A GIS fault gas classification and identification system and method based on spectral absorption principle

Abstract

The present invention discloses a GIS fault gas classification and identification system and method based on the principle of spectral absorption, which relates to the technical field of GIS fault identification. A GIS fault gas classification and identification system based on the principle of spectral absorption includes: a spectrum acquisition module, a spectrum correction module and a fault identification module. The present invention makes fault diagnosis more accurate and dynamic through continuous monitoring and environmental correction; steps S1 and S2 ensure that accurate GIS local discharge gas spectra are obtained at different time nodes, and continuously record environmental parameters through the environmental sensor array. Through environmental correction, the interference of environmental factors on spectral data is eliminated, and the accuracy of fault gas identification is enhanced; the three-level fault classification diagnosis in step S3 helps to more accurately locate the fault type by distinguishing single faults, compound faults and inherent material faults, and provides targeted solutions for equipment maintenance.

Description

Technical Field

[0001] The present invention relates to the technical field of GIS fault identification, and in particular to a GIS fault gas classification and identification system and method based on the spectral absorption principle. Background Art

[0002] In the existing field of GIS partial discharge diagnosis, the severity of GIS faults can be effectively assessed by sampling and analyzing the gas in the interval after the partial discharge occurs. Currently, the identification of typical defects mainly relies on gas decomposition product analysis, that is, detecting four to eight characteristic gases produced by the decomposition of SF6 gas, such as H2S, SO2, and CO. However, analysis based only on these typical defects and a limited number of characteristic gases often makes the diagnostic algorithm overly sensitive to parameters, easily leading to overfitting, and is only applicable to typical GIS partial discharge types, making it difficult to accurately identify complex or compound defects.

[0003] Therefore, it is necessary to introduce spectral imaging technology and combine it with an improved feature recognition algorithm to conduct a more comprehensive and integrated diagnosis of partial discharge; by establishing a feature recognition model for fault gas, the classification and identification of single defects, composite defects and inherent material defects can be realized, thereby improving the accuracy and adaptability of GIS partial discharge fault diagnosis. Summary of the Invention

[0004] The present invention aims to provide a GIS fault gas classification and identification system and method based on the spectral absorption principle, so as to improve the accuracy and adaptability of GIS partial discharge fault diagnosis.

[0005] Step S1: obtaining a GIS partial discharge gas spectrum; recording an initial test node; recording the GIS partial discharge gas spectrum of the initial test node as GIS partial discharge gas spectrum F1;

[0006] Step S2: Continuously obtain the GIS partial discharge gas spectrum F based on the initial test node and dynamic time interval i , i=1, 2, ..., I, i represents the order of the time series nodes, I is the total number of all monitoring time series nodes; Based on the deployed environmental sensor array, the GIS environmental parameters C i ; GIS partial discharge gas spectrum F i Perform environmental correction to obtain the corrected GIS partial discharge gas spectrum J i ;

[0007] Step S3: Correcting the GIS partial discharge gas spectrum J i Perform three-level fault classification diagnosis and obtain the GIS fault gas classification identification result G i ; Among them, GIS fault gas classification identification result G iContains single fault identification results, composite fault identification results and inherent material fault identification results; dynamic time interval is based on the GIS fault gas classification identification result G of the current time series time node i Adaptively adjust the next time series node.

[0008] As a preferred technical solution of the present invention, in step S2, the GIS partial discharge gas spectrum F i Specific steps for environmental remediation include:

[0009] Obtaining GIS Partial Discharge Gas Spectrum F Using Fourier Transform Infrared Spectroscopy i ;

[0010] Based on GIS partial discharge gas spectrum F i and GIS environmental parameters C i Establish a time-space coordinate system and obtain the spectrum-environment fusion spectrum imaging X i ;

[0011] Spectral imaging based on spectrum-environment fusion i Extract dB amplitude attenuation characteristics, phase coherence coefficient and time-frequency energy distribution;

[0012] Based on dB amplitude attenuation characteristics, phase coherence coefficient and time-frequency energy distribution and GIS environmental parameters C i The pre-trained baseline-correction mapping model is used for feature enhancement to obtain the corrected GIS partial discharge gas spectrum J i .

