temperature measurement point measured in real time. According to the invention, the temperature change can be accurately detected, the temperature change can be quickly responded, and the real-time performance and the accuracy are ensured.

Description

Technical Field

[0001] This invention relates to the field of fiber optic temperature measurement technology, and in particular to a fluorescent fiber optic temperature measurement method and system for transformers. Background Technology

[0002] During operation, the temperature of a transformer may change due to factors such as load and environmental conditions. Abnormal temperature fluctuations may be an early signal of equipment failure. In particular, overheating may cause the transformer's insulation materials to age, leading to serious equipment failure. Therefore, transformers need to have their temperature changes monitored in real time to ensure their stable and safe operation.

[0003] Currently, traditional temperature monitoring methods may not be able to provide high-sensitivity, high-precision real-time data, and it is difficult to maintain stable performance in variable working environments. Traditional sensor technologies, such as thermocouples or RTDs, can monitor temperature, but their response speed and sensitivity are relatively low, and they are easily affected by environmental interference, making it difficult to fully reflect the dynamic changes in temperature distribution. Summary of the Invention

[0004] Therefore, it is necessary to provide a fluorescent fiber optic temperature measurement method and system for transformers to address the aforementioned technical problems.

[0005] In a first aspect, the present invention provides a fluorescent fiber optic temperature measurement method for transformers, the method comprising: S1. Calibrate the fluorescent optical fiber at different preset temperatures, and place the calibrated fluorescent optical fiber at the preset temperature measurement point of the transformer. S2. Use a fluorescent fiber optic sensor to acquire the fluorescence signal and fluorescence time sequence signal at the preset temperature measurement point, and analyze the attenuation characteristics of the fluorescence time sequence signal to obtain the attenuation signal. S3. Perform signal interaction between fluorescence signal, fluorescence timing signal and attenuation signal to obtain different types of fluorescence interaction signals; S4. Fuse the fluorescence interaction signals to obtain the fused attention signal, and identify the temperature signal of the preset temperature measurement point based on the fused attention signal; S5. Extract topological features from the temperature signals at the preset temperature measurement points to obtain temperature curves. Analyze the dynamic correlation between the temperature curve coordinates and the internal radial position of the transformer using a pre-built circumferential positioning mapping model, and identify the spatial distribution characteristics of the transformer temperature by combining multi-physics field coupling. S6. Monitor abnormal temperature points in the temperature curve and the abnormal coordinate intervals where the abnormal temperature points are located, and mark the temperature abnormal areas corresponding to the abnormal coordinate intervals in combination with the spatial distribution characteristics. S7. Based on the temperature signals of each preset temperature measurement point measured in real time, generate a temperature gradient distribution map.

[0006] Furthermore, a fluorescence fiber optic sensor is used to acquire fluorescence signals and fluorescence time-series signals at preset temperature measurement points. The attenuation characteristics of the fluorescence time-series signals are then analyzed to obtain the attenuation signals, which include: S21. Integrate fluorescence signals from at least one consecutive time point into a fluorescence time-series signal, and simultaneously acquire transformer partial discharge signals and oil chromatography data; the partial discharge signal includes acoustic signals acquired by an ultrasonic sensor, and the oil chromatography data includes the concentrations of key characteristic components; S22. By analyzing the acoustic-optical correlation features, the attenuation dynamic features of the fluorescence timing signal and the timing correlation features of the partial discharge acoustic signal are extracted to eliminate the interference of electromagnetic pulses on the fluorescence timing signal. S23. Based on the factors affecting the attenuation of optical fiber aging and oil quality changes, a multi-factor attenuation characteristic correction model is constructed to systematically correct the deviation of the original fluorescence attenuation signal in the fluorescence time series signal. S24. Extract the core features of the decay characteristics of the fluorescence timing signal after interference removal and correction to obtain the decay signal; the core features include the decay time constant, half-life and steady-state residual amplitude.

[0007] Furthermore, based on the influencing factors of fiber optic aging and oil quality changes, a multi-factor attenuation characteristic correction model is constructed, including: S231. Conduct aging tests on fluorescent optical fibers, record the fluorescence decay time constant under different service durations, and establish an optical fiber aging compensation function based on the aging test monitoring data to characterize the quantitative relationship between the service duration of fluorescent optical fibers and the attenuation characteristic deviation. S232. Regularly collect transformer oil samples and test key oil quality parameters, synchronously record the attenuation signal of the corresponding time period, and construct an oil quality-attenuation correlation model based on the mapping relationship between the concentration of characteristic components of oil samples and the fluorescence attenuation rate. S233. Integrate the fiber optic aging compensation function with the oil quality-attenuation correlation model to form a multi-factor attenuation characteristic correction model.

[0008] Furthermore, by interacting the fluorescence signal, fluorescence timing signal, and attenuation signal, different types of fluorescence interaction signals are obtained, including: S31. Extract at least one fluorescence signal segment of the fluorescence time sequence signal according to the preset first window, and subtract it from the adjacent fluorescence signal segment to obtain the time sequence difference signal segment data. S32. Divide each fluorescence signal segment and time-difference signal segment into at least one signal block to obtain the fluorescence signal block and time-difference signal block of each fluorescence signal segment, and combine the fluorescence signal blocks and time-difference signal blocks of all fluorescence signal segments to form fluorescence signal segment data. S33. Normalize the attenuation signal to obtain attenuation signal data; S34. Input the fluorescence signal segment data into the first signal feature extraction model to obtain the fluorescence signal features; input the time-series differential signal segment data into the second signal characteristic extraction model to obtain the time-series signal features; input the attenuation signal data into the third signal feature extraction model to obtain the attenuation signal features. S35. Input the fluorescence signal features, time-series signal features and attenuation signal features into the signal interaction fusion model to obtain different types of fluorescence interaction signals; the fluorescence interaction signals include the first interaction signal, the second interaction signal and the third interaction signal.

[0009] Furthermore, the fluorescence signal segment data is input into the first signal feature extraction model to obtain fluorescence signal features including: S341. Extract the fluorescence signal block features of each fluorescence signal block in the fluorescence signal segment data using the first signal feature extraction model; S342. Calculate the average value of the fluorescence signal block characteristics of each fluorescence signal block as the fluorescence signal characteristics.

