Transformer energy absorbing material state online monitoring method and system based on multi-mode sensing

Abstract

The invention discloses an online monitoring method and system for the state of a transformer energy absorption material based on multi-mode sensing. The method comprises the steps that multi-source data of an energy absorption material are synchronously collected through multiple channels of a multi-mode sensor arranged on the transformer energy absorption material in advance, and the multi-source data comprise the vacuum degree, the temperature and strain data; carrying out preprocessing operation on the multi-source data and carrying out feature extraction to obtain a feature vector; inputting the feature vector into a pre-trained neural network, and outputting the failure probability of the energy-absorbing material; and determining an alarm mechanism of the energy-absorbing material according to the failure probability of the energy-absorbing material and a preset threshold value.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment safety protection, and more particularly to a method and system for online monitoring the state of transformer energy-absorbing materials based on multimodal sensing. Background Art

[0002] With the rapid development of transformer technology towards high voltage and ultra-high voltage, the limitations of existing technologies and methods have become increasingly apparent, specifically in the following aspects: First, traditional energy-absorbing material performance evaluation mainly relies on laboratory destructive testing methods, such as drop hammer impact tests. Although these methods can provide performance data of materials under extreme conditions, they cannot achieve real-time monitoring of the performance degradation of materials in actual long-term operation, which limits a comprehensive understanding of the health status of the materials. In addition, in the existing technology, some methods only monitor the status of the equipment through a single parameter (such as temperature or pressure sensor). For example, patent CN113432764A only uses temperature or pressure sensors for monitoring. This single-parameter evaluation method is difficult to fully reflect the comprehensive performance of the energy-absorbing material and may lead to misjudgment of the material status. Moreover, the current monitoring methods fail to integrate dynamic working condition data (such as explosion impact energy) with multiple parameters such as material deformation and vacuum degree for analysis, lack intelligent data processing and prediction capabilities, and cannot achieve predictive maintenance of energy-absorbing materials.

[0003] In high-voltage and ultra-high-voltage environments, energy-absorbing materials may be affected by complex environmental factors such as temperature, humidity, and radiation, which can accelerate material aging and performance degradation. Existing monitoring technologies often overlook the combined impact of these environmental factors, resulting in inaccurate material condition assessments. Even when data on multiple parameters is collected in some systems, the lack of effective data processing and analysis methods makes it impossible to extract valuable information from large amounts of data, resulting in the inability to promptly identify potential problems. Currently, the technical specifications for online monitoring devices for substation equipment are not yet fully unified, and compatibility issues may exist between different manufacturers and between different devices, making data sharing and comprehensive analysis difficult. Summary of the Invention

[0004] In view of the deficiencies in the prior art, the present invention provides a method and system for online monitoring of the status of energy-absorbing materials in transformers based on multimodal sensing.

[0005] According to one aspect of the present invention, a method for online monitoring of the state of transformer energy-absorbing materials based on multimodal sensing is provided, comprising:

[0006] Multi-source data of the energy-absorbing material are collected synchronously through multi-channel multi-modal sensors pre-arranged on the energy-absorbing material of the transformer, wherein the multi-source data includes vacuum degree, temperature and strain data;

[0007] Perform preprocessing operations on multi-source data and perform feature extraction to obtain feature vectors;

[0008] The feature vector is input into a pre-trained neural network to output the failure probability of the energy-absorbing material;

[0009] The alarm mechanism of the energy-absorbing material is determined based on the failure probability of the energy-absorbing material and the preset threshold.

[0010] Optionally, the method further includes: determining a vacuum degree decreasing rate of the energy-absorbing material according to the real-time monitored vacuum degree of the energy-absorbing material, and replenishing air to the energy-absorbing material when the vacuum degree decreasing rate exceeds a preset rate threshold.

[0011] Optionally, multi-source data of the energy absorbing material is collected by using a multi-modal sensor pre-arranged on the energy absorbing material of the transformer, including:

[0012] The vacuum degree of the energy absorbing material is collected by a high-frequency dynamic pressure sensor integrated in the sandwich layer of the energy absorbing material in a flexible sealed form;

[0013] The temperature of the energy absorbing material is collected by using fiber Bragg grating temperature sensors distributed along the surface of the energy absorbing material;

[0014] The strain data of the energy-absorbing material is collected by strain sensors arranged in stress concentration areas or key structural parts of the energy-absorbing material.

