Coal quality detection method, electronic equipment and storage medium

Abstract

The invention discloses a coal quality detection method, electronic equipment and a storage medium. The method comprises the following steps: acquiring a spectral coal quality data set of a training coal sample; based on the spectral coal quality data set, a plurality of coal quality index prediction models are established for different coal quality indexes through multiple algorithms, and the coal quality index prediction models take spectral data as input to predict predicted coal quality index data corresponding to the spectral data; comparing the plurality of coal quality index prediction models, and selecting an optimal coal quality index prediction model of each coal quality index; and acquiring spectral data of a to-be-detected coal sample, and predicting predicted coal quality index data corresponding to the spectral data of the to-be-detected coal sample by adopting the optimal coal quality index prediction model. According to the method, independent modeling is carried out for specific coal quality indexes of specific coal sample types, the technical problems of low prediction accuracy, wrong judgment and the like of a wide model are solved, and the prediction accuracy of the specific coal sample is improved.

Description

Technical Field

[0001] The present invention relates to the technical field related to coal, and in particular to a coal quality detection method, electronic equipment and storage medium. Background Art

[0002] Coal is still the most important energy source in my country's energy structure. In the country's future energy policies and plans, efficient and clean application of coal is the only way to transform coal. Coal is a mineral that can be used as fuel or industrial raw material. Its composition and structure are very complex, which is the most critical factor affecting the efficient use of coal. If the key components in coal quality can be accurately measured, the corresponding coal utilization process can be adjusted according to its composition and characteristics to achieve efficient and clean application.

[0003] Coal quality testing refers to the technical method of obtaining coal composition and property indicators by using physical and / or chemical methods to test coal samples in accordance with national standards or relevant methods. This process often takes at least one day. The existing coal quality analysis methods mostly use chemical testing methods for coal quality analysis, and the results are highly accurate and representative. However, the implementation process requires systematic sampling and sample preparation of coal quality samples, resulting in a large time granularity and low timeliness of coal quality test results, making it difficult for process personnel to adjust operating parameters in real time according to the current application coal quality conditions, and even more difficult to meet the current needs of intelligent adjustment of coal applications.

[0004] Near-infrared analysis technology is one of the most promising and widely used rapid analysis methods. Near-infrared spectroscopy mainly records the frequency-doubled and frequency-combined absorption information of hydrogen groups (OH, CH, NH, SH) in organic molecules to near-infrared light of different wavelengths. Hydrogen-containing groups are the basic units of the molecular structure of most organic substances. Different spectra are formed by different absorption of near-infrared light. The composition of the substance can be quickly determined by analyzing the spectrum. Compared with other rapid detection technologies, near-infrared spectroscopy technology has the characteristics of mature technology, no radioactive sources, simple operation, low cost, and strong field applicability.

[0005] For coal samples, the near-infrared spectrum can contain the structure and composition information of most organic molecules, but because different organic substances have different compositions and contain different groups, different groups will absorb near-infrared light differently. Even the same group will absorb near-infrared light differently in different physical and chemical environments. This makes the near-infrared spectra of different substances often different. This difference in the spectrum enables the near-infrared spectrum to carry useful information about the composition structure of the sample, and thus realize the quantitative detection and analytical research of important coal quality indicators such as calorific value, sulfur content, total moisture, ash content, volatile matter and fixed carbon.

[0006] However, the existing coal quality rapid detection technology based on near-infrared spectroscopy generally only uses a single algorithm to build a model, which is a broad model. However, there are many types of coal samples, and the broad model built by a single algorithm has technical problems such as low prediction accuracy and misjudgment when used for different coal samples. Summary of the invention

[0007] Based on this, it is necessary to provide a coal quality detection method, electronic device and storage medium to address the technical problems that the existing technology uses a broad model with a single algorithm to model different coal samples, which has technical problems such as low prediction accuracy and wrong judgment.

[0008] The present invention provides a coal quality detection method, comprising:

[0009] Acquire a spectral coal quality data group of a training coal sample, wherein the spectral coal quality data group includes a plurality of spectral coal quality data pairs, each pair of the spectral coal quality data pairs includes spectral data of the training coal sample and corresponding experimental coal quality index data, and the training coal sample and the coal sample to be tested are the same type of coal samples;

[0010] Based on the spectral coal quality data group, a plurality of coal quality index prediction models are respectively established for different coal quality indexes by using a plurality of algorithms, wherein the coal quality index prediction models take the spectral data as input and predict the predicted coal quality index data corresponding to the spectral data;

[0011] Compare multiple coal quality index prediction models and select the optimal coal quality index prediction model for each coal quality index;

[0012] The spectral data of the coal sample to be tested is obtained, and the optimal coal quality index prediction model of the coal quality index to be predicted is used to predict the predicted coal quality index data corresponding to the spectral data of the coal sample to be tested.

[0013] Furthermore, the experimental coal quality index data of the training coal sample and the predicted coal quality index data required to be predicted for the coal sample to be tested are data of the same coal quality index.

[0014] Furthermore, the comparison of multiple coal quality index prediction models and selection of the optimal coal quality index prediction model for each coal quality index specifically includes:

[0015] Calculate the verification parameters of each coal quality index prediction model, compare the verification parameters of multiple coal quality index prediction models for the same coal quality index, and select the optimal coal quality index prediction model for each coal quality index.

[0016] Furthermore, based on the spectral coal quality data set, multiple coal quality index prediction models are established respectively using multiple algorithms for different coal quality indexes, specifically including:

[0017] Preprocessing the spectral coal quality data set using a plurality of preprocessing methods to obtain a plurality of preprocessed spectral coal quality data sets;

[0018] For different coal quality indicators, multiple algorithms are used to establish multiple coal quality indicator prediction models based on multiple groups of preprocessed spectral coal quality data groups. Each coal quality indicator prediction model corresponds to a coal quality indicator, an algorithm and a group of preprocessed spectral coal quality data groups.

