Coal type replacement path generation method based on hidden coal blending strategy identification

Abstract

The invention provides a coal type replacement path generation method based on hidden coal blending strategy identification, and relates to the technical field of intelligent coal type replacement. The coal type replacement path generation method based on recessive coal blending strategy recognition comprises the steps that historical successful coal blending schemes are collected through data collection and preprocessing; an unsupervised clustering algorithm is adopted, and a hidden coal blending strategy is mined from the multi-dimensional historical coal blending data; carrying out feasibility verification on the coal blending replacement scheme based on coal blending attribute balance analysis; carrying out refined quantitative analysis on the feasible coal blending replacement scheme; and outputting a quantized coal type replacement report. According to the method, all complex formula modes which are proved to be successful in history are objectively concluded and summarized, view limitation and subjective prejudice of manual searching are overcome, the discovery process of the replacement opportunities is more comprehensive and systematic, and implicit knowledge of enterprises is dominated and structured.

Description

Technical Field

[0001] This invention relates to the field of intelligent coal type replacement technology, and in particular to a method for generating coal type replacement paths based on implicit coal blending strategy identification. Background Technology

[0002] Metallurgical coke, serving as the backbone, fuel, and reducing agent in blast furnace ironmaking, is crucial for ensuring smooth steel production and achieving energy conservation and emission reduction due to its stable and superior quality. To address the increasing scarcity of high-quality prime coking coal resources and market price fluctuations, optimizing coal blending from multiple coal sources has become the norm in modern coking production. Developing scientific and reliable coal substitution strategies to maximize the utilization of cost-effective economical coal types while ensuring coke quality is a core technical challenge for coking enterprises to reduce costs, increase efficiency, and enhance market competitiveness.

[0003] Existing methods for developing coal type replacement schemes suffer from significant technical bottlenecks:

[0004] 1. Limitations of the traditional expert experience method: At present, the adjustment of coal blending schemes and the introduction of new coal types largely depend on the personal experience of coal blending engineers. The effectiveness of this method depends heavily on the expert's knowledge reserves and familiarity with specific coal types, and has the following defects: (1) Subjectivity and non-replicability: The expert's decision-making process is difficult to quantify and standardize, making it difficult to accumulate and pass on excellent coal blending knowledge. (2) High trial and error cost: When facing new coal types with little experience or drastic changes in the market environment, relying purely on experience is prone to errors. It is necessary to verify through multiple expensive and time-consuming small coke oven experiments or industrial tests, resulting in a long research and development cycle and high risk.

[0005] 2. Dimensional limitations of traditional statistical analysis methods: Some enterprises try to use data analysis tools, such as calculating the pairwise correlation of different coal types (e.g., Pearson correlation coefficient) to find substitution relationships. Although this method is more objective than pure empirical methods, its limitations are: (1) It cannot discover complex patterns: It can only identify simple "one-to-one" linear substitution relationships, but it is powerless against the more common and valuable "many-to-one" (e.g., replacing coal C with a combination of coals A and B) or "many-to-many" complex and nonlinear substitution patterns in production practice. (2) It is susceptible to interference from confounding variables: "spurious correlation" is common in industrial data. Simple correlation analysis cannot eliminate the interference of other factors and may lead to incorrect conclusions.

[0006] 3. The "black box" and "vulnerability" problems of traditional machine learning prediction models: In recent years, although some studies have tried to use models such as neural networks and random forests to predict coke quality, they generally have the following problems: (1) "Black box" problem: Although these models may predict accurately, their internal decision-making logic is not transparent, making it difficult to explain "why" the result is this, and failing to give engineers enough trust to make production decisions. (2) Correlation trap: These models are essentially learning the statistical correlation in the data, rather than the real causal relationship. When the production operating conditions and raw material sources change, the correlation on which the model depends may fail, causing its prediction performance to drop sharply, making the model "vulnerable" and with poor generalization ability.

[0007] In summary, existing technologies lack a method that can objectively and automatically extract, systematically, and practically tested complex coal blending strategy patterns from massive amounts of historical data, and provide scientifically validated, high-potential replacement solutions with clear quantitative indicators before conducting expensive physical experiments. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this invention proposes a method for generating coal type replacement paths based on implicit coal blending strategy identification, and specifically discloses a method for intelligent discovery and quantitative evaluation of coal type replacement paths.

