A method and system for online monitoring of decomposition products of GIL / GIS mixed gas

Abstract

This application discloses an online monitoring method and system for decomposition products of GIL / GIS mixed gases, relating to the field of GIL / GIS mixed gas detection technology. This method combines mid-infrared TDLAS and ultraviolet DOAS for GIL / GIS mixed gas decomposition product detection, utilizing the complementary information of the two bands and significantly reducing the uncertainty of single-band inversion through co-component consistency constraints and weighted fusion correction. An iterative weighted least squares algorithm is used for adaptive correction: the weights of each sampling point are dynamically adjusted through a weight function to suppress noise and local interference, improving the robustness of trace component inversion. Furthermore, this method employs a dynamic absorption cross-section matrix expansion strategy based on residual principal component analysis: by analyzing the principal components of the residual vector, unmodeled interfering components are automatically identified, and the absorption cross-section matrix is ​​temporarily expanded for compensation. This strategy solves the problem of mismatch between fixed models and actual operating conditions, and has strong engineering practical value.

Description

Technical Field

[0001] This application belongs to the field of GIL (Gas Insulated Transmission Line) / GIS (Gas Insulated Switchgear) mixed gas detection technology, specifically relating to an online monitoring method and system for GIL / GIS mixed gas decomposition products. Background Technology

[0002] The insulation and arc-quenching media used in GIL / GIS are transitioning from pure SF6 gas to environmentally friendly mixtures such as SF6 / N2 and C4F7N / CO2 / O2. During long-term operation of GIL / GIS, insulation defects such as internal partial discharge and localized overheating can lead to the decomposition of the insulating gas, generating various characteristic decomposition products. Online monitoring of the types and concentrations of these decomposition products can effectively assess the insulation condition of the equipment and provide early warnings of faults.

[0003] Optical-based gas detection technologies, such as tunable diode laser absorption spectroscopy (TDLAS) and differential ultraviolet absorption spectroscopy (DOAS), have become mainstream methods for online monitoring of decomposition products due to their advantages of being non-contact, highly sensitive, and having a fast response. However, the spectroscopic detection of decomposition products in mixed gases faces a long-standing core technical challenge: cross-interference caused by severe overlap of absorption bands among multiple components. Specifically, this manifests as follows:

[0004] (1) Mid-infrared band cross-interference: The absorption peaks of SO2 (7.3μm), CF4 (7.8μm), C3F6 (8.4μm) and COF2 (5.2μm) overlap; the NF3 (8.0μm) generated by the discharge of SF6 / N2 mixed gas partially overlaps with the CF4 spectrum, making it impossible to distinguish them by conventional single-wavelength detection methods.

[0005] (2) Cross-interference in the ultraviolet band: The differential absorption cross-sections of CS2 and SO2 in the 194nm-225nm band are highly similar, making it difficult to separate them using traditional least squares fitting.

[0006] (3) Inconsistency of dual-spectrum fusion: The results of independent inversion of SO2 in the mid-infrared and ultraviolet bands may contradict each other, and there is a lack of effective fusion correction mechanism.

[0007] (4) Model mismatch problem: Existing methods assume that the absorption cross section matrix is ​​fixed and complete, but in actual GIL / GIS operation, unknown unconventional decomposition products may be generated (such as characteristic gases generated when solid insulating materials participate in discharge), resulting in serious mismatch of the inversion model. Summary of the Invention

[0008] To overcome the shortcomings of existing technologies, this application proposes an online monitoring method and system for decomposition products of GIL / GIS mixed gases.

[0009] This application is achieved through the following technical solution:

[0010] A method for online monitoring of decomposition products of GIL / GIS mixed gases includes:

[0011] The mid-infrared and ultraviolet absorption spectra of the mixed gas to be monitored are acquired in real time and preprocessed. At the same time, the corresponding standard absorption cross section matrix is ​​compensated for on-site conditions to obtain the mid-infrared absorption cross section matrix and the ultraviolet absorption cross section matrix.

[0012] A dual-spectral joint linear superposition model is established, and a concentration consistency constraint is applied to the co-components that absorb in both the mid-infrared and ultraviolet bands. The dual-spectral joint linear superposition model includes: a mid-infrared linear model, which is expressed as: the mid-infrared absorption spectrum equals the product of the concentration vector of the mid-infrared band and the mid-infrared absorption cross-section matrix plus the residual vector of the mid-infrared band; and an ultraviolet model, which is expressed as: the ultraviolet absorption spectrum equals the product of the concentration vector of the ultraviolet band and the ultraviolet absorption cross-section matrix plus the residual vector of the ultraviolet band.

[0013] Model solution steps: The dual-spectral joint linear superposition model is solved using the non-negative constraint adaptive iterative weighted least squares algorithm to obtain the concentration vectors of the mid-infrared band and the ultraviolet band, as well as the corresponding residual vectors.

[0014] Back-substitution correction step: The co-component concentrations in the mid-infrared and ultraviolet bands are weighted and fused. The fusion weight is determined by the reciprocal of the sum of squares of the average fitting residuals retrieved from each band. The weighted and fused concentrations are then used back-substituted into the dual-spectrum joint linear superposition model to correct the concentrations of the remaining components in each band.

[0015] Unknown interference identification steps: Calculate the final residual vector, perform principal component analysis on the residual vectors of multiple consecutive detections, if the sum of the variance contribution rates of the first two principal components exceeds the preset threshold, it is determined that there is an unmodeled interference component, and the first principal component vector is normalized and used as the absorption cross section of the temporary pseudo-component to expand into the absorption cross section matrix, and the model solving step, back substitution correction step and unknown interference identification step are repeated; otherwise, proceed to the next step.

[0016] Output the concentration values ​​of each decomposition product.

