LIBS (laser-induced breakdown spectroscopy) coal quality quantitative analysis method and related equipment

Abstract

The invention discloses an LIBS (Laser-induced Breakdown Spectroscopy) coal quality quantitative analysis method and related equipment, and the method comprises the following steps: collecting LIBS spectrums of the same coal sample through different instruments, and dividing a training set of a training migration model and a training set of a training quantitative analysis model; performing spectrum pretreatment on the spectrums respectively; carrying out wavelength offset self-correction on the spectrum to realize that the spectrum collected by the slave instrument corresponds to the spectrum collected by the master instrument in wavelength; performing feature selection on the spectrum, and taking the spectrum after feature selection as the input of a migration model; establishing and training a transfer learning model based on feature mapping and feature representation, and utilizing the trained transfer learning model to realize feature transfer from the instrument spectrum to the main instrument spectrum; and training a quantitative analysis model by utilizing the spectrum of the master instrument and the migrated spectrum of the slave instrument, and predicting the coal quality index of the slave instrument by adopting the trained quantitative analysis model. The coal quality quantitative analysis model constructed by the method can stably and efficiently run across instruments, and the modeling cost of different instruments is reduced.

Description

Technical Field

[0001] The present invention relates to the field of online coal quality detection, and in particular to a LIBS coal quality quantitative analysis method and related equipment. Background Art

[0002] The large-scale absorption of renewable energy power places higher demands on the flexible and deep peak-shaving capabilities of coal-fired power plants. Coal quality directly impacts key processes such as coal blending, pulverizing, combustion, and pollutant emission control. Therefore, online coal quality monitoring has become a key technology for the flexible, low-carbon, and safe operation of coal-fired power plants. Laser-induced breakdown spectroscopy (LIBS), an atomic emission spectrometric analysis technique based on laser ablation, has shown great potential for online coal quality monitoring. However, due to factors such as varying equipment aging, performance differences between different instruments, and changes in the measurement environment, stable and effective quantitative models used on the master instrument are no longer applicable to other slave instruments, leading to a sharp decline in quantitative analysis accuracy. This is a major obstacle to the further commercialization and standardization of LIBS online coal quality monitoring technology. Currently, transfer learning (TL) is the primary approach to addressing these issues. Feature-based transfer learning methods are widely used, leveraging shared features between master and slave instrument spectra to reduce inter-domain disparities. These methods can be categorized into feature mapping-based transfer learning and feature representation-based transfer learning, depending on their processing approach. The former is to map the spectral data into the same high-dimensional space, in which the spectral characteristics of the master and slave instruments are close or the data distribution tends to be consistent; the latter is generally to characterize the spectral data of the master and slave instruments by learning the effective characteristics, reducing the difference between the two, and achieving the effect of spectral characteristic calibration of the master and slave instruments.

[0003] At present, the existing technology has the following shortcomings: 1) It ignores the problem of wavelength offset in the spectra collected by the master and slave instruments. The features extracted directly from the non-corresponding spectral data matrices have low correlation, which makes it impossible to effectively improve the generalization ability of the quantitative analysis model. 2) The transfer learning method based on feature mapping requires the mapping function to reduce the difference in data distribution while retaining the original data characteristics. When the correlation between the spectral data of the master and slave instruments is insufficient, noise will be introduced, resulting in a decrease in the prediction accuracy of the quantitative analysis model. 3) In the case of large data differences, transfer learning based on feature representation will have the problem that the learned feature representation cannot effectively capture the common characteristics of the data. During the training process, the model over-learns the characteristics of the master instrument data, which will lead to a decrease in the generalization ability of the model, especially when there is a nonlinear relationship in the data. Summary of the Invention

[0004] In order to at least partially solve one of the technical problems existing in the prior art, the present invention aims to provide a LIBS coal quality quantitative analysis method and related equipment based on wavelength offset self-correction combined with feature transfer learning.

[0005] The first technical solution adopted by the present invention is:

[0006] A LIBS coal quality quantitative analysis method comprises the following steps:

[0007] LIBS spectra of the same coal sample were collected using different instruments and divided into training sets for training the migration model and training sets and test sets for training the quantitative analysis model;

[0008] The collected spectra are preprocessed to reduce the influence of spectral noise;

[0009] The pre-processed spectrum is self-calibrated for wavelength offset, so that the spectrum collected by the slave instrument corresponds to the spectrum collected by the main instrument in wavelength;

[0010] Feature selection is performed on the spectrum after wavelength shift self-correction to reduce the impact of irrelevant characteristic variables in the spectrum on model performance, and the spectrum after feature selection is used as the input of the migration model;

[0011] Establish and train a transfer learning model based on feature mapping and feature representation, and use the trained transfer learning model to achieve feature migration from instrument spectra to master instrument spectra;

[0012] The master instrument spectrum and the migrated slave instrument spectrum are used to train the quantitative analysis model, and the trained quantitative analysis model is used to predict the coal quality indicators of the slave instrument.