[0019] Furthermore, after comparing the multiple coal quality index prediction models and selecting the optimal coal quality index prediction model for each coal quality index, the method further includes:

[0020] The preprocessed spectral coal quality data set corresponding to the optimal coal quality index prediction model for each coal quality index is selected as the optimal spectral coal quality data set for the coal quality index.

[0021] Further, the acquiring of the spectral data of the coal sample to be tested and the use of the optimal coal quality index prediction model for the coal quality index to be predicted to predict the predicted coal quality index data corresponding to the spectral data of the coal sample to be tested specifically include:

[0022] Acquiring spectrum data of the coal sample to be tested;

[0023] The spectrum preprocessing method for generating the optimal spectrum coal quality data group is used to preprocess the spectrum data of the coal sample to be tested, so as to obtain preprocessed spectrum data of the coal sample to be tested;

[0024] The optimal coal quality index prediction model is used to predict the predicted coal quality index data corresponding to the preprocessed spectral data of the coal sample to be tested.

[0025] Furthermore, the use of multiple preprocessing methods to preprocess the spectral coal quality data set to obtain multiple groups of preprocessed spectral coal quality data sets specifically includes:

[0026] Determine a plurality of spectral pre-processing methods and / or a plurality of characteristic spectrum selection methods;

[0027] Each of the spectral preprocessing methods is used to preprocess the spectral coal quality data group, and / or each of the characteristic spectrum selection methods is used to perform spectral selection on the spectral data of each pair of the spectral coal quality data pairs in the spectral coal quality data group, so as to obtain multiple groups of preprocessed spectral coal quality data groups.

[0028] Further, the acquiring of the spectral data of the coal sample to be tested and the use of the optimal coal quality index prediction model for the coal quality index to be predicted to predict the predicted coal quality index data corresponding to the spectral data of the coal sample to be tested specifically include:

[0029] Acquiring spectrum data of the coal sample to be tested;

[0030] Preprocessing the spectral data of the coal sample to be tested by using the spectral preprocessing method for generating the optimal spectral coal quality data group, and / or performing spectral selection on the spectral data of the coal sample to be tested by using the characteristic spectrum selection method for generating the optimal spectral coal quality data group, so as to obtain preprocessed spectral data of the coal sample to be tested;

[0031] The optimal coal quality index prediction model is used to predict the predicted coal quality index data corresponding to the preprocessed spectral data of the coal sample to be tested.

[0032] The present invention provides an electronic device, comprising:

[0033] at least one processor; and,

[0034] a memory communicatively connected to at least one of the processors; wherein,

[0035] The memory stores instructions that can be executed by at least one of the processors. The instructions are executed by at least one of the processors so that the at least one of the processors can execute the coal quality detection method as described above.

[0036] The present invention provides a storage medium, which stores computer instructions. When a computer executes the computer instructions, it is used to execute all the steps of the coal quality detection method as described above.

[0037] The present invention adopts multiple algorithms to model each coal quality index based on the same type of training coal samples for different coal samples to be tested, and selects the optimal coal quality index prediction model from them. Therefore, the present invention independently models specific coal quality indicators of specific coal sample types, solves technical problems such as low prediction accuracy and misjudgment of broad models, and improves the accuracy of prediction of specific coal samples. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a workflow diagram of a coal quality detection method according to an embodiment of the present invention;

[0039] Figure 2 This is a workflow diagram of a coal quality detection method according to another embodiment of the present invention;

[0040] Figure 3 A schematic diagram of modeling of an embodiment of the present invention;

[0041] Figure 4 A working flow chart of a coal quality detection method according to the best embodiment of the present invention;

[0042] Figure 5 Schematic diagram of near infrared spectrometer irradiating coal sample;

[0043] Figure 6 The figure is a schematic diagram of the hardware structure of an electronic device of the present invention. DETAILED DESCRIPTION

[0044] The specific embodiments of the present invention are further described below in conjunction with the accompanying drawings. The same components are represented by the same reference numerals. It should be noted that the words "front", "rear", "left", "right", "upper" and "lower" used in the following description refer to directions in the accompanying drawings, and the words "inner" and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0045] like Figure 1 The figure shows a working flow chart of a coal quality detection method according to an embodiment of the present invention, comprising:

[0046] Step S101, obtaining a spectral coal quality data group of a training coal sample, wherein the spectral coal quality data group includes a plurality of spectral coal quality data pairs, each pair of the spectral coal quality data pairs includes spectral data of the training coal sample and corresponding experimental coal quality index data, and the training coal sample and the coal sample to be tested are the same type of coal samples;

[0047] Step S102, based on the spectral coal quality data group, multiple algorithms are used to respectively establish multiple coal quality index prediction models for different coal quality indexes, wherein the coal quality index prediction models use the spectral data as input to predict the predicted coal quality index data corresponding to the spectral data;

[0048] Step S103, comparing multiple coal quality index prediction models, and selecting the optimal coal quality index prediction model for each coal quality index;

[0049] Step S104, obtaining the spectral data of the coal sample to be tested, and using the optimal coal quality index prediction model for the coal quality index to be predicted to predict the predicted coal quality index data corresponding to the spectral data of the coal sample to be tested.

[0050] Specifically, the present invention can be applied to electronic devices having processing capabilities.

[0051] First, step S101 is executed to obtain a spectral coal quality data set of a training coal sample.

[0052] Specifically, the coal sample is processed by repeated reduction, crushing, mixing, air drying and screening operations to gradually reduce its quantity and particle size to prepare a sample. The sample is tested in the laboratory to obtain experimental coal quality index data, which is used to build a coal quality index database. The spectral data of the training coal sample is obtained by performing a spectrum sampling experiment on the sample using an interference near-infrared spectrometer, and the spectral data is matched with the aforementioned experimental coal quality index data to build a spectrum-coal quality index database. When executing step S101, a spectral coal quality data group of the training coal sample is selected from the spectrum-coal quality index database. The training coal sample is selected according to the coal sample to be tested, and the training coal sample and the coal sample to be tested are the same type of coal sample. The spectral coal quality data group includes multiple pairs of spectral coal quality data pairs, and each pair of the spectral coal quality data pairs includes the spectral data of the training coal sample and the corresponding experimental coal quality index data.

