A coking coal resource comprehensive evaluation method based on unsupervised machine learning

Abstract

The present application relates to the field of coal resource management, and discloses a coking coal resource comprehensive evaluation method based on unsupervised machine learning, comprising the following steps: step 1, establishing a coal quality database based on a GIS platform; in the GIS, using a grid calculator to remove coal types that do not belong to coking coal blending; step 2, selecting ash content, sulfur content, phosphorus content, volatile matter content and caking index as key indicators of coking coal, and extracting them through the grid calculator tool in the GIS; step 3, calling coalfield drilling data in the coalfield drilling coal quality database, and displaying the coordinate points of each drilling on the GIS map in sequence to eliminate the influence of dimension on the results. The present application uses a hierarchical clustering algorithm based on python to perform multi-angle and multi-dimensional classification processing on the coal quality indicators of coking coal, and combines the ISO algorithm to perform similar clustering on the coal quality indicator data, which greatly expands the use mode of coal quality data.

Description

Technical Field

[0001] This invention relates to the field of coal resource management, and in particular to a comprehensive evaluation method for coking coal resources based on unsupervised machine learning. Background Technology

[0002] Coking coal refers to bituminous coal with a high degree of metamorphism. It is now widely used in metallurgy, coal chemical industry, power industry and other industries, and has an important impact on the long-term development of the country's entire industrial system and people's lives. In the past three years, the large-scale development of blast furnaces and coke ovens has put forward higher requirements for coke quality. However, the resources of low-sulfur high-quality coking coal that can be mined and utilized year-round are already very limited. Coking enterprises must make the most of the limited coking coal resources to produce high-quality coke to meet the needs of blast furnaces.

[0003] Coal quality evaluation is an objective and accurate assessment of coal quality based on coal quality analysis results. Through coal quality evaluation, the composition and properties of coal can be understood, providing a scientific basis for the rational development and utilization of coal resources. Currently, the coal quality evaluation systems for liquefaction coal and gasification coal are relatively complete; however, the research level on the quality grading of coking coal is relatively low.

[0004] As machine learning continues to advance, its applications are becoming increasingly diverse across various industries. However, its use in coal resource management remains in its early stages. Different big data analytics methods can optimize coal resource analysis models, offering better accuracy compared to traditional methods. This enables big data analysis for the green mining of coal resources.

[0005] This paper focuses on the integration of machine learning and coal resource management, giving full play to the enormous advantages of machine learning. In a specific research area, based on massive local coal quality data, a coal quality grading standard for coking coal is established, and the advantageous range is delineated. This has great significance for the clean and refined exploration and development of coking coal resources in the future.

[0006] Chinese invention patent CN 111423894 A discloses a method and apparatus for delineating coal resources for coking. It delineates a first research area based on selected borehole flotation indicators, then selects boreholes with phosphorus content meeting preset conditions to delineate another area. Finally, the overlapping area of ​​the first and second delineated areas is taken as the area of ​​coal resources for coking in the exploration area.

[0007] Chinese invention patent CN 115247075 A discloses a method for classifying coking coal. It classifies coking coal based on its vitrinite random reflectance and mineral catalytic index.

[0008] A method for classifying the quality of coking coal—a method for blending coking coal—is disclosed in Chinese invention patent announcement number CN 109294608 A (Document 3). The method determines the coking properties of coking coal by measuring the dry ash-free volatile matter, the Oya expansion degree b, the average maximum reflectance of the vitrinite group, the maximum flowability of the Gibbs free, and the optical microstructure of the coke.

[0009] Regarding the aforementioned technologies, the inventors believe that the following defects exist:

[0010] Based on the aforementioned relevant literature, the inventors, after comprehensive consideration, believe that:

[0011] 1. The research methodology has problems and fails to reflect the current advantages of machine learning algorithms.

[0012] 2. The problem of cumbersome steps and time-consuming operation (traditional evaluation methods cannot quickly cluster coking coal and involve subjective judgment. Existing methods are either too simple or too cumbersome to operate, and most of them are human-computer interaction operations. GIS spatial data processing can greatly reduce manual operation steps and improve evaluation efficiency and accuracy.)