[0017] In some implementations, the process of constructing the standard absorption cross-section matrix includes:

[0018] For the type of mixed gas to be monitored, high-resolution absorption cross-sections are measured in advance using pure component gases under standard laboratory conditions: for the mid-infrared band, the absorption cross-section of each component at each wavenumber is measured, and the measurement range covers all characteristic absorption peaks of the target component; for the ultraviolet band, the differential absorption cross-section of each component at each wavelength is measured.

[0019] The measurement data is stored in matrix form to obtain the mid-infrared standard absorption cross section matrix and the ultraviolet standard absorption cross section matrix; wherein the mid-infrared standard absorption cross section matrix is ​​composed of the absorption cross sections of all mid-infrared absorbing components at different wavenumbers; and the ultraviolet standard absorption cross section matrix is ​​composed of the differential absorption cross sections of all ultraviolet absorbing components at different wavelengths.

[0020] In some embodiments, the real-time acquisition and preprocessing of the mid-infrared and ultraviolet absorption spectra of the mixed gas to be monitored, along with on-site condition compensation for the pre-constructed corresponding standard absorption cross-section matrix, includes:

[0021] The spectral data of the gas mixture to be tested, including mid-infrared absorption spectrum and ultraviolet absorption spectrum, are collected through an online monitoring system.

[0022] The collected mid-infrared and ultraviolet absorption spectra were preprocessed to obtain the measured mid-infrared absorbance vector and the measured ultraviolet absorbance vector, respectively.

[0023] Temperature and pressure normalization compensations were performed on the pre-constructed mid-infrared standard absorption cross section matrix and ultraviolet standard absorption cross section matrix respectively to obtain the mid-infrared absorption cross section matrix and ultraviolet absorption cross section matrix.

[0024] In some implementations, the model solving steps include:

[0025] Initial solution: The non-negative least squares algorithm without considering weights is used to solve the mid-infrared band and the ultraviolet band independently, so as to obtain the initial solution of the concentration vector of the mid-infrared band and the initial solution of the concentration vector of the ultraviolet band, as well as the corresponding initial solution of the residual vector.

[0026] Iterative weighted solution: Calculate the residual of each sampling point in the previous iteration; calculate the standard deviation of the residual of the previous iteration based on the residuals of all sampling points; update weights: for each sampling point, if the absolute value of the residual in the previous iteration is less than or equal to a threshold, the weight of the sampling point in the current iteration is set to 1; if the absolute value of the residual in the previous iteration is greater than the threshold, the weight of the sampling point in the current iteration is set to the ratio of the threshold to the absolute value of the residual in the previous iteration; where the threshold is 1.5 times the standard deviation; use a weighted non-negative least squares algorithm to solve the mid-infrared and ultraviolet bands independently to obtain the concentration vectors of the mid-infrared and ultraviolet bands corresponding to the current iteration; determine convergence: if the relative change between the concentration vector obtained in the current iteration and the concentration vector obtained in the previous iteration is less than the convergence tolerance or the current iteration number is greater than or equal to the maximum iteration number, stop the iteration and output the concentration vector obtained in the current iteration; otherwise, repeat the iterative weighted solution steps for the next iterative weighted solution.

[0027] Integration of co-component consistency constraints: Based on the concentration vectors of the mid-infrared and ultraviolet bands obtained by iterative weighted solution, the co-component concentrations are averaged, and the co-component concentrations in the mid-infrared and ultraviolet bands are fixed as the average co-component concentrations. The concentrations of the remaining components in the mid-infrared and ultraviolet bands are obtained by using a non-negative least squares algorithm without considering weights.

[0028] Determine whether the convergence criterion is met. If it is, output the final concentration vectors of the mid-infrared band and the ultraviolet band, as well as the corresponding residual vectors; otherwise, repeat the integrated steps of iterative weighted solution and co-component consistency constraint.

[0029] In some implementations, the non-negative least squares algorithm that does not consider weights specifically refers to:

[0030] First, the normalized and compensated absorption cross-section matrix is ​​multiplied by the concentration vector to be solved to obtain the model absorbance vector;

[0031] Then, the square of the Euclidean norm of the difference between the measured absorbance vector and the model absorbance vector is used as the objective function;

[0032] Find the concentration vector that minimizes the objective function, where each element in the concentration vector must be greater than or equal to 0.

[0033] The difference between the weighted non-negative least squares algorithm and the weightless non-negative least squares algorithm lies only in the objective function. The objective function in the weighted non-negative least squares algorithm is the square of the Euclidean norm of the product of the difference between the measured absorbance vector and the model absorbance vector multiplied by the square root of the weight matrix.

[0034] In some implementations, the back-substitution correction step includes:

[0035] The reciprocal of the sum of squared residuals in the mid-infrared and ultraviolet bands are used as their respective fusion weights;

[0036] For each co-component, its weighted fusion concentration is obtained;

[0037] The weighted fusion concentration of co-components is taken as a known true value and substituted into the mid-infrared linear model and the ultraviolet model respectively. The non-negative least squares method without considering the weights is used to solve for the concentrations of the remaining components in the mid-infrared and ultraviolet bands respectively.

[0038] The concentrations of the remaining components and the co-components are combined by weighted fusion to obtain the final concentration vectors for the mid-infrared band and the ultraviolet band.

[0039] In some implementations, the unknown interference identification step includes:

[0040] The most recent residual vectors are combined into a matrix, with each column of the matrix being a residual vector.

[0041] Perform singular value decomposition on the matrix to obtain a singular value diagonal matrix; the principal components in the singular value diagonal matrix are arranged in descending order.

[0042] Calculate the variance contribution rate of the first two principal components in the singular valued diagonal matrix;

[0043] If the sum of the variance contribution rates of the first two principal components is greater than the preset threshold, it is determined that there is a significant unmodeled interference component and proceeds to the next step; otherwise, the final concentration vectors of the mid-infrared and ultraviolet bands are output.

[0044] The first principal component vector is normalized, and the normalized principal component vector is used as the absorption cross section of the temporarily added pseudo-component to expand the original absorption cross section matrix.

