Transformer fault remote monitoring system and monitoring method
By establishing a nonlinear causal model based on multimodal data, the problem of unclear distinction between temperature and vibration effects in traditional transformer fault monitoring is solved, and early fault prediction and risk avoidance are achieved.
Patent Information
- Application Number
- CN202510831987.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-20
- Publication Date
- 2025-09-12
AI Technical Summary
Traditional transformer fault monitoring methods fail to effectively distinguish the independent effects of temperature and vibration, resulting in unreasonable weight distribution and potentially distorted prediction results.
Multimodal data is used to establish a nonlinear causal model, and the gas concentration is predicted through vibration drive and temperature drive. The vibration chain and temperature chain construction modules are combined to optimize the impact of electrical quantity data on material temperature. The vibration-concentration and temperature-concentration causal fusion models are constructed to generate a comprehensive prediction value of gas concentration and monitor anomalies in real time.
It can detect abnormal gas concentrations at the early stages of a fault, predict faults in advance, avoid line aging risks, and prevent visible damage or power outages.
Smart Images
Figure CN120636112A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of remote monitoring, and in particular to a transformer fault remote monitoring system and monitoring method. Background Art
[0002] As power systems continue to expand in size and become more intelligent, the operating status of transformers, as critical components in these systems, directly impacts the safe and stable operation of the power grid. To monitor transformer operating conditions and improve fault warning and resolution efficiency, remote transformer fault monitoring technology has emerged. This technology uses sensors to collect key transformer parameters, such as temperature, oil level, voltage, and current. This data is then transmitted to a monitoring center via a communication network, enabling real-time monitoring and analysis of transformer operating status, effectively safeguarding the safe operation of the power system.
[0003] As transformer operating environments become increasingly complex and load demands continue to rise, their internal state is influenced by a variety of factors, particularly temperature, vibration, and changes in electrical parameters. While traditional monitoring methods can achieve basic condition monitoring, they still have limitations in multi-parameter collaborative analysis and in-depth identification of fault mechanisms. This creates new requirements for the development of more accurate and intelligent remote monitoring technologies.
[0004] The existing technology does not distinguish the independent effects of temperature on vibration drive and temperature drive, resulting in unreasonable weight distribution. High temperature may enhance vibration drive and thermal decomposition drive at the same time, but traditional methods may superimpose their dual effects, resulting in distorted prediction results. Therefore, a transformer fault remote monitoring system and monitoring method are provided. Summary of the Invention
[0005] The purpose of the present invention is to provide a remote monitoring system and monitoring method for transformer faults to solve the problem proposed in the above background technology that high temperature may simultaneously enhance vibration drive and thermal decomposition drive, but traditional methods may superimpose their dual effects, causing distortion of prediction results.
[0006] To achieve the above objectives, on the one hand, the present invention aims to provide a transformer fault remote monitoring system, comprising:
[0007] As a further improvement of the present technical solution, the multimodal data includes electrical quantity data, vibration spectrum data, gas concentration data and material temperature data.
[0008] As a further improvement of the present technical solution, the multi-source causal chain modeling unit includes a vibration chain building module and a temperature chain building module; The vibration chain construction module is used to construct a vibration-concentration causal fusion model based on electrical quantity data, vibration spectrum data, gas concentration data, and material temperature data. This vibration-concentration causal fusion model can predict the gas concentration driven by vibration in the future. In the process of constructing the vibration-concentration causal fusion model, the influence of electrical quantity data on material temperature data is considered and optimized. The temperature chain construction module is used to construct a total temperature model by decomposing the sources of temperature rise. Based on the total temperature model, combined with the modified thermal decomposition rate equation and the modified Henry's law, a temperature-concentration causal fusion model between temperature and gas concentration is constructed. This temperature-concentration causal fusion model can predict the gas concentration driven by temperature in the future.
[0009] As a further improvement of this technical solution, in the vibration chain construction module, the specific steps of jointly constructing the vibration-concentration causal fusion model based on material temperature data, electrical quantity data, and vibration spectrum data are as follows: S1. Real-time extraction of electrical harmonic distortion rate from electrical quantity data , based on electrical harmonic distortion Get vibration amplitude , and calculate the energy density of the vibration spectrum by combining the power spectrum density ; S2. Real-time acquisition of the elastic modulus of the insulation material based on the material temperature data , and then according to the elastic modulus The shift of the resonant frequency is calculated by the change of ; S3. Energy density based on vibration spectrum and the offset of the resonant frequency Obtain real-time vibration impact energy ; S4, based on real-time vibration impact energy Get the rate of gas release and predict the time Vibration-driven gas concentration .