[0013] 3. It mainly relies on national industry standards and lacks a quality grade evaluation method based on big data of coal quality, which makes it impossible to deeply explore the correlation between coking coal quality data. Summary of the Invention

[0014] To address the challenge of traditional methods failing to rapidly identify, classify, and evaluate favorable areas of high-quality coking coal resources from large-scale borehole coal quality data at the coalfield scale, this invention, based on unsupervised machine learning principles and combined with the ISO clustering algorithm, utilizes massive amounts of borehole coal quality data within a coalfield to quickly uncover hidden homogeneous information within the big data, achieving a comprehensive evaluation of coking coal quality based on multiple attributes and indicators. This invention provides a method for classifying and comprehensively evaluating the quality of coking coal resources based on unsupervised machine learning.

[0015] This invention is achieved using the following technical solution: a comprehensive evaluation method for coking coal resources based on unsupervised machine learning, comprising the following steps:

[0016] Step 1: Establish a coal type and quality database based on a GIS platform; use a raster calculator in the GIS to remove coal types that do not belong to coking coal blending; in Step 1, the data to be removed includes coal quality data that does not belong to coking coal blending and coal quality data that are obviously entered incorrectly in the original data.

[0017] Step 2: Select ash content (Ad), sulfur content (Std), phosphorus content (Pd), volatile matter (Vdaf), and caking index (G) as key indicators for coking coal, and extract them using the GIS raster calculator tool;

[0018] Step 3: Call the coalfield boreholes in the coalfield borehole coal quality database, display the coordinates of each borehole on the GIS map in sequence, eliminate the influence of dimensions on the results, and draw thematic maps for ash (Ad), sulfur (Std), phosphorus (Pd), volatile matter (Vdaf), and bonding index (G).

[0019] In step three, the dimensions are eliminated using the following formulas: Formula 1 is used for inverse indices, and Formula 2 is used for positive indices. Ash (Ad), sulfur (Std), phosphorus (Pd), and volatile matter (Vdaf) are inverse indices reflecting the quality of coking coal, while the caking index (G) is a positive indices reflecting the quality of coking coal.

[0020] (1)

[0021] (2)

[0022] Step 4: Perform cluster analysis on borehole coal quality data within the coalfield area using the unsupervised learning ISO algorithm;

[0023] Step 4.1: Select a hierarchical clustering algorithm to perform clustering operations on the ash content (Ad), sulfur content (Std), phosphorus content (Pd), volatile matter (Vdaf), and adhesion index (G) of the selected boreholes from a purely data perspective;

[0024] Step 4.2: Establish a hierarchical clustering algorithm model based on Python, and iteratively calculate the selected borehole coal quality data using the contour coefficient method; where the contour coefficient is an indicator that describes the clarity of the contour of each category after clustering, and the larger the contour coefficient, the better the clustering effect. The contour coefficient can be iterated continuously using formula (3), and the number of iterations can be set by yourself. The maximum number of clusters corresponding to the maximum contour coefficient is the maximum number of clusters of the coal seam data in this borehole.

[0025] S(i)=(b(i)−a(i)) / max{a(i),b(i)} (3)

[0026] Step 4.3: Based on the results calculated in Step 4.2, a comprehensive cluster analysis is performed on all borehole coal quality data in the coalfield by combining the hierarchical clustering learning algorithm with the ISO algorithm in GIS software.

[0027] Step 4.4: To more accurately reflect the differences between clustering results, the selected data are projected according to coordinate positions. Based on the results of hierarchical clustering and the ISO algorithm, borehole points on each result are extracted in GIS using mask extraction and spatial connection tools. Each clustering result is then evaluated and analyzed separately.

[0028] Step 5: Calculate the weights of each index (Ad, Std, Pd, Vdaf, G) on the technical conditions for using coking coal using the entropy weight method, laying a reference foundation for the operation of subsequent methods;

[0029] Step 5 includes the following sub-steps:

[0030] Step 5.1, the operation steps are as follows: First, a judgment matrix with m samples and n evaluation indicators needs to be constructed: X = (xij)m×n (i=1,2,…..m;j=1,2,…,n);

[0031] Step 5.2: Since the units of ash (Ad), sulfur (Std), phosphorus (Pd), volatile matter (Vdaf), and adhesion index (G) are different, the matrix needs to be normalized and dimensionless to obtain the normalized matrix B = (bij)m×n (i=1,2,…..m;j=1,2,…,n) and the new matrix B´ = (bij´)m×n (i=1,2,…..m;j=1,2,…,n) with the same trend.