[0015] Optionally, the frequency band coverage of the high-frequency dynamic pressure sensor is 0.1 to 10 kHz, and the arrangement spacing is 50 mm.

[0016] Optionally, the wavelength of the fiber grating temperature sensor is 1550 nm, and the fiber grating temperature sensor is fixed on the surface of the energy absorbing material by using an adhesive or mechanical clamping.

[0017] Optionally, the strain sensor has a detection frequency band of 0 to 10 kHz and a measurement range of ±5000 με.

[0018] Optionally, preprocessing operations are performed on the multi-source data and feature extraction is performed to obtain feature vectors, including:

[0019] Denoising of multi-source data using a hybrid method of sampled wavelet transform and variational mode analysis;

[0020] Interpolation compensation algorithm and dynamic time warping algorithm are used to asynchronously process the denoised multi-source data;

[0021] The principal component analysis method is used to construct the spatiotemporal correlation matrix based on the preprocessed multi-source data;

[0022] The key characteristic parameters of the spatiotemporal correlation matrix are extracted to obtain the characteristic vector.

[0023] Optionally, the number of hidden layer neurons in the neural network is 128, and the learning rate is 0.001.

[0024] According to another aspect of the present invention, there is provided a transformer energy-absorbing material state online monitoring system based on multimodal sensing, comprising:

[0025] The sensor module and data acquisition module are used to collect multi-source data of energy-absorbing materials on the transformer;

[0026] An intelligent analysis module is used to analyze the failure probability of the energy-absorbing material based on multi-source data to obtain the failure probability of the energy-absorbing material;

[0027] The execution module is used to take corresponding execution measures according to the failure probability of the energy-absorbing material and multi-source data.

[0028] Optionally, the sensor module includes a high-frequency dynamic pressure sensor integrated in a flexible sealed form inside the interlayer of the energy-absorbing material; a fiber optic Bragg grating temperature sensor laid in a distributed manner along the surface of the energy-absorbing material; and a strain sensor arranged in stress concentration areas or key structural parts of the energy-absorbing material.

[0029] Optionally, the data acquisition module includes:

[0030] A high-frequency signal processing unit, used to achieve multi-channel synchronous acquisition of multi-source data according to a preset sampling rate and pre-processing circuit;

[0031] The wireless transmission unit collects multi-source data based on LoRa and 5G dual-mode wireless transmission.

[0032] Therefore, the present invention uses multimodal sensing of vacuum, temperature, and strain sensing technologies to achieve all-round monitoring of the state of energy-absorbing materials. High-frequency data acquisition with a high sampling rate ensures the capture of rapid dynamic changes to provide real-time feedback on material stress and environmental changes. By combining intelligent analysis and prediction with Kalman filtering, PCA, and LSTM neural networks, predictive warning of material failure is achieved, thereby significantly improving maintenance efficiency and safety. In addition, the automatic execution response of the vacuum pump truck linked to the sound and light alarm can automatically activate the air replenishment and alarm mechanism when an anomaly is detected to achieve intelligent fault response. It has a wide range of applicable scenarios, covering energy-absorbing material monitoring in high-voltage and ultra-high-voltage transformers, real-time monitoring and data analysis of material status during explosion impact energy transfer, and online health management and predictive maintenance of long-term equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] A more complete understanding of exemplary embodiments of the present invention may be obtained by referring to the following drawings:

[0034] Figure 1It is a flow chart of a method for online monitoring the state of transformer energy-absorbing materials based on multimodal sensing provided by an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0035] Below, the exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0036] It should be noted that the relative arrangement of components and steps, the numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless specifically stated otherwise.

[0037] Those skilled in the art will understand that the terms "first" and "second" in the embodiments of the present invention are only used to distinguish different steps, devices or modules, and neither represent any specific technical meaning nor indicate the necessary logical order between them.

[0038] It should also be understood that, in the embodiments of the present invention, “a plurality of” may refer to two or more than two, and “at least one” may refer to one, two or more than two.