[0010] As a further improvement of this technical solution, the vibration impact energy obtained in S3 Specifically: based on the offset of the resonance frequency, the key frequency band is obtained. Integrate the key frequency band around the resonant frequency and introduce the weight of the resonant frequency offset Optimize.
[0011] As a further improvement of this technical solution, in the vibration chain construction module, the influence of electrical quantity data on material temperature is considered and optimized during the construction of the vibration-concentration causal fusion model. The specific steps after optimization are as follows: S5. Real-time extraction of current effective value from electrical quantity data and winding resistance , calculate the material temperature rise caused by the current ; S6. Material heating caused by current , obtain the optimized real-time material temperature , and then get the temperature difference of the optimized reference temperature ; S7, temperature difference based on optimized reference temperature , get the optimized real-time vibration impact energy , and finally obtain the optimized Vibration-driven gas concentration .
[0012] As a further improvement of this technical solution, in the temperature chain construction module, the specific steps of constructing a temperature-concentration causal fusion model between temperature and gas concentration are as follows: S8, based on optimized material temperature And build a total temperature model based on the ambient temperature around the material ; S9, based on the thermal decomposition rate equation combined with the total temperature model , and introduce real-time dynamic saturation concentration Instead of a fixed threshold, calculate the current concentration of thermal decomposition gas ; S10. Establishing a temperature-solubility model by modifying Henry's law , real-time calculation of changes in gas solubility in oil , dynamically balancing the concentration of gases released by thermal decomposition and dissolution; S11, integrated thermal decomposition gas generation concentration Changes in solubility , output temperature driven initial total gas concentration , and finally the integral method is used to predict the Temperature-driven gas concentrations .
[0013] As a further improvement of the present technical solution, the concentration mixing prediction unit includes a concentration analysis and prediction module and a concentration anomaly alarm module; The concentration analysis and prediction module is used to convert the predicted vibration-driven gas concentration into and predicted temperature-driven gas concentrations Perform fusion analysis to generate comprehensive predictions of gas concentrations in future time periods ; The concentration abnormality alarm module is used to provide the comprehensive gas concentration prediction value based on the concentration analysis and prediction module. , monitor in real time whether it exceeds the preset concentration threshold, divide the alarm level according to the amplitude of exceeding the concentration threshold and the concentration rising rate, and generate multi-level alarm signals.
[0014] As a further improvement of this technical solution, the concentration analysis and prediction module generates a comprehensive prediction value of the gas concentration in the future time period. The specific steps are as follows: S12. Divide the time series into three scales: short-term, medium-term, and long-term. Construct an attention module for each scale and calculate the temperature response. and Contribution of S13. Quantify the contribution to obtain the gas concentration driven by temperature Total contribution ratio and temperature-driven gas concentrations Total contribution ratio ; S14, based on and Get the vibration-driven gas concentration Comprehensive prediction value of gas concentration Weight and temperature-driven gas concentrations Comprehensive prediction value of gas concentration Weight , thus obtaining the comprehensive prediction value of gas concentration in the future time period .
[0015] On the other hand, the present invention provides a method for remote monitoring of transformer faults, which is used in any one of the above-mentioned remote monitoring systems for transformer faults, comprising the following steps: S10.1. Collect multimodal data of the transformer in real time and construct a vibration-concentration causal fusion model and a temperature-concentration causal fusion model based on the multimodal data; S10.2. Predicting the vibration-driven gas concentration in the future based on the vibration-concentration causal fusion model, and predicting the temperature-driven gas concentration in the future based on the temperature-concentration causal fusion model; S10.3. Fusion analysis of the predicted vibration-driven gas concentration and temperature-driven gas concentration to obtain a comprehensive prediction of the gas concentration in the future time period; S10.4. Real-time monitoring of the comprehensive predicted value of gas concentration to determine whether it exceeds the preset concentration threshold, and the alarm level is divided according to the magnitude of the concentration threshold and the rate of concentration increase, generating a multi-level alarm signal.