[0009] The implicit coal blending strategies described in this invention are formula patterns or paradigms that exist in data but are not obvious, are composed of multiple coal types in different proportions, and have been repeatedly proven to successfully produce qualified coke. These strategies are the culmination of empirical knowledge from production practice. Coal substitution is a process in which one or more types of coal can be substituted for one type of coal while the quality of the resulting coke remains unchanged.

[0010] This invention applies unsupervised machine learning, data mining, and statistical analysis techniques to automatically identify multiple coexisting and successful "hidden coal blending strategies" from a large amount of historical production observation data of coking enterprises. By comparing the structural differences between these strategies and combining them with the coal blending attributes of coking coal, the invention verifies their feasibility and ultimately generates coal type replacement suggestions that have been verified by historical data, can directly guide small coke oven experiments, and have clear quantitative indicators. This provides a scientific and efficient data-driven decision support system for coking enterprises to reduce costs and increase efficiency.

[0011] In a first aspect, the present invention provides a method for generating coal type replacement paths based on implicit coal blending strategy identification, comprising the following steps:

[0012] S1. Data acquisition and preprocessing; including:

[0013] S11. Data Acquisition: Collect historical coal blending data of the enterprise;

[0014] S12. Data Filtering: Retain coal blending schemes with complete proportions and compliant coke, and remove duplicate data;

[0015] S13. Data Reconstruction: Reconstruct the filtered data into a tabular data frame for coal type structure.

[0016] S2. Breadth-based exploration: Uncovering hidden coal blending strategies from multi-dimensional data; including:

[0017] S21. Through correlation analysis, the correlation, synergy, or substitution relationships between the proportions of different coal types were discovered.

[0018] S22. Data Standardization: Perform Z-score standardization on the coal type structure data frame;

[0019] S23. Implicit Strategy Clustering: For the data obtained in S22, an unsupervised clustering algorithm is used to cluster the most common coal blending schemes. The average scheme of each cluster is taken as the implicit coal blending strategy that has been verified in history, i.e., the strategy paradigm. By comparing the ratio differences between strategy paradigms, coal blending replacement schemes are identified.

[0020] S3. Based on the balance analysis of the properties of blended coal, verify the feasibility of the coal replacement scheme; including:

[0021] S31. Construct a coal type attribute database, including the average physicochemical properties of each type of coal;

[0022] S32. For all strategy paradigm transformations and combinations, verify the property balance before and after the transformation;

[0023] S4. Conduct a detailed and quantitative analysis of feasible coal blending replacement schemes; specifically, based on the historical data in the coal type structure data frame, analyze the historical replacement rate, the proportion distribution of the replaced coal types, and the historical safe usage of the coal blending replacement schemes.

[0024] S5. Output Report: Output a quantitative coal type replacement report.

[0025] As a further improvement of the present invention, in step S11, data is collected from the enterprise's MES or LIMS system. The data includes at least: coal blending scheme ID, the name of each type of coal and its proportion in the scheme, the key quality indicators of the final coke produced by the scheme, and the detailed physicochemical properties of the raw coal entering the plant.

[0026] As a further improvement of the present invention, step S12 specifically includes the following steps:

[0027] S121. Completeness Screening: Summing up the proportions of all coal types in each coal blending scheme to obtain the total proportion, and eliminating invalid scheme data with incomplete proportions;

[0028] S122. Success Screening: Screening coal blending schemes based on coke quality indicators;

[0029] S123. Representative screening: For the same scheme with multiple test records, the test results are averaged.

[0030] As a further improvement of the present invention, the specific screening method of step S121 is: only retain the records of schemes with a total ratio between 99% and 101%.

[0031] As a further improvement of the present invention, the specific screening method of step S122 is: only retaining coal blending schemes whose strength after coke reaction is not less than a set threshold.

[0032] As a further improvement of the present invention, the coal type structure data frame in step S13 is composed of: each row represents a successful coal blending scheme, each column represents a type of coal, and the value in the cell is the proportion of that coal type in the scheme; the coal blending schemes are summarized to form a monthly structural trend.

[0033] As a further improvement of the present invention, the specific steps of step S21 include:

[0034] The coal blending data obtained in step S1 is used to obtain a correlation matrix for the proportion of different coal types; among which,

[0035] There is a substitution relationship between coal types that exhibit a strong negative correlation;

[0036] There is a synergistic relationship between coal types that show a strong positive correlation.

[0037] As a further improvement of the present invention, the processing method of step S22 is as follows:

[0038] ;

[0039] in, It is a standardized formula. This is the original formula. This is the average value obtained by statistically analyzing all the original formulas. This represents the standard deviation obtained by statistically analyzing all original formulations.