[0010] Furthermore, by inputting fluorescence signal features, time-series signal features, and attenuation signal features into the signal interaction fusion model, different types of fluorescence interaction signals are obtained, including: S351. Perform feature interaction between fluorescence signal features, time series signal features and attenuation signal features to obtain the first interaction signal; S352. Add the fluorescence signal features to the first interactive signal to obtain the first fluorescence fusion signal; add the timing signal features to the first interactive signal to obtain the first timing fusion signal; add the attenuation signal features to the first interactive signal to obtain the first attenuation fusion signal; perform feature interaction on the first fluorescence fusion signal, the first timing fusion signal, and the first attenuation fusion signal to obtain the second interactive signal; S353. Add the first fluorescence fusion signal to the second interaction signal to obtain the second fluorescence fusion signal; add the first time-series fusion signal to the second interaction signal to obtain the second time-series fusion signal; add the first attenuation fusion signal to the second interaction signal to obtain the second attenuation fusion signal; perform feature interaction on the second fluorescence fusion signal, the second time-series fusion signal and the second attenuation fusion signal to obtain the third interaction signal.

[0011] Furthermore, the various fluorescence interaction signals are fused to obtain a fused attention signal, and the temperature signal of the preset temperature measurement point is identified based on the fused attention signal, including: S41. Perform matrix transformations on the first interaction signal, the second interaction signal, and the third interaction signal to obtain the corresponding attention query vector, attention key vector, and attention value vector. S42. Perform self-attention mechanism processing based on the attention query vector, attention key vector, and attention value vector to obtain the fused attention signal; S43. Perform linear mapping on the fused attention signal to obtain the temperature feature sequence; S44. Perform feature matching in the temperature calibration database based on the temperature feature sequence to obtain the feature matching result, and identify the temperature signal of the preset temperature measurement point based on the feature matching result.

[0012] Furthermore, a pre-constructed circumferential positioning mapping model is used to analyze the dynamic correlation between the temperature curve coordinates and the internal radial position of the transformer, and multi-physics coupling is combined to identify the spatial distribution characteristics of the transformer temperature, including: S51. Based on the geometric structure of the transformer and the layout of the fluorescent sensor, a circumferential positioning mapping model of the transformer is established to map the real-time monitored temperature signal with the actual position inside the transformer, so as to establish the correlation between the temperature signal and the actual position inside the transformer. S52. By using the circumferential positioning mapping model to minimize the reconstruction error between the spatial coordinates of the measuring point and the temperature curve, the dynamic correlation between the horizontal coordinate of the temperature curve and the radial position inside the transformer is obtained. S53. By introducing leakage flux density and oil flow velocity as constraints, a set of partial differential equations coupled with electromagnetic, fluid and thermal fields is constructed. The spatial distribution characteristics of the set of partial differential equations are solved by finite element method to identify the spatial distribution of electromagnetic field, oil flow and temperature inside the transformer.

[0013] Furthermore, the abnormal temperature points in the monitoring temperature curve and the abnormal coordinate intervals where these abnormal temperature points are located include: S61. Slide the preset second sliding window along the temperature curve from one side to the other according to the preset step size; S62. Based on the preset abnormal temperature judgment conditions, set the temperature threshold of abnormal temperature, and identify whether there is an abnormal temperature point in the preset sliding window at each sliding position by comparing the threshold. S63. Record the corresponding horizontal coordinate interval within the preset sliding window for each abnormal temperature point as the abnormal coordinate interval.

[0014] Secondly, the present invention also provides a fluorescent fiber optic temperature measurement system for transformers, the system comprising: The fluorescence calibration module is used to calibrate the fluorescent optical fiber at different preset temperatures, and the calibrated fluorescent optical fiber is placed at the preset temperature measurement point of the transformer. The signal analysis module is used to acquire the fluorescence signal and fluorescence timing signal at the preset temperature measurement point using a fluorescence fiber optic sensor, and to analyze the attenuation characteristics of the fluorescence signal to obtain the attenuation signal. The signal interaction module is used to interact with fluorescence signals, fluorescence timing signals, and attenuation signals to obtain different types of fluorescence interaction signals. The fiber optic temperature measurement module is used to fuse various fluorescence interaction signals to obtain a fused attention signal, and to identify the temperature signal of a preset temperature measurement point based on the fused attention signal. The positioning and fitting module is used to extract topological features from the temperature signals of preset temperature measurement points to obtain temperature curves. It also uses a pre-built circumferential positioning mapping model to analyze the dynamic correlation between the temperature curve coordinates and the internal radial position of the transformer in order to identify the spatial distribution characteristics of the transformer temperature. The anomaly analysis module is used to monitor the abnormal coordinate intervals where abnormal temperature points are located in the temperature curve, and, in combination with spatial distribution characteristics, mark the temperature anomaly areas corresponding to the abnormal coordinate intervals. The temperature modeling module is used to generate a temperature gradient distribution map based on the temperature signals of each preset temperature measurement point measured in real time.

[0015] The beneficial effects of this invention are as follows: 1. The fluorescence fiber calibration unit is calibrated at different preset temperatures and the fluorescence attenuation characteristics are recorded to ensure that the system can provide accurate temperature data in different temperature ranges. The fluorescence fiber technology has high sensitivity and can achieve accurate detection of temperature changes. Furthermore, the fluorescence signal and fluorescence time sequence signal are acquired in real time through the signal analysis unit, and the temperature change information is obtained by analyzing the attenuation characteristics of the fluorescence signal. This enables the system to respond quickly to temperature changes and ensures real-time performance and accuracy.

[0016] 2. This invention has the ability to automatically identify and locate abnormal temperature points. By judging abnormal points in the temperature curve and combining factors such as heat flow and electromagnetic fields, it can accurately predict possible fault areas and issue alarms in a timely manner. This helps to prevent equipment from malfunctioning due to overheating, thereby improving the safety and stability of the equipment.

[0017] 3. This invention can collect and analyze multiple monitoring data of transformers. Through sound-light correlation, signal interaction and fusion models, the analysis of temperature anomalies is more refined and matched with the causal relationship of transformer faults, providing more accurate fault diagnosis. Furthermore, through real-time monitoring based on temperature gradient distribution maps, the diffusion trend of local overheating areas can be identified in a timely manner. Through causal inference model analysis, the root cause behind temperature anomalies can be determined, thereby providing effective fault warnings and maintenance suggestions. Combined with material failure prediction models, by analyzing historical data of temperature anomaly areas and insulation paper performance, the remaining service life of the transformer can be effectively assessed. By outputting maintenance priority suggestions, timely maintenance of the transformer can be ensured, extending its service life. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a flowchart of a fluorescent fiber optic temperature measurement method for a transformer according to an embodiment of the present invention; Figure 2 This is a system principle block diagram of a fluorescent fiber optic temperature measurement system for transformers according to an embodiment of the present invention.

[0019] The reference numerals are as follows: 1. Fluorescence calibration module; 2. Signal analysis module; 3. Signal interaction module; 4. Fiber optic temperature measurement module; 5. Positioning fitting module; 6. Anomaly analysis module; 7. Temperature modeling module. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0021] Please see Figure 1 A fluorescent fiber optic temperature measurement method for transformers is provided, the method comprising: S1. Calibrate the fluorescent optical fiber at different preset temperatures, and place the calibrated fluorescent optical fiber at the preset temperature measurement point of the transformer.