[0016] Compared with the prior art, the present invention has the following beneficial effects: A remote monitoring system and method for transformer faults combines multimodal data to establish a nonlinear causal model. Gas concentration is predicted through vibration drive and temperature drive, and transformer faults are determined based on the gas concentration. This allows abnormal gas concentration to be captured at the early stage of a fault, long before visible damage or power outages occur. This allows faults to be predicted in advance and the risk of line aging to be avoided. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 It is the overall flow chart of the present invention; Figure 2 is a flow chart of the overall method of the present invention; The meaning of each number in the figure is: 1. Multimodal data acquisition unit; 2. Multi-source causal chain modeling unit; 21. Vibration chain construction module; 22. Temperature chain construction module; 3. Concentration mixing prediction unit; 31. Concentration analysis and prediction module; 32. Concentration anomaly alarm module. DETAILED DESCRIPTION
[0018] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.
[0019] See also Figure 1 As shown, a remote monitoring system for transformer faults is provided. This embodiment monitors gas concentration to identify transformer faults. Internal transformer faults (such as partial discharge, overheating, and arcing) accelerate the decomposition of insulating materials and gases in the oil, generating specific gases. The accumulation of CO and CO2 reflects the aging of the paper insulation. Abnormal gas concentrations can be detected in the early stages of a fault, long before visible damage or power outages occur. Furthermore, this embodiment combines multimodal data such as electrical quantities, vibration spectra, and temperature gradients to establish a nonlinear causal model, quantifying the weight of each factor's influence on gas generation. This allows for early prediction of faults and mitigates the risk of line aging. The system includes a multimodal data acquisition unit 1, a multi-source causal chain modeling unit 2, and a concentration hybrid prediction unit 3.
[0020] The multimodal data acquisition unit 1 is used to collect multimodal data of the transformer in real time; the multimodal data includes electrical quantity data, vibration spectrum data, gas concentration data and material temperature data; The multi-source causal chain modeling unit 2 is used to construct a nonlinear causal chain between multimodal data and quantify the contribution of each path. The nonlinear causal chain includes a vibration chain and a temperature chain. The multi-source causal chain modeling unit 2 includes a vibration chain construction module 21 and a temperature chain construction module 22. The vibration chain construction module 21 is used to construct a vibration-concentration causal fusion model based on electrical quantity data, vibration spectrum data, gas concentration data, and material temperature data. The vibration-concentration causal fusion model can predict the gas concentration driven by vibration in the future, and optimize the vibration-concentration causal fusion model by considering the influence of electrical quantity data on material temperature data during the construction process. The specific logic of the vibration chain building module 21 is: electrical quantity abnormalities lead to mechanical vibration changes, temperature affects material property changes, material property changes affect vibration response changes, mechanical vibrations lead to fault evolution, and then lead to gas generation. In the vibration chain construction module 21, the specific steps of constructing the vibration-concentration causal fusion model based on material temperature data, electrical quantity data, and vibration spectrum data are as follows: S1. Real-time extraction of electrical harmonic distortion rate from electrical quantity data , based on electrical harmonic distortion Get vibration amplitude , and calculate the energy density of the vibration spectrum by combining the power spectrum density ; It can quantify the nonlinear driving effect of electrical quantities on the vibration spectrum and reflect the amplification effect of high harmonic distortion rate on vibration energy.
[0021] Vibration amplitude and Exponential relationship: Where, is the proportional coefficient of electrical harmonic distortion rate to vibration amplitude; is the sensitivity coefficient, describing the harmonic distortion rate Exponential amplification effect on vibration amplitude; Energy density of the vibration spectrum It is represented by the power spectral density: Where, is the starting frequency of the frequency interval; is the upper limit frequency of the frequency interval; is the power spectral density; Where, is the reference power spectral density, at the reference frequency PSD value at ; is the reference frequency, the reference point used to normalize the PSD; is the current frequency; is the attenuation index, which describes the slope of PSD changing with frequency; S2. Real-time acquisition of the elastic modulus of the insulation material based on the material temperature data , and then according to the elastic modulus The shift of the resonant frequency is calculated by the change of ; The relationship between material elastic modulus and temperature: Where, is the temperature difference from the reference temperature; is the temperature coefficient, which describes the proportional relationship between the elastic modulus and temperature; The elastic modulus at the reference temperature is usually the value under standard test conditions; ; is the current measured temperature; is the normalized reference temperature; The relationship between the elastic modulus of the material and the magnitude of the resonant frequency shift: Where, is the reference temperature The fundamental resonant frequency of S3. Energy density based on vibration spectrum and the offset of the resonant frequency Obtain real-time vibration impact energy ; Obtain vibration impact energy in S3 Specifically: based on the offset of the resonance frequency, the key frequency band is obtained. Integrate the key frequency band around the resonant frequency and introduce the weight of the resonant frequency offset Optimize; S4, based on real-time vibration impact energy Get the rate of gas release and predict the time Vibration-driven gas concentration , , , ; Gas release rate: Where, is the gas concentration at the initial moment; is the proportionality coefficient, reflecting the influence of material properties and environmental conditions on gas generation; is the vibration weight index (ranging from 0.5 to 1.5), reflecting the nonlinear amplification effect of vibration on gas generation; The activation energy describes the minimum energy required for a chemical reaction to proceed and affects the difficulty of the gas generation reaction. is the ideal gas constant; For time Material temperature at the moment; is the time variable representing the time from the initial moment to the current moment.