[0045] The concentration of the newly added pseudo-component is used as the variable to be solved. The model solving steps, back substitution correction steps, and unknown interference identification steps are re-executed to obtain the final concentration vector.

[0046] In some implementations, the method further includes:

[0047] According to the preset cycle or when the system restarts, the absorption cross-sections of newly added pseudo-components are cleared, and the original absorption cross-section matrix is ​​restored.

[0048] On the other hand, this application also proposes an online monitoring system for the decomposition products of GIL / GIS mixed gases, comprising:

[0049] The gas circulation module is used to extract the mixed gas to be tested from the GIL / GIS gas chamber to the gas pool for testing. After the test is completed, the gas is purified and then refilled into the GIL / GIS gas chamber to achieve zero-emission closed-loop circulation.

[0050] The acquisition module includes a mid-infrared TDLAS detection unit and an ultraviolet DOAS detection unit; the two detection units share the same flow-through long optical path gas cell; it also includes a temperature sensor and a pressure sensor installed in the gas cell for real-time acquisition of environmental parameters.

[0051] An edge computing module, equipped with an industrial real-time operating system, is used to execute the online monitoring method for decomposition products of GIL / GIS mixed gas as described in any of the above embodiments;

[0052] Additionally, a remote monitoring platform is used to receive the concentration data uploaded by the edge computing module, and to perform visualization and historical trend analysis.

[0053] In some implementations, the edge computing module performs periodic detection according to a preset time period, and the concentration of each decomposition product obtained in each detection is stored in a local database and uploaded to the remote monitoring platform;

[0054] The remote monitoring platform plots a concentration-time curve based on the uploaded data and calculates the gas production rate.

[0055] This application proposes an online monitoring method and system for decomposition products of GIL / GIS mixed gases. It combines mid-infrared TDLAS and ultraviolet DOAS for GIL / GIS mixed gas decomposition product detection, utilizing the complementary information from the two bands and significantly reducing the uncertainty of single-band inversion through co-component consistency constraints and weighted fusion correction. An iterative weighted least squares algorithm is employed for adaptive correction: the weights of each sampling point are dynamically adjusted through a weight function to suppress noise and local interference, improving the robustness of trace component inversion. This algorithm can be extended to any quantitative analysis scenario with overlapping multi-component spectra. Furthermore, the method employs a dynamic absorption cross-section matrix expansion strategy based on residual principal component analysis: by analyzing the principal components of the residual vector, unmodeled interfering components are automatically identified, and the absorption cross-section matrix is ​​temporarily expanded for compensation. This strategy solves the problem of mismatch between fixed models and actual operating conditions, demonstrating strong engineering practical value. Attached Figure Description

[0056] The accompanying drawings, which are included to provide a further understanding of the embodiments of this application and form part of this application, do not constitute a limitation on the embodiments of this application. In the drawings:

[0057] Figure 1 This is a schematic diagram of the online monitoring method proposed in the embodiments of this application;

[0058] Figure 2 This is a schematic diagram of the model solution process proposed in the embodiments of this application;

[0059] Figure 3 This is a schematic diagram of the online monitoring system architecture proposed in an embodiment of this application. Detailed Implementation

[0060] In the following, the terms “comprising” or “may include” as used in the various embodiments of this application indicate the presence of a function, operation, or element of the invention and do not limit the addition of one or more functions, operations, or elements. Furthermore, as used in the various embodiments of this application, the terms “comprising,” “having,” and their cognates are intended only to indicate a specific feature, number, step, operation, element, component, or combination of the foregoing and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing, or adding one or more combinations of the foregoing.

[0061] In various embodiments of this application, the expression "or" or "at least one of A and / or B" includes any combination or all combinations of the words listed simultaneously. For example, the expression "A or B" or "at least one of A and / or B" may include A, may include B, or may include both A and B.

[0062] The terms used in the various embodiments of this application (such as "first," "second," etc.) may modify various constituent elements in the various embodiments, but do not limit the corresponding constituent elements. For example, the above terms do not limit the order and / or importance of the elements. The above terms are only used for the purpose of distinguishing one element from other elements. For example, a first user device and a second user device refer to different user devices, although both are user devices. For example, without departing from the scope of the various embodiments of this application, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element.

[0063] It should be noted that if a description is made of "connecting" one component to another, then the first component can be directly connected to the second component, and a third component can be "connected" between the first and second components. Conversely, when a component is "directly connected" to another component, it can be understood that there is no third component between the first and second components.

[0064] The terminology used in the various embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the various embodiments of this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of this application pertain. The terms (such as those defined in a generally used dictionary) are to be interpreted as having the same meaning as in the context of the relevant technical field and are not to be interpreted as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of this application.

[0065] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the embodiments and accompanying drawings. The illustrative embodiments and descriptions of this application are only for explaining this application and are not intended to limit this application.

[0066] Existing mixed gas detection technologies either ignore interference (resulting in high errors) or employ conventional least squares or partial least squares regression, failing to consider the ill-conditioned conditions of the absorption cross-section matrix and the non-Gaussian distribution of measured noise, and also neglecting to introduce dual-band consistency constraints. To address this, this application proposes an online monitoring method for the decomposition products of GIL / GIS mixed gases. This method enables high-precision, low-cross-interference quantitative inversion of multi-component decomposition products of mixed gases through TDLAS and DOAS dual-spectral fusion detection under GIL / GIS powered operation, combined with iterative weighted least squares adaptive correction, co-component consistency constraints, and dynamic absorption cross-section matrix correction.

[0067] Specifically, such as Figure 1 As shown, the online monitoring method proposed in this application includes the following steps:

[0068] Step 1: Real-time acquisition and preprocessing of the mid-infrared and ultraviolet absorption spectra of the mixed gas to be monitored; Simultaneously, on-site condition compensation is performed on the pre-constructed corresponding standard absorption cross-section matrix to obtain the mid-infrared absorption cross-section matrix and the ultraviolet absorption cross-section matrix.