[0053] Among them, the detectable coal quality indicators are coal ash (Ash), moisture (Mt), fixed carbon (FC), coal calorific value (Q), coal sulfur (St) and coal volatile matter (V).

[0054] Then, step S102 is executed to respectively establish a plurality of coal quality index prediction models for different coal quality indexes using a plurality of algorithms based on the spectral coal quality data set.

[0055] Specifically, each algorithm establishes a coal quality index prediction model for different coal quality indexes based on the spectral coal quality data set obtained in step S101. For each coal quality index, multiple algorithms have multiple coal quality index prediction models for the coal quality index. The coal quality index prediction model takes the spectral data as input and predicts the predicted coal quality index data corresponding to the spectral data.

[0056] Finally, step S103 is executed to compare multiple coal quality index prediction models and select the optimal coal quality index prediction model for each coal quality index.

[0057] Specifically, based on the comparison between the predicted coal quality index data and the experimental coal quality index data, one or more verification parameters of the coal quality index prediction model are compared to select the optimal coal quality index prediction model.

[0058] After determining the optimal coal quality index prediction model, step S104 is executed to use the optimal prediction model to predict the coal quality index data corresponding to the spectral data of the coal sample to be tested for the coal quality index to be predicted.

[0059] Specifically, the interferometric near-infrared spectrometer is also used to irradiate the sample to be tested, obtain the spectral results, and use the optimal prediction model to make a rapid prediction to display the coal quality results.

[0060] The present invention adopts multiple algorithms to model each coal quality index based on the same type of training coal samples for different coal samples to be tested, and selects the optimal coal quality index prediction model from them. Therefore, the present invention independently models specific coal quality indicators of specific coal sample types, solves technical problems such as low prediction accuracy and misjudgment of broad models, and improves the accuracy of prediction of specific coal samples.

[0061] like Figure 2 The figure shows a working flow chart of a coal quality detection method in another embodiment of the present invention, comprising:

[0062] Step S201, obtaining a spectral coal quality data group of a training coal sample, the spectral coal quality data group comprising a plurality of spectral coal quality data pairs, each pair of the spectral coal quality data pairs comprising spectral data of the training coal sample and corresponding experimental coal quality index data, the training coal sample and the coal sample to be tested are the same type of coal samples, the experimental coal quality index data of the training coal sample and the predicted coal quality index data required to be predicted for the coal sample to be tested are data of the same coal quality index.

[0063] Step S202, using a plurality of preprocessing methods to preprocess the spectral coal quality data set to obtain a plurality of preprocessed spectral coal quality data sets.

[0064] In one embodiment, the preprocessing of the spectral coal quality data set by using multiple preprocessing methods to obtain multiple groups of preprocessed spectral coal quality data sets specifically includes:

[0065] Determine a plurality of spectral pre-processing methods and / or a plurality of characteristic spectrum selection methods;

[0066] Each of the spectral preprocessing methods is used to preprocess the spectral coal quality data group, and / or each of the characteristic spectrum selection methods is used to perform spectral selection on the spectral data of each pair of the spectral coal quality data pairs in the spectral coal quality data group, so as to obtain multiple groups of preprocessed spectral coal quality data groups.

[0067] Step S203, for different coal quality indicators, multiple algorithms are used to establish multiple coal quality indicator prediction models based on multiple groups of pre-processed spectral coal quality data groups, each of the coal quality indicator prediction models corresponds to a coal quality indicator, an algorithm and a group of pre-processed spectral coal quality data groups, and the coal quality indicator prediction model takes the spectral data as input to predict the predicted coal quality indicator data corresponding to the spectral data.

[0068] Step S204, calculating the verification parameters of each of the coal quality index prediction models, comparing the verification parameters of multiple coal quality index prediction models of the same coal quality index, and selecting the optimal coal quality index prediction model for each coal quality index.

[0069] Step S205, selecting the preprocessed spectral coal quality data set corresponding to the optimal coal quality index prediction model for each coal quality index as the optimal spectral coal quality data set for the coal quality index.

[0070] Step S206, obtaining the spectral data of the coal sample to be tested, and using the optimal coal quality index prediction model for the coal quality index to be predicted to predict the predicted coal quality index data corresponding to the spectral data of the coal sample to be tested.

[0071] In one embodiment, the step of acquiring the spectral data of the coal sample to be tested and using the optimal coal quality index prediction model of the coal quality index to be predicted to predict the predicted coal quality index data corresponding to the spectral data of the coal sample to be tested specifically includes:

[0072] Acquiring spectrum data of the coal sample to be tested;

[0073] The spectrum preprocessing method for generating the optimal spectrum coal quality data group is used to preprocess the spectrum data of the coal sample to be tested, so as to obtain preprocessed spectrum data of the coal sample to be tested;

[0074] The optimal coal quality index prediction model is used to predict the predicted coal quality index data corresponding to the preprocessed spectral data of the coal sample to be tested.

[0075] In one embodiment, the step of acquiring the spectral data of the coal sample to be tested and using the optimal coal quality index prediction model of the coal quality index to be predicted to predict the predicted coal quality index data corresponding to the spectral data of the coal sample to be tested specifically includes:

[0076] Acquiring spectrum data of the coal sample to be tested;

[0077] Preprocessing the spectral data of the coal sample to be tested by using the spectral preprocessing method for generating the optimal spectral coal quality data group, and / or performing spectral selection on the spectral data of the coal sample to be tested by using the characteristic spectrum selection method for generating the optimal spectral coal quality data group, so as to obtain preprocessed spectral data of the coal sample to be tested;

[0078] The optimal coal quality index prediction model is used to predict the predicted coal quality index data corresponding to the preprocessed spectral data of the coal sample to be tested.