[0032] In the formula, bij is the j-th index value of the i-th evaluation object;

[0033] According to the definition of information entropy, the formula for calculating entropy weight is:

[0034] (i=1,2,…..m;j=1,2,…,n) (4)

[0035] In the formula: Hj is the corresponding transitional variable value of evaluation index j in the entropy weight calculation process; fij is the transitional variable in the entropy weight calculation process, and

[0036] The information entropy calculation vector is obtained from Formula 3:

[0037] (j=1,2…..,n) (5)

[0038] In the formula, W is the entropy weight variable obtained in the final calculation; wj is the entropy weight value corresponding to evaluation index j, and

[0039] (6)

[0040] Step 5.3 Finally, based on the entropy weight method calculation principle and using the raster calculator tool in GIS, calculate the weights of ash (Ad), sulfur (Std), phosphorus (Pd), volatile matter (Vdaf), and adhesion index (G) according to the above formulas.

[0041] Step 6: Classify the quality of coking coal into grades to obtain coking coal resources of different grades and different coal quality characteristics under each cluster.

[0042] Step 6 includes the following sub-steps:

[0043] Step 6.1: From the perspective of economic and environmental indicators of coking coal, the coal quality of coking coal is classified into individual indicators. Based on the quality standards of coking coal and the actual situation of the study area, coking coal can be divided into three grades: high-quality coal, medium-quality coal, and low-quality coal. Among them, sulfur (Std) and phosphorus (Pd) are environmental indicators, while ash (Ad), volatile matter (Vdaf), and caking index (G) are economic indicators.

[0044] Step 6.2: Referring to the comprehensive index evaluation method, assign values ​​to each indicator and calculate the comprehensive evaluation value of the borehole according to the weights in Table 1 corresponding to the relevant steps; the comprehensive index evaluation method classifies each quality indicator into grades and assigns values ​​to them, where: high-quality coal is assigned a value of "3.5"; medium-quality coal is assigned a value of "2.5"; and low-quality coal is assigned a value of "1.5". Then, calculate the comprehensive quality evaluation value of each coal sample according to the weights of each indicator and the following formula (whether the indicator boundaries need to be explained).

[0045] E j = ×X ij , (i = 1, 2, 3,..., n; j = 1, 2, 3,..., m)

[0046] Where Xij corresponds to the assigned value of the quality index, and Xij = 1, 2, 3, (i represents the quality number of each index, and j represents the coal sample number); Kj represents the assigned value of each quality index, Kj = 0.1 or 0.2.

[0047] Step 6.3: Calculate the comprehensive evaluation value according to the above formula, and classify the borehole coking coal of different clusters into high-quality coal, medium-quality coal, and low-quality coal. According to the grid calculator function in GIS, when 3≤Ej≤3.5, the coal type belongs to high-quality coking coal; when 2.5≤Ej<3, the coal type belongs to medium-quality coking coal; when Ej<2.5, the coal type belongs to low-quality coking coal.

[0048] Step 6.4: Project the borehole results processed in Step 6.3 into the GIS. Use the inverse distance weighting method and the natural discontinuity grading method to divide the coal quality data of the two clusters. Based on different color areas, the coking coal resources of different grades and different coal quality characteristics under each cluster can be clearly distinguished.

[0049] Step 6.5: Compare the differences in the results of individual indicators of different clustering boreholes, and conduct a comparative analysis of the two types of data boreholes in accordance with national standards and specifications.

[0050] Step 7: Based on cluster analysis of a large amount of coal quality data, classification of coking coal grades, definition of coal quality index grades, and comprehensive consideration, delineate the development areas of high-quality coking coal and provide reasonable development and utilization plans.

[0051] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0052] (1) This invention uses Python to perform multi-angle and multi-dimensional classification of coal quality indicators of coking coal using hierarchical clustering algorithm, and combines ISO algorithm to perform similar clustering of coal quality indicator data, which greatly expands the ways of using coal quality data.