[0069] Step 2: Establish a dual-spectral joint linear superposition model and apply concentration consistency constraints to the co-components that absorb in both the mid-infrared and ultraviolet bands. The dual-spectral joint linear superposition model includes: a mid-infrared linear model, which is expressed as: the mid-infrared absorption spectrum equals the product of the concentration vector in the mid-infrared band and the mid-infrared absorption cross-section matrix plus the residual vector in the mid-infrared band; and an ultraviolet model, which is expressed as: the ultraviolet absorption spectrum equals the product of the concentration vector in the ultraviolet band and the ultraviolet absorption cross-section matrix plus the residual vector in the ultraviolet band; the concentration vector is a vector composed of the concentrations of each component.

[0070] Step 3: The dual-spectral joint linear superposition model is solved using the non-negative constraint adaptive iterative weighted least squares algorithm to obtain the concentration vectors of the infrared band and the ultraviolet band, as well as the corresponding residual vectors.

[0071] Step 4: Weighted fusion of co-component concentrations in the mid-infrared and ultraviolet bands is performed. The fusion weight is determined by the reciprocal of the sum of squares of the average fitting residuals retrieved from each band. The weighted fused concentrations are then used to substitute the dual-spectral joint linear superposition model to correct the concentrations of the remaining components in each band.

[0072] Step 5: Calculate the final residual vector. Perform principal component analysis on the residual vectors of multiple consecutive detections. If the sum of the variance contribution rates of the first two principal components exceeds the preset threshold, it is determined that there is an unmodeled interference component. The first principal component vector is normalized and used as the absorption cross section of the temporary pseudo-component, which is then extended into the absorption cross section matrix. Steps 3 to 5 are repeated. Otherwise, proceed to step 6.

[0073] Step 6: Output the concentration values ​​of each decomposition product.

[0074] Furthermore, in this embodiment, the standard absorption cross-section matrix is ​​constructed as follows:

[0075] For the type of GIL / GIS mixed gas to be monitored (taking SF6 / N2 mixed gas as an example), it is pre-tested under standard laboratory conditions (temperature) =296.15K, pressure =101.325 kPa) High-resolution absorption cross section measured using pure component gas: For the mid-infrared band, each component In wavenumber absorption cross section at the location The measurement range covers all characteristic absorption peaks of the target component; for the ultraviolet band, each component... At wavelength Differential absorption cross section at This was obtained by removing the slowly changing background.

[0076] The measurement data is stored in matrix form to obtain the mid-infrared standard absorption cross section matrix. and UV standard absorption cross section matrix , respectively represented as:

[0077] ;

[0078] in, Effective optical path; This represents the number of mid-infrared sampling points; This represents the number of mid-infrared absorbing components. Components In wavenumber The absorption cross section at that location.

[0079] ;

[0080] in, This represents the number of ultraviolet sampling points. This represents the number of ultraviolet absorbing components. Components At wavelength The differential absorption cross section at that location.

[0081] Furthermore, in step 1 of this application embodiment, the mid-infrared absorption spectrum and ultraviolet absorption spectrum of the mixed gas to be monitored are acquired in real time and preprocessed. Simultaneously, the pre-constructed corresponding standard absorption cross-section matrix is ​​corrected for on-site conditions. The specific implementation process includes:

[0082] The raw spectral signals collected from the GIL / GIS gas cell are converted into absorbance vectors that can be used for quantitative inversion. Specifically, the spectral data of the gas mixture to be tested are first collected through an online monitoring system, including:

[0083] Mid-infrared absorption spectrum ;in The light intensity after passing through the gas to be measured. Background light intensity (acquired and stored during system initialization);

[0084] Ultraviolet absorption spectrum ;in, The light intensity after passing through the gas to be measured. Background light intensity (acquired and stored during system initialization).

[0085] The original spectra are then preprocessed, including:

[0086] For noise reduction, Savitzky-Golay smoothing filter (window width 11 points, polynomial order 3) is used to smooth the mid-infrared absorption spectrum and ultraviolet absorption spectrum respectively to remove high-frequency random noise.

[0087] Baseline correction: For mid-infrared absorption spectra, there may be slowly changing baseline drift due to factors such as optical interference and particle scattering. Therefore, it is necessary to use a second-order polynomial to fit the baseline (selecting regions without absorption characteristics) and then subtract the baseline from the original absorbance.

[0088] Through the above preprocessing steps, the measured mid-infrared absorbance vector is obtained. and the measured UV absorbance vector .

[0089] Considering that the on-site temperature and pressure may differ from the laboratory standard conditions, it is necessary to perform temperature and pressure normalization compensation on the pre-constructed standard absorption cross-section matrix to ensure that the coefficients in the subsequent linear model are under the same thermodynamic conditions as the measured data. The specific implementation method is as follows:

[0090] Normalized compensated mid-infrared absorption cross section matrix (The normalized and compensated matrix has the same size as the standard matrix) is represented as:

[0091] ;

[0092] Normalized compensated UV absorption cross section matrix (The normalized and compensated matrix has the same size as the standard matrix) is represented as:

[0093] ;

[0094] in, and These are the on-site temperature and on-site pressure, respectively. and These are laboratory temperature and laboratory pressure, respectively.

[0095] Furthermore, in step 2 of this application embodiment, a dual-spectral joint linear superposition model is established, and a concentration consistency constraint is applied to the co-components that absorb in both the mid-infrared and ultraviolet bands. The specific implementation process includes:

[0096] First, the measured absorbance vector is represented as a linear combination of the absorption cross-sections of each component, and a co-component consistency constraint is introduced to provide a mathematical model for subsequent inversion. Specifically, according to the Lambert-Beer law, the total absorbance of the gas mixture is equal to the sum of the absorbances of each component, resulting in a mid-infrared linear model and an ultraviolet model. The mid-infrared linear model is expressed as:

[0097] ;

[0098] in, This represents the mid-infrared absorbance vector. This is the concentration vector in the mid-infrared band (composed of the concentrations of each absorbing component in the mid-infrared band, with a magnitude of...). ); The residual vector in the mid-infrared band (size is...) This includes measurement noise, model error, and unmodeled absorption.