[0079] This embodiment proposes a method for constructing a model for rapid detection of coal quality using a near-infrared spectrometer. A spectrum-coal quality database is constructed through near-infrared spectrum irradiation and national standard coal quality detection. According to specific coal quality indicators, a single or multiple pre-processing methods are automatically selected to process the spectrum; characteristic spectra are selected for specific coal quality indicators; a coal quality indicator prediction model is established using multiple algorithms, including conventional PCA methods and deep learning algorithms; for unknown coal quality indicators, a near-infrared spectrometer is used to irradiate the sample, and the coal quality indicator prediction model is used to quickly predict the result. The modeling process is completed according to a certain built-in system to ensure the accuracy of the prediction. The modeling ideas of this embodiment are as follows: Figure 3 As shown:

[0080] The near-infrared coal quality rapid detection modeling process can be divided into two parts: a modeling data processing part 310 based on mathematical methods, and a quantitative model analysis part 320 based on mathematical regression.

[0081] A coal sample 330 is obtained from raw coal 360 , and a national standard analysis result 340 of the coal sample is obtained through national standard analysis, and a near infrared spectrum 350 is obtained through near infrared analysis.

[0082] The quantitative model analysis (regression prediction) part 320 is based on the intrinsic connection between the spectral data of the near-infrared spectrum 350 of the coal sample and the various coal quality index data of the national standard analysis results 340 of the coal sample. By establishing a coal quality index prediction model between the two, the coal quality prediction index model is a regression model (quantitative relationship). The process of using mathematical methods to continuously approximate the true correspondence between the two to obtain the predicted values ​​of various coal quality indicators of unknown coal samples is the core of rapid coal quality detection and a direct reflection of the effectiveness and reliability of the analysis results.

[0083] The modeling data processing part 310 is a supplement to the quantitative model analysis part 320, and its purpose is to improve the prediction performance of the coal quality index prediction model. In terms of content, it can be divided into national standard benchmark optimization 311 for the national standard analysis results 340 of coal samples, spectral data preprocessing 312 for near-infrared spectra 350, characteristic spectrum extraction 313 and abnormal sample screening 314 for the "spectrum-coal quality" data set.

[0084] Specifically, first, step S201 is executed to obtain a spectral coal quality data set of a training coal sample.

[0085] Specifically, a spectral coal quality data group of a training coal sample is selected from a spectral-coal quality index database obtained in advance through experiments. The training coal sample is selected according to the coal sample to be tested, and the training coal sample and the coal sample to be tested are the same type of coal samples. The spectral coal quality data group includes a plurality of pairs of spectral coal quality data pairs, and each pair of the spectral coal quality data pairs includes the spectral data of the training coal sample and the corresponding experimental coal quality index data. At the same time, the experimental coal quality index data of the training coal sample and the predicted coal quality index data required to be predicted for the coal sample to be tested are data of the same coal quality index, that is, the experimental coal quality index data of the training coal sample is the data of the coal quality index required to be measured for the coal sample to be tested.

[0086] Then, step S202 is executed to preprocess the spectral coal quality data set using a plurality of preprocessing methods to obtain a plurality of preprocessed spectral coal quality data sets.

[0087] Specifically, preprocessing includes eliminating random noise in the spectrum, enhancing the characteristic information of "interest" in the spectrum, and weakening scattering interference through spectral preprocessing methods and related spectral data preprocessing techniques to obtain spectral data with high signal-to-noise ratio, thereby ensuring that subsequent modeling and analysis are accurate and effective. In the near-infrared spectral analysis model, commonly used spectral preprocessing methods can be divided into three categories: smoothing methods to eliminate random noise, multivariate scattering correction methods to reduce scattering interference, and data transformation methods and derivation methods to enhance the difference between spectra.

[0088] Preprocessing also includes feature spectrum selection, and the methods include correlation coefficient method, uninformative variable elimination method, Monte Carlo uninformative variable elimination method and competitive adaptive reweighted sampling method.

[0089] For the spectral coal quality data set, a variety of preprocessing methods are used to preprocess it, and multiple groups of preprocessed spectral coal quality data sets are obtained.

[0090] In one embodiment, the preprocessing of the spectral coal quality data set by using multiple preprocessing methods to obtain multiple groups of preprocessed spectral coal quality data sets specifically includes:

[0091] Determine various spectral pre-processing methods;

[0092] Each of the spectral preprocessing methods is used to preprocess the spectral coal quality data group, and / or each of the characteristic spectrum selection methods is used to perform spectral selection on the spectral data of each pair of the spectral coal quality data pairs in the spectral coal quality data group, so as to obtain multiple groups of preprocessed spectral coal quality data groups.

[0093] In this embodiment, for a group of spectral coal quality data sets, a plurality of spectral preprocessing methods and / or characteristic spectrum selection methods are used to perform preprocessing and / or spectral processing, thereby obtaining a plurality of groups of preprocessed spectral coal quality data sets.

[0094] In some embodiments, the spectral coal quality data set is pre-processed by using each of the spectral pre-processing methods, and / or the spectral data of each pair of the spectral coal quality data pairs in the spectral coal quality data set are spectrally selected by using each of the characteristic spectral selection methods, to obtain multiple groups of pre-processed spectral coal quality data sets, including:

[0095] For a set of spectral coal quality data sets, a plurality of spectral pre-processing methods are used to perform pre-processing respectively, so as to obtain a plurality of pre-processed spectral coal quality data sets; or

[0096] For a set of spectral coal quality data sets, a plurality of characteristic spectrum selection methods are used to select the spectrum, thereby obtaining a plurality of pre-processed spectral coal quality data sets; or

[0097] For a group of spectral coal quality data sets, a variety of spectral preprocessing methods are used for preprocessing to obtain multiple groups of preprocessed spectral coal quality data sets, and then a variety of characteristic spectrum selection methods are used to perform spectral selection on each group of processed spectral coal quality data sets to obtain multiple groups of preprocessed spectral coal quality data sets.

[0098] Then, execute step S203, and for different coal quality indicators, use multiple algorithms to establish multiple coal quality indicator prediction models based on multiple groups of the preprocessed spectral coal quality data groups. Each of the coal quality indicator prediction models corresponds to a coal quality indicator, an algorithm and a group of the preprocessed spectral coal quality data groups. The coal quality indicator prediction model takes the spectral data as input to predict the predicted coal quality indicator data corresponding to the spectral data.