[0039] It should also be understood that any component, data or structure mentioned in the embodiments of the present invention can generally be understood as one or more, unless explicitly limited or otherwise indicated in the context.

[0040] In addition, the term "and / or" in this invention merely describes an association relationship between related objects, indicating that three possible relationships exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Furthermore, the character " / " in this invention generally indicates that the related objects are in an "or" relationship.

[0041] It should also be understood that the description of the various embodiments of the present invention focuses on the differences between the various embodiments, and the same or similar aspects thereof can be referenced with each other. For the sake of brevity, they will not be described one by one.

[0042] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0043] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0044] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, they should be considered part of the specification.

[0045] It should be noted that like reference numerals and letters refer to like items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0046] Figure 1 FIG1 is a flow chart of a method for online monitoring of transformer energy-absorbing material status based on multimodal sensing provided by the first aspect of the embodiment of the present invention. Figure 1 As shown, the method 100 for online monitoring of transformer energy-absorbing material status based on multimodal sensing includes the following steps:

[0047] Step 101, using a multi-modal sensor pre-arranged on the transformer energy-absorbing material to synchronously collect multi-source data of the energy-absorbing material through multiple channels, wherein the multi-source data includes vacuum degree, temperature, and strain data;

[0048] Step 102: preprocessing the multi-source data and extracting features to obtain feature vectors;

[0049] Step 103: input the feature vector into a pre-trained neural network and output the failure probability of the energy-absorbing material;

[0050] Step 104 : determining an alarm mechanism of the energy absorbing material according to the failure probability of the energy absorbing material and a preset threshold value.

[0051] Specifically, the purpose of the present invention is to be able to monitor the multi-parameter status of energy-absorbing materials in real time online. This intelligent system combines dynamic operating data to perform multi-parameter fusion analysis, and predicts failure risks through dynamic feedback to achieve comprehensive evaluation and predictive maintenance of energy-absorbing materials, ensuring the safe and reliable operation of transformers under high voltage and ultra-high voltage conditions.

[0052] In response to the shortcomings of existing monitoring, the present invention provides online monitoring of the status of transformer energy-absorbing materials based on multimodal sensing, which realizes real-time monitoring of multiple parameters of energy-absorbing materials under explosion impact, improves the anti-interference ability and prediction accuracy of monitoring, and reduces maintenance costs and the need for manual intervention.

[0053] In order to achieve the above object, the present invention adopts the following technical solutions:

[0054] 1. Sensor module

[0055] (1) Vacuum voltage sensor

[0056] Selection and specifications: MEMS micro high-frequency dynamic pressure sensor is used, with a response frequency band of 0.1 to 10 kHz, which can capture transient pressure fluctuations.

[0057] Installation: The sensor is integrated into the sandwich layer of the energy-absorbing material in a flexible package to ensure a close fit with the material. The sensor is small in size and low in thickness (usually ≤ 0.5mm) to avoid interfering with the mechanical properties of the energy-absorbing material itself.

[0058] Functional Description: Real-time monitoring of vacuum changes within energy-absorbing materials, providing accurate pressure data to assess the sealing status and stability of the material's internal environment. Under extreme impact or thermal stress, the sensor can provide timely feedback on the decrease in vacuum level, providing a basis for subsequent refilling and maintenance.

[0059] Data calibration: Before the device is installed, it is calibrated using a standard pressure chamber to ensure that the measurement data has high accuracy and repeatability.

[0060] (2) Fiber Bragg Grating Temperature Sensor

[0061] Selection and specifications: FOT-T-SS optical fiber temperature sensor is used, and a distributed fiber Bragg grating temperature sensor array is arranged along the surface of the energy-absorbing material. The central wavelength is 1550nm, and the temperature measurement accuracy reaches ±0.3℃.

[0062] Installation: The sensors are encased in heat-resistant, corrosion-resistant optical fiber and distributed across the surface of the material, ensuring coverage of the entire critical monitoring area. The sensors are secured to the energy-absorbing material with a high-strength adhesive to prevent displacement despite temperature fluctuations or mechanical vibration.