[0022] It should be noted that the fluorescent fiber was calibrated at different preset temperatures to determine its fluorescence attenuation characteristics; the fluorescence signal attenuation was recorded at different time periods by placing the fluorescent fiber under different temperature conditions.

[0023] Attenuation characteristics refer to the effect of temperature changes on fluorescence signals, i.e., how much the intensity of the fluorescence signal decreases with temperature changes. For example, in the laboratory, fluorescent optical fibers are exposed to different temperature environments, such as 30°C, 50°C, and 70°C, and the changes in fluorescence signals at each temperature are recorded. The calibrated fluorescent optical fiber sensor is installed at multiple preset temperature measurement points on the transformer to monitor the temperature in real time.

[0024] Fluorescent fiber optic sensors are installed in specific locations within a transformer to detect temperature changes during operation by collecting fluorescence signals from those locations; for example, fluorescent fiber optic sensors are installed in hot spots such as the transformer windings and core.

[0025] S2. Use a fluorescent fiber optic sensor to acquire the fluorescence signal and fluorescence timing signal at the preset temperature measurement point, and analyze the attenuation characteristics of the fluorescence timing signal to obtain the attenuation signal.

[0026] It should be noted that the signal data acquired from the fluorescence fiber optic sensor includes real-time fluorescence signals and time-varying fluorescence time-series signals; the fluorescence time-series signals contain fluorescence signals at continuous time points, which reflect the dynamic process of temperature change over time.

[0027] The sensor collects fluorescence signals caused by temperature changes in real time over a period of time, forming a signal sequence that changes over time. The fluorescence signal is analyzed to obtain its attenuation characteristics, and the attenuation signal is further extracted. The attenuation signal refers to the attenuation process of the fluorescence signal, which can be used to analyze the temperature response behavior of the fluorescent fiber. The degree of attenuation usually increases with the increase of temperature. During the process of temperature rise, the degree of attenuation of the fluorescence signal becomes greater, so the increase of temperature can be inferred.

[0028] In the description of this invention, a fluorescent fiber optic sensor is used to acquire the fluorescence signal and fluorescence timing signal at a preset temperature measurement point, and the attenuation characteristics of the fluorescence timing signal are analyzed to obtain the attenuation signal, which includes: S21. Integrate fluorescence signals from at least one continuous time point into a fluorescence time-series signal, and simultaneously acquire transformer partial discharge signals and oil chromatography data; the partial discharge signal includes acoustic signals acquired by an ultrasonic sensor, and the oil chromatography data includes the concentrations of key characteristic components.

[0029] Specifically, fluorescence sensors are typically used to monitor the temperature or gas concentration inside transformers; changes in fluorescence signals may be related to equipment aging, malfunctions, etc.; fluorescence signals are generally acquired through fiber optic sensors.

[0030] Partial discharge refers to the phenomenon of partial electrical discharge inside a transformer, which usually causes aging of the equipment's insulation materials; partial discharge signals are collected by ultrasonic sensors, which can capture the sound waves generated during the discharge process.

[0031] Oil chromatography analysis determines the quality of transformer oil and the health status of the transformer by analyzing the concentration changes of various chemical components in transformer oil, especially some key characteristic components (such as acidity, carbides, etc.). For example, when monitoring a high-voltage transformer, a fluorescence sensor can provide real-time temperature signals, an ultrasonic sensor can capture the sound waves generated during partial discharge, and an oil chromatography analyzer can monitor the chemical composition of the oil.

[0032] These data, working together, can help analyze whether there are potential faults in the transformer. Electromagnetic pulses (EMPs) are a common source of interference during the operation of electrical equipment, especially high-voltage equipment, and may interfere with fiber optic sensors, affecting the accuracy of fluorescence signals. To remove EMP interference from fluorescence signals, the validity of the signals can be determined by analyzing the correlation characteristics between acoustic and optical signals. For example, during partial discharge, the timing of acoustic and fluorescence signals has a certain correlation. By comparing the dynamic characteristics of these two signals, it is possible to determine which fluorescence signals are caused by EMP interference, and thus remove the interference.

[0033] S22. By analyzing the acoustic-optical correlation features, the attenuation dynamic features of the fluorescence timing signal and the timing correlation features of the partial discharge acoustic signal are extracted to eliminate the interference of electromagnetic pulses on the fluorescence timing signal.

[0034] Specifically, the dynamic characteristics of fluorescence signal decay (such as decay rate) and the temporal correlation characteristics of acoustic signals of partial discharge can be analyzed together to eliminate electromagnetic interference; for example, if electromagnetic pulse interference occurs when collecting fluorescence signals, the fluorescence signal may show abrupt changes or irregular fluctuations.

[0035] By simultaneously analyzing the partial discharge acoustic signal, a correlation can be observed between the timing of the acoustic signal and the attenuation characteristics of the fluorescence signal, thereby eliminating the influence of electromagnetic interference.

[0036] S23. Based on the factors affecting fiber optic aging and oil quality changes, a multi-factor attenuation characteristic correction model is constructed to systematically correct the deviation of the original fluorescence attenuation signal in the fluorescence time series signal.

[0037] Specifically, over time, optical fibers may age due to environmental or usage conditions, leading to signal attenuation. Aging optical fibers may cause inaccurate measurements when transmitting fluorescent signals; the quality of transformer oil may also affect signal propagation, and changes in oil quality (such as temperature, acidity, etc.) may cause variations in the degree of signal attenuation.

[0038] The multi-factor attenuation characteristic correction model establishes a correction function by taking into account the influence of fiber optic aging and oil quality changes, to compensate for signal attenuation caused by these factors.

[0039] The calibration model typically includes an optical fiber aging compensation function and an oil quality-attenuation correlation model. For example, if a transformer has been used for many years, the performance of the optical fiber may deteriorate, leading to fluorescence signal attenuation. If the quality of the transformer oil deteriorates and the acidity of the oil increases, the attenuation of the fluorescence signal may also be aggravated. In order to compensate for these attenuations, a calibration model can be established to adjust the signal and ensure that the measured fluorescence data is more accurate.

[0040] After interference in the fluorescence signal is removed and corrected by multi-factor attenuation characteristics, the obtained fluorescence signal will be more accurate. At this time, the core features of the attenuated signal can be further extracted. The attenuation time constant is a parameter that represents the signal attenuation rate and is usually used to describe the exponential attenuation characteristics in a physical process. The half-life refers to the time required for the signal intensity to decay to half of its initial value. The steady-state residual amplitude represents the amplitude that eventually stabilizes after the signal has decayed to a certain extent. For example, assuming that the fluorescence signal has undergone interference removal and correction, the remaining signal shows certain attenuation characteristics.