[0099] Specifically, the spectral data and coal quality index data in the spectral coal quality data group are used for modeling, and principal component analysis (PCA), partial least squares (PLS), machine learning algorithm, deep learning algorithm, etc. are used to establish multiple coal quality index prediction models.

[0100] Then, step S204 is executed to calculate the verification parameters of each coal quality index prediction model, compare the verification parameters of multiple coal quality index prediction models of the same coal quality index, and select the optimal coal quality index prediction model for each coal quality index.

[0101] Specifically, for each specific measurement index, i.e., coal quality index, the verification parameters of multiple coal quality index prediction models of this coal quality index can be compared to obtain an optimal model; for another measurement index, an optimal model can also be obtained, and different coal quality indexes independently obtain the corresponding optimal coal quality index prediction model. The optimal coal quality index prediction models for different coal quality indexes may be different.

[0102] In some embodiments, the verification parameters include reproducibility critical value corrected mean square error, prediction mean square error, model correlation coefficient and model variation coefficient, and the verification parameters of multiple coal quality index prediction models are compared to select the optimal coal quality index prediction model, specifically including:

[0103] The reproducibility critical values, corrected mean square error, predicted mean square error, model correlation coefficient and model variation coefficient of the multiple coal quality index prediction models are compared to select the optimal coal quality index prediction model.

[0104] 1) Root Mean Square Error of Calibration (RMSEC)

[0105] The corrected mean square error RMSEC is used for internal model testing, and its calculation method is shown in the formula:

[0106]

[0107] Where n is the number of samples in the calibration set; y i is the value determined by the national standard analysis method of sample i in the calibration set, i.e., the calibration value; It is the predicted value of sample i by the regression model built for the calibration set.

[0108] 2) Root Mean Square Error of Prediction (RMSEP) is used for external verification. Its calculation method is shown in the formula:

[0109]

[0110] Where m is the number of samples in the prediction set; is the value obtained by predicting sample i in the prediction set using the regression model established in the calibration set; is the calibration (national standard measurement) value of sample i in the prediction set.

[0111] In the actual modeling process, the smaller the RMSEP is, the higher the accuracy of the model is; the closer the RMSEC and RMSEP are, the higher the stability of the model is. In the selection process of actual regression models and optimization algorithms, it is necessary to tend to reduce RMSEC and RMSEP to improve the stability and accuracy of the model.

[0112] 3) Correlation coefficient (R)

[0113] The correlation coefficient is used to characterize the fitting effect of the prediction set samples, and its calculation method is:

[0114]

[0115] where y i is the calibration value (national standard analysis value) of sample set i; is the prediction value matrix; for y i and The covariance of Var(y i ) is y i The variance of for The correlation coefficient indicates that, under the premise of the same sample range, the closer it is to 1, the better the regression effect or prediction effect.

[0116] (4) Coefficient of Variation (CV)

[0117] The coefficient of variation reflects the absolute value of the discreteness of the data. By calculating the coefficient of variation, the influence of unit or mean differences when evaluating the model can be eliminated. The specific calculation formula is as follows:

[0118]

[0119] In the formula, RMSEP is the predicted mean square error of the prediction set, and MEAN is the mean of the prediction set. The correlation coefficient (R) and the coefficient of variation (CV) can be used together as an important reference information for evaluating the predicted mean square error (RMSEP) of the auxiliary prediction set to characterize the prediction ability of the established regression model for external samples.

[0120] (5) Reproducibility critical difference

[0121] The critical difference of reproducibility refers to the critical value of the difference between the average values ​​of the results obtained by repeated measurements of the same sample in different tests (under 95% probability). That is, when the method of the present invention is used to test the same sample, the difference between the results obtained and the results measured by the national standard method (under 95% probability) meets the requirements of the national standard for corresponding data.

[0122] Specifically, the model is verified internally, and the same spectral data is compared with the predicted coal quality index data of the coal quality index prediction model and the experimental coal quality index data from the laboratory, and multiple verification parameters are compared to select the optimal coal quality index prediction model. The verification parameters include but are not limited to: reproducibility critical value, correction mean square error, prediction mean square error, model correlation coefficient and model variation coefficient.

[0123] For example, with the national standard as the comparison standard, the reproducibility critical value of the coal quality index prediction model is required to meet the national standard. In addition, the combined evaluation criteria are the corrected mean square error, the predicted mean square error, the model correlation coefficient and the model variation coefficient, and finally determine the rapid prediction model of the same type and coal quality index as the coal sample to be tested, that is, obtain the rapid prediction model of the specific coal quality index of the coal sample to be tested.

[0124] Then, step S205 is executed to select the pre-processed spectral coal quality data set corresponding to the optimal coal quality index prediction model for each coal quality index as the optimal spectral coal quality data set for the coal quality index.

[0125] Finally, step S206 is executed to obtain the spectral data of the coal sample to be tested, and the optimal coal quality index prediction model for the coal quality index to be predicted is used to predict the predicted coal quality index data corresponding to the spectral data of the coal sample to be tested.

[0126] Specifically, for the same type of coal, the same interferometric near-infrared spectrometer used in the experiment is used to irradiate the sample to obtain the spectral results. After the same preprocessing method, the optimal coal quality index prediction model is input for rapid prediction, and the predicted coal quality index data corresponding to the spectral data of the coal sample to be tested output by the optimal coal quality index prediction model is obtained, and the coal quality results are displayed.

[0127] In one embodiment, the step of acquiring the spectral data of the coal sample to be tested and using the optimal coal quality index prediction model of the coal quality index to be predicted to predict the predicted coal quality index data corresponding to the spectral data of the coal sample to be tested specifically includes:

[0128] Acquiring spectrum data of the coal sample to be tested;

[0129] The spectrum preprocessing method for generating the optimal spectrum coal quality data group is used to preprocess the spectrum data of the coal sample to be tested, so as to obtain preprocessed spectrum data of the coal sample to be tested;

[0130] The optimal coal quality index prediction model is used to predict the predicted coal quality index data corresponding to the preprocessed spectral data of the coal sample to be tested.