[0013] As a preferred technical solution of the present invention, the specific steps of performing three-level fault classification diagnosis include:

[0014] Three-level fault classification diagnosis is performed based on the GIS partial discharge fault classification and identification model; the GIS partial discharge fault classification and identification model includes a feature extraction layer, a component screening layer, and a three-level fault diagnosis layer;

[0015] The feature extraction layer is used to correct the GIS partial discharge gas spectrum J i Extract all characteristic gas fingerprint peak features;

[0016] For any characteristic gas fingerprint peak feature, extract the gas concentration matrix in the component screening layer; extract the time evolution characteristics and spatial distribution characteristics based on the gas concentration matrix;

[0017] The three-level fault diagnosis layer is used to perform three-level fault diagnosis based on all gas concentration matrices and the corresponding time evolution characteristics and spatial distribution characteristics, and obtain the GIS fault gas classification and identification result G i .

[0018] As a preferred technical solution of the present invention, the specific steps of determining the dynamic time interval include:

[0019] At the time node i, based on the GIS fault gas classification identification result G i Identify the gas concentration gradient vectors of all characteristic gases; based on GIS environmental parameter C i Calculate the current environmental severity index;

[0020] The pre-trained fault gas-environment variation model is used to predict and analyze the current environmental severity index and gas concentration gradient vector to obtain gas scale prediction factors and environmental scale prediction factors.

[0021] The dynamic time interval when the time series time node is i+1 is determined based on the time series adaptive calculation function using the gas concentration gradient vector, the current environmental severity index, the gas scale prediction factor and the environmental scale prediction factor.

[0022] As a preferred technical solution of the present invention, the three-level fault diagnosis layer includes a first fault diagnosis forest, a second fault diagnosis forest, and a third fault diagnosis forest, and n decision trees with a depth of m are constructed in each forest;

[0023] In the first fault diagnosis forest, a single special spectral feature is set as a decision tree splitting node; in the second fault diagnosis forest, a composite special spectral feature is set as a decision tree splitting node; in the third fault diagnosis forest, an inherent special spectral feature is set as a decision tree splitting node;

[0024] The gas concentration matrix and the corresponding time evolution characteristics and spatial distribution characteristics are input into the three-level fault diagnosis layer for fault identification, and the single fault identification results, composite fault identification results and inherent material fault identification results are obtained;

[0025] Combine the single fault identification results, compound fault identification results and inherent material fault identification results to obtain the GIS fault gas classification identification result G i .

[0026] As a preferred technical solution of the present invention, the basic model of the GIS partial discharge fault classification and identification model is a random forest model.

[0027] A GIS fault gas classification and identification system based on spectral absorption principle, including:

[0028] The spectrum acquisition module includes a spectrum acquisition unit and an interval adjustment unit. The spectrum acquisition unit is used to obtain the GIS partial discharge gas spectrum; record the initial test node; record the GIS partial discharge gas spectrum of the initial test node as the GIS partial discharge gas spectrum F1; the interval adjustment unit is used to adjust the dynamic time interval, which is based on the GIS fault gas classification and identification result G at the current time sequence time node. i Adaptively adjust the next time series node;

[0029] The spectrum correction module includes a spectrum correction unit for continuously acquiring the GIS partial discharge gas spectrum F based on the initial test node and dynamic time interval i , i=1, 2, ..., I, i represents the order of the time series nodes, I is the total number of all monitoring time series nodes; Based on the deployed environmental sensor array, the GIS environmental parameters C i ; GIS partial discharge gas spectrum F i Perform environmental correction to obtain the corrected GIS partial discharge gas spectrum J i ;

[0030] Fault identification module, including classification identification unit, used to correct GIS partial discharge gas spectrum J i Perform three-level fault classification diagnosis and obtain the GIS fault gas classification identification result G i ; Among them, GIS fault gas classification identification result G i It includes single fault identification results, composite fault identification results and inherent material fault identification results.

[0031] The present invention has the following advantages:

[0032] 1. The present invention makes fault diagnosis more accurate and dynamic through continuous monitoring and environmental correction. Steps S1 and S2 ensure that accurate GIS partial discharge gas spectra are obtained at different time points, and environmental parameters are continuously recorded through the environmental sensor array. Through environmental correction, the interference of environmental factors on spectral data is eliminated, and the accuracy of fault gas identification is enhanced. The three-level fault classification diagnosis in step S3 helps to more accurately locate the fault type by distinguishing between single faults, compound faults and inherent material faults, and provides targeted solutions for equipment maintenance.