[0022] In the vibration chain construction module 21, the influence of electrical quantity data on material temperature is considered and optimized during the construction of the vibration-concentration causal fusion model. The specific steps after optimization are as follows: S5. Real-time extraction of current effective value from electrical quantity data and winding resistance , calculate the material temperature rise caused by the current ; S6. Material heating caused by current , obtain the optimized real-time material temperature , and then get the temperature difference of the optimized reference temperature ; Where, For The material heats up due to the current at that moment; S7, temperature difference based on optimized reference temperature , get the optimized real-time vibration impact energy , and finally obtain the optimized Vibration-driven gas concentration ; The temperature chain construction module 22 is used to construct a total temperature model by decomposing the temperature rise source, and to construct a temperature-concentration causal fusion model between temperature and gas concentration based on the total temperature model combined with the modified thermal decomposition rate equation and the modified Henry's law. The temperature-concentration causal fusion model can predict the gas concentration driven by temperature in the future. Specifically, by quantifying the direct effect of temperature on gas generation (changes in gas solubility in oil), the evolution of gas concentration is dynamically predicted. Based on real-time material temperature data (including the contribution of electrical temperature rise) and material characteristic parameters, combined with the modified Henry's law model and thermal decomposition rate equation, the temperature-driven gas release rate is calculated, and the gas concentration forecast for the future time period is generated. .
[0023] In the temperature chain construction module 22, the specific steps of constructing the temperature-concentration causal fusion model between temperature and gas concentration are as follows: S8, based on optimized material temperature And build a total temperature model based on the ambient temperature around the material ; S9, based on the thermal decomposition rate equation combined with the total temperature model , and introduce real-time dynamic saturation concentration Instead of a fixed threshold, calculate the current concentration of thermal decomposition gas , which can reflect the nonlinear change of gas solubility in oil with temperature; Where, is the pre-exponential factor, which is related to the type of insulation material; is the concentration of decomposable substances in oil (updated through oil quality testing); S10. Establishing a temperature-solubility model by modifying Henry's law , real-time calculation of changes in gas solubility in oil , dynamically balancing the concentration of gases released by thermal decomposition and dissolution; Where, is the temperature-dependent Henry's constant, which describes the solubility of gas in oil and generally decreases with increasing temperature; The partial pressure of gas in oil can be inferred from the gas concentration or measured directly by a pressure sensor; is the solubility temperature coefficient, which reflects the sensitivity of solubility to temperature changes and requires experimental calibration (usually negative, indicating that solubility decreases with increasing temperature); is the reference temperature Gas solubility under ; Current total temperature Gas solubility under ; The base temperature for solubility calculation (such as the standard operating temperature of 25°C or the initial temperature) is used to define the reference point for solubility changes with temperature.
[0024] S11, integrated thermal decomposition gas generation concentration Changes in solubility , output temperature driven initial total gas concentration , and finally the integral method is used to predict the Temperature-driven gas concentrations .