[0131] Specifically, after the spectral coal quality data group of the training coal samples is processed by the spectral preprocessing method and the optimal spectral coal quality data group is selected, the spectral data of the coal sample to be tested is preprocessed by the same spectral preprocessing method as that used to obtain the optimal spectral coal quality data group to obtain the preprocessed spectral data of the coal sample to be tested, and the optimal coal quality index prediction model is used for prediction.

[0132] This embodiment adopts the same preprocessing method for the training coal samples and the coal samples to be tested, thereby ensuring the accuracy of the predicted coal quality index data.

[0133] In one embodiment, the step of acquiring the spectral data of the coal sample to be tested and using the optimal coal quality index prediction model of the coal quality index to be predicted to predict the predicted coal quality index data corresponding to the spectral data of the coal sample to be tested specifically includes:

[0134] Acquiring spectrum data of the coal sample to be tested;

[0135] Preprocessing the spectral data of the coal sample to be tested by using the spectral preprocessing method for generating the optimal spectral coal quality data group, and / or performing spectral selection on the spectral data of the coal sample to be tested by using the characteristic spectrum selection method for generating the optimal spectral coal quality data group, so as to obtain preprocessed spectral data of the coal sample to be tested;

[0136] The optimal coal quality index prediction model is used to predict the predicted coal quality index data corresponding to the preprocessed spectral data of the coal sample to be tested.

[0137] Specifically, after the spectral coal quality data group of the training coal sample is pre-processed and / or spectrally selected by the characteristic spectrum selection method, and the optimal spectral coal quality data group is selected, the spectral data of the coal sample to be tested is subjected to the same pre-processing as that used to obtain the optimal spectral coal quality data group and / or spectrally selected by the characteristic spectrum selection method to obtain the pre-processed spectral data of the coal sample to be tested, and the optimal coal quality index prediction model is used for prediction.

[0138] In some embodiments, the spectral preprocessing method for generating the optimal spectral coal quality data group is used to preprocess the spectral data of the coal sample to be tested, and / or the characteristic spectrum selection method for generating the optimal spectral coal quality data group is used to perform spectrum selection on the spectral data of the coal sample to be tested, so as to obtain the preprocessed spectral data of the coal sample to be tested, including:

[0139] Pre-processing the spectral data of the coal sample to be tested by using the spectral pre-processing method for generating the optimal spectral coal quality data group to obtain pre-processed spectral data of the coal sample to be tested; or

[0140] Using the characteristic spectrum selection method for generating the optimal spectrum coal quality data group, spectrum selection is performed on the spectrum data of the coal sample to be tested to obtain pre-processed spectrum data of the coal sample to be tested; or

[0141] The spectral preprocessing method for generating the optimal spectral coal quality data group is used to pre-process the spectral data of the coal sample to be tested to obtain pre-processed spectral data of the coal sample to be tested. The characteristic spectrum selection method for generating the optimal spectral coal quality data group is used to perform spectral selection on the pre-processed spectral data of the coal sample to be tested to obtain pre-processed spectral data of the coal sample to be tested.

[0142] This embodiment adopts the same spectrum selection method for the training coal samples and the coal samples to be tested, thereby ensuring the accuracy of the predicted coal quality index data.

[0143] This embodiment uses near-infrared spectroscopy to quickly detect and model coal quality, without over-grinding samples. The method is simple, fast, without radioactive sources, and the process is safe. An interferometric near-infrared spectrometer is used, which has high resolution, signal-to-noise ratio, sensitivity and measurement accuracy, fast scanning speed, and wide measurement range. This embodiment uses a training coal sample of the same type as the coal sample to be tested, and uses the same coal quality index to obtain coal quality index data, so as to train the optimal coal quality index prediction model for the coal sample to be tested and the specific coal quality index, ensure the accuracy of the prediction, and realize the practical application value. This embodiment independently selects a single or multiple pre-processing methods to pre-process the spectrum and extract characteristic peaks for different coal quality indicators, combines mechanism knowledge to determine the range of coal quality indicators, removes abnormal samples, and ensures the accuracy of the coal quality prediction model modeling samples. This embodiment has a variety of prediction model algorithms to choose from, including PCA, PLS, machine learning algorithms, deep learning algorithms, etc., which can analyze the influence of coal quality indicators on spectral superposition, and the model prediction accuracy is high. This embodiment independently models characteristic coal types or coal qualities to solve common problems such as low prediction accuracy and misjudgment of broad models. Finally, the highly recognized method for judging model accuracy meets national standards and facilitates industry personnel to judge and correct the model.

[0144] like Figure 4 The following is a working flow chart of a coal quality detection method according to a preferred embodiment of the present invention, comprising:

[0145] Step S401, prepare an interferometric near-infrared spectrometer and calibrate the equipment. The measured wavenumber range is from 4000-12800cm-1, which can cover the entire near-infrared spectrum;

[0146] Step S402: According to the provisions of "GB 474-2008 Coal Sample Preparation Method", the coal sample is processed by repeated reduction, crushing, mixing, air drying and screening operations to gradually reduce its quantity and particle size to prepare a test sample;

[0147] Step S403: Perform laboratory testing on the sample in step S402 to obtain national standard test values ​​of coal sample coal quality information as experimental coal quality index data for building a coal quality index database;

[0148] Step S404: Use the instrument in step S401 to perform a spectrum sampling experiment on the coal quality in step S402 to obtain spectrum data corresponding to the experimental coal quality in step S403, and construct a near infrared spectrum-coal quality index database;

[0149] Step S405: Pre-process the spectral data in the spectrum-coal quality index database; through the relevant spectral data pre-processing technology, eliminate the random noise in the spectrum, enhance the "interesting" characteristic information in the spectrum, and weaken the scattering interference to obtain spectral data with high signal-to-noise ratio, thereby ensuring that the subsequent modeling analysis is accurate and effective. In the near-infrared spectral analysis model, the commonly used spectral pre-processing methods can be divided into three categories: smoothing methods to eliminate random noise, multivariate scattering correction methods to reduce scattering interference, and data transformation methods and derivation methods to enhance the difference between spectra;