[0013] Furthermore, the collected spectra are subjected to spectral preprocessing, including:

[0014] Perform effective spectrum screening and spectrum noise reduction on the collected spectra;

[0015] Among them, the effective spectrum screening method includes but is not limited to the effective spectrum screening method based on the standard deviation (SD) value of the characteristic peak intensity, and the spectrum noise reduction method includes but is not limited to data normalization processing.

[0016] Furthermore, the wavelength offset self-correction method is implemented based on the identification and correction of the peak SD value: any coal sample is selected as the standard sample, the effective spectrum of the coal sample is collected on different instruments, the offset of each spectral channel between the instruments is calculated and aligned and corrected to the main instrument, thereby realizing the wavelength offset self-correction of the instrument spectrum.

[0017] Furthermore, the wavelength offset self-correction is performed on the pre-processed spectrum to achieve wavelength correspondence between the spectrum collected by the slave instrument and the spectrum collected by the master instrument, including:

[0018] Select a band with relatively high signal-to-background ratio in each channel of the main instrument spectrum. In this band, there is only one spectrum line with obvious excitation as the reference spectrum line, and the wavelength point where it is located is the reference point A. Sa (a is the number of spectrometer channels), with the reference point as the center, select the wavelength range L with the preset window a (a is the number of spectrometer channels);

[0019] Select the same wavelength range L from the instrument spectrum a ';

[0020] In the interval L a Combined with the peak intensity SD value method, the wavelength point where the reference spectrum line is located is selected as the offset point A Ta (a is the number of spectrometer channels);

[0021] Set reference point A Sa With offset point A Ta The wavelength difference is taken as the wavelength offset S between the master and slave instruments of the channel. a (a is the number of spectrometer channels).

[0022] Furthermore, in the interval L a Combined with the peak intensity SD value method, the wavelength point where the reference spectrum line is located is selected as the offset point A Ta ,include:

[0023] Set a window of length 2N+1 and a 'Move this window with a step size of 1;

[0024] In the jth movement, the size of the spectral intensity corresponding to the wavelength point in the window is compared; when the spectral intensity I N+j When it is the maximum value in the interval and is greater than the spectral intensity corresponding to the previous and next wavelength points, the wavelength point is identified as the peak point within the window length, recorded as m k , k=1,2,3…, until L is found a 'All the peak points in the window; where the maximum value of k is the total number of peak points; j is the number of window moves, j = 1, 2, 3, ..., b-2N, b is the selected window length;

[0025] m k As the center, N is the radius and the wavelength interval Q is selected k , calculate the peak m k SD k , the calculation formula is as follows:

[0026]

[0027] Among them, I n is the spectral intensity at the nth wavelength point, For interval Q k The mean value of the spectral intensity corresponding to all wavelength points within;

[0028] Compare all calculated SDs k , m corresponding to the maximum value k That is, the reference point A Sa Corresponding offset point A Ta .

[0029] Furthermore, the feature selection of the spectrum after the wavelength shift self-correction includes:

[0030] The input features for analyzing coal quality indicators are determined by combining the coal chemical analysis mechanism with a feature selection algorithm. The feature selection algorithm includes but is not limited to a competitive adaptive reweighted sampling (CARS) algorithm, which selects features with high correlation as input features of the migration model.

[0031] Furthermore, the feature selection method combines the coal chemical analysis mechanism with the CARS feature selection algorithm to determine the input features for analyzing coal quality indicators. The correlation and interaction mechanism between each coal quality indicator and element content are analyzed, and the element spectral lines and related bands with the greatest impact on coal quality indicator analysis are selected as empirical characteristic peak bands. Furthermore, CARS is combined with Monte Carlo sampling and an exponentially decreasing function (EDF) to adaptively adjust the selection probability of each band. Ultimately, the optimal band combination that contributes most to model performance is selected based on coal quality indicators.

[0032] Furthermore, feature mapping methods include but are not limited to kernel principal component analysis (KPCA), and feature representation methods include but are not limited to piecewise direct standardization (PDS); a mapping of master and slave instrument spectra is established in a high-dimensional space through the feature mapping method, and the nonlinear features in the data are converted into linear features; at the same time, parameters are optimized according to the objective function combined with optimization algorithms such as steepest descent, grid search, and Bayesian optimization to minimize the distance between mappings; then, the feature representation method is used to calculate the kernel migration matrix between instruments to realize the migration of the master and slave instrument spectral features.

[0033] Furthermore, the training of the transfer learning model includes:

[0034] The migration model was trained using the spectra after feature selection (including spectra collected by the master instrument and the slave instrument). The objective function was to minimize the maximum mean discrepancy (MMD). The Gaussian kernel function parameter σ was optimized using the steepest descent algorithm, and the appropriate number of KPCA principal components j and PDS window width c were selected.