[0063] Functional Description: Leveraging the temperature-dependent characteristics of fiber Bragg grating reflection wavelengths, this system monitors the temperature distribution on a material surface in real time. This data can be used to analyze the material's thermal gradient distribution and dynamic response in the operating environment, providing data support for preventing thermally induced performance degradation.

[0064] Integration: Through dedicated fiber grating distributed thermometer, the entire temperature sensing network is interrogated and continuous temperature curve data is output.

[0065] (3) Strain sensor

[0066] Selection and specifications: It uses a broadband fiber Bragg grating strain sensor with a detection frequency band covering 0 to 10 kHz, suitable for monitoring mechanical strain changes at low to medium frequencies.

[0067] Installation method: The sensor is installed in the stress concentration area or key structural parts of the energy-absorbing material, such as the weld of the fuel tank. The sensor and the material surface are installed by embedded or adhesive method to ensure accurate transmission of strain signals.

[0068] Functional Description: This device records in real time the minute deformations of materials under impact or load, capturing both elastic and plastic deformation. This data reflects potential fatigue and crack growth during long-term operation, providing a crucial basis for predicting material failure.

[0069] 2. Data acquisition module

[0070] (1) High-frequency signal processing unit

[0071] Sampling requirements: The sampling rate should be set to no less than 10kHz to ensure complete capture of high-speed dynamically changing signals.

[0072] Signal processing flow:

[0073] Pre-processing circuit: including amplification, filtering and analog-to-digital conversion (ADC) to ensure signal quality.

[0074] Multi-channel synchronous acquisition: Each sensor signal is synchronously acquired through an independent channel to ensure timing consistency and provide accurate data for subsequent multi-parameter fusion.

[0075] Hardware implementation: Use a high-performance microcontroller or FPGA platform to implement real-time signal processing, supporting high data throughput and low-latency transmission.

[0076] (2) Wireless transmission module

[0077] Communication mode: Supports LoRa and 5G dual-mode wireless transmission, flexibly responding to different monitoring scenarios.

[0078] Data transmission characteristics:

[0079] Low power consumption: Optimizes transmission protocols to extend operating time.

[0080] Secure encryption: Data encryption technology is used to ensure that data is not tampered with or leaked during transmission.

[0081] Network integration: The wireless module can be seamlessly connected to a remote monitoring center or cloud platform to achieve real-time remote monitoring and data backup.

[0082] 3. Intelligent analysis module

[0083] (1) Data preprocessing

[0084] Algorithm selection: Hybrid denoising using wavelet transform and variational modal analysis. The former separates the signal into high-frequency and low-frequency components, while the latter extracts the effective frequency band. Normalization is also performed to unify data from different sources into the same dimension.

[0085] Preprocessing is done by filtering out high-frequency noise and occasional interference signals, eliminating the range differences of each sensor to facilitate subsequent feature extraction and model training.

[0086] Due to the sampling rate differences when collecting multi-source data, in order to solve this asynchronous problem, some interpolation compensation algorithms can be used, such as cubic spline interpolation and other methods (which can interpolate low-frequency signals such as temperature signals to match high-frequency signals). Combined with the dynamic time warping algorithm, a time-space correlation matrix can be constructed for the timing phase offset of this multi-source data, which can effectively solve this asynchronous problem.

[0087] (2) Feature extraction

[0088] Principal component analysis (PCA) was used to construct a spatiotemporal correlation matrix from preprocessed vacuum, temperature, and strain data, extracting key characteristic parameters. First, a data matrix was constructed, building a multidimensional data matrix based on the time series data. The covariance matrix was then calculated and the eigenvalues ​​and eigenvectors were solved to filter the principal components and retain the majority of the data information. Finally, the characteristic parameters were output to form eigenvectors, which served as input to the LSTM network.

[0089] (3) Failure warning

[0090] Based on a long short-term memory (LSTM) neural network, failure probability prediction is performed using historical data and real-time monitoring data. The LSTM network is trained using extensive experimental data to form a stable prediction model. During real-time monitoring, when the predicted failure probability reaches or exceeds a preset threshold (e.g., 0.8), the model outputs an alarm signal. The model is adaptively updated and supports online learning mechanisms, continuously updating model parameters based on the latest monitoring data to improve prediction accuracy.