[0150] Step S406: Select characteristic spectra from the spectrum pre-processed in step S405 by using a correlation coefficient method, a non-information variable elimination method, a Monte Carlo non-information variable elimination method, a competitive adaptive reweighted sampling method, etc.;

[0151] Step S407: selecting training set samples for the spectral data obtained in step S406, and removing abnormal samples in combination with mechanism analysis;

[0152] Step S408: using the experimental coal quality index data in the spectrum-coal quality index database and the spectrum data obtained in step S407 to build models to obtain multiple coal quality index prediction models, the basic model selected is PCA (principal component analysis), PLS (partial least squares), machine learning algorithm, deep learning algorithm, etc., the coal quality index prediction model predicts the predicted coal quality index data corresponding to the spectrum data;

[0153] Step S409: Compare the predicted coal quality index data of each coal quality index prediction model with the experimental coal quality index data in step S403 to perform model internal verification, the comparison standard is the national standard detection standard, and the reproducibility critical value meets the national standard; in addition, the combined evaluation standards are the corrected mean square error, the predicted mean square error, the model correlation coefficient and the model variation coefficient; finally determine the optimal coal quality index prediction model for the specific coal quality index;

[0154] Step S410: for the same type of coal, use the instrument in step S401 to irradiate the sample to obtain the spectrum result, and after the same pre-processing method, use the optimal coal quality index prediction model obtained in step S409 to perform rapid prediction, and display the predicted coal quality index data as the coal quality result; this step process is automatically processed by the spectrometer processing software;

[0155] During use, sample laboratory tests are carried out regularly and randomly, and the index results are measured using the existing national standard laboratory test methods. The reproducibility critical value is compared with the predicted coal quality index data predicted by the optimal coal quality index prediction model. If it meets the national standard, it passes the test; if it does not meet the national standard, the model is re-modeled according to steps S401 to S409.

[0156] Compared with the prior art, this embodiment has the following advantages:

[0157] 1. This embodiment conducts rapid detection modeling for raw coal and uses a near-infrared spectrometer to irradiate the raw coal. On the one hand, the probe will not be inserted into the sample and will not be worn; on the other hand, the near-infrared spectrometer can better reflect the organic components in the coal quality. Figure 5 The figure shows a schematic diagram of a coal sample being irradiated by a near infrared spectrometer, wherein the irradiation light 51 is diffusely reflected.

[0158] 2. This embodiment fully explains a logical method for constructing a model for rapid coal quality detection using a near-infrared spectrometer, including data processing and modeling algorithms for coal quality indicators, as well as standards for selecting appropriate processing methods and modeling models, etc. It is separated from the predictive modeling method for a certain indicator or several types of coal; the model evaluation method is established in accordance with the national standard to make the model evaluation more accurate, reasonable, and easier to understand.

[0159] Using the method of this embodiment to test and model the rapid detection of coal quality can ensure that the user can smoothly and quickly complete the modeling and apply it, ensure the accuracy of the rapid prediction results, and realize the practical application value.

[0160] As an example, for a certain type of coal used in a gasifier, near-infrared spectroscopy technology was used to build and apply a rapid detection and prediction model for coal quality indicators. The fixed carbon index of coal quality was used as a specific coal quality indicator, and a total of 321 spectral-coal quality data were obtained.

[0161] Through trying the pre-processing methods individually and in combination, it was finally determined that spectral data normalization (ZN) combined with the first-order derivative (1D) and standard normal variable transformation (SNV) could achieve the best prediction performance.

[0162] Competitive adaptive reweighted sampling is used to screen spectral feature peaks, remove wavelength points with small weights, and use cross-validation to select the subset with the lowest model cross-validation mean square error value, which can effectively select the optimal wavelength combination related to the measured properties.

[0163] Outliers are removed according to the standard score (Z-Score). The standard score refers to the conversion score of the original characteristic value of a certain type of value. With the help of this standard value, the originally incomparable values ​​can be made comparable. The specific definition is as follows:

[0164]

[0165] Among them, Z is the standard score, x i is the sample value, x is the sample mean, and σ is the sample standard deviation.

[0166] The Z score is the deviation from the mean in units of standard deviation. When the Z score is greater than 3, that is, the deviation from the mean exceeds 3σ, the sample is considered to be an abnormal sample. After eliminating the abnormal samples, 291 spectral-coal quality data of valid samples were obtained. A rapid prediction model was established by hybrid PCA+PLS algorithm. The absolute error of the final model with the national standard algorithm in internal verification was 0.265%, which met the national standard reproducibility critical value of <0.35%, and the applicable dry ash-free basis range was 56.62% to 65.34%. For coal types within this range, the error between the results predicted by near-infrared technology and the national standard prediction results is less than 0.3%, meeting the national standard requirements. The time to obtain the test results does not exceed 5 minutes. In contrast, the test time for laboratory measurements using the national standard method is at least 6 hours.

[0167] like Figure 6 The figure shows a hardware structure diagram of an electronic device of the present invention, including:

[0168] at least one processor 601; and,

[0169] A memory 602 is communicatively connected to at least one of the processors 601; wherein,

[0170] The memory 602 stores instructions that can be executed by at least one of the processors. The instructions are executed by at least one of the processors so that the at least one of the processors can execute the coal quality detection method as described above.

[0171] Figure 6 A processor 601 is taken as an example.

[0172] The electronic device may further include: an input device 603 and a display device 604 .

[0173] The processor 601, the memory 602, the input device 603 and the display device 604 may be connected via a bus or other means, and the figure takes the connection via a bus as an example.

[0174] The memory 602 is a non-volatile computer-readable storage medium that can be used to store non-volatile software programs, non-volatile computer executable programs and modules, such as program instructions / modules corresponding to the coal quality detection method in the embodiment of the present application, for example, Figure 1 , Figure 2 The processor 601 executes various functional applications and data processing by running the non-volatile software programs, instructions and modules stored in the memory 602, that is, the coal quality detection method in the above embodiment is implemented.