[0035] According to the parameter σ, the spectrum V collected by the main instrument after feature selection is S and the spectrum V collected from the instrument T , mapped to high-dimensional space by Gaussian kernel function and PCA dimensionality reduction processing, the corresponding eigenvector ω is obtained S and ω T , and the high-dimensional mapping K G (V S ), K G (V T ); Combined with the window width c, use PDS to calculate K G (V S ) and K G (V T ) and obtain the migration matrix F between the master and slave instruments.

[0036] The second technical solution adopted by the present invention is:

[0037] An electronic device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, at least one program, or the code set or instruction set is loaded and executed by the processor to implement a LIBS coal quality quantitative analysis method as described above.

[0038] The third technical solution adopted by the present invention is:

[0039] A computer-readable storage medium stores at least one instruction, at least one program, a code set, or an instruction set, wherein the at least one instruction, at least one program, the code set, or the instruction set is loaded and executed by a processor to implement a LIBS coal quality quantitative analysis method as described above.

[0040] The fourth technical solution adopted by the present invention is:

[0041] A computer program product or computer program includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the aforementioned LIBS coal quality quantitative analysis method.

[0042] The beneficial effects of the present invention are as follows: the present invention solves the problem of the influence of spectral wavelength drift of different instruments on transfer learning and coal quality quantitative analysis through the wavelength offset self-correction method, and captures the global relationship and nonlinear relationship of the data through the feature mapping method combined with the feature representation method, and effectively calibrates in the local range, effectively improving the accuracy and generalization ability of the quantitative analysis model. The coal quality quantitative analysis model constructed by this method can run robustly and efficiently across instruments, reducing the modeling cost of different instruments. Compared with the traditional model without transfer algorithm, the coal quality quantitative analysis model constructed based on the method proposed by the present invention reduces the average absolute error of calorific value, carbon content and ash content predicted on different instruments by 69.26%, 78.91% and 86.13% respectively. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following introduction is made to the drawings of the embodiments of the present invention or the related technical solutions in the prior art. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative work.

[0044] Figure 1 This is a specific flow chart for modeling the LIBS coal quality analysis migration model based on wavelength offset self-correction and feature transfer learning in an embodiment of the present invention;

[0045] Figure 2 This is a flow chart for constructing a coal quality quantitative analysis model in combination with a transfer learning algorithm in an embodiment of the present invention;

[0046] Figure 3 2 is a schematic diagram comparing the heat generation prediction results in an embodiment of the present invention;

[0047] Figure 4 2 is a schematic diagram comparing carbon content prediction results in an embodiment of the present invention;

[0048] Figure 5 2 is a schematic diagram comparing ash content prediction results in an embodiment of the present invention;

[0049] Figure 6 This is a flowchart of the steps of a LIBS coal quality quantitative analysis method in an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application and are not to be construed as limiting the present application. For the step numbers in the following embodiments, they are provided only for the convenience of explanation and are not intended to limit the order of the steps. The order of execution of the steps in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0051] The terms used in the embodiments of the present application are only for the purpose of describing specific embodiments and are not intended to limit the embodiments of the present application. The singular forms of "a", "said", and "the" used in the embodiments of the present application and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. In addition, unless otherwise clearly defined, words such as setting, installing, and connecting should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above words in the present invention in combination with the specific content of the technical solution.

[0052] In the description of this application, it should be understood that descriptions involving orientations, such as up, down, front, back, left, right, etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, they cannot be understood as limitations on this application.

[0053] In the description of this application, "several" means one or more, "many" means more than two, "greater than," "less than," and "exceed" are understood to exclude the number itself, while "above," "below," and "within" are understood to include the number itself. The terms "first" and "second" are used solely to distinguish technical features and are not to be construed as indicating or implying relative importance, or as implicitly specifying the number or order of the technical features indicated.

[0054] In the description of this application, "and / or" describes the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, and B exists alone. The character " / " generally indicates that the associated objects are in an "or" relationship.

[0055] Example 1

[0056] like Figure 1 and Figure 6 As shown, this embodiment provides a LIBS coal quality quantitative analysis method, comprising the following steps:

[0057] S1. LIBS spectra of the same coal sample are collected using different instruments, and the spectra are divided into a training set for training the migration model and a training set and a test set for training the quantitative analysis model.

[0058] S2. Perform spectral preprocessing on the collected spectra to reduce the influence of spectral noise.

[0059] Specifically, the spectrum preprocessing method includes effective spectrum screening and spectrum noise reduction. The effective spectrum screening method includes but is not limited to an effective spectrum screening method based on the standard deviation (SD) value of the characteristic peak intensity, and the spectrum noise reduction method includes but is not limited to data normalization.