[0091] 4. Execution module

[0092] (1) Vacuum pump truck

[0093] Automatic Air Replenishment Mechanism: When the intelligent analysis module detects that the rate of vacuum drop exceeds a set threshold, the vacuum pump truck automatically activates. Through feedback control and interaction with the data acquisition module, the vacuum pump truck automatically replenishes the working fluid and restores the vacuum level within the energy-absorbing material to the designed range. Furthermore, overload protection and emergency shutdown functions ensure that the air replenishment process does not cause other faults.

[0094] (2) Sound and light alarm device

[0095] When the LSTM network predicts a failure probability of 0.8 or greater, an audible and visual alarm is triggered. Early warning, alarm, and emergency alarm levels are set based on the risk of failure, providing graded response prompts. The audible and visual alarm system issues an on-site alarm signal and simultaneously transmits the alarm information to a remote monitoring platform in real time via a wireless transmission module, ensuring timely response by maintenance personnel. Automatic alarm logging and self-diagnosis capabilities facilitate subsequent fault analysis and maintenance management.

[0096] Through the collaborative work between various modules, this solution can not only monitor the status of energy-absorbing materials in real time, but also provide early warning of possible failure risks, providing a comprehensive monitoring solution for transformers and other high-risk equipment that integrates high-precision measurement, intelligent data analysis and automated response.

[0097] In an exemplary embodiment of the present invention: sensor deployment

[0098] 1.1 Vacuum sensor deployment

[0099] The Honeywell 26PC series vacuum sensor was selected. This sensor has high-frequency dynamic response capability and a response frequency band covering 0.1 to 10 kHz. It is suitable for detecting subtle pressure changes inside energy-absorbing materials. An embedded installation method is adopted to embed the vacuum sensor directly into the interlayer of the energy-absorbing material through a micro-flexible package. The sensor is arranged at a spacing of 50 mm, which not only ensures sufficient spatial resolution but also does not interfere with the overall structure of the material. During the embedding process, ensure good adhesion between the sensor and the material interlayer to avoid loosening or displacement caused by factors such as temperature and vibration. At the same time, the sensor is calibrated using pre-calibration equipment (such as a standard vacuum chamber) to ensure that the sensing data after installation is highly accurate and consistent.

[0100] 1.2 Fiber Bragg Grating Temperature Sensor Deployment

[0101] Fiber Bragg grating temperature sensors with a wavelength of 1550nm achieve a temperature measurement accuracy of ±0.3°C. They are distributed along the surface of the energy-absorbing material and secured to the material using adhesives or mechanical clamps. The entire temperature sensor array covers all critical areas, enabling continuous monitoring of temperature gradients and localized hotspots. The optical signal output by the sensors is analyzed by a dedicated fiber optic thermometer and transmitted to the data acquisition system via a fiber optic communication interface, ensuring synchronous acquisition of temperature data with other sensor data.

[0102] 1.3 Strain sensor deployment

[0103] A broadband fiber Bragg grating strain sensor with a detection frequency band of 0 to 10 kHz and a range of ±5000 με is used, making it suitable for capturing tiny deformations of energy-absorbing materials in areas of high stress concentration. Precise placement is performed on stress-concentrated areas of the energy-absorbing material, such as material edges or pre-set weak points. The sensor is fixed to the target area using high-performance adhesives or pre-made embedded grooves to ensure accurate strain response data when the material is subjected to stress. After installation, the strain sensor is calibrated by loading standard mechanical tests (such as tension and compression tests) to ensure its linear response and repeatability under working conditions.

[0104] In an exemplary embodiment of the present invention: algorithm training

[0105] 2.1 Data collection and experimental design

[0106] A simulated explosion environment was constructed in the laboratory, with impact energies ranging from 10 to 100 kJ. A total of 1,000 sets of experimental data were collected. Each set of data included a continuous record of the time-varying changes in three parameters: vacuum, temperature, and strain. The collected raw data was preprocessed (e.g., noise reduction and normalization). Each data set was annotated based on the experimental results to distinguish between normal conditions and failure signs, forming a training dataset.