[0033] 2. The present invention divides the fault diagnosis process into three stages, which can achieve accurate step-by-step classification of fault types. This hierarchical structure helps to comprehensively consider various possible fault causes from multiple angles, reducing misjudgments and missed judgments; single fault diagnosis can accurately identify abnormal changes in a specific gas, composite fault diagnosis can consider the synergistic effects between multiple gases, and inherent material fault diagnosis can reveal the degradation characteristics of equipment materials; through composite spectral features and inherent features, complex fault modes can be processed and separated. Since these fault modes may produce similar gas types, a single fault diagnosis method is difficult to effectively distinguish, and composite features can reveal chemical reactions or physical coupling effects between gases, helping to achieve higher-precision classification; the step-by-step diagnosis mechanism from single fault to composite fault to inherent material fault can improve the resolution of fault diagnosis layer by layer and enhance the reliability of the final classification results. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a structural diagram of a GIS fault gas classification and identification system based on the spectral absorption principle adopted in an embodiment of the present invention.

[0035] Figure 2 The figure is a flow chart of a method for classifying and identifying GIS fault gases based on the principle of spectral absorption, which is adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0036] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0037] Example 1, a GIS fault gas classification and identification method based on the principle of spectral absorption, see Figure 2 As shown, the following steps are included:

[0038] Step S1: obtaining a GIS partial discharge gas spectrum; recording an initial test node; recording the GIS partial discharge gas spectrum of the initial test node as GIS partial discharge gas spectrum F1;

[0039] The common method to obtain the GIS partial discharge gas spectrum is to use Fourier transform infrared spectrometer and laser absorption spectroscopy technology. Spectral absorption refers to the absorption of light energy by gas molecules under the irradiation of light of a specific wavelength and the generation of characteristic absorption spectral lines. Each gas molecule has its own specific absorption spectral line, and these spectral lines correspond to different vibration and rotation energy level transitions. By analyzing these spectral lines, the type and concentration of the gas can be identified; by irradiating the laser into the gas to be tested, different gases have characteristic absorption of the laser light wave. When the light wave passes through the gas, the intensity of the light is weakened due to absorption. By measuring the change in light intensity, the concentration of the gas can be inferred. This is especially true for gases generated during partial discharge; FTIR (Fourier transform infrared spectroscopy) uses the absorption of infrared light on gas to accurately measure the intensity change of infrared light of different wavelengths after passing through the gas sample. Through FTIR, the vibration mode of the gas molecules can be analyzed to derive the type and concentration of the gas;

[0040] In this embodiment, fault classification and identification are performed on the gas products produced by the SF6 gas reaction after partial discharge in GIS. The main components of the gas produced by the decomposition reaction of SF6 gas are: H2S, HF, SO2, SiF4, SF4, CF4, CO, COS, CO2, SOF4, SOF2, SO2F2 and S2F 10 ;

[0041] Step S2: Continuously obtain the GIS partial discharge gas spectrum F based on the initial test node and dynamic time interval i , i=1, 2, ..., I, i represents the order of the time series nodes, I is the total number of all monitoring time series nodes; Based on the deployed environmental sensor array, the GIS environmental parameters C i ; GIS partial discharge gas spectrum F i Perform environmental correction to obtain the corrected GIS partial discharge gas spectrum J i ;

[0042] The GIS environmental sensor array includes temperature sensors, humidity sensors, air pressure sensors, and vibration sensors;

[0043] In step S2, the GIS partial discharge gas spectrum F i Specific steps for environmental remediation include:

[0044] Obtaining GIS Partial Discharge Gas Spectrum F Using Fourier Transform Infrared Spectroscopy i Fourier transform infrared spectroscopy (FTIR) obtains the absorption spectrum of the gas by analyzing the absorption spectrum of the gas sample in the infrared light region. Different gas molecules have unique absorption characteristics under specific wavelengths of infrared light. The GIS partial discharge gas spectrum F obtained by the FTIR instrument iContains information about gas concentration and spectral characteristics; converts the gas absorption signal to the frequency domain, analyzes the absorption peaks of gas at different wavelengths through mathematical transformation, and uses the least squares fitting method to fit the gas concentration based on the known gas type and its characteristic absorption wavelength;