[0175] The memory 602 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required by at least one function; the data storage area may store data created according to the use of the coal quality detection method, etc. In addition, the memory 602 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 602 may optionally include a memory remotely arranged relative to the processor 601, and these remote memories may be connected to the device for executing the coal quality detection method via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0176] The input device 603 can receive user clicks and generate signal inputs related to user settings and function control of the coal quality detection method. The display device 604 can include display devices such as a display screen.

[0177] The one or more modules are stored in the memory 602, and when executed by the one or more processors 601, the coal quality detection method in any of the above method embodiments is executed.

[0178] The present invention adopts multiple algorithms to model each coal quality index based on the same type of training coal samples for different coal samples to be tested, and selects the optimal coal quality index prediction model from them. Therefore, the present invention independently models specific coal quality indicators of specific coal sample types, solves technical problems such as low prediction accuracy and misjudgment of broad models, and improves the accuracy of prediction of specific coal samples.

[0179] An embodiment of the present invention provides a storage medium, wherein the storage medium stores computer instructions. When a computer executes the computer instructions, the storage medium is used to execute all steps of the coal quality detection method as described above.

[0180] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A coal quality detection method, characterized in that: include: Acquire a spectral coal quality data group of a training coal sample, wherein the spectral coal quality data group includes a plurality of spectral coal quality data pairs, each pair of the spectral coal quality data pairs includes spectral data of the training coal sample and corresponding experimental coal quality index data, and the training coal sample and the coal sample to be tested are the same type of coal samples; Based on the spectral coal quality data group, a plurality of coal quality index prediction models are respectively established for different coal quality indexes by using a plurality of algorithms, wherein the coal quality index prediction models take the spectral data as input and predict the predicted coal quality index data corresponding to the spectral data; Compare multiple coal quality index prediction models and select the optimal coal quality index prediction model for each coal quality index; The spectral data of the coal sample to be tested is obtained, and the optimal coal quality index prediction model of the coal quality index to be predicted is used to predict the predicted coal quality index data corresponding to the spectral data of the coal sample to be tested.

2. The coal quality detection method according to claim 1, characterized in that: The experimental coal quality index data of the training coal sample and the predicted coal quality index data required to be predicted for the tested coal sample are data of the same coal quality index.

3. The coal quality detection method according to claim 1, characterized in that: The step of comparing a plurality of coal quality index prediction models and selecting the optimal coal quality index prediction model for each coal quality index specifically includes: Calculate the verification parameters of each coal quality index prediction model, compare the verification parameters of multiple coal quality index prediction models for the same coal quality index, and select the optimal coal quality index prediction model for each coal quality index.

4. The coal quality detection method according to claim 1, characterized in that: Based on the spectral coal quality data set, a plurality of coal quality index prediction models are established respectively for different coal quality indexes by using a plurality of algorithms, specifically including: Preprocessing the spectral coal quality data set using a plurality of preprocessing methods to obtain a plurality of preprocessed spectral coal quality data sets; For different coal quality indicators, multiple algorithms are used to establish multiple coal quality indicator prediction models based on multiple groups of preprocessed spectral coal quality data groups. Each coal quality indicator prediction model corresponds to a coal quality indicator, an algorithm and a group of preprocessed spectral coal quality data groups.

5. The coal quality detection method according to claim 4, characterized in that: After comparing the multiple coal quality index prediction models and selecting the optimal coal quality index prediction model for each coal quality index, the method further includes: The preprocessed spectral coal quality data set corresponding to the optimal coal quality index prediction model for each coal quality index is selected as the optimal spectral coal quality data set for the coal quality index.

6. The coal quality detection method according to claim 5, characterized in that: The step of acquiring the spectral data of the coal sample to be tested and using the optimal coal quality index prediction model of the coal quality index to be predicted to predict the predicted coal quality index data corresponding to the spectral data of the coal sample to be tested specifically includes: Acquiring spectrum data of the coal sample to be tested; The spectrum preprocessing method for generating the optimal spectrum coal quality data group is used to preprocess the spectrum data of the coal sample to be tested, so as to obtain preprocessed spectrum data of the coal sample to be tested; The optimal coal quality index prediction model is used to predict the predicted coal quality index data corresponding to the preprocessed spectral data of the coal sample to be tested.

7. The coal quality detection method according to claim 4, characterized in that: The preprocessing of the spectral coal quality data set by using a plurality of preprocessing methods to obtain a plurality of preprocessed spectral coal quality data sets specifically includes: Determine a plurality of spectral pre-processing methods and / or a plurality of characteristic spectrum selection methods; Each of the spectral preprocessing methods is used to preprocess the spectral coal quality data group, and / or each of the characteristic spectrum selection methods is used to perform spectral selection on the spectral data of each pair of the spectral coal quality data pairs in the spectral coal quality data group, so as to obtain multiple groups of preprocessed spectral coal quality data groups.

8. The coal quality detection method according to claim 7, characterized in that: The step of acquiring the spectral data of the coal sample to be tested and using the optimal coal quality index prediction model of the coal quality index to be predicted to predict the predicted coal quality index data corresponding to the spectral data of the coal sample to be tested specifically includes: Acquiring spectrum data of the coal sample to be tested; Preprocessing the spectral data of the coal sample to be tested by using the spectral preprocessing method for generating the optimal spectral coal quality data group, and / or performing spectral selection on the spectral data of the coal sample to be tested by using the characteristic spectrum selection method for generating the optimal spectral coal quality data group, so as to obtain preprocessed spectral data of the coal sample to be tested; The optimal coal quality index prediction model is used to predict the predicted coal quality index data corresponding to the preprocessed spectral data of the coal sample to be tested.

9. An electronic device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by at least one of the processors, and the instructions are executed by at least one of the processors so that at least one of the processors can execute the coal quality detection method according to any one of claims 1 to 8.

10. A storage medium, characterized in that: The storage medium stores computer instructions, and when a computer executes the computer instructions, it is used to execute all steps of the coal quality detection method according to any one of claims 1 to 8.

Images