[0107] 2.2 Dataset Division

[0108] The 1,000 data sets were divided into training, validation, and test sets in a 70%:15%:15% ratio. The training set was used for preliminary learning of model parameters; the validation set was used to adjust model parameters and prevent overfitting; and the test set was used to independently evaluate model performance.

[0109] 2.3LSTM network model setting

[0110] A dynamic failure warning model based on an LSTM neural network was constructed. The number of hidden layer neurons was set to 128 to capture long-term dependencies in the data. A learning rate of 0.001 was set to ensure smooth convergence of the model on larger datasets. The number of iterations was set to 200, and network weights were optimized through continuous iterative training until the loss function reached a stable state. During training, the model's performance on the validation set was monitored in real time, and cross-validation was used to adjust network parameters. Finally, the test set data was used to verify the model's failure prediction accuracy, and the prediction error and failure probability distribution were calculated to ensure the model's reliability in practical applications.

[0111] In an exemplary embodiment of the present invention: On-site verification

[0112] 3.1 Installation and Integration

[0113] This online monitoring system was installed on a 110kV transformer. Its modules include a sensor module (embedded vacuum, temperature, and strain sensors), a data acquisition module, a high-frequency signal processing unit, a wireless transmission module, and an intelligent analysis and execution module. Sensors were pre-placed in key areas of the transformer's energy-absorbing material according to the design drawings. All sensor signals were aggregated to a data acquisition center via a wired or wireless network. This system is linked to the transformer control system to enable real-time interaction between monitoring data and on-site control.

[0114] 3.2 On-site monitoring and data feedback

[0115] During on-site operation, the energy-absorbing material status is monitored in real time. Data shows that vacuum abnormalities can be detected 30 minutes in advance, allowing the vacuum pump to automatically replenish air, achieving preventive maintenance.

[0116] When the failure probability output by the LSTM prediction model reaches a set threshold (e.g., 0.8), an audible and visual alarm system is immediately triggered. Simultaneously, the abnormal data is uploaded to a remote monitoring center via a wireless transmission module, enabling remote fault diagnosis and early warning. During field verification, the false alarm rate was statistically controlled at 3.2%, fully complying with the power equipment reliability standard (GB / T 32507-2016), demonstrating high accuracy and stability under actual operating conditions.

[0117] Therefore, the present invention uses multimodal sensing of vacuum, temperature, and strain sensing technologies to achieve all-round monitoring of the state of energy-absorbing materials. High-frequency data acquisition with a high sampling rate ensures the capture of rapid dynamic changes to provide real-time feedback on material stress and environmental changes. By combining intelligent analysis and prediction with Kalman filtering, PCA, and LSTM neural networks, predictive warning of material failure is achieved, thereby significantly improving maintenance efficiency and safety. In addition, the automatic execution response of the vacuum pump truck linked to the sound and light alarm can automatically activate the air replenishment and alarm mechanism when an anomaly is detected to achieve intelligent fault response. It has a wide range of applicable scenarios, covering energy-absorbing material monitoring in high-voltage and ultra-high-voltage transformers, real-time monitoring and data analysis of material status during explosion impact energy transfer, and online health management and predictive maintenance of long-term equipment.

[0118] In addition, a second aspect of the present invention provides an online monitoring system for transformer energy-absorbing material status based on multimodal sensing, comprising:

[0119] The sensor module and data acquisition module are used to collect multi-source data of energy-absorbing materials on the transformer;

[0120] An intelligent analysis module is used to analyze the failure probability of the energy-absorbing material based on multi-source data to obtain the failure probability of the energy-absorbing material;

[0121] The execution module is used to take corresponding execution measures according to the failure probability of the energy-absorbing material and multi-source data.

[0122] Optionally, the sensor module includes a high-frequency dynamic pressure sensor integrated in a flexible sealed form inside the interlayer of the energy-absorbing material; a fiber optic Bragg grating temperature sensor laid in a distributed manner along the surface of the energy-absorbing material; and a strain sensor arranged in stress concentration areas or key structural parts of the energy-absorbing material.