[0045] Based on GIS partial discharge gas spectrum F i and GIS environmental parameters C i Establish a time-space coordinate system and obtain the spectrum-environment fusion spectrum imaging X i Combine GIS environmental parameters with the partial discharge gas spectrum to create a spatiotemporal coordinate system. By fusing the gas spectrum and environmental data, an environmentally corrected spectral imaging is obtained. Environmental parameters can affect the behavior and absorption characteristics of gas molecules. Therefore, fusing the spectrum and environmental data can more accurately reflect the true concentration and properties of the gas. This step can effectively combine environmental information, reduce the interference of environmental changes on spectral data, and provide more accurate data support for subsequent feature extraction and correction.

[0046] Spectral imaging based on spectrum-environment fusion i Extract dB amplitude attenuation characteristics, phase coherence coefficient and time-frequency energy distribution;

[0047] Based on dB amplitude attenuation characteristics, phase coherence coefficient and time-frequency energy distribution and GIS environmental parameters C i The pre-trained baseline-correction mapping model is used for feature enhancement to obtain the corrected GIS partial discharge gas spectrum J i ;

[0048] Among them, in the pre-trained benchmark-correction mapping model, feature recognition is performed based on dB amplitude attenuation characteristics, phase coherence coefficient and time-frequency energy distribution to obtain temperature influencing factors, humidity influencing factors and air pressure influencing factors. The multivariate weighted regression formula is used for feature enhancement to obtain the corrected GIS partial discharge gas spectrum J. i ;

[0049] The dB amplitude attenuation feature is used to enhance the amplitude dimension, the phase coherence coefficient is used to enhance the phase dimension, and the time-frequency energy distribution is used to enhance the frequency domain dimension. The dB amplitude attenuation feature reflects the concentration change of the gas by analyzing the amplitude attenuation of the spectral signal. The phase coherence coefficient obtains the coherence change between different gases by analyzing the phase information of the spectral signal, reflecting the physical state of the gas. The time-frequency energy distribution obtains the signal distribution in time and frequency by performing time-frequency analysis on the spectral signal, revealing the energy characteristics of the partial discharge process. By extracting these features, the gas behavior during the GIS partial discharge process can be more comprehensively described, and the necessary input data for the next step of feature enhancement and correction can be provided.

[0050] The spectral characteristics of gas may change under different environmental conditions. For example, an increase in temperature may increase the molecular motion speed of the gas, thereby affecting the characteristic peak position and intensity of its absorption spectrum. By performing environmental correction on the spectral data, the interference of environmental factors on the spectral signal can be effectively eliminated, thereby improving the accuracy of the spectral data. By combining environmental parameters with spectral information, the true concentration of the gas can be more accurately reflected, avoiding misjudgment caused by environmental fluctuations. This helps to reduce the interference of the environment on spectral analysis and improve the accuracy of fault gas detection. dB amplitude attenuation can help to estimate the gas concentration more accurately, especially when the gas concentration is low. Enhancing the amplitude dimension helps to improve Signal-to-noise ratio; Phase coherence coefficient can reveal the phase behavior of gas molecules and help distinguish different gas types, especially in complex gas mixtures; Time-frequency energy distribution helps identify the dynamic behavior of gas molecules, especially the instantaneous changes generated during the discharge process; Enhanced multi-dimensional features not only improve the accuracy of each dimension, but also make spectral features more recognizable in different analysis models, thereby improving the performance of fault diagnosis and gas prediction; Corrected spectral data can accurately identify the gas composition caused by partial discharge, because partial discharge produces specific gases with unique absorption characteristics in the spectrum. Corrected spectral data can more effectively reveal the characteristics of these fault gases;

[0051] Step S3: Correcting the GIS partial discharge gas spectrum J i Perform three-level fault classification diagnosis and obtain the GIS fault gas classification identification result G i ; Among them, GIS fault gas classification identification result G i Contains single fault identification results, composite fault identification results and inherent material fault identification results; dynamic time interval is based on the GIS fault gas classification identification result G of the current time series time node i Adaptively adjust the next time series node;

[0052] The specific steps for three-level fault classification diagnosis include:

[0053] A three-level fault classification diagnosis is performed based on the GIS partial discharge fault classification and identification model. The GIS partial discharge fault classification and identification model includes a feature extraction layer, a component screening layer, and a three-level fault diagnosis layer. The basic model of the GIS partial discharge fault classification and identification model is the random forest model.