[0041] By analyzing the attenuation curve, features such as attenuation time constant and half-life can be extracted. For example, an attenuation time constant of 10 seconds means that the signal attenuates to about 37% (i.e., 1 / e times) within 10 seconds; a half-life of 5 seconds means that the signal strength drops to half of its original value.

[0042] In the description of this invention, a multi-factor attenuation characteristic correction model is constructed based on the attenuation influencing factors of optical fiber aging and oil quality changes, including: S231. Conduct aging tests on fluorescent optical fibers, record the fluorescence decay time constant under different usage durations, and establish an optical fiber aging compensation function based on the aging test monitoring data to characterize the quantitative relationship between the usage duration and attenuation characteristic deviation of fluorescent optical fibers.

[0043] Specifically, by long-term monitoring of the attenuation characteristics of fluorescent optical fibers, the attenuation time constant (i.e., fluorescence attenuation rate) under different usage durations is recorded. As the usage time of the optical fiber increases, the signal attenuation may accelerate due to aging. The monitoring data records the attenuation time constant of the fluorescence signal under different usage durations to obtain the attenuation trend of the optical fiber.

[0044] Based on the monitored data, a mathematical model (compensation function) is established. This function can quantify the deviation relationship between the fiber's usage time and its attenuation characteristics. For example, as the fiber ages, the attenuation time constant may become longer, and this deviation can be corrected by the compensation function. For example, if the fiber has been used for 5 years and the attenuation time constant becomes 1.5 times its original value, then based on this trend, a fiber aging compensation function can be constructed to help with attenuation correction in subsequent signal analysis.

[0045] S232. Periodically collect transformer oil samples and test key oil quality parameters, synchronously record the attenuation signal for the corresponding time period, and construct an oil quality-attenuation correlation model based on the mapping relationship between the concentration of characteristic components of the oil sample and the fluorescence attenuation rate.

[0046] Specifically, the quality of transformer oil directly affects signal attenuation; for example, the acidity, water content, and gas composition of the oil can all lead to changes in the fluorescence signal attenuation rate.

[0047] Regularly collect oil samples and test these key parameters, recording the fluorescence decay signal at the corresponding time period each time an oil sample is collected; this data will support subsequent data analysis and modeling; by analyzing the relationship between oil quality parameters and fluorescence decay signals, the specific impact of oil quality characteristics on fluorescence decay can be revealed; for example, in oil quality analysis, if the water content in the oil increases, it may lead to a faster fluorescence decay rate.

[0048] This change can be observed by synchronously recording fluorescence decay signals, and the relationship between oil quality and decay rate can be further analyzed. The oil quality-decay correlation model is used to reflect the relationship between the concentration of characteristic components in the oil (such as acidity, water content, dissolved gases, etc.) and fluorescence decay rate.

[0049] Data analysis can establish a mapping relationship to quantify the impact of oil quality changes on fluorescence decay rate; based on oil quality changes, the rate of fluorescence signal decay can be predicted, thereby making a more accurate assessment of the transformer's health status; for example, assuming that the water concentration in the oil is linearly related to the fluorescence decay rate, the signal decay rate can be predicted based on the water content, and the monitoring results can be adjusted accordingly.

[0050] S233. Integrate the fiber optic aging compensation function with the oil quality-attenuation correlation model to form a multi-factor attenuation characteristic correction model.

[0051] Specifically, the fiber aging compensation function and the oil quality-attenuation correlation model are combined to form a comprehensive correction model. This model can simultaneously correct the effects of fiber attenuation and oil quality changes on the fluorescence signal. Through this model, the original fluorescence attenuation signal is systematically corrected to ensure that the fluorescence signal more accurately reflects the health status of the transformer.

[0052] S24. Extract the core features of the decay characteristics of the fluorescence timing signal after interference removal and correction to obtain the decay signal. The core features include the decay time constant, half-life, and steady-state residual amplitude.

[0053] Specifically, by comprehensively considering both fiber optic aging and oil quality changes, fluorescence signal analysis can better reflect the actual operating state of the transformer. For example, at a certain moment, the aging effect of the fiber optic cable and the change in oil quality work together to affect the fluorescence attenuation signal. By using a multi-factor attenuation characteristic correction model, the effects of fiber optic attenuation and oil quality on the signal can be considered simultaneously, accurately correcting the attenuation signal, thereby improving the reliability of the signal and the accuracy of the monitoring results.

[0054] S3. Perform signal interaction between the fluorescence signal, fluorescence timing signal, and attenuation signal to obtain different types of fluorescence interaction signals.

[0055] It should be noted that different signals (fluorescence signal, fluorescence timing signal, and decay signal) are processed interactively to obtain multiple interactive signals. For example, the fluorescence signal intensity may be high at a certain moment, but its decay signal may be weak. By processing the interactive signals, this information can be effectively integrated to enhance the perception of temperature changes.

[0056] In the description of this invention, fluorescence signals, fluorescence timing signals, and attenuation signals are interacted to obtain different types of fluorescence interaction signals, including: S31. Extract at least one fluorescence signal segment of the fluorescence timing signal according to the preset first window, and subtract it from the adjacent fluorescence signal segment to obtain the timing difference signal segment data.

[0057] Specifically, a fixed-size sliding window is set to perform sliding extraction on the time series of fluorescence signals; each sliding window extracts a fluorescence signal segment; differential operations are performed on adjacent fluorescence signal segments to obtain time-series differential signal segment data; the purpose of differential operations is to highlight the changing parts of the signal, which helps to analyze the dynamic change characteristics of the signal.

[0058] S32. Divide each fluorescence signal segment and time-difference signal segment into at least one signal block to obtain the fluorescence signal block and time-difference signal block of each fluorescence signal segment, and combine the fluorescence signal blocks and time-difference signal blocks of all fluorescence signal segments to form fluorescence signal segment data.

[0059] Specifically, each fluorescence signal segment and time-difference signal segment is divided into several signal blocks. These signal blocks are used to further refine the signal data and facilitate subsequent feature extraction. Each of the divided fluorescence signal segments and time-difference signal segments contains multiple signal blocks. The signal blocks of all signal segments constitute the final fluorescence signal segment data.

[0060] S33. Normalize the attenuation signal to obtain attenuation signal data.

[0061] Specifically, normalization of attenuated signals aims to eliminate scale differences between different signals, making signal processing more comparable; the normalized signal data is called attenuated signal data.