[0123] Optionally, the data acquisition module includes:

[0124] A high-frequency signal processing unit, used to achieve multi-channel synchronous acquisition of multi-source data according to a preset sampling rate and pre-processing circuit;

[0125] The wireless transmission unit collects multi-source data based on LoRa and 5G dual-mode wireless transmission.

[0126] The above description has been presented for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of the present invention to the forms disclosed herein. Although a number of example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A method for online monitoring of transformer energy-absorbing material status based on multimodal sensing, characterized in that: include: Synchronously collecting multi-source data of the energy-absorbing material through multi-channel multi-modal sensors pre-arranged on the transformer energy-absorbing material, wherein the multi-source data includes vacuum degree, temperature and strain data; Performing preprocessing operations on the multi-source data and performing feature extraction to obtain feature vectors; Inputting the feature vector into a pre-trained neural network to output the failure probability of the energy absorbing material; An alarm mechanism of the energy absorbing material is determined according to the failure probability of the energy absorbing material and a preset threshold.

2. The method according to claim 1, characterized in that Also includes: The vacuum degree decreasing rate of the energy absorbing material is determined according to the real-time monitored vacuum degree of the energy absorbing material, and the energy absorbing material is replenished with air when the vacuum degree decreasing rate exceeds a preset rate threshold.

3. The method according to claim 1, characterized in that The multi-source data of the energy absorbing material is collected by using a multi-modal sensor pre-arranged on the energy absorbing material of the transformer, including: The vacuum degree of the energy absorbing material is collected by a high-frequency dynamic pressure sensor integrated into the interlayer of the energy absorbing material in a flexible sealing form; collecting the temperature of the energy absorbing material by using fiber grating temperature sensors laid in a distributed manner along the surface of the energy absorbing material; The strain data of the energy absorbing material is collected by strain sensors arranged in stress concentration areas or key structural parts of the energy absorbing material.

4. The method according to claim 3, characterized in that The frequency band of the high-frequency dynamic pressure sensor is 0.1-10 kHz, and the arrangement interval is 50 mm.

5. The method according to claim 3, characterized in that The wavelength of the fiber grating temperature sensor is 1550 nm, and the sensor is fixed on the surface of the energy absorbing material by using an adhesive or mechanical clamping.

6. The method according to claim 3, characterized in that The strain sensor has a detection frequency band of 0 to 10 kHz and a measuring range of ±5000 με.

7. The method according to claim 1, characterized in that Preprocessing the multi-source data and extracting features to obtain feature vectors includes: Denoising the multi-source data using a hybrid method of sampling wavelet transform and variational mode analysis; Interpolation compensation algorithm and dynamic time warping algorithm are used to asynchronously process the denoised multi-source data; The principal component analysis method is used to construct the spatiotemporal correlation matrix based on the preprocessed multi-source data; The key characteristic parameters of the spatiotemporal correlation matrix are extracted to obtain the characteristic vector.

8. The method according to claim 1, characterized in that The number of neurons in the hidden layer of the neural network is 128, and the learning rate is 0.

001.

9. A system for online monitoring the state of energy-absorbing materials of transformers based on multimodal sensing, for implementing the method for online monitoring the state of energy-absorbing materials of transformers based on multimodal sensing according to any one of claims 1 to 8, characterized in that: include: The sensor module and data acquisition module are used to collect multi-source data of energy-absorbing materials on the transformer; an intelligent analysis module, configured to perform a failure probability analysis on the energy-absorbing material based on the multi-source data to obtain a failure probability of the energy-absorbing material; An execution module is used to take corresponding execution measures according to the failure probability of the energy-absorbing material and multi-source data.

10. The system according to claim 9, characterized in that The sensor module includes a high-frequency dynamic pressure sensor integrated in the interlayer of the energy-absorbing material in a flexible and sealed form; a fiber grating temperature sensor laid in a distributed manner along the surface of the energy-absorbing material; and a strain sensor arranged in the stress concentration area or key structural position in the energy-absorbing material.

11. The system according to claim 9, wherein: The data acquisition module includes: A high-frequency signal processing unit, configured to implement multi-channel synchronous acquisition of the multi-source data according to a preset sampling rate and a pre-processing circuit; The wireless transmission unit collects multi-source data based on LoRa and 5G dual-mode wireless transmission.

Images