[0054] This step involves using the GIS partial discharge fault classification and identification model to classify partial discharge faults. The model is typically composed of multiple layers, including a feature extraction layer, a component screening layer, and a three-level fault diagnosis layer. Each layer is responsible for different tasks, extracting important information from the spectral data and processing it to ultimately derive the fault diagnosis results. The input data is the corrected GIS partial discharge gas spectrum data. Through the three layers of feature extraction, component screening, and fault diagnosis, the model gradually extracts and analyzes features, and finally outputs the fault classification results.

[0055] The feature extraction layer is used to correct the GIS partial discharge gas spectrum J i Extract all characteristic gas fingerprint peak features;

[0056] The goal of the feature extraction layer is to extract all characteristic gas fingerprint peaks from the corrected GIS partial discharge gas spectrum. The gases generated by partial discharge have unique absorption characteristics within a specific wavelength range. The characteristic gas fingerprint peaks can provide key information for subsequent fault classification. The gas fingerprint peaks in the spectrum are extracted through Fourier transform or other spectral line analysis methods. Each gas molecule has its own specific absorption peak. By extracting the wavelengths corresponding to these peaks, characteristic information of each gas can be obtained.

[0057] For any characteristic gas fingerprint peak feature, the gas concentration matrix is extracted in the component screening layer; the time evolution characteristics and spatial distribution characteristics are extracted based on the gas concentration matrix; the component screening layer is used to extract the gas concentration matrix corresponding to the characteristic gas fingerprint peak feature, and further analyze the time evolution characteristics and spatial distribution characteristics of the gas. By deeply exploring these characteristics, the change pattern of gas concentration over time and its distribution at different spatial locations can be revealed; by extracting the time evolution characteristics and spatial distribution characteristics, the dynamic process of gas concentration change can be fully understood and the development trend of faults can be identified; the time and spatial distribution characteristics are crucial for distinguishing different types of partial discharge faults; for example, some types of faults may cause a rapid increase in gas concentration, while other types may cause long-term low concentration changes;

[0058] The three-level fault diagnosis layer is used to perform three-level fault diagnosis based on all gas concentration matrices and the corresponding time evolution characteristics and spatial distribution characteristics, and obtain the GIS fault gas classification and identification result G i The goal of the third-level fault diagnosis layer is to classify faults based on all gas concentration matrices and their corresponding temporal evolution characteristics and spatial distribution characteristics.

[0059] In the three-level fault diagnosis layer, there are the first fault diagnosis forest, the second fault diagnosis forest and the third fault diagnosis forest. In each forest, n decision trees with a depth of m are constructed.

[0060] In the first fault diagnosis forest, a single special spectral feature is set as a decision tree splitting node; in the second fault diagnosis forest, a composite special spectral feature is set as a decision tree splitting node; in the third fault diagnosis forest, an inherent special spectral feature is set as a decision tree splitting node;

[0061] The gas concentration matrix and its corresponding temporal evolution and spatial distribution characteristics are input into the three-level fault diagnosis layer for fault identification. Single fault identification results, composite fault identification results, and inherent material fault identification results are obtained. Among them, the fault identification results include free conductive particles, sharp electrodes, corona discharge, surface-fixed copper particles, surface-fixed aluminum particles, corona discharge + free conductive particles, sharp electrodes + surface-fixed copper particles, and sharp electrodes + free conductive particles.

[0062] Combine the single fault identification results, compound fault identification results and inherent material fault identification results to obtain the GIS fault gas classification identification result G i ;

[0063] A single special spectral feature splitting node can use the strongest absorption peak of a single gas as the first splitting feature in the decision tree. For example, the absorption peak intensity of SO2 at 4.3μm, the absorption peak slope of HF at 2.5μm, or the peak signal-to-noise ratio of H2S at 3.7μm. This allows rapid identification of key fault gases through a single significant feature, eliminating interference from non-critical bands, and improving the efficiency of subsequent processing. A composite special spectral feature splitting node can use dynamic combined features based on the synergistic effects of multiple gases, such as the H2S / SO2 concentration ratio, the CO / (CO2+SOF2) energy ratio, or the time-frequency correlation between HF and vibration signals. This is used to capture chemical reactions or physical coupling effects between gases and separate similar fault modes, such as corona discharge or suspended discharge, through combined features. An inherent special spectral feature splitting node can use physical and chemical markers combined with material degradation fingerprints, such as the characteristic absorption integral of AlF3 at 10.6μm or the coupling coefficient between SF6 decomposition products and gas pressure, to determine the degradation mechanism of associated gas features and equipment materials.