[0060] S3. Perform wavelength offset self-calibration on the pre-processed spectrum to achieve wavelength correspondence between the spectrum collected by the slave instrument and the spectrum collected by the main instrument.

[0061] Specifically, the wavelength offset self-correction method is based on the identification and correction of peak SD values. A single coal sample is selected as a standard. Its effective spectrum is collected on different instruments. The offset of each channel of the inter-instrument spectrum is calculated and corrected for alignment with the master instrument, thus achieving wavelength offset self-correction of the slave instrument's spectrum.

[0062] As an implementation manner, step S3 specifically includes the following steps:

[0063] S31. Select a band with relatively high signal-to-background ratio in each channel of the main instrument spectrum. In this band, there is only one spectrum line with obvious excitation as the reference spectrum line, and the wavelength point where it is located is the reference point A. Sa (a is the number of spectrometer channels), with the reference point as the center, select the wavelength range L with a suitable window a (a is the number of spectrometer channels).

[0064] S32, select the same wavelength range L from the instrument spectrum a '. L a and L a ' is defined as follows:

[0065]

[0066] Where b is the selected window length.

[0067] S33, in L a Combined with the peak intensity SD value method, the wavelength point where the reference spectrum line is located is selected as the offset point A Ta (a is the number of spectrometer channels), the specific process is as follows:

[0068] S331, set a window with a length of 2N+1, and a 'Move this window with a step size of 1;

[0069] S332, in the jth movement, compare the magnitudes of the spectral intensities corresponding to the wavelength points in the window. N+j When it is the maximum value in the interval and is greater than the spectral intensity corresponding to the previous and next wavelength points, the wavelength point can be considered as the peak point within the window length, recorded as m k , k=1,2,3…, until L is found a '. The maximum value of k is the total number of peak points; j is the number of window moves, j = 1, 2, 3, ..., b-2N.

[0070] S333, m k As the center, N is the radius and the wavelength interval Q is selected k , calculate the peak m k SD k The calculation formula is as follows:

[0071]

[0072] Among them, I n is the spectral intensity at the nth wavelength point, For interval Q k The mean value of the spectral intensity corresponding to all wavelength points within the .

[0073] S34, the reference point A Sa With offset point A Ta The wavelength difference is taken as the wavelength offset S between the master and slave instruments of the channel. a (a is the number of spectrometer channels)S a =|A Sa -A Ta |.

[0074] S4. Perform feature selection on the spectrum after wavelength shift self-correction to reduce the impact of irrelevant feature variables in the spectrum on model performance, and use the spectrum after feature selection as the input of the migration model.

[0075] Specifically, the feature selection method determines the input features for analyzing coal quality indicators by combining coal chemical analysis mechanisms with feature selection algorithms. Feature selection algorithms include, but are not limited to, the Competitive Adaptive Reweighted Sampling (CARS) algorithm, which selects highly correlated features as input features for the migration model.

[0076] S5. Establish and train a transfer learning model based on feature mapping and feature representation, and use the trained transfer learning model to achieve feature migration from the instrument spectrum to the master instrument spectrum.

[0077] Specifically, the feature transfer learning algorithm combines feature mapping methods and feature representation methods. Feature mapping methods include but are not limited to kernel principal component analysis (KPCA), and feature representation methods include but are not limited to piecewise direct standardization (PDS). The feature mapping method establishes a mapping of the master and slave instrument spectra in a high-dimensional space, converting nonlinear features in the data into linear features. Simultaneously, the parameters are optimized based on the objective function using optimization algorithms such as steepest descent, grid search, and Bayesian optimization to minimize the distance between the mappings. Feature representation methods are then used to calculate the kernel transfer matrix between instruments, enabling the migration of master and slave instrument spectral features.

[0078] S6. Use the master instrument spectrum and the migrated slave instrument spectrum to train a quantitative analysis model, and use the trained quantitative analysis model to predict the coal quality indicators of the slave instrument.

[0079] Exemplarily, training a coal quality quantitative analysis model (i.e., a quantitative analysis model) refers to using the sample spectral data of the training set to train coal quality quantitative analysis models such as calorific value, carbon content, and ash content based on machine learning algorithms such as partial least squares (PLS), multiple linear regression (MLR), and principal component regression (PCR), and determining the optimal model parameters through ten-fold cross-validation.

[0080] In some embodiments, the trained quantitative analysis model is also verified by inputting the spectral data of the test set sample into the trained model to predict the coal quality index and calculating the determination coefficient (R t 2 ), prediction root mean square error (RMSE P ), mean absolute error (MAE P ) and other indicators to evaluate the model performance.

[0081] The method of this embodiment is described in detail below with reference to the accompanying drawings and specific embodiments.