[0053] (2) This invention analyzes the weights of coal quality indicators for coking coal by comprehensively applying the entropy weight method and the comprehensive index evaluation method, highlighting the degree of influence of each indicator on the quality of coking coal, and making the evaluation results more objective.

[0054] (3) This paper considers the quality grade of coking coal from both environmental protection and economic perspectives, delineates the coking coal occurrence range in the mining area from the source, and controls the cost of using coking coal and environmental pollution factors, which can provide technical guidance for the later production and development of coking coal in the mining area. Attached Figure Description

[0055] Figure 1 This is a coal quality database map of the Huaibei coalfield proposed in this invention;

[0056] Figure 2 This is a thematic map showing the borehole distribution and various indicators in an embodiment of the present invention;

[0057] Figure 3 The contour coefficients in this embodiment of the invention are iterated using formula (3) to obtain the following results:

[0058] Figure 4 This is a comprehensive cluster analysis diagram of coal quality data from all boreholes in the coalfield, as presented in this embodiment of the invention.

[0059] Figure 5 In this embodiment of the invention, the inverse distance weighting method and the natural discontinuity grading method are used to divide the two types of coal quality data to obtain a coal quality grade classification map;

[0060] Figure 6 This is a comparison chart of the first type of data and the second type of data regarding coal quality environmental indicators after cluster analysis in an embodiment of the present invention;

[0061] Figure 7 As described in the embodiments of the present invention Figure 7 This is a comparison chart of the first and second categories of data regarding coal quality economic indicators after clustering.

[0062] Figure 8 This is a flowchart illustrating the process of the comprehensive evaluation method for coking coal resources proposed in this embodiment of the invention. Detailed Implementation

[0063] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments. It should be noted that, without conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0064] Example

[0065] This proposal presents a comprehensive evaluation method for coking coal resources based on unsupervised machine learning, comprising the following steps:

[0066] Step 1: Establish a coal type and quality database based on a GIS platform; use a raster calculator in the GIS to remove coal types that do not belong to coking coal blending; in Step 1, the data to be removed includes coal quality data that does not belong to coking coal blending and coal quality data that are obviously entered incorrectly in the original data.

[0067] First, coal quality data from all boreholes within the Huaibei coalfield were collected and organized to obtain borehole coal quality index data. Borehole tables, coal quality tables, lithology tables, beneficial element tables, and harmful element tables were then established for each mine. Finally, all data was imported into a GIS platform to establish a coal type and quality database, which forms the foundation for subsequent operational steps. Figure 1 .

[0068] Ash content is a harmful impurity in coke. Higher ash content results in lower coke stickiness, lower fixed carbon content, lower calorific value, and reduced coke strength. Therefore, ash content is one of the key indicators for coking coal. Volatile matter in coal determines the porosity of coke. Higher volatile matter content reduces coke strength, and the amount of volatile matter determines the degree of coal metamorphism; lower volatile matter indicates a higher degree of coal metamorphism. Therefore, volatile matter is the second key indicator for coking coal. Sulfur is a harmful element in coal. Excessive sulfur content leads to excessive sulfur content in pig iron, reducing its quality. Therefore, sulfur content in coke should be as low as possible and stable. Therefore, sulfur is the third key indicator for evaluating the quality of coking coal. Phosphorus in coal will all enter the coke during coking, making the smelted steel brittle. Therefore, phosphorus is the fourth key indicator for coking coal. The caking index is a key indicator for coking coal. A higher caking index indicates better caking properties, higher coking ability, and higher coke quality. Therefore, the caking index is the fifth key indicator for coking coal. Besides the above parameters, other coal quality indicators such as moisture and calorific value also have a certain impact on coking coal and should be considered as important indicators based on specific circumstances.