[0064] By dividing the fault diagnosis process into three stages, fault types can be accurately classified step by step. This hierarchical structure helps to comprehensively consider various possible fault causes from multiple perspectives, reducing misdiagnosis and missed diagnosis. For example, single fault diagnosis can accurately identify abnormal changes in a specific gas, composite fault diagnosis can consider the synergistic effects between multiple gases, and inherent material fault diagnosis can reveal the degradation characteristics of equipment materials. The composite spectral characteristics and inherent characteristics can be used to process and separate complex fault modes such as corona discharge and suspended discharge. Because these fault modes may produce similar gas types, single fault diagnosis methods are difficult to effectively distinguish. Composite characteristics can reveal chemical reactions or physical coupling effects between gases, helping to achieve more accurate classification. The step-by-step diagnosis mechanism from single fault to composite fault to inherent material fault can gradually improve the resolution of fault diagnosis and enhance the reliability of the final classification results.

[0065] The specific steps to determine the dynamic time interval include:

[0066] At the time node i, based on the GIS fault gas classification identification result G i Identify the gas concentration gradient vectors of all characteristic gases; based on GIS environmental parameter C i Calculate the current environmental severity index;

[0067] The pre-trained fault gas-environment variation model is used to predict and analyze the current environmental severity index and gas concentration gradient vector to obtain gas scale prediction factors and environmental scale prediction factors.

[0068] Determine the dynamic time interval when the time series time node is i+1 based on the time series adaptive calculation function using the gas concentration gradient vector, the current environmental severity index, the gas scale prediction factor and the environmental scale prediction factor;

[0069] The time series adaptive function is ΔT i+1 =ΔT min +(ΔT max -ΔT min )e -λRi ; ΔT i+1 is the dynamic time interval when the time node is i+1; ΔT max is the maximum value of the dynamic time interval, ΔT min is the minimum value of the dynamic time interval; λ is the sensitivity coefficient, which is set by professional technicians according to actual conditions; Ri is the comprehensive adjustment index at the time node i, which is determined by weighting the gas concentration gradient vector, the current environmental severity index, the gas scale prediction factor, and the environmental scale prediction factor;

[0070] The environmental severity index is used to quantify the comprehensive severity of the environment in which the GIS equipment is located, reflecting the impact of parameters such as temperature, humidity, and air pressure on the equipment failure rate, and is calculated using a weighted nonlinear model; the gas scale prediction factor is used to characterize the amplification effect of specific gas concentrations on fault development, and is used to dynamically adjust the diagnostic threshold, using a staged dynamic calculation; the environmental scale prediction factor is used to quantify the degree of interference of environmental parameters on gas detection results, and is used for spectral data compensation, and is calculated based on a pre-trained fault gas-environment variation model. The training process of the pre-trained fault gas-environment variation model includes the following steps: collecting a large amount of historical data, including gas concentrations and environmental parameters, these Data will serve as the input and output of the model; through data processing and feature extraction, gas concentration gradients, environmental severity indexes, and related time series features are extracted, which will serve as the basis for model training; a multi-level model is constructed, usually based on machine learning methods such as regression models, neural networks, etc., combining environmental parameters with gas concentration data for predictive analysis. The goal of the model is to learn the complex relationship between environmental factors and gas concentration changes, and then predict gas scale predictors and environmental scale predictors; the model is trained using historical data, and model parameters are optimized through methods such as backpropagation and gradient descent so that the model performs best on the training data. Cross-validation and other methods are usually used to evaluate the generalization ability of the model; during the training process, the validation set data is used to evaluate the model, adjust hyperparameters, and perform model tuning to ensure the model's predictive ability for new data;