[0062] S34. Input the fluorescence signal segment data into the first signal feature extraction model to obtain the fluorescence signal features. Input the time-series differential signal segment data into the second signal characteristic extraction model to obtain the time-series signal features. Input the attenuation signal data into the third signal feature extraction model to obtain the attenuation signal features.

[0063] Specifically, the fluorescence signal segment data is input into the first signal feature extraction model to extract the features of the fluorescence signal; the features include fluorescence intensity.

[0064] The time-series differential signal segment data is input into the second signal feature extraction model to extract the features of the time-series signal; the features of the time-series signal may reflect the dynamic laws of signal changes.

[0065] Inputting the attenuation signal data into the third signal feature extraction model to extract the features of the attenuation signal helps to understand the attenuation law of the fluorescence signal.

[0066] Features extracted from fluorescence signals, time-series signals, and decay signals are input into a signal interaction fusion model; the interaction fusion model can comprehensively consider the relationships and interactions between different signals, resulting in a more comprehensive signal understanding.

[0067] In the description of this invention, the fluorescence signal segment data is input into a first signal feature extraction model to obtain fluorescence signal features including: S341. Use the first signal feature extraction model to extract the fluorescence signal block features of each fluorescence signal block in the fluorescence signal segment data.

[0068] S342. Calculate the average value of the fluorescence signal block characteristics of each fluorescence signal block as the fluorescence signal characteristics.

[0069] S35. Input the fluorescence signal characteristics, time-series signal characteristics, and attenuation signal characteristics into the signal interaction fusion model to obtain different types of fluorescence interaction signals. The fluorescence interaction signals include the first interaction signal, the second interaction signal, and the third interaction signal.

[0070] In the description of this invention, fluorescence signal features, time-series signal features, and attenuation signal features are input into a signal interaction fusion model to obtain different types of fluorescence interaction signals, including: S351. Perform feature interaction between fluorescence signal features, time-series signal features and attenuation signal features to obtain the first interactive signal.

[0071] S352. Add the fluorescence signal features to the first interactive signal to obtain the first fluorescence fusion signal. Add the timing signal features to the first interactive signal to obtain the first timing fusion signal. Add the attenuation signal features to the first interactive signal to obtain the first attenuation fusion signal. Perform feature interaction on the first fluorescence fusion signal, the first timing fusion signal, and the first attenuation fusion signal to obtain the second interactive signal.

[0072] S353. Add the first fluorescence fusion signal to the second interactive signal to obtain the second fluorescence fusion signal. Add the first time-series fusion signal to the second interactive signal to obtain the second time-series fusion signal. Add the first attenuation fusion signal to the second interactive signal to obtain the second attenuation fusion signal. Perform feature interaction on the second fluorescence fusion signal, the second time-series fusion signal, and the second attenuation fusion signal to obtain the third interactive signal.

[0073] S4. Fuse the various fluorescence interaction signals to obtain the fused attention signal, and identify the temperature signal of the preset temperature measurement point based on the fused attention signal.

[0074] It should be noted that different interactive signals are fused to form a fused attention signal, thereby accurately determining the temperature signal.

[0075] Through signal fusion, the system can integrate the characteristics of various signals, enhance the accuracy of temperature signals, and use them for final temperature estimation. It can also weightedly fuse the interaction results of fluorescence signals, attenuation signals, and memory signals from different time periods to produce a more accurate temperature calculation result.

[0076] In the description of this invention, fusing the various fluorescence interaction signals to obtain a fused attention signal, and identifying the temperature signal of a preset temperature measurement point based on the fused attention signal, includes: S41. Perform matrix transformations on the first interaction signal, the second interaction signal, and the third interaction signal to obtain the corresponding attention query vector, attention key vector, and attention value vector.

[0077] Specifically, the first, second, and third interaction signals are transformed into matrix vectors to obtain the corresponding attention query vector, attention key vector, and attention value vector, respectively. These three vectors represent different feature information of the signals and can capture the correlation between signals through the attention mechanism.

[0078] S42. Perform self-attention mechanism processing based on the attention query vector, attention key vector, and attention value vector to obtain the fused attention signal.

[0079] Specifically, weights are obtained by calculating the similarity between the query vector (Q) and the key vector (K). A dot product operation is typically used to calculate the similarity between Q and K. The resulting attention score is then scaled. To avoid gradient explosion, a common practice is to divide by a constant (usually the square root of the key vector's dimension). A softmax operation is then performed on the resulting attention score to calculate the weight of each attention value. Softmax converts the weight values ​​into a probability distribution, ensuring that the sum of all weights is 1, thus obtaining the weight of each value vector. The weights determine how information from different signals is fused. The calculated attention weights are then used to perform a weighted summation of the attention value vector (V). This step weights the contributions of different signals according to their similarity and synthesizes them into a fused attention signal.

[0080] The final result of the above steps is the fused attention signal (often called the weighted value), which contains the most relevant information among the three interaction signals and reflects their mutual influence.

[0081] S43. Perform linear mapping on the fused attention signal to obtain the temperature feature sequence.

[0082] S44. Perform feature matching in the temperature calibration database based on the temperature feature sequence to obtain the feature matching result, and identify the temperature signal of the preset temperature measurement point based on the feature matching result.

[0083] S5. Extract topological features from the temperature signals at the preset temperature measurement points to obtain temperature curves. Analyze the dynamic correlation between the temperature curve coordinates and the internal radial position of the transformer using a pre-constructed circumferential positioning mapping model, and identify the spatial distribution characteristics of the transformer temperature by combining multi-physics field coupling.

[0084] It should be noted that multiple temperature measurement points may be set up inside the transformer, and sensors (such as fluorescent fiber optic sensors) are used to collect the temperature signal at each measurement point. Here, "topology feature extraction" refers to extracting data representing the temperature distribution characteristics from the measured temperature signals. This means processing the original temperature signal, such as filtering and smoothing, to extract a more representative temperature curve. The temperature curve refers to the temperature change trend of each temperature measurement point of the transformer over a certain period of time. The temperature curve reflects the temperature distribution inside the transformer and can provide basic data for subsequent analysis. For example, suppose multiple fluorescent fiber optic sensors are deployed inside the transformer to measure the temperature signals at different locations.

[0085] Through data processing, temperature curves are obtained at each location, which describe the temperature changes of various parts inside the transformer. The transformer has specific geometric structures, such as the oil tank, windings, and core, and the arrangement of fluorescent fiber optic sensors is closely related to these structures.