[0069] Step 2: Select ash content (Ad), sulfur content (Std), phosphorus content (Pd), volatile matter (Vdaf), and caking index (G) as key indicators for coking coal, and extract them using the GIS raster calculator tool;

[0070] Step 3: Access the coalfield boreholes from the coalfield borehole quality database, and sequentially display the coordinates of each borehole on the GIS map to eliminate the influence of dimensions on the results. Furthermore, create thematic maps for ash (Ad), sulfur (Std), phosphorus (Pd), volatile matter (Vdaf), and bonding index (G). The borehole distribution and thematic maps for each indicator are shown below. Figure 2 .

[0071] In step three, the dimensions are eliminated using the following formulas: Formula 1 is used for inverse indices, and Formula 2 is used for positive indices. Ash (Ad), sulfur (Std), phosphorus (Pd), and volatile matter (Vdaf) are inverse indices reflecting the quality of coking coal, while the caking index (G) is a positive indices reflecting the quality of coking coal.

[0072] (1)

[0073] (2)

[0074] Step 4: Perform cluster analysis on borehole coal quality data within the coalfield using the unsupervised learning ISO algorithm. This clustering method is actually an iterative process. It calculates the minimum Euclidean distance when assigning each candidate pixel to a cluster. Then, during each iteration, all samples are assigned to the existing cluster centers and a new average value is recalculated for each class. Finally, the optimal number of classes needs to be manually specified. This method can perform rapid cluster analysis on various coal quality indicators and has significant advantages in processing large data samples.

[0075] Step 4.1: Select a hierarchical clustering algorithm to perform clustering operations on the ash (Ad), sulfur (Std), phosphorus (Pd), volatile matter (Vdaf), and adhesion index (G) of the selected boreholes from a purely data perspective. This can reduce the influence of human subjective factors on the clustering results, and its results are more scientific and objective, and are not affected by traditional human subjective factors.

[0076] Step 4.2: Establish a hierarchical clustering algorithm model based on Python, and iteratively calculate the selected borehole coal quality data using the silhouette coefficient method; where the silhouette coefficient is an indicator describing the clarity of the silhouette of each category after clustering, and the larger the silhouette coefficient, the better the clustering effect. The silhouette coefficient can be iterated continuously using formula (3), and the number of iterations can be set by yourself. The iteration results are as follows. Figure 3 The number of clusters corresponding to the maximum value of the contour coefficient is the maximum number of clusters in the coal seam data of this borehole;

[0077] S(i)=(b(i)−a(i)) / max{a(i),b(i)} (3)

[0078] Step 4.3: Based on the results calculated in Step 4.2, a comprehensive cluster analysis is performed on all borehole coal quality data in the coalfield, combining the hierarchical clustering learning algorithm with the ISO algorithm in GIS software; for example... Figure 4

[0079] Step 4.4: To more accurately reflect the differences between different clustering results, the selected data are projected according to coordinate positions. Based on the results of hierarchical clustering and ISO algorithm, borehole points on each result are extracted in GIS using mask extraction and spatial connection tools. Each clustering result is evaluated and analyzed separately.

[0080] Step 5: In this scheme, in order to fully understand the characteristics and variation patterns of coal quality of coking coal in the clustering results, and adhering to the principles of clear primary and secondary indicators, clear evaluation scale, independence and availability among indicators, the entropy weight method is used to calculate the weights of each indicator (Ad, Std, Pd, Vdaf, and caking index) on the technical conditions of coking coal use, so as to lay a reference foundation for the operation of subsequent methods.

[0081] The entropy weighting method determines the weights of indicators based on the degree of variation in their values. As an objective weighting method, it avoids biases caused by human factors to a certain extent, making it suitable for calculating the weight coefficients of indicators in large data samples. When used in conjunction with the ISO clustering method, it can better uncover the hidden correlations between data indicators.

[0082] Step 5.1, the operation steps are as follows: First, a judgment matrix with m samples and n evaluation indicators needs to be constructed: X = (xij)m×n (i=1,2,…..m;j=1,2,…,n);

[0083] Step 5.2: Since the units of ash (Ad), sulfur (Std), phosphorus (Pd), volatile matter (Vdaf), and adhesion index (G) are different, the matrix needs to be normalized and dimensionless to obtain the normalized matrix B = (bij)m×n (i=1,2,…..m;j=1,2,…,n) and the new matrix B´ = (bij´)m×n (i=1,2,…..m;j=1,2,…,n) with the same trend.