[0040] As a further improvement of the present invention, the specific steps of the unsupervised clustering algorithm in step S23 include:

[0041] The standardized coal blending scheme data is divided into K clusters;

[0042] The optimal clustering scheme and the centroids of the K clusters are obtained with the goal of minimizing the sum of squares within each cluster.

[0043] By denormalizing the centroid, we obtain the proportioning schemes within the original proportioning space, namely K strategy paradigms.

[0044] As a further improvement of the present invention, the unsupervised clustering algorithm adopts the K-Means algorithm, and the specific method includes:

[0045] N formula samples Divided into K clusters In, each sample It is a vector containing the proportions of M types of coal;

[0046] Cluster analysis is performed with the goal of minimizing the sum of squares within each cluster, i.e.:

[0047] ;

[0048] To obtain the optimal clustering scheme, and each cluster in it. The centroids are the average proportions of all formulations within the cluster. Each centroid is then denormalized to obtain K formulation schemes within the original formulation space. This refers to the strategy paradigm, which represents the implicit coal blending strategy paradigm that has been repeatedly proven successful throughout history.

[0049] As a further improvement of the present invention, the number of clusters K is preset according to business needs, or determined by the elbow rule or the contour coefficient method.

[0050] As a further improvement of the present invention, in step S23, the method for identifying coal blending replacement schemes includes: identifying the main, high-frequency coal blending replacement schemes by comparing the differences of K paradigm center points.

[0051] As a further improvement of the present invention, the specific steps of step S32 include:

[0052] S321. Determine the reduction and addition packages for different coal types during the strategy paradigm shift;

[0053] S322. Calculate the equivalent coal type attribute changes corresponding to the reduction and increase of packages. and ;

[0054] S323, with and The feasibility of conversion is judged based on the similarity.

[0055] As a further improvement of the present invention, the specific steps of step S321 include:

[0056] Calculate the difference vector Δ between the center points of the two strategy paradigms involved in the transformation;

[0057] In vector Δ, coal types with values ​​less than a set threshold constitute a reduction package, while coal types with values ​​greater than the set threshold constitute an increase package.

[0058] As a further improvement of the present invention, the specific steps of step S322 include:

[0059] Calculate the reduction of packages The reduction in the proportion of each type of coal in the internal blending.

[0060] Calculate the increase in packages The increase in the proportion of each type of coal in the internal blending;

[0061] Calculate the weighted sum property of decreasing and increasing packages. and :

[0062] ;

[0063] in, It refers to the coal quality properties of coal type j; It means reducing or increasing the vector component corresponding to coal type j in the package.

[0064] As a further improvement of the present invention, step S323 specifically involves calculating... and The conversion matching degree of each attribute is calculated by the following steps:

[0065] Computational attribute differences, i.e. and The difference in each attribute;

[0066] Calculate the properties of the blended coal, that is, the value of the property in the overall coal blending scheme;

[0067] Calculate the transformation match degree of this attribute using the following formula:

[0068] ;

[0069] When the average conversion matching degree of each attribute is higher than the preset threshold, the conversion is deemed feasible.

[0070] As a further improvement of the present invention, the specific steps of step S4 include:

[0071] S41. Data Focus: From the coal type structure data frame, filter out all coal blending schemes that simultaneously use all coal types in both the reduction package and the increase package;

[0072] S42. Conditional Relationship Analysis: Based on the amount of the replaced coal type, the coal blending scheme obtained in step S41 is stratified. By performing conditional statistics on different strata, the quantitative relationship between the replacement combination and the replaced coal type is obtained, namely the historical replacement rate.

[0073] S43. Internal Structure Analysis: For each coal blending scheme obtained in step S41, calculate the proportion distribution of each coal type in the current replacement path within the added package, analyze its mean and stability, and obtain the golden ratio within the added package as a key parameter that must be strictly followed.

[0074] S44. Safety Usage Boundary Statistics: Statistically analyze the maximum ratio of this increased package that has been successfully used in history, as a reference for safety usage.

[0075] As a further improvement of the present invention, the method for conditional statistics in step S42 includes:

[0076] ;

[0077] in, This is a quantitative relationship between the replacement combination and the coal type being replaced. To reduce the number of bags, To add packages; and These represent the average proportions of coal type j in the lower and higher usage strata, respectively.