[0025] The concentration mixing prediction unit 3 is used to predict the gas concentration change in the future time period based on the results of the vibration chain and the temperature chain, and generate an alarm signal when the gas concentration exceeds a preset threshold; The concentration mixing prediction unit 3 includes a concentration analysis and prediction module 31 and a concentration abnormality alarm module 32; The concentration analysis and prediction module 31 is used to convert the predicted vibration-driven gas concentration into and predicted temperature-driven gas concentrations Perform fusion analysis to generate comprehensive predictions of gas concentrations in future time periods ; The concentration analysis and prediction module 31 generates a comprehensive prediction value of the gas concentration in the future time period The specific steps are as follows: S12. Divide the time series into three scales: short-term, medium-term, and long-term. Construct an attention module for each scale and calculate the temperature response. and Contribution of S13. Quantify the contribution to obtain the gas concentration driven by temperature Total contribution ratio and temperature-driven gas concentrations Total contribution ratio ; Where, To quantify the weight of temperature on vibration driving at different time scales; is the coupled eigenvector and the eigenvector of multi-source data; To quantify the weight of temperature on temperature driving at different time scales; is the time scale; S14, based on and Get the vibration-driven gas concentration Comprehensive prediction value of gas concentration Weight and temperature-driven gas concentrations Comprehensive prediction value of gas concentration Weight , thus obtaining the comprehensive prediction value of gas concentration in the future time period ; The concentration abnormality alarm module 32 is used to generate the comprehensive gas concentration prediction value based on the concentration analysis and prediction module 31. , monitor in real time whether it exceeds the preset concentration threshold, divide the alarm level according to the amplitude of exceeding the concentration threshold and the concentration rising rate, generate multi-level alarm signals, and trigger the local sound and light alarm device at the same time; the concentration abnormality alarm module 32 also supports adaptive threshold adjustment function, dynamically optimizes the alarm threshold according to the equipment operation age, historical fault records and environmental conditions.
[0026] Example 2: See also Figure 1 As shown, a transformer fault remote monitoring method is provided, which is used in any one of the above transformer fault remote monitoring systems, comprising the following steps: S10.1. Collect multimodal data of the transformer in real time and construct a vibration-concentration causal fusion model and a temperature-concentration causal fusion model based on the multimodal data; S10.2. Predicting the vibration-driven gas concentration in the future based on the vibration-concentration causal fusion model, and predicting the temperature-driven gas concentration in the future based on the temperature-concentration causal fusion model; S10.3. Fusion analysis of the predicted vibration-driven gas concentration and temperature-driven gas concentration to obtain a comprehensive prediction of the gas concentration in the future time period; S10.4. Real-time monitoring of the comprehensive predicted value of gas concentration to determine whether it exceeds the preset concentration threshold, and the alarm level is divided according to the magnitude of the concentration threshold and the rate of concentration increase, generating a multi-level alarm signal.
[0027] The basic principles, main features, and advantages of the present invention are shown and described above. It should be understood by those skilled in the art that the present invention is not limited to the above-described embodiments. The above-described embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention, and such changes and modifications fall within the scope of the invention claimed.
Claims
1. A transformer fault remote monitoring system, characterized by: include: A multimodal data acquisition unit (1), the multimodal data acquisition unit (1) being used to acquire multimodal data of the transformer in real time; A multi-source causal chain modeling unit (2), wherein the multi-source causal chain modeling unit (2) is used to construct a nonlinear causal chain between multimodal data and quantify the contribution of each path, wherein the nonlinear causal chain includes a vibration chain and a temperature chain; A concentration mixing prediction unit (3) is used to predict the change of gas concentration in a future time period based on the results of the vibration chain and the temperature chain, and to generate an alarm signal when the gas concentration exceeds a preset threshold.
2. The transformer fault remote monitoring system according to claim 1, characterized in that: The multimodal data includes electrical quantity data, vibration spectrum data, gas concentration data, and material temperature data.
3. The transformer fault remote monitoring system according to claim 2, characterized in that: The multi-source causal chain modeling unit (2) includes a vibration chain building module (21) and a temperature chain building module (22); Among them, the vibration chain construction module (21) is used to construct a vibration-concentration causal fusion model based on electrical quantity data, vibration spectrum data, gas concentration data and material temperature data. The vibration-concentration causal fusion model can predict the gas concentration driven by vibration in the future time, and consider the influence of electrical quantity data on material temperature data in the process of constructing the vibration-concentration causal fusion model for optimization; The temperature chain building module (22) is used to construct a total temperature model by decomposing the temperature rise sources. Based on the total temperature model combined with the modified thermal decomposition rate equation and the modified Henry's law, a temperature-concentration causal fusion model between temperature and gas concentration is constructed. The temperature-concentration causal fusion model can predict the gas concentration driven by temperature in the future.
4. The transformer fault remote monitoring system according to claim 3 is characterized in that: In the vibration chain construction module (21), the specific steps of constructing the vibration-concentration causal fusion model based on material temperature data, electrical quantity data, and vibration spectrum data are as follows: S1. Real-time extraction of electrical harmonic distortion rate from electrical quantity data , based on electrical harmonic distortion Get vibration amplitude , and calculate the energy density of the vibration spectrum by combining the power spectrum density ; S2. Real-time acquisition of the elastic modulus of the insulation material based on the material temperature data , and then according to the elastic modulus The shift of the resonant frequency is calculated by the change of ; S3. Energy density based on vibration spectrum and the offset of the resonant frequency Obtain real-time vibration impact energy ; S4, based on real-time vibration impact energy Get the rate of gas release and predict the time Vibration-driven gas concentration .