[0084] In the formula, bij is the j-th index value of the i-th evaluation object;

[0085] According to the definition of information entropy, the formula for calculating entropy weight is:

[0086] (i=1,2,…..m;j=1,2,…,n) (4)

[0087] In the formula: Hj is the corresponding transitional variable value of evaluation index j in the entropy weight calculation process; fij is the transitional variable in the entropy weight calculation process, and

[0088] The information entropy calculation vector is obtained from Formula 3:

[0089] (j=1,2…..,n) (5)

[0090] In the formula, W is the entropy weight variable obtained in the final calculation; wj is the entropy weight value corresponding to evaluation index j, and

[0091] (6)

[0092] Step 5.3: Finally, based on the entropy weight method calculation principle and using the raster calculator tool in GIS, calculate the weights of ash (Ad), sulfur (Std), phosphorus (Pd), volatile matter (Vdaf), and adhesion index (G) according to the formulas above. See the table below.

[0093]

[0094] Step 6: Classify the quality of coking coal into grades to obtain coking coal resources of different grades and different coal quality characteristics under each cluster.

[0095] Step 6 includes the following sub-steps:

[0096] Step 6.1: From the perspective of economic and environmental indicators of coking coal, the coal quality of coking coal is classified into individual indicators. Based on the quality standards of coking coal and the actual situation of the study area, coking coal can be divided into three grades: high-quality coal, medium-quality coal, and low-quality coal. Among them, sulfur (Std) and phosphorus (Pd) are environmental indicators, while ash (Ad), volatile matter (Vdaf), and caking index (G) are economic indicators.

[0097] The table above shows the quality grades of coking coal.

[0098] Step 6.2: Referring to the comprehensive index evaluation method, assign values ​​to each indicator and calculate the comprehensive evaluation value of the borehole according to the weights in Table 1 corresponding to the relevant steps. The comprehensive index evaluation method classifies each quality indicator into grades and assigns values ​​to them, where: high-quality coal is assigned a value of "3.5"; medium-quality coal is assigned a value of "2.5"; and low-quality coal is assigned a value of "1.5". Then, calculate the comprehensive quality evaluation value of each coal sample according to the weights of each indicator and the following formula. Values ​​(3.5, 2.5, 1.5) This value represents the process of quantifying coal quality grades based on comprehensive data results, while simplifying the calculation model to make the process more efficient. It is concise and clear, with no other special meaning.

[0099] E j = ×X ij , (i = 1, 2, 3,..., n; j = 1, 2, 3,..., m)

[0100] Where Xij corresponds to the assigned value of the quality index, and Xij = 1, 2, 3, (i represents the quality number of each index, and j represents the coal sample number); Kj represents the assigned value of each quality index, Kj = 0.1 or 0.2.

[0101] Step 6.3: Calculate the comprehensive evaluation value according to the above formula, and classify the borehole coking coal of different clusters into high-quality coal, medium-quality coal, and low-quality coal. According to the grid calculator function in GIS, when 3≤Ej≤3.5, the coal type belongs to high-quality coking coal; when 2.5≤Ej<3, the coal type belongs to medium-quality coking coal; when Ej<2.5, the coal type belongs to low-quality coking coal.

[0102] Step 6.4: Project the borehole results processed in Step 6.3 into the GIS. Use the inverse distance weighting method and the natural discontinuity grading method to divide the coal quality data of the two clusters. Based on different color regions, coking coal resources of different grades and coal quality characteristics under each cluster can be clearly distinguished. Figure 5 It is quite evident from the data that there are significant differences between the two clustering results. In the first cluster, medium-quality coking coal is dominant, followed by high-quality coking coal, with low-quality coking coal scattered throughout. In the second cluster, the quantity of high-quality coking coal has increased significantly, while low-quality coking coal still exists sporadically. Therefore, according to the clustering results, the quality of coking coal in the second cluster is significantly better than that in the first cluster. The specific distribution area proportions are shown in the table below.