[0078] As a further improvement of the present invention, the specific steps for analyzing stability in step S43 include:

[0079] Calculate the coefficient of variation, 25%, 50%, and 75% quantiles. The ratios with a coefficient of variation less than a set threshold and a 25%-75% quantile interval narrower than the set threshold are determined to be stable ratios, i.e., the golden ratio.

[0080] The coefficient of variation is calculated as: standard deviation / mean.

[0081] As a further improvement of the present invention, the report output in step S5 provides quantitative recommendations based on historical replacement rates, golden ratios, and safety boundaries for coal type replacement paths, in conjunction with the conclusions of step S4.

[0082] In a second aspect, the present invention provides a computer device including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in the first aspect.

[0083] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0084] Fourthly, the present invention provides a computer program product that, when executed by a processor, implements the steps of the method described in the first aspect.

[0085] This patent addresses several technical problems in the development of existing coal type replacement schemes, including over-reliance on subjective experience, limited data mining dimensions, inability to detect complex replacement patterns, susceptibility to misleading data spurious correlations, and lack of reliable quantitative guidance before physical experiments. Through the method described in this patent, the following beneficial effects can be achieved without relying on any "black box" quality prediction models:

[0086] Achieving objectivity, systematization, and depth in the discovery of replacement opportunities: This invention elevates the analytical dimension from simple "pairwise relationships between coal types" to the level of "multiple coal blending strategy paradigms" through statistical analysis, data mining, and unsupervised machine learning algorithms. It can automatically and objectively summarize and generalize all historically proven, complex blending patterns, overcoming the limitations of manual searching and subjective biases. This makes the discovery of replacement opportunities more comprehensive and systematic, and makes the company's tacit knowledge explicit and structured.

[0087] Significantly improving the success rate and R&D efficiency of physical experiments while reducing costs: The "alternative paths" output by this method are not based on guesswork, but rather on a deep summary of numerous historical successful cases. Furthermore, after obtaining the hypothesis, the introduction of a "property balance verification" step automatically filters out "pseudo-related" solutions that are not feasible in metallurgical principles. This ensures that only high-value, high-probability-of-success "targets" are submitted to the small coke oven experiments. It reduces blind trial-and-error experiments, shortens the R&D cycle of new solutions, and significantly saves valuable experimental and production resources.

[0088] Provides quantifiable and directly operable experimental design guidelines: This invention completely changes the previous "vague" experimental model. It can not only propose qualitative directions of "what can be replaced by what" (breadth exploration), but also provide a set of precise quantitative operational suggestions for high-potential directions (depth exploration), specifically including: (1) historical replacement rate: for example, how much proportion of B and C coal combination needs to be added on average for every 1% reduction of A coal; (2) golden ratio of replacement combination: the optimal ratio of B and C coal within the combination package; (3) verified safety boundary: what is the highest proportion used in this replacement combination in history. These indicators provide unprecedented clarity and operability for experimental design. Attached Figure Description

[0089] Figure 1 This is a flowchart of a method for generating coal type replacement paths based on implicit coal blending strategy identification, as disclosed in this invention.

[0090] Figure 2 This is an example of a heat map showing the correlation matrix of coal type proportions.

[0091] Figure 3The heat map of the core coal blending strategy paradigm discovered by K-Means clustering.

[0092] Figure 4 Here is a radar chart comparison example of several core coal blending strategy paradigms. Detailed Implementation

[0093] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Steps S1, S2… in the described embodiments of the present invention do not limit the scope of execution of the present invention; the various models, simulation environments, and software described in the present invention are not considered as the only limiting methods of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0094] In this invention, computer device / equipment / system refers to a related entity applied to a computer, such as hardware, a combination of hardware and software, software, or software in execution. More specifically, for example, software includes, but is not limited to, a process running on a processor, a processor, an object, executable software, an execution thread, a program, and / or a computer. Furthermore, an application program or script running on a server, and the server itself, can also be software. One or more software programs may be in an execution process and / or thread, and the software may be localized on one computer and / or distributed across two or more computers, and may be run on various computer-readable media.

[0095] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0096] In a first aspect, the present invention provides an embodiment of a coal type replacement path generation method based on implicit coal blending strategy identification, such as... Figure 1 As shown, the specific process can be as follows:

[0097] S1. Data Acquisition and Preprocessing;

[0098] This step aims to build a high-quality, highly reliable dataset of "historical successful recipes" as the foundation for subsequent analysis. Specific steps include:

[0099] S11. Data Acquisition: Collect historical coal blending data from the enterprise's MES or LIMS system. The collected data should include at least: coal blending scheme ID, the name of