[0099] The ultraviolet radiation model is represented as follows:

[0100] ;

[0101] in, This is the ultraviolet absorbance vector; This is the concentration vector in the ultraviolet band (composed of the concentrations of each absorbing component in the ultraviolet band, with a magnitude of...). ); The residual vector in the ultraviolet band (size is...) This includes measurement noise, model error, and unmodeled absorption.

[0102] In real gases, a component gas may exhibit characteristic absorption in both the mid-infrared and ultraviolet bands, but its physical concentration is unique. Therefore, this application imposes a concentration consistency constraint on the co-components: co-components that absorb in both the mid-infrared and ultraviolet bands have the same concentration. This constraint couples two independent linear models, increasing the information content of the inversion problem and reducing ambiguity.

[0103] It is understood that step 2 in this embodiment does not solve for the concentration, but rather defines the problem. The solution process is performed in subsequent step 3. The output of step 2 is essentially the entire model and its constraints, which will serve as the objective function and constraints for the optimization problem in step 3.

[0104] Furthermore, in step 3 of the embodiments of this application, as follows: Figure 2 As shown, the bispectral joint linear superposition model is solved using an adaptive iterative weighted least squares algorithm with non-negativity constraints. The specific implementation process includes:

[0105] Initial solution (i.e., the concentration vector estimate in the 0th iteration): First, ignoring weights, multiply the normalized and compensated absorption cross-section matrix with the concentration vector to be solved to obtain the model absorbance vector. Then, using the square of the Euclidean norm of the difference between the measured absorbance vector and the model absorbance vector as the objective function, solve for the concentration vector that minimizes the objective function, where each element in the concentration vector must be greater than or equal to 0. Following the above initial solution solution steps (i.e., the non-negative least squares algorithm without considering weights), solve independently for the mid-infrared and ultraviolet bands respectively, obtaining the initial solutions for the concentration vectors of the mid-infrared and ultraviolet bands, as well as the corresponding initial solutions for the residual vectors.

[0106] Iterative weighted solution: For the th In the next iteration (i.e., the previous iteration), the residual for each sampling point is calculated, expressed as follows:

[0107] ;

[0108] in, For the first In the nth iteration The residuals of each sampling point; For the first Measured absorbance at each sampling point; For the normalized first The sampling point Absorption cross section of each component; For the first The second iteration The concentration of each component; This represents the number of absorbing components in the mid-infrared or ultraviolet band. The residuals for each sampling point in the previous iteration in the mid-infrared and ultraviolet bands are calculated using the residual calculation method described above.

[0109] The standard deviation of the residuals is calculated as follows:

[0110] ;

[0111] in, For the first The standard deviation of the residuals in the next iteration; This represents the number of sampling points; For the first The mean of the residuals from each iteration. The standard deviations of the residuals from the previous iteration are calculated using the standard deviation calculation method described above for the mid-infrared and ultraviolet bands respectively.

[0112] Update weights: For each sampling point, if the absolute value of the residual of the sampling point in the previous iteration is less than or equal to the threshold, then the weight of that sampling point in the current iteration is set to 1. If the absolute value of the residual of the sampling point in the previous iteration is greater than the threshold, then the weight of that sampling point in the current iteration is set to the ratio of the threshold to the absolute value of the residual of the sampling point in the previous iteration, specifically expressed as follows:

[0113] ;

[0114] in, As the threshold, and ; For the first In the nth iteration (i.e., the current nth iteration) The weight of each sampling point is determined by the following: when the fitting error is small, the sampling point is considered reliable and has a weight of 1; when the fitting error exceeds 1.5 times the standard deviation, the sampling point is considered to be affected by noise, baseline residuals, or unmodeled interference, and its weight is reduced inversely to reduce its impact on the objective function.

[0115] Construct and solve the weighted least squares problem: Define the first... The weight diagonal matrix of the next iteration :

[0116] ;

[0117] in, To construct a diagonal matrix function.

[0118] Construct the objective function and solve it:

[0119] ;

[0120] in, For the first The concentration vector estimate for the next iteration; The vector represents the variable that minimizes the objective function, and each element in the vector must be greater than or equal to 0. The square of the Euclidean norm; This is the measured absorbance vector; This is the absorption cross-section matrix after normalization and compensation; Let be the concentration vector to be solved, and be the optimization variable. The only difference between this objective function and the objective function in the initial solution process is that the objective function in the initial solution process does not consider weights (i.e., it is a non-negative least squares algorithm without weights), while the objective function in the iterative weighted solution process considers weights. This objective function is equivalent to multiplying the original data by the square root of the weights and then performing the standard non-negative least squares method (i.e., a non-negative least squares algorithm considering weights).

[0121] Finally, to determine if convergence has occurred, calculate the relative change, as follows:

[0122] ;

[0123] in, It should be a very small integer to prevent the denominator from being zero; It is the Euclidean norm.

[0124] If the relative change is less than the convergence tolerance or the current iteration number is greater than or equal to the maximum iteration number, the iteration stops, and the concentration vector estimate obtained in the current iteration is output; otherwise, the above process of residual calculation, standard deviation calculation, weight update, construction and solution of the weighted least squares problem is repeated for the next iteration of weighted solution. Through the above iterative weighted solution, the concentration vector estimates and corresponding residual vectors for the mid-infrared and ultraviolet bands are obtained respectively.