[0103] Statistical table of area of ​​coking coal quality grades by cluster

[0104]

[0105] Step 6.5: Compare the differences in individual index results from different clustering boreholes. Refer to national standards and specifications to conduct a comparative analysis of the two types of data boreholes. Specifically, coal ash content is classified according to GN / T15224.1-2018 "Coal Quality Classification"; coal volatile matter is classified according to MT / T849-2000 "Volatile Matter Yield Classification of Coal"; coal caking index is classified according to MT / T596-2008 "Bitcoin Coal Caking Index Classification"; coal sulfur content is classified according to GB / T15224.2-2021 "Coal Quality Classification"; and coal phosphorus content is classified according to GB / T20475.1-2006 "Classification of Harmful Element Content in Coal".

[0106] The clustering results were compared according to the above standards and specifications, such as... Figure 6 , 7 Among environmental indicators, the sulfur content varies significantly in the Linhuan mining area. The first cluster is dominated by low-sulfur to medium-sulfur coal, while the second cluster is dominated by ultra-low-sulfur to low-sulfur coal. The average phosphorus content of the second cluster is slightly higher than that of the first cluster. In terms of economic indicators, the first cluster is dominated by medium-ash to medium-high-ash coal, while the second cluster is dominated by low-ash to medium-ash coal. The first cluster is dominated by strongly caking coal, while the second cluster is dominated by medium-caking to strongly caking coal. The average volatile matter content of the second cluster is slightly lower than that of the first cluster. Based on the comprehensive analysis of the above coal quality results, the second type of coking coal is of superior quality to the first type.

[0107] Step 7: Based on cluster analysis of a large amount of coal quality data, classification of coking coal grades, definition of coal quality index grades, and comprehensive consideration, delineate the development areas of high-quality coking coal and provide reasonable development and utilization plans.

[0108] The above embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of protection of the present invention. Any non-substantial changes and substitutions made by those skilled in the art based on the present invention shall fall within the scope of protection claimed by the present invention.