[0082] See also Figure 1 This embodiment provides a LIBS coal quality quantitative analysis method based on wavelength shift self-correction combined with feature transfer learning, including the following steps:

[0083] Step 1: 69 coal samples with known coal quality indicators were selected and LIBS spectra were collected on two instruments. The primary instrument tested 58 samples and the secondary instrument tested 27 samples.

[0084] Step 2: Randomly select 16 sample spectra (V S ) and select 16 coal sample spectra from the instrument detection coal samples (V T) constitute the training set of the migration model. The remaining 53 samples were systematically clustered according to ash and volatile matter, and 42 coal samples detected by the main instrument were selected according to the classification results as the training set for the quantitative analysis model (C S ), 11 coal samples detected by the instrument were used as the test set for testing the quantitative analysis model (T T ).

[0085] Step 3: Preprocess the spectra of 69 coal samples. The specific process is as follows:

[0086] Step 3.1: Use the characteristic peak SD value method to screen for valid spectra. Select a characteristic peak and calculate the SD value of the spectrum intensity for five wavelengths, centered on the wavelength corresponding to the maximum characteristic peak intensity. Set an appropriate threshold and exclude spectra with SD values ​​below the threshold as invalid. In this specific implementation example, CI 247.85 nm was selected as the characteristic peak for valid spectrum screening, and the threshold was set to 500.

[0087] Step 3.2: Use the channel normalization method to normalize the spectrum after valid data screening. Add the intensity of the spectrum of the same channel to get the sum of the intensity of each channel. The intensity of each wavelength point I i (i is the serial number of each wavelength point i = 1, 2, ..., 8192) and then divided by the sum of the corresponding channel intensities to obtain the normalized intensity I i '(i=1,2,…,8192).

[0088] Step 4: Perform wavelength offset correction on the pre-processed spectrum from the instrument using a wavelength offset self-correction method based on the peak standard deviation. The specific process is as follows:

[0089] Step 4.1: Randomly select a coal sample as the standard sample, and its spectral data is averaged after validity data screening to obtain the effective average spectrum of the master and slave instruments. The offset is calculated based on the effective average spectrum, and the calculation result is applied to all coal sample spectra after preprocessing.

[0090] Step 4.2: Select a band with high signal-to-background ratio in each channel of the main instrument spectrum. There is only one spectrum line with obvious excitation in this band as the reference spectrum line, and the wavelength point where it is located is the reference point A. Sa (a is the number of spectrometer channels), with the reference point as the center and the appropriate window width to determine the wavelength range L a (a is the number of spectrometer channels). The reference spectral lines selected in this specific embodiment are CI 247.86nm, Fe 1 393.351nm, Na 1 588.916nm, O 1777.697nm, corresponding to the reference point serial number A. SaThey are 785, 3454, 5543, and 7384 respectively.

[0091] Step 4.3: Select the same wavelength range L from the instrument spectrum i '. L i and L i ' is defined as follows:

[0092]

[0093] In this specific embodiment, the window length b is selected as 200, and the wavelength point number interval L of each channel is i They are [685,885], [3354,3554], [5443,5643], and [7214,7414] respectively.

[0094] Step 4.4: In L a Combined with the peak intensity SD value method, the wavelength point where the reference spectrum line is located is selected as the offset point A Ta (a is the number of spectrometer channels), the specific process is as follows:

[0095] Step 4.4.1: Set a window of length 2N+1 and a The window is moved with a step length of 1. In this specific implementation case, N=2, that is, the window length is 5.

[0096] Step 4.4.2: In the jth movement, compare the spectral intensities corresponding to the wavelength points in the window. When the spectral intensity I N+j When it is the maximum value in the interval and is greater than the spectral intensity corresponding to the previous and next wavelength points, the wavelength point N+j can be considered as the peak point, which is recorded as m k (k=1,2,3,...) until all peak points in L' are found. The maximum value of k is the total number of peak points; j is the number of window moves, j=1,2,3,...,b-2N.

[0097] Step 4.4.3: Take m k As the center, N is the radius to divide the wavelength interval Q k , calculate m k SD k The calculation formula is as follows:

[0098]

[0099] Among them, I n is the spectral intensity at the nth wavelength point, For interval Q k The mean value of the spectral intensity corresponding to all wavelength points within the .

[0100] Step 4.4.4: Compare all calculated SDs k , m corresponding to the maximum value k That is the offset point A Ta In this specific embodiment, the offset point A of each channel is obtained Ta The wavelength offset S of each channel is calculated as 761, 3424, 5494, and 7314. a They are 24, 30, 49 and 70 respectively.