[0071] The GIS fault gas classification and identification method based on the spectral absorption principle makes fault diagnosis more accurate and dynamic through continuous monitoring and environmental correction; steps S1 and S2 ensure that accurate GIS partial discharge gas spectra are obtained at different time nodes, and the environmental parameters are continuously recorded through the environmental sensor array, providing comprehensive support for the spectral data; through environmental correction, the interference of environmental factors on the spectral data is eliminated, and the accuracy of fault gas identification is enhanced; the three-level fault classification diagnosis in step S3 helps to more accurately locate the fault type by distinguishing between single faults, complex faults and inherent material faults, and provides targeted solutions for equipment maintenance; in addition, the adaptive adjustment of the dynamic time interval ensures that the fault diagnosis system can respond to changes in equipment status in real time, making the fault prediction and maintenance decision of GIS equipment more intelligent, and improving the equipment's operating stability and maintenance efficiency. This method not only enhances real-time performance, but also improves the accuracy of GIS equipment fault diagnosis, helping to reduce the risk of faults and extend the service life of the equipment.

[0072] Example 2, a GIS fault gas classification and identification system based on the principle of spectral absorption, see Figure 1 Shown, including:

[0073] The spectrum acquisition module includes a spectrum acquisition unit and an interval adjustment unit. The spectrum acquisition unit is used to obtain the GIS partial discharge gas spectrum; record the initial test node; record the GIS partial discharge gas spectrum of the initial test node as the GIS partial discharge gas spectrum F1; the interval adjustment unit is used to adjust the dynamic time interval, which is based on the GIS fault gas classification and identification result G at the current time sequence time node. i Adaptively adjust the next time series node;

[0074] The spectrum correction module includes a spectrum correction unit for continuously acquiring the GIS partial discharge gas spectrum F based on the initial test node and dynamic time interval i , i=1, 2, ..., I, i represents the order of the time series nodes, I is the total number of all monitoring time series nodes; Based on the deployed environmental sensor array, the GIS environmental parameters C i ; GIS partial discharge gas spectrum F i Perform environmental correction to obtain the corrected GIS partial discharge gas spectrum J i ;

[0075] Fault identification module, including classification identification unit, used to correct GIS partial discharge gas spectrum J i Perform three-level fault classification diagnosis and obtain the GIS fault gas classification identification result G i ; Among them, GIS fault gas classification identification result G i It includes single fault identification results, composite fault identification results and inherent material fault identification results.

[0076] It should be understood that those skilled in the art may make improvements or modifications based on the above description, and all such improvements and modifications shall fall within the scope of protection of the appended claims. Any portion of this specification not described in detail is prior art known to those skilled in the art.

Claims

1. A GIS fault gas classification and identification method based on spectral absorption principle, characterized in that: The following steps are involved: Step S1: Acquire GIS partial discharge gas spectrum; Record the initial test node; record the GIS partial discharge gas spectrum at the initial test node as GIS partial discharge gas spectrum F1; Step S2: Continuously obtain the GIS partial discharge gas spectrum F based on the initial test node and dynamic time interval i , i=1, 2, ..., I, i represents the order of the time series nodes, I is the total number of all monitoring time series nodes; Based on the deployed environmental sensor array, the GIS environmental parameters C i ; GIS partial discharge gas spectrum F i Perform environmental correction to obtain the corrected GIS partial discharge gas spectrum J i ; Step S3: Correcting the GIS partial discharge gas spectrum J i Perform three-level fault classification diagnosis and obtain the GIS fault gas classification identification result G i ; Among them, GIS fault gas classification identification result G i Contains single fault identification results, composite fault identification results and inherent material fault identification results; dynamic time interval is based on the GIS fault gas classification identification result G of the current time series time node i Adaptively adjust the next time series node; In step S2, the GIS partial discharge gas spectrum F i Specific steps for environmental remediation include: Obtaining GIS Partial Discharge Gas Spectrum F Using Fourier Transform Infrared Spectroscopy i ; Based on GIS partial discharge gas spectrum F i and GIS environmental parameters C i Establish a time-space coordinate system and obtain the spectrum-environment fusion spectrum imaging X i ; Spectral imaging based on spectrum-environment fusion i Extract dB amplitude attenuation characteristics, phase coherence coefficient and time-frequency energy distribution; Based on dB amplitude attenuation characteristics, phase coherence coefficient and time-frequency energy distribution and GIS environmental parameters C i The pre-trained baseline-correction mapping model is used for feature enhancement to obtain the corrected GIS partial discharge gas spectrum J i .