[0125] Integration of co-component consistency constraints: Since the mid-infrared and ultraviolet inversion problems are coupled through co-components, simple independent iteration cannot guarantee consistency. Therefore, this application's embodiment uses the alternating projection method to achieve a constrained global solution. The specific process includes:

[0126] Based on the concentration vector estimates of the mid-infrared and ultraviolet bands obtained by the above iterative weighted solution, the cocomponent concentrations are averaged, that is, the arithmetic mean of the cocomponent concentrations retrieved from the two bands:

[0127] ;

[0128] in, This represents the average concentration of the co-components. This is an estimated value for the co-component concentration in the mid-infrared band; This is an estimated value for the co-component concentration in the ultraviolet band.

[0129] The cocomponent concentrations in the mid-infrared and ultraviolet bands were fixed at [value missing]. The concentrations of the remaining components in the mid-infrared and infrared bands were obtained by re-applying the initial solution solution process (i.e., the solution method without considering weights).

[0130] Determine if convergence: Let there be... The convergence criterion for the co-components is:

[0131] ;

[0132] in, This represents the average concentration of co-component j in the current iteration; This represents the average concentration of co-component j from the previous iteration. The initial value is the average of the initial values ​​of cocomponent concentration in the mid-infrared band (cocomponent concentration in the initial solution of the concentration vector in the mid-infrared band) and the initial values ​​of cocomponent concentration in the ultraviolet band (cocomponent concentration in the initial solution of the concentration vector in the ultraviolet band) obtained from the initial solution. This is the convergence tolerance for co-components.

[0133] If the convergence criterion is met, the final concentration vectors of the mid-infrared band and the ultraviolet band, as well as the corresponding residual vector (i.e., the difference between the measured absorbance vector and the product of the final concentration vector obtained by iterative solution and the normalized compensated absorption cross-section matrix), are output, and the process proceeds to the next step. If the convergence criterion is not met, the integrated steps of iterative weighted solution and co-component consistency constraint are repeated until the change in co-components is less than the threshold.

[0134] Furthermore, in step 4 of this embodiment, the co-component concentrations in the mid-infrared and ultraviolet bands are weighted and fused, and the weighted and fused concentrations are then used to substitute the dual-spectral joint linear superposition model to correct the concentrations of the remaining components in each band. The specific implementation process includes:

[0135] Calculate the fusion weights: The fusion weights are inversely proportional to the inversion accuracy and are expressed as the reciprocal of the sum of squared residuals.

[0136] ;

[0137] in, and These are the residual sums of squares for the mid-infrared and ultraviolet bands, respectively. The residual sum of squares is equal to the mean of the residual sums of squares at each sampling point in the residual vector obtained in the previous step. and These are the fusion weights for the mid-infrared and ultraviolet bands, respectively.

[0138] Co-component concentration-weighted fusion: For each co-component, its weighted fusion concentration is obtained, expressed as:

[0139] ;

[0140] This weighted fusion value utilizes both the high sensitivity of the infrared band and the high selectivity of the ultraviolet band, combining the advantages of both.

[0141] Back-substitution correction, resolving the concentrations of other components in each band: Treating the weighted fused concentration of co-components as known true values, and substituting them into the mid-infrared linear model and the ultraviolet model respectively, yields:

[0142] ;

[0143] in, Column vector of absorption cross sections for co-components; This is the absorption cross-section matrix for the remaining components (excluding the co-components); This represents the concentration vector of the remaining components; This represents the weighted fusion concentration vector of the co-components; This is the residual vector. Then, the initial solution solution process (i.e., the non-negative least squares method without considering weights) is used again to solve for the concentrations of the remaining components.

[0144] The corrected concentrations of the remaining components and the weighted fusion concentrations of the co-components are combined to obtain the final concentration vectors for the mid-infrared and ultraviolet bands.

[0145] Furthermore, in step 5 of this embodiment, the process of dynamically correcting the absorption cross-section matrix includes:

[0146] Residual vector collection: A joint judgment strategy is adopted, that is, the residual vectors of the mid-infrared and ultraviolet bands obtained in step 4 are concatenated into a long vector and then principal component analysis is performed to uniformly determine whether there are unmodeled interference components. The final residual vector of this detection is stored in a queue of length K, following a first-in, first-out (FIFO) principle.

[0147] Perform principal component analysis on the residual vectors: construct a matrix from the K most recent residual vectors. (Each column in the matrix is ​​a residual vector of length M), perform singular value decomposition:

[0148] ;

[0149] in, These are left singular vectors, with each column representing a characteristic spectral direction; It is a singular value diagonal matrix, and , is a singular value, dimensionless; It is a right singular vector; the superscript T is the vector transpose.

[0150] Determine if there are unmodeled interference components: Calculate the variance contribution rates of the first two principal components.

[0151] ;

[0152] ;

[0153] If the sum of the variance contribution rates of the first two principal components is greater than a preset threshold (e.g., 0.8), it is determined that there is a significant unmodeled interference component (the residual vector has structured characteristics, rather than random noise), and subsequent steps (expanding the absorption cross-section matrix) are performed; otherwise, the final concentration vector of each band is output.

[0154] Extended absorption cross-section matrix: Take the first principal component vector (That is, the first column of the left singular vector matrix, with a length of M), normalized to obtain:

[0155] ;

[0156] The normalized principal component vector is used as the absorption cross section of the temporarily added pseudo-component, and the original absorption cross section matrix is ​​expanded accordingly:

[0157] ;

[0158] ;

[0159] in, and These are the extended absorption cross-section matrices for the mid-infrared and ultraviolet bands, respectively. and The absorption cross sections of the pseudo-components in the mid-infrared and ultraviolet bands are respectively calculated using the above normalization formula.

[0160] The concentration of the newly added pseudo-component is used as the variable to be determined. Steps 3-5 are repeated to obtain the final concentration vector.

[0161] Furthermore, step 5 of this embodiment also includes:

[0162] The matrix reset procedure, in order to avoid excessive matrix expansion, involves periodically clearing the expanded pseudo-components and restoring the original absorption cross-section matrix after a system restart.

[0163] Furthermore, embodiments of this application also propose an online monitoring system for the decomposition products of GIL / GIS mixed gases, such as... Figure 3 As shown, it includes:

[0164] The gas circulation module is used to extract the mixed gas to be tested from the GIL / GIS gas chamber to the gas pool for testing. After the test is completed, the gas is purified and returned to the GIL / GIS gas chamber to achieve a zero-emission closed-loop circulation.

[0165] The acquisition module includes a mid-infrared TDLAS detection unit (at least four tunable lasers, with center wavelengths corresponding to target components such as SO2, CF4, NF3, and COF2, respectively) and an ultraviolet DOAS detection unit (deuterium lamp light source and spectrometer); the two detection units share the same flow-through long optical path gas cell (effective optical path L: 5~30m, with an inert coating on the inner wall); it also includes temperature and pressure sensors installed in the gas cell for real-time acquisition of environmental parameters.

[0166] The edge computing module, equipped with an industrial real-time operating system, is used to execute any of the above-described implementation methods of the online monitoring method proposed in the embodiments of this application.

[0167] Additionally, a remote monitoring platform receives concentration data uploaded by edge computing modules via 5G or fiber optic Ethernet, and performs visualization and historical trend analysis.

[0168] Furthermore, the edge computing module repeatedly executes steps 1-5 according to a preset time period to achieve periodic detection. The concentration values ​​of each decomposition product obtained from each detection are stored in the local database and uploaded to the remote monitoring platform. The concentration-time curve is plotted on the remote monitoring platform, and the gas production rate is calculated, providing a basis for equipment insulation status assessment and fault early warning.

[0169] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0170] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0171] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0172] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0173] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0174] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above description is only a specific embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for online monitoring of decomposition products of GIL / GIS mixed gas, characterized in that, include: The mid-infrared and ultraviolet absorption spectra of the mixed gas to be monitored are acquired in real time and preprocessed. At the same time, the corresponding standard absorption cross section matrix is ​​compensated for on-site conditions to obtain the mid-infrared absorption cross section matrix and the ultraviolet absorption cross section matrix. A dual-spectral joint linear superposition model was established, and a concentration consistency constraint was applied to the co-components that have absorption in both the mid-infrared and ultraviolet bands. The dual-spectral joint linear superposition model includes: a mid-infrared linear model, which is expressed as: the mid-infrared absorption spectrum equals the product of the concentration vector of the mid-infrared band and the mid-infrared absorption cross-section matrix plus the residual vector of the mid-infrared band; and an ultraviolet model, which is expressed as: the ultraviolet absorption spectrum equals the product of the concentration vector of the ultraviolet band and the ultraviolet absorption cross-section matrix plus the residual vector of the ultraviolet band. Model solution steps: The dual-spectral joint linear superposition model is solved using the non-negative constraint adaptive iterative weighted least squares algorithm to obtain the concentration vectors of the mid-infrared band and the ultraviolet band, as well as the corresponding residual vectors. Back-substitution correction step: The co-component concentrations in the mid-infrared and ultraviolet bands are weighted and fused. The fusion weight is determined by the reciprocal of the sum of squares of the average fitting residuals retrieved from each band. The weighted and fused concentrations are then used back-substituted into the dual-spectrum joint linear superposition model to correct the concentrations of the remaining components in each band. Unknown interference identification steps: Calculate the final residual vector, perform principal component analysis on the residual vectors of multiple consecutive detections, if the sum of the variance contribution rates of the first two principal components exceeds the preset threshold, it is determined that there is an unmodeled interference component, and the first principal component vector is normalized and used as the absorption cross section of the temporary pseudo-component to expand into the absorption cross section matrix, and the model solving step, back substitution correction step and unknown interference identification step are repeated; otherwise, proceed to the next step. Output the concentration values ​​of each decomposition product.

2. The method for online monitoring of decomposition products of GIL / GIS mixed gas according to claim 1, characterized in that, The process of constructing the standard absorption cross-section matrix includes: For the type of mixed gas to be monitored, high-resolution absorption cross-sections are measured in advance using pure component gases under standard laboratory conditions: for the mid-infrared band, the absorption cross-section of each component at each wavenumber is measured, and the measurement range covers all characteristic absorption peaks of the target component; for the ultraviolet band, the differential absorption cross-section of each component at each wavelength is measured. The measurement data is stored in matrix form to obtain the mid-infrared standard absorption cross section matrix and the ultraviolet standard absorption cross section matrix; wherein the mid-infrared standard absorption cross section matrix is ​​composed of the absorption cross sections of all mid-infrared absorbing components at different wavenumbers; and the ultraviolet standard absorption cross section matrix is ​​composed of the differential absorption cross sections of all ultraviolet absorbing components at different wavelengths.

3. The method for online monitoring of decomposition products of GIL / GIS mixed gas according to claim 1, characterized in that, The real-time acquisition and preprocessing of the mid-infrared and ultraviolet absorption spectra of the mixed gas to be monitored, along with on-site condition compensation for the pre-constructed standard absorption cross-section matrix, includes: The spectral data of the gas mixture to be tested, including mid-infrared absorption spectrum and ultraviolet absorption spectrum, are collected through an online monitoring system. The collected mid-infrared and ultraviolet absorption spectra were preprocessed to obtain the measured mid-infrared absorbance vector and the measured ultraviolet absorbance vector, respectively. Temperature and pressure normalization compensations were performed on the pre-constructed mid-infrared standard absorption cross section matrix and ultraviolet standard absorption cross section matrix respectively to obtain the mid-infrared absorption cross section matrix and ultraviolet absorption cross section matrix.

4. The method for online monitoring of decomposition products of GIL / GIS mixed gas according to claim 3, characterized in that, The model solution steps include: Initial solution: The non-negative least squares algorithm without considering weights is used to solve the mid-infrared band and the ultraviolet band independently, so as to obtain the initial solution of the concentration vector of the mid-infrared band and the initial solution of the concentration vector of the ultraviolet band, as well as the corresponding initial solution of the residual vector. Iterative weighted solution: Calculate the residual of each sampling point in the previous iteration; calculate the standard deviation of the residual of the previous iteration based on the residuals of all sampling points; update weights: for each sampling point, if the absolute value of the residual in the previous iteration is less than or equal to a threshold, the weight of the sampling point in the current iteration is set to 1; if the absolute value of the residual in the previous iteration is greater than the threshold, the weight of the sampling point in the current iteration is set to the ratio of the threshold to the absolute value of the residual in the previous iteration; where the threshold is 1.5 times the standard deviation; use a weighted non-negative least squares algorithm to solve the mid-infrared and ultraviolet bands independently to obtain the concentration vectors of the mid-infrared and ultraviolet bands corresponding to the current iteration; determine convergence: if the relative change between the concentration vector obtained in the current iteration and the concentration vector obtained in the previous iteration is less than the convergence tolerance or the current iteration number is greater than or equal to the maximum iteration number, stop the iteration and output the concentration vector obtained in the current iteration; otherwise, repeat the iterative weighted solution steps for the next iterative weighted solution. Integration of co-component consistency constraints: Based on the concentration vectors of the mid-infrared and ultraviolet bands obtained by iterative weighted solution, the co-component concentrations are averaged, and the co-component concentrations in the mid-infrared and ultraviolet bands are fixed as the average co-component concentrations. The concentrations of the remaining components in the mid-infrared and ultraviolet bands are obtained by using a non-negative least squares algorithm without considering weights. Determine whether the convergence criterion is met. If it is, output the final concentration vectors of the mid-infrared band and the ultraviolet band, as well as the corresponding residual vectors; otherwise, repeat the integrated steps of iterative weighted solution and co-component consistency constraint.

5. The method for online monitoring of decomposition products of GIL / GIS mixed gas according to claim 4, characterized in that, The non-negative least squares algorithm that does not consider weights is specifically as follows: First, the normalized and compensated absorption cross-section matrix is ​​multiplied by the concentration vector to be solved to obtain the model absorbance vector; Then, the square of the Euclidean norm of the difference between the measured absorbance vector and the model absorbance vector is used as the objective function; Find the concentration vector that minimizes the objective function, where each element in the concentration vector must be greater than or equal to 0. The difference between the weighted non-negative least squares algorithm and the weightless non-negative least squares algorithm lies only in the objective function. The objective function in the weighted non-negative least squares algorithm is the square of the Euclidean norm of the product of the difference between the measured absorbance vector and the model absorbance vector multiplied by the square root of the weight matrix.

6. The method for online monitoring of decomposition products of GIL / GIS mixed gas according to claim 5, characterized in that, The back-substitution correction step includes: The reciprocal of the sum of squared residuals in the mid-infrared and ultraviolet bands are used as their respective fusion weights; For each co-component, its weighted fusion concentration is obtained; The weighted fusion concentration of co-components is taken as a known true value and substituted into the mid-infrared linear model and the ultraviolet model respectively. The non-negative least squares method without considering the weights is used to solve for the concentrations of the remaining components in the mid-infrared and ultraviolet bands respectively. The concentrations of the remaining components and the co-components are combined by weighted fusion to obtain the final concentration vectors for the mid-infrared band and the ultraviolet band.

7. A method for online monitoring of decomposition products of GIL / GIS mixed gas according to any one of claims 1-6, characterized in that, The unknown interference identification step includes: The most recent residual vectors are combined into a matrix, with each column of the matrix being a residual vector. Perform singular value decomposition on the matrix to obtain a singular value diagonal matrix; the principal components in the singular value diagonal matrix are arranged in descending order. Calculate the variance contribution rate of the first two principal components in the singular valued diagonal matrix; If the sum of the variance contribution rates of the first two principal components is greater than the preset threshold, it is determined that there is a significant unmodeled interference component and proceeds to the next step; otherwise, the final concentration vectors of the mid-infrared and ultraviolet bands are output. The first principal component vector is normalized, and the normalized principal component vector is used as the absorption cross section of the temporarily added pseudo-component to expand the original absorption cross section matrix. The concentration of the newly added pseudo-component is used as the variable to be solved. The model solving steps, back substitution correction steps, and unknown interference identification steps are re-executed to obtain the final concentration vector.

8. The method for online monitoring of decomposition products of GIL / GIS mixed gas according to claim 7, characterized in that, Also includes: According to the preset cycle or when the system restarts, the absorption cross section of the newly added pseudo-components is cleared and the original absorption cross section matrix is ​​restored.

9. An online monitoring system for decomposition products of GIL / GIS mixed gas, characterized in that, include: The gas circulation module is used to extract the mixed gas to be tested from the GIL / GIS gas chamber to the gas pool for testing. After the test is completed, the gas is purified and then refilled into the GIL / GIS gas chamber to achieve zero-emission closed-loop circulation. The acquisition module includes a mid-infrared TDLAS detection unit and an ultraviolet DOAS detection unit; the two detection units share the same flow-through long optical path gas cell; it also includes a temperature sensor and a pressure sensor installed in the gas cell for real-time acquisition of environmental parameters. An edge computing module, equipped with an industrial real-time operating system, is used to execute the online monitoring method for decomposition products of GIL / GIS mixed gas as described in any one of claims 1-8; Additionally, a remote monitoring platform is used to receive the concentration data uploaded by the edge computing module, and to perform visualization and historical trend analysis.

10. The online monitoring system for decomposition products of GIL / GIS mixed gas according to claim 9, characterized in that, The edge computing module performs periodic detection according to a preset time period. The concentration values ​​of each decomposition product obtained in each detection are stored in the local database and uploaded to the remote monitoring platform. The remote monitoring platform plots a concentration-time curve based on the uploaded data and calculates the gas production rate.

Images