[0101] Step 4.4.5: Use the channel-by-channel combination of wavelength offset to intercept the spectrum bands of the master and slave instruments to achieve wavelength offset self-calibration. Sa and A Ta The wavelength number can be seen from the offset point A of each channel of the instrument spectrum. Ta The wavelength numbers are all smaller than the corresponding reference point A from the instrument spectrum Sa The wavelength sequence number indicates that the slave instrument spectrum is offset to the left relative to the master instrument spectrum. Therefore, the left band of the slave instrument spectrum is intercepted as follows: [1:2024], [2049:4066], [4097:6095], [6145:8122]; the right band of the master instrument spectrum is intercepted as follows: [25:2048], [2079:4096], [4146:6144], [6215:8192]. Finally, by aligning the master and slave instrument bands by channel, wavelength offset correction for the slave instrument spectrum is achieved.

[0102] Step 5: The feature selection method combines the coal chemical analysis mechanism with the CARS feature selection algorithm to determine the input features for analyzing coal quality indicators. The correlation and interaction mechanism between each coal quality indicator and element content are analyzed, and the element spectral lines and related bands that have a significant impact on coal quality indicator analysis are selected as empirical characteristic peak bands. CARS is further combined with Monte Carlo sampling and an exponentially decreasing function (EDF) to adaptively adjust the selection probability of each band. Ultimately, the optimal band combination that contributes most to model performance is selected based on the coal quality indicator. In this specific implementation case, a total of 230 features were selected based on the coal chemical analysis mechanism, including the bands containing 46 empirical characteristic peaks, including CI 247.86nm, Fe I 393.351nm, Si I 288.124nm, and MgI 285.161nm. Through the CARS feature selection algorithm, the optimal input band feature number of the calorific value model is 179, the optimal input band feature number of the carbon content model is 52, and the optimal input band feature number of the ash content model is 17.

[0103] Step 6: Use the spectral data V processed by steps 1-5 S and V TThe transfer model was trained with the objective function of minimizing the Maximum Mean Discrepancy (MMD). The steepest descent algorithm was used to optimize the Gaussian kernel parameter σ, and the appropriate number of KPCA principal components j and PDS window width c were selected. In this specific implementation, the Gaussian kernel parameter σ for the calorific value model was 1.23, the number of KPCA principal components was 16, and the window width c was 8; the Gaussian kernel parameter σ for the carbon content model was 0.65, the number of KPCA principal components was 16, and the window width c was 4; and the Gaussian kernel parameter σ for the ash content model was 0.64, the number of KPCA principal components was 16, and the window width c was 8.

[0104] Step 7: See Figure 2 , combined with the parameter σ, V S 、V T The two sets of spectral data are mapped to high-dimensional space through Gaussian kernel function and PCA dimensionality reduction is performed to obtain the corresponding eigenvector ω S and ω T , and the high-dimensional mapping K G (V S ), K G (V T ), and use PDS to calculate K in combination with parameter c G (V S ) and K G (V T ) between the master and slave instruments to obtain the migration matrix F between the master and slave instruments. The specific calculation process is as follows:

[0105] K G (V S ) i =K G (V T ) i-k,i+w ×f i T

[0106] Among them, K G (V S ) i is the intensity value of the i-th column of the main instrument spectrum after mapping, K G (V T ) i-k,i+w The column vector f of the migration matrix F is the interval of the intensity value of the i-th column of the spectrum of the slave instrument after mapping and the master instrument spectrum, and the window width is c = k + w + 1. i T The calculation process is as follows:

[0107]

[0108] f i T=[β i,0 ,β i,1 ,...,β i,k+w+1 ] T

[0109] Among them, K G (V T ) t is the intensity value of the tth column from the instrument spectrum after mapping, β i,t-i+k+1 is the regression coefficient, β i,0 is the residual coefficient.

[0110] The final migration matrix can be expressed as

[0111] Step 8: According to σ, ω S and ω T , the spectral data C of the coal quality quantitative analysis model training set after wavelength offset self-correction S and T T Get K through KPCA G (C S ) and K G (T T ).

[0112] Step 9: For the data K from the instrument G (T T ), combined with the migration matrix F, the characteristic migration spectrum K from the instrument to the main instrument is calculated G (T T_S The specific calculation formula is as follows:

[0113] K G (T T_S )=K G (T T )×F

[0114] Step 10: Utilize K G (C S ) and its corresponding coal quality index (calorific value, carbon content and ash content) reference value Y S , based on the partial least squares (PLS) method to train the quantitative analysis model and use ten-fold cross validation to find the best number of principal components of the model. In this specific embodiment, the optimal number of principal components of the calorific value quantitative analysis model is 8, the optimal number of principal components of the carbon content quantitative analysis model is 14, and the optimal number of principal components of the ash content quantitative analysis model is 8. The performance of the model is tested by inputting the instrument spectrum data into the above-mentioned coal quality quantitative analysis model to predict the coal quality indicators. Compared with the traditional model without transfer learning, the RMSE of calorific value is 0. P 、MAE PThe carbon content of R t 2 , RMSE P 、MAE P The ash content R was optimized from 0.738, 6.434wt.%, 6.149wt.% to 0.937, 1.324wt.%, 1.202wt.%, respectively. t 2 , RMSE P 、MAE P The results were optimized from 0.894, 10.031wt.%, 9.393wt.% to 0.911, 1.481wt.%, 1.303wt.% respectively.