2. The GIS fault gas classification and identification method based on the spectral absorption principle according to claim 1 is characterized in that: The specific steps for three-level fault classification diagnosis include: Three-level fault classification diagnosis is performed based on the GIS partial discharge fault classification and identification model; the GIS partial discharge fault classification and identification model includes a feature extraction layer, a component screening layer, and a three-level fault diagnosis layer; The feature extraction layer is used to correct the GIS partial discharge gas spectrum J i Extract all characteristic gas fingerprint peak features; For any characteristic gas fingerprint peak feature, extract the gas concentration matrix in the component screening layer; extract the time evolution characteristics and spatial distribution characteristics based on the gas concentration matrix; The three-level fault diagnosis layer is used to perform three-level fault diagnosis based on all gas concentration matrices and the corresponding time evolution characteristics and spatial distribution characteristics, and obtain the GIS fault gas classification and identification result G i .

3. The GIS fault gas classification and identification method based on the spectral absorption principle according to claim 2 is characterized in that: The specific steps to determine the dynamic time interval include: At the time node i, based on the GIS fault gas classification identification result G i Identify the gas concentration gradient vectors of all characteristic gases; based on GIS environmental parameter C i Calculate the current environmental severity index; The pre-trained fault gas-environment variation model is used to predict and analyze the current environmental severity index and gas concentration gradient vector to obtain gas scale prediction factors and environmental scale prediction factors. The dynamic time interval when the time series time node is i+1 is determined based on the time series adaptive calculation function using the gas concentration gradient vector, the current environmental severity index, the gas scale prediction factor and the environmental scale prediction factor.

4. The GIS fault gas classification and identification method based on the spectral absorption principle according to claim 3 is characterized in that: In the three-level fault diagnosis layer, there are the first fault diagnosis forest, the second fault diagnosis forest and the third fault diagnosis forest. In each forest, n decision trees with a depth of m are constructed. In the first fault diagnosis forest, a single special spectral feature is set as a decision tree splitting node; in the second fault diagnosis forest, a composite special spectral feature is set as a decision tree splitting node; in the third fault diagnosis forest, an inherent special spectral feature is set as a decision tree splitting node; The gas concentration matrix and the corresponding time evolution characteristics and spatial distribution characteristics are input into the three-level fault diagnosis layer for fault identification, and the single fault identification results, composite fault identification results and inherent material fault identification results are obtained; Combine the single fault identification results, compound fault identification results and inherent material fault identification results to obtain the GIS fault gas classification identification result G i .

5. The GIS fault gas classification and identification method based on the spectral absorption principle according to claim 4 is characterized in that: The basic model of the GIS partial discharge fault classification and identification model is the random forest model.

6. A GIS fault gas classification and identification system based on spectral absorption principle, characterized in that: The system is a method for classifying and identifying GIS fault gases based on the spectral absorption principle as described in any one of claims 1 to 5, comprising: The spectrum acquisition module includes a spectrum acquisition unit and an interval adjustment unit. The spectrum acquisition unit is used to obtain the GIS partial discharge gas spectrum; record the initial test node; record the GIS partial discharge gas spectrum of the initial test node as the GIS partial discharge gas spectrum F1; the interval adjustment unit is used to adjust the dynamic time interval, which is based on the GIS fault gas classification and identification result G at the current time sequence time node. i Adaptively adjust the next time series node; The spectrum correction module includes a spectrum correction unit for continuously acquiring the GIS partial discharge gas spectrum F based on the initial test node and dynamic time interval i , i=1, 2, ..., I, i represents the order of the time series nodes, I is the total number of all monitoring time series nodes; Based on the deployed environmental sensor array, the GIS environmental parameters C i ; GIS partial discharge gas spectrum F i Perform environmental correction to obtain the corrected GIS partial discharge gas spectrum J i ; Fault identification module, including classification identification unit, used to correct GIS partial discharge gas spectrum J i Perform three-level fault classification diagnosis and obtain the GIS fault gas classification identification result G i ; Among them, GIS fault gas classification identification result G i It includes single fault identification results, composite fault identification results and inherent material fault identification results.

Images