5. The transformer fault remote monitoring system according to claim 4 is characterized in that: Vibration impact energy is obtained in S3 Specifically: based on the offset of the resonance frequency, the key frequency band is obtained. Integrate the key frequency band around the resonant frequency and introduce the weight of the resonant frequency offset Optimize.
6. The transformer fault remote monitoring system according to claim 5, characterized in that: In the vibration chain construction module (21), the influence of electrical quantity data on material temperature is considered and optimized in the process of constructing the vibration-concentration causal fusion model. The specific steps after optimization are as follows: S5. Real-time extraction of current effective value from electrical quantity data and winding resistance , calculate the material temperature rise caused by the current ; S6. Material heating caused by current , obtain the optimized real-time material temperature , and then get the temperature difference of the optimized reference temperature ; S7, temperature difference based on optimized reference temperature , get the optimized real-time vibration impact energy , and finally obtain the optimized Vibration-driven gas concentration .
7. The transformer fault remote monitoring system according to claim 6, characterized in that: In the temperature chain construction module (22), the specific steps of constructing the temperature-concentration causal fusion model between temperature and gas concentration are as follows: S8, based on optimized material temperature And build a total temperature model based on the ambient temperature around the material ; S9, based on the thermal decomposition rate equation combined with the total temperature model , and introduce real-time dynamic saturation concentration Instead of a fixed threshold, calculate the current concentration of thermal decomposition gas ; S10. Establishing a temperature-solubility model by modifying Henry's law , real-time calculation of changes in gas solubility in oil , dynamically balancing the concentration of gases released by thermal decomposition and dissolution; S11, integrated thermal decomposition gas concentration Changes in solubility , output temperature driven initial total gas concentration , and finally the integral method is used to predict the Temperature-driven gas concentrations .
8. The transformer fault remote monitoring system according to claim 7, characterized in that: The concentration mixing prediction unit (3) includes a concentration analysis prediction module (31) and a concentration abnormality alarm module (32); The concentration analysis and prediction module (31) is used to convert the predicted vibration-driven gas concentration into and predicted temperature-driven gas concentrations Perform fusion analysis to generate comprehensive predictions of gas concentrations in future time periods ; The concentration abnormality alarm module (32) is used to calculate the comprehensive gas concentration prediction value provided by the concentration analysis and prediction module (31). , monitor in real time whether it exceeds the preset concentration threshold, divide the alarm level according to the amplitude of exceeding the concentration threshold and the concentration rising rate, and generate multi-level alarm signals.
9. The transformer fault remote monitoring system according to claim 8, characterized in that: The concentration analysis and prediction module (31) generates a comprehensive prediction value of the gas concentration in the future time period The specific steps are as follows: S12. Divide the time series into three scales: short-term, medium-term, and long-term. Construct an attention module for each scale and calculate the temperature response. and Contribution of S13. Quantify the contribution to obtain the gas concentration driven by temperature Total contribution ratio and temperature-driven gas concentrations Total contribution ratio ; S14, based on and Get the vibration-driven gas concentration Comprehensive prediction value of gas concentration Weight and temperature-driven gas concentrations Comprehensive prediction value of gas concentration Weight , thus obtaining the comprehensive prediction value of gas concentration in the future time period .
10. A transformer fault remote monitoring method, used in a transformer fault remote monitoring system according to any one of claims 1 to 9, characterized in that: The steps include: S10.
1. Collect multimodal data of the transformer in real time and construct a vibration-concentration causal fusion model and a temperature-concentration causal fusion model based on the multimodal data; S10.
2. Predict vibration-driven gas concentrations in the future based on the vibration-concentration causal fusion model, and simultaneously predict temperature-driven gas concentrations in the future based on the temperature-concentration causal fusion model; S10.
3. Fusion analysis of the predicted vibration-driven gas concentration and temperature-driven gas concentration to obtain a comprehensive prediction of gas concentration in the future time period; S10.
4. Real-time monitoring of the comprehensive predicted value of gas concentration to determine whether it exceeds the preset concentration threshold, and the alarm level is divided according to the magnitude of the concentration threshold and the rate of concentration increase, generating a multi-level alarm signal.