[0115] See also Figure 3 , Figure 3 Comparison of the heat generation prediction results between the traditional non-transfer learning model and the model of the present invention, where Figure 3 (a) is the heat generated by the traditional non-transfer learning model. Figure 3 (b) is the calorific value of the model of the present invention. Figure 4 , Figure 4 Comparison of carbon content prediction results between the traditional non-transfer learning model and the proposed model, where Figure 4 (a) is the comparison of carbon content prediction results of the traditional non-transfer learning model. Figure 4 (b) is a comparison of the carbon content prediction results of the model of the present invention. Figure 5 , Figure 5 Comparison of ash content prediction results between the traditional non-transfer learning model and the proposed model, where Figure 5 (a) is the comparison of the ash content prediction results of the traditional non-transfer learning model. Figure 5 (b) is a comparison of the ash content prediction results of the model of the present invention.

[0116] Example 2

[0117] An embodiment of the present invention further provides an electronic device, comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the following Figure 6 A LIBS coal quality quantitative analysis method is shown.

[0118] It is understood that the memory may include random access memory (RAM) or read-only memory (ROM). Optionally, the memory includes a non-transitory computer-readable storage medium. The memory may be used to store instructions, programs, codes, code sets, or instruction sets. The memory may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function, instructions for implementing the various method embodiments described above, etc.; the data storage area may store data created based on the use of the server, etc.

[0119] The processor may include one or more processing cores. The processor utilizes various interfaces and circuits to connect various components within the server. It executes various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory, as well as accessing data stored in memory. Optionally, the processor may be implemented using at least one of the following hardware forms: digital signal processing (DSP), field-programmable gate array (FPGA), and programmable logic array (PLA). The processor may integrate one or a combination of a central processing unit (CPU) and a modem. The CPU primarily processes the operating system and application programs, while the modem handles wireless communications. It is understood that the modem may not be integrated into the processor and may be implemented separately via a single chip.

[0120] Since the electronic device is an electronic device corresponding to a LIBS coal quality quantitative analysis method in an embodiment of the present invention, and the principle of solving the problem by the electronic device is similar to that of the method, the implementation of the electronic device can refer to the implementation process of the above-mentioned method embodiment, and the repeated parts will not be repeated.

[0121] Example 3

[0122] An embodiment of the present invention further provides a computer-readable storage medium, wherein the storage medium stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, the at least one program, the code set or instruction set is loaded and executed by a processor to implement the following Figure 6 A LIBS coal quality quantitative analysis method is shown.

[0123] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program. The program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0124] Since the storage medium is a storage medium corresponding to a LIBS coal quality quantitative analysis method in an embodiment of the present invention, and the principle of solving the problem by the storage medium is similar to that of the method, the implementation of the storage medium can refer to the implementation process of the above-mentioned method embodiment, and the repeated parts will not be repeated.

[0125] Example 4

[0126] In some possible implementations, various aspects of the methods of the embodiments of the present invention may also be implemented in the form of a program product, which includes program code. When the program product is executed on a computer device, the program code is used to cause the computer device to perform the steps of the LIBS coal quality quantitative analysis method according to various exemplary embodiments of the present application as described above in this specification. The executable computer program code or "code" used to perform the various embodiments may be written in a high-level programming language such as C, C++, Python, Smalltalk, Java, JavaScript, Visual Basic, Structured Query Language (e.g., Transact-SQL), Perl, or various other programming languages.

[0127] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0128] In the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0129] The above embodiments are intended only to illustrate the technical concepts and features of the present invention. Their purpose is to enable those skilled in the art to understand the contents of the present invention and implement them accordingly. They are not intended to limit the scope of protection of the present invention. Any equivalent changes or modifications made based on the essence of the present invention are intended to be covered by the scope of protection of the present invention.

Claims

1. A LIBS coal quality quantitative analysis method, characterized in that: The following steps are involved: LIBS spectra of the same coal sample were collected using different instruments, and the spectra were divided into two groups: a training set for training the migration model and a training set for training the quantitative analysis model. Perform spectral preprocessing on the collected spectra; The pre-processed spectrum is self-calibrated for wavelength offset, so that the spectrum collected by the slave instrument corresponds to the spectrum collected by the main instrument in wavelength; Perform feature selection on the spectrum after wavelength shift self-correction, and use the spectrum after feature selection as the input of the migration model; Establish and train a transfer learning model based on feature mapping and feature representation, and use the trained transfer learning model to achieve feature migration from instrument spectra to master instrument spectra; The master instrument spectrum and the migrated slave instrument spectrum are used to train the quantitative analysis model, and the trained quantitative analysis model is used to predict the coal quality indicators of the slave instrument.