[0086] Furthermore, by understanding the transformer's geometry (e.g., the size, shape, and spacing of its parts) and the sensor layout (sensor positions, arrangement angles, etc.), a mapping model can be established to correlate temperature signals with actual locations inside the transformer. A circumferential positioning mapping model means mapping temperature signals to specific locations inside the transformer (e.g., radial positions from the outside in) through the sensor layout, thereby pinpointing the specific location of the temperature signal inside the transformer. For example, assuming sensors are uniformly distributed along the outer circumference of the transformer and the transformer's geometric dimensions are known, a model can be created based on this information to correlate the temperature curve at a specific sensor location with the temperature at a specific location inside the transformer.

[0087] In practical applications, the measurement points of the temperature curve cannot always completely cover the entire interior of the transformer. Therefore, it is necessary to use mathematical models to correlate these limited measurement points with the temperature distribution in other locations inside the transformer. In the description of this invention, the dynamic correlation between the temperature curve coordinates and the internal radial position of the transformer is analyzed using a pre-constructed circumferential positioning mapping model, and the spatial distribution characteristics of the transformer temperature are identified by combining multi-physics coupling, including: S51. Based on the geometric structure of the transformer and the layout of the fluorescent sensor, a circumferential positioning mapping model of the transformer is established to map the real-time monitored temperature signal with the actual position inside the transformer, so as to establish the correlation between the temperature signal and the actual position inside the transformer.

[0088] S52. By using the circumferential positioning mapping model to minimize the reconstruction error between the spatial coordinates of the measuring point and the temperature curve, the dynamic correlation between the horizontal coordinate of the temperature curve and the radial position inside the transformer is obtained.

[0089] Specifically, minimizing reconstruction error means adjusting the model through optimization algorithms to make the temperature distribution reconstructed from the measured temperature curves as close as possible to the actual temperature distribution. By minimizing the error, a precise relationship is obtained between the abscissa of the temperature curve (i.e., the time or position coordinates of the temperature change curve) and the radial position inside the transformer (i.e., the distance from the outer shell to the center). For example, suppose there are 10 temperature measurement points on the outside of the transformer, but there are actually more temperature measurement points inside. By minimizing the error of the measured temperature points, the model is dynamically adjusted so that the temperature curves represented by these points can accurately reflect the temperature distribution inside the entire transformer.

[0090] S53. By introducing leakage flux density and oil flow velocity as constraints, a set of partial differential equations coupled with electromagnetic, fluid and thermal fields is constructed. The spatial distribution characteristics of the set of partial differential equations are solved by finite element method to identify the spatial distribution of electromagnetic field, oil flow and temperature inside the transformer.

[0091] Specifically, leakage flux density refers to the distribution of the magnetic field within a transformer. Leakage flux density has a certain impact on the transformer's operating state and temperature distribution, especially under high load conditions, where the leakage flux effect may lead to localized overheating. The oil flow inside the transformer has a significant impact on temperature distribution; the oil flow can carry away heat, affecting temperature uniformity. Oil flow velocity is an important physical quantity affecting temperature distribution, usually described by fluid dynamics equations. For example, assuming a higher oil flow velocity inside the transformer, the temperature distribution may be more uniform, while a lower oil flow velocity may lead to overheating in localized areas. By introducing these physical quantities as constraints, the temperature changes inside the transformer can be predicted more accurately.

[0092] Electromagnetic fields, fluid fields, and thermal fields are interconnected. During transformer operation, electromagnetic fields generate heat, oil flow affects heat conduction and distribution, and temperature changes affect fluid flow. By coupling these three physical fields, the temperature distribution and other physical characteristics inside the transformer can be accurately simulated. Partial differential equations are used to describe the interaction between electromagnetic, fluid, and thermal fields. These equations typically include heat conduction equations, fluid dynamics equations, and electromagnetic field equations. By solving these equations, the spatial distribution of temperature, oil flow, and electromagnetic fields inside the transformer can be obtained. For example, assuming that the magnetic field generated by the current inside the transformer causes heat to be generated in the windings, and the oil flow carries away some of the heat, by establishing a set of partial differential equations with three coupled fields, the temperature distribution inside the transformer can be solved, thereby predicting overheated areas and oil flow paths.

[0093] Finite element analysis (FEA) is a numerical computation method used to solve complex partial differential equations. By dividing a complex geometric structure into small elements and using a computer to simulate the physical behavior of these elements, the temperature distribution, oil flow velocity distribution, and electromagnetic field distribution at various points inside a transformer can be obtained. The results of finite element analysis can reveal the spatial distribution characteristics of temperature, oil flow, leakage flux density, etc., at various locations inside the transformer, helping to analyze the transformer's operating state, optimize the design, and avoid problems such as overheating. For example, by calculating the temperature distribution at different locations inside the transformer through finite element analysis, it is possible to clearly identify which areas may overheat, thereby guiding engineers to take measures to adjust the oil flow velocity or enhance the cooling system.

[0094] S6. Monitor abnormal temperature points in the temperature curve and the abnormal coordinate intervals where the abnormal temperature points are located, and mark the temperature abnormal areas corresponding to the abnormal coordinate intervals based on the spatial distribution characteristics.

[0095] In the description of this invention, the abnormal temperature points in the monitored temperature curve and the abnormal coordinate intervals where the abnormal temperature points are located include: S61. Slide the preset second sliding window along the temperature curve from one side to the other according to the preset step size.

[0096] S62. Based on the preset abnormal temperature judgment conditions, set the temperature threshold of abnormal temperature, and identify whether there is an abnormal temperature point in the preset sliding window at each sliding position by comparing the threshold.

[0097] Specifically, the abnormal temperature judgment criteria include whether there is a temperature signal greater than the preset temperature threshold.

[0098] S63. Record the corresponding horizontal coordinate interval within the preset sliding window for each abnormal temperature point as the abnormal coordinate interval.

[0099] S7. Based on the temperature signals of each preset temperature measurement point measured in real time, generate a temperature gradient distribution map.

[0100] Specifically, during the operation of a transformer, temperature changes in different parts can reflect the operating status of the equipment. By collecting real-time temperature signals from various preset temperature measurement points of the transformer (such as windings, oil temperature, core, etc.) and calculating the temperature change rate between each measurement point, a temperature gradient distribution map can be generated.

[0101] Temperature gradient distribution maps show the temperature differences and distribution of various parts of the transformer, which is particularly helpful in identifying the spread trend of local overheating areas (e.g., winding overheating). Overheating areas may be precursors to faults. Real-time monitoring of temperature changes can detect local overheating phenomena in a timely manner. By using temperature gradient maps, potential problem areas of the transformer can be located, and measures can be taken to prevent faults. For example, during transformer operation, if the temperature rise rate at a certain temperature measuring point is faster than that at adjacent temperature measuring points, the calculated temperature change rate will be significantly increased.

[0102] Please see Figure 2 Furthermore, a fluorescent fiber optic temperature measurement system for transformers is also provided, the system comprising: Fluorescence calibration module 1 is used to calibrate fluorescent optical fibers at different preset temperatures, and the calibrated fluorescent optical fibers are placed at preset temperature measurement points on the transformer.

[0103] Signal analysis module 2 is used to acquire fluorescence signals and fluorescence timing signals at preset temperature measurement points using a fluorescence fiber optic sensor, and to analyze the attenuation characteristics of the fluorescence signals to obtain attenuation signals.

[0104] Signal interaction module 3 is used to perform signal interaction between fluorescence signal, fluorescence timing signal and attenuation signal to obtain different types of fluorescence interaction signals.

[0105] The fiber optic temperature measurement module 4 is used to fuse the various fluorescence interaction signals to obtain a fused attention signal, and to identify the temperature signal of the preset temperature measurement point based on the fused attention signal.

[0106] The positioning and fitting module 5 is used to extract topological features from the temperature signals of the preset temperature measurement points to obtain temperature curves, and to analyze the dynamic correlation between the temperature curve coordinates and the internal radial position of the transformer using a pre-built circumferential positioning mapping model, so as to identify the spatial distribution characteristics of the transformer temperature.

[0107] Anomaly analysis module 6 is used to monitor the abnormal coordinate intervals where abnormal temperature points are located in the temperature curve, and, in combination with spatial distribution characteristics, mark the temperature anomaly areas corresponding to the abnormal coordinate intervals.

[0108] Temperature modeling module 7 is used to generate a temperature gradient distribution map based on the temperature signals of each preset temperature measurement point measured in real time.

[0109] In summary, by utilizing the technical solution of this invention, the fluorescence fiber calibration unit performs calibration at different preset temperatures and records fluorescence attenuation characteristics, ensuring that the system can provide accurate temperature data within different temperature ranges. Fluorescence fiber technology has high sensitivity, enabling precise detection of temperature changes. Furthermore, the signal analysis unit acquires fluorescence signals and fluorescence timing signals in real time, and by analyzing the attenuation characteristics of the fluorescence signals, temperature change information is obtained. This allows the system to respond quickly to temperature changes, ensuring real-time performance and accuracy. This invention possesses the ability to automatically identify and locate abnormal temperature points. By judging abnormal points in the temperature curve and combining factors such as thermal fluid and electromagnetic fields, it can accurately predict possible fault areas and issue timely alarms. This helps prevent equipment failure due to overheating, thereby improving equipment safety and stability. This invention can collect and analyze multiple monitoring data from transformers. Through acoustic-optical correlation, signal interaction, and fusion models, it enables more refined analysis of temperature anomalies and matches them with the causal relationships of transformer faults, providing more accurate fault diagnosis. Furthermore, through real-time monitoring based on temperature gradient distribution maps, it can promptly identify the diffusion trend of local overheating areas. Through causal inference model analysis, it can determine the root cause behind temperature anomalies, thereby providing effective fault warnings and maintenance suggestions. Combined with material failure prediction models, by analyzing historical data of temperature anomaly areas and insulation paper performance, it can effectively assess the remaining service life of the transformer. By outputting maintenance priority suggestions, it ensures that the transformer receives timely maintenance and extends its service life.

[0110] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

Claims

1. A fluorescent fiber optic temperature measurement method for transformers, characterized in that, The method includes: The fluorescent optical fiber was calibrated at different preset temperatures, and the calibrated fluorescent optical fiber was placed at the preset temperature measurement point of the transformer. Fluorescent fiber optic sensors are used to acquire fluorescence signals and fluorescence time-series signals at preset temperature measurement points, and the attenuation characteristics of the fluorescence time-series signals are analyzed to obtain the attenuation signals. By interacting fluorescence signals, fluorescence timing signals, and attenuation signals, different types of fluorescence interaction signals are obtained. The fluorescence interaction signals are fused to obtain a fused attention signal, and the temperature signal of the preset temperature measurement point is identified based on the fused attention signal. The temperature signals from the preset temperature measurement points are extracted using topological features to obtain temperature curves. A pre-built circumferential positioning mapping model is used to analyze the dynamic correlation between the temperature curve coordinates and the radial position inside the transformer. The spatial distribution characteristics of the transformer temperature are also identified by combining multi-physics field coupling. Monitor abnormal temperature points in the temperature curve and the abnormal coordinate intervals where the abnormal temperature points are located, and mark the temperature abnormal areas corresponding to the abnormal coordinate intervals based on the spatial distribution characteristics. A temperature gradient distribution map is generated based on the temperature signals from each preset temperature measurement point measured in real time.

2. The fluorescent fiber optic temperature measurement method for transformers according to claim 1, characterized in that, The process involves using a fluorescent fiber optic sensor to acquire fluorescence signals and fluorescence time-series signals at preset temperature measurement points, and then performing attenuation characteristic analysis on the fluorescence time-series signals to obtain attenuation signals including: The fluorescence signals at at least one consecutive time point are integrated into a fluorescence time-series signal, and transformer partial discharge signals and oil chromatography data are acquired simultaneously. By analyzing the acoustic-optical correlation features, the attenuation dynamic characteristics of the fluorescence timing signal and the timing correlation characteristics of the partial discharge acoustic signal are extracted to eliminate the interference of electromagnetic pulses on the fluorescence timing signal. Based on the factors affecting fiber optic aging and oil quality changes, a multi-factor attenuation characteristic correction model is constructed to systematically correct the deviation of the original fluorescence attenuation signal in the fluorescence time series signal. The core features of the decay characteristics of the fluorescence timing signal after interference removal and correction are extracted to obtain the decay signal; the core features include decay time constant, half-life and steady-state residual amplitude.

3. The fluorescent fiber optic temperature measurement method for transformers according to claim 2, characterized in that, The multi-factor attenuation characteristic correction model constructed based on the attenuation factors of optical fiber aging and oil quality changes includes: The fluorescent optical fiber was subjected to aging tests to monitor the fluorescence decay time constant under different usage durations. Based on the aging test monitoring data, an optical fiber aging compensation function was established to characterize the quantitative relationship between the usage duration and the attenuation characteristic deviation of the fluorescent optical fiber. Transformer oil samples were collected regularly and key oil quality parameters were tested. Attenuation signals were recorded synchronously for the corresponding time periods, and an oil quality-attenuation correlation model based on the mapping relationship between the concentration of characteristic components of the oil sample and the fluorescence attenuation rate was constructed. By integrating the fiber aging compensation function with the oil quality-attenuation correlation model, a multi-factor attenuation characteristic correction model is formed.

4. The fluorescent fiber optic temperature measurement method for transformers according to claim 1, characterized in that, The process of interacting fluorescence signals, fluorescence timing signals, and attenuation signals to obtain different types of fluorescence interaction signals includes: At least one fluorescence signal segment of the fluorescence timing signal is extracted according to the preset first window, and the difference is obtained by subtracting it from the adjacent fluorescence signal segment to obtain the timing difference signal segment data. Each fluorescence signal segment and time-difference signal segment is divided into at least one signal block to obtain the fluorescence signal block and time-difference signal block of each fluorescence signal segment, and the fluorescence signal blocks and time-difference signal blocks of all fluorescence signal segments are combined to form fluorescence signal segment data. The attenuation signal is normalized to obtain attenuation signal data; The fluorescence signal segment data is input into the first signal feature extraction model to obtain fluorescence signal features; the time-series differential signal segment data is input into the second signal characteristic extraction model to obtain time-series signal features; and the attenuation signal data is input into the third signal feature extraction model to obtain attenuation signal features. The fluorescence signal features, time-series signal features, and attenuation signal features are input into the signal interaction fusion model to obtain different types of fluorescence interaction signals; the fluorescence interaction signals include a first interaction signal, a second interaction signal, and a third interaction signal.

5. The fluorescent fiber optic temperature measurement method for transformers according to claim 4, characterized in that, The step of inputting the fluorescence signal segment data into the first signal feature extraction model to obtain fluorescence signal features includes: The fluorescence signal block features of each fluorescence signal block in the fluorescence signal segment data are extracted using the first signal feature extraction model; Calculate the average value of the fluorescence signal block features for each fluorescence signal block, and use it as the fluorescence signal feature.

6. The fluorescent fiber optic temperature measurement method for transformers according to claim 4, characterized in that, The process of inputting the fluorescence signal features, time-series signal features, and attenuation signal features into the signal interaction fusion model to obtain different types of fluorescence interaction signals includes: The fluorescence signal features, time-series signal features, and attenuation signal features are interacted to obtain the first interactive signal; The fluorescence signal feature is added to the first interactive signal to obtain a first fluorescence fusion signal; the timing signal feature is added to the first interactive signal to obtain a first timing fusion signal; the attenuation signal feature is added to the first interactive signal to obtain a first attenuation fusion signal; the first fluorescence fusion signal, the first timing fusion signal, and the first attenuation fusion signal are subjected to feature interaction to obtain a second interactive signal. The first fluorescence fusion signal is added to the second interaction signal to obtain the second fluorescence fusion signal; the first time-series fusion signal is added to the second interaction signal to obtain the second time-series fusion signal; the first attenuation fusion signal is added to the second interaction signal to obtain the second attenuation fusion signal; and the second fluorescence fusion signal, the second time-series fusion signal, and the second attenuation fusion signal are subjected to feature interaction to obtain the third interaction signal.

7. The fluorescent fiber optic temperature measurement method for transformers according to claim 6, characterized in that, The step of fusing the various fluorescence interaction signals to obtain a fused attention signal, and identifying the temperature signal of a preset temperature measurement point based on the fused attention signal, includes: Perform matrix transformations on the first, second, and third interaction signals to obtain the corresponding attention query vector, attention key vector, and attention value vector; The attention query vector, the attention key vector, and the attention value vector are processed by a self-attention mechanism to obtain a fused attention signal. A temperature feature sequence is obtained by linearly mapping the fused attention signal; Based on the temperature feature sequence, feature matching is performed in the temperature calibration database to obtain the feature matching result, and the temperature signal of the preset temperature measurement point is identified based on the feature matching result.

8. The fluorescent fiber optic temperature measurement method for transformers according to claim 1, characterized in that, The method of analyzing the dynamic correlation between temperature curve coordinates and the internal radial position of the transformer using a pre-constructed circumferential positioning mapping model, and identifying the spatial distribution characteristics of transformer temperature by combining multi-physics coupling, includes: Based on the transformer's geometric structure and the layout of the fluorescent sensors, a circumferential positioning mapping model of the transformer is established to map the real-time monitored temperature signal with the actual position inside the transformer, so as to establish the correlation between the temperature signal and the actual position inside the transformer. By using the circumferential positioning mapping model to minimize the reconstruction error between the spatial coordinates of the measuring point and the temperature curve, the dynamic correlation between the horizontal coordinate of the temperature curve and the radial position inside the transformer is obtained. By introducing leakage flux density and oil flow velocity as constraints, a set of partial differential equations coupling electromagnetic, fluid, and thermal fields is constructed. The spatial distribution characteristics of the partial differential equations are solved by finite element method to identify the spatial distribution of electromagnetic field, oil flow, and temperature inside the transformer.

9. The fluorescent fiber optic temperature measurement method for transformers according to claim 1, characterized in that, The abnormal temperature points in the monitored temperature curve and the abnormal coordinate intervals where the abnormal temperature points are located include: The preset second sliding window is slid along the temperature curve from one side to the other according to a preset step size; Based on preset abnormal temperature judgment conditions, a temperature threshold for abnormal temperature is set, and the presence of an abnormal temperature point within the preset sliding window at each sliding position is identified by threshold comparison. Record the corresponding horizontal coordinate interval within the preset sliding window for each abnormal temperature point as the abnormal coordinate interval.

10. A fluorescent fiber optic temperature measurement system for transformers, used to implement the fluorescent fiber optic temperature measurement method for transformers according to any one of claims 1-9, characterized in that, The system includes: The fluorescence calibration module is used to calibrate the fluorescent optical fiber at different preset temperatures, and the calibrated fluorescent optical fiber is placed at the preset temperature measurement point of the transformer. The signal analysis module is used to acquire the fluorescence signal and fluorescence timing signal at the preset temperature measurement point using a fluorescence fiber optic sensor, and to analyze the attenuation characteristics of the fluorescence signal to obtain the attenuation signal. The signal interaction module is used to interact with fluorescence signals, fluorescence timing signals, and attenuation signals to obtain different types of fluorescence interaction signals. The fiber optic temperature measurement module is used to fuse various fluorescence interaction signals to obtain a fused attention signal, and to identify the temperature signal of a preset temperature measurement point based on the fused attention signal. The positioning and fitting module is used to extract topological features from the temperature signals of preset temperature measurement points to obtain temperature curves. It also uses a pre-built circumferential positioning mapping model to analyze the dynamic correlation between the temperature curve coordinates and the internal radial position of the transformer in order to identify the spatial distribution characteristics of the transformer temperature. The anomaly analysis module is used to monitor the abnormal coordinate intervals where abnormal temperature points are located in the temperature curve, and, in combination with spatial distribution characteristics, mark the temperature anomaly areas corresponding to the abnormal coordinate intervals. The temperature modeling module is used to generate a temperature gradient distribution map based on the temperature signals of each preset temperature measurement point measured in real time.

Images