Claims

1. A comprehensive evaluation method for coking coal resources based on unsupervised machine learning, characterized in that, Includes the following steps: Step 1: Establish a coal type and quality database based on a GIS platform; use a raster calculator in the GIS to remove coal types that do not belong to coking coal blending. Step 2: Select ash content, sulfur content, phosphorus content, volatile matter, and caking index as key indicators for coking coal, and extract them using the GIS raster calculator tool; Step 3: Call the coalfield boreholes in the coalfield borehole coal quality database, display the coordinates of each borehole on the GIS map in sequence, eliminate the influence of dimensions on the results, and draw thematic maps of ash, sulfur, phosphorus, volatile matter and bonding index. Step 4: Perform cluster analysis on borehole coal quality data within the coalfield area using the unsupervised learning ISO algorithm; Step 5: Calculate the weights of each index (ash content, sulfur content, phosphorus content, volatile matter, and caking index) on the technical conditions for using coking coal using the entropy weight method, laying a reference foundation for the operation of subsequent methods. Step 6: Classify the quality of coking coal into grades to obtain coking coal resources of different grades and different coal quality characteristics under each cluster; Step 4 includes the following sub-steps: Step 4.1: Select a hierarchical clustering algorithm to perform clustering operations on the ash content, sulfur content, phosphorus content, volatile matter, and bonding index of the selected boreholes from a purely data perspective; Step 4.2: Establish a hierarchical clustering algorithm model based on Python, and iteratively calculate the selected borehole coal quality data using the contour coefficient method; where the contour coefficient is an indicator that describes the clarity of the contour of each category after clustering, and the larger the contour coefficient, the better the clustering effect. The contour coefficient can be iterated continuously using formula (3), and the number of iterations can be set by yourself. The maximum number of clusters corresponding to the maximum contour coefficient is the maximum number of clusters of the coal seam data in this borehole. （3）； Step 4.3: Based on the results calculated in Step 4.2, a comprehensive cluster analysis is performed on all borehole coal quality data in the coalfield by combining the hierarchical clustering learning algorithm with the ISO algorithm in GIS software. Step 4.4: Project the selected data according to the coordinate position, and based on the results of hierarchical clustering and ISO algorithm, extract the borehole points on each result in GIS by using the mask extraction and spatial connection tools, and evaluate and analyze each clustering result separately. Step 5 includes the following sub-steps: Step 5.1, the operation steps are as follows: First, a judgment matrix with m samples and n evaluation indicators needs to be constructed: ； Step 5.2: Since the units of ash content, sulfur content, phosphorus content, volatile matter, and viscosity index are different, the matrix needs to be normalized and dimensionless to obtain a normalized matrix. ； With the new trend matrix ； In the formula, b ij Let j be the index value of the i-th evaluation object; According to the definition of information entropy, the formula for calculating entropy weight is: （4）； Where: H j For the corresponding transitional variable value of evaluation index j in the entropy weight calculation process; This is a transitional variable in the entropy weight calculation process, and ; The information entropy calculation vector is obtained from Formula 4: （5）； In the formula, W is the entropy weight variable obtained from the final calculation; w j Let j be the entropy weight corresponding to the evaluation index, and （6）； Step 5.3: Finally, based on the entropy weight method calculation principle and using the raster calculator tool in GIS, calculate the weights of ash, sulfur, phosphorus, volatile matter, and bonding index according to the formulas above. Step 6 includes the following sub-steps: Step 6.1: Classify the quality of coking coal by individual indicators. Based on the quality standards for coking coal and the actual conditions of the study area, coking coal can be divided into three grades: high-quality coal, medium-quality coal, and low-quality coal. Among them, sulfur and phosphorus are environmental indicators, while ash, volatile matter, and caking index are economic indicators. Step 6.2: Referring to the comprehensive index evaluation method, assign values ​​to each indicator and calculate the comprehensive evaluation value of the borehole according to the weights in Table 1 corresponding to the relevant steps; the comprehensive index evaluation method classifies each quality indicator into grades and assigns values ​​to them, where: the value for high-quality coal is "3.5"; the value for medium-quality coal is "2.5"; and the value for low-quality coal is "1.5". Then, calculate the comprehensive quality evaluation value of each coal sample according to the weights of each indicator and the following formula. E q = ×X vq ,(v=1,2,3,…,m;q=1,2,3,…,m); Among them, X qv The corresponding quality indicator is assigned a value, and X qv =1.5, 2.5, 3.5, (q represents the quality number of each indicator, and v represents the number of each coal sample); K q K represents the weight assignment of each quality indicator. q =0.1 or 0.2; Step 6.3: Calculate the comprehensive evaluation value according to the above formula, and classify the borehole coking coal of different clusters into high-quality coal, medium-quality coal, and low-quality coal. According to the raster calculator function in GIS, when 3≤E q When 2.5 ≤ E, this type of coal is considered high-quality coking coal; when 2.5 ≤ E q When E < 3, the coal type belongs to medium-quality coking coal; when E q If the content is less than 2.5, the coal is considered low-quality coking coal. Step 6.4: Project the borehole results processed in Step 6.3 into the GIS. Use the inverse distance weighting method and the natural discontinuity grading method to divide the coal quality data of the two clusters. Based on different color areas, the coking coal resources of different grades and different coal quality characteristics under each cluster can be clearly distinguished. Step 6.5: Compare the differences in the results of individual indicators of different clustering boreholes, and conduct a comparative analysis of the two types of data boreholes in accordance with national standards and specifications.

2. The comprehensive evaluation method for coking coal resources based on unsupervised machine learning as described in claim 1, characterized in that, In step one, the data to be removed includes coal quality data that does not belong to coking coal blending and coal quality data that was obviously entered incorrectly in the original data.

3. The comprehensive evaluation method for coking coal resources based on unsupervised machine learning as described in claim 1, characterized in that, In step three, the dimensions are eliminated using the following formulas: Formula 1 is used for inverse indices, and Formula 2 is used for positive indices. Ash, sulfur, phosphorus, and volatile matter are inverse indices reflecting the quality of coking coal, while the caking index is a positive indices reflecting the quality of coking coal. （1）； （2）。 4. The comprehensive evaluation method for coking coal resources based on unsupervised machine learning as described in claim 1, characterized in that, The process also includes the following steps: based on cluster analysis of a large amount of coal quality data, classification of coking coal grades, definition of coal quality index grades, and comprehensive consideration, delineation of high-quality coking coal development areas, and provision of reasonable development and utilization plans.

Images