2. A LIBS coal quality quantitative analysis method according to claim 1, characterized in that: The collected spectra are respectively subjected to spectral preprocessing, including: Perform effective spectrum screening and spectrum noise reduction on the collected spectra; Among them, the effective spectrum screening method includes an effective spectrum screening method based on the standard deviation value of the characteristic peak intensity, and the spectrum noise reduction method includes data normalization.

3. A LIBS coal quality quantitative analysis method according to claim 1, characterized in that: The wavelength offset self-correction method is based on the recognition and correction of peak SD values: any coal sample is selected as the standard sample, the effective spectrum of the coal sample is collected on different instruments, the offset of each channel of the spectrum between instruments is calculated and aligned with the main instrument for correction, thereby realizing wavelength offset self-correction of the spectrum from the instrument.

4. A LIBS coal quality quantitative analysis method according to claim 1, characterized in that: The wavelength offset self-correction of the pre-processed spectrum to achieve wavelength correspondence between the spectrum collected by the slave instrument and the spectrum collected by the main instrument includes: Select a band with relatively high signal-to-background ratio in each channel of the main instrument spectrum. In this band, there is only one spectrum line with obvious excitation as the reference spectrum line, and the wavelength point where it is located is the reference point A. Sa , with the reference point as the center, select the wavelength range L with the preset window a ; Select the same wavelength range L from the instrument spectrum a '; In the interval L a Combined with the peak intensity SD value method, the wavelength point where the reference spectrum line is located is selected as the offset point A Ta ; Set reference point A Sa With offset point A Ta The wavelength difference is taken as the wavelength offset S between the master and slave instruments of the channel. a .

5. A LIBS coal quality quantitative analysis method according to claim 4, characterized in that: In the interval L a Combined with the peak intensity SD value method, the wavelength point where the reference spectrum line is located is selected as the offset point A Ta ,include: Set a window of length 2N+1 and a 'Move this window with a step size of 1; In the jth movement, the size of the spectral intensity corresponding to the wavelength point in the window is compared; when the spectral intensity I N+j When it is the maximum value in the interval and is greater than the spectral intensity corresponding to the previous and next wavelength points, the wavelength point is identified as the peak point within the window length, recorded as m k , k=1,2,3…, until L is found a 'All the peak points in the window; where the maximum value of k is the total number of peak points; j is the number of window moves, j = 1, 2, 3, ..., b-2N, b is the selected window length; m k As the center, N is the radius and the wavelength interval Q is selected k , calculate the peak m k SD k , the calculation formula is as follows: Among them, I n is the spectral intensity at the nth wavelength point, For interval Q k The mean value of the spectral intensity corresponding to all wavelength points within; Compare all calculated SDs k , m corresponding to the maximum value k That is, the reference point A Sa Corresponding offset point A Ta .

6. A LIBS coal quality quantitative analysis method according to claim 1, characterized in that: The feature selection of the spectrum after the wavelength shift self-correction includes: The input features for analyzing coal quality indicators are determined by combining the coal chemical analysis mechanism with the feature selection algorithm. The feature selection algorithm includes a competitive adaptive reweighted sampling algorithm, which selects features with high correlation as input features of the migration model.

7. A LIBS coal quality quantitative analysis method according to claim 1, characterized in that: The feature mapping method includes kernel principal component analysis, and the feature representation method includes segmented direct normalization. The feature mapping method is used to establish a mapping of the master and slave instrument spectra in a high-dimensional space, and the nonlinear features in the data are converted into linear features. At the same time, the parameters are optimized according to the objective function combined with the optimization algorithm to minimize the distance between the mappings. The feature representation method is then used to calculate the kernel migration matrix between instruments to realize the migration of the master and slave instrument spectral features.

8. A LIBS coal quality quantitative analysis method according to claim 1, characterized in that: The training of the transfer learning model includes: The spectrum V collected by the main instrument after feature selection S and the spectrum V collected from the instrument T The transfer model is trained with the objective function of minimizing the maximum mean difference, optimizing the Gaussian kernel function parameter σ, and selecting the number of KPCA principal components j and the PDS window width c; According to the parameter σ, the spectrum V collected by the main instrument after feature selection is S and the spectrum V collected from the instrument T , mapped to high-dimensional space by Gaussian kernel function and PCA dimensionality reduction processing, the corresponding eigenvector ω is obtained S and ω T , and the high-dimensional mapping K G (V S ), K G (V T ); Combined with the window width c, use PDS to calculate K G (V S ) and K G (V T ) and obtain the migration matrix F between the master and slave instruments.

9. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that The storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the method according to any one of claims 1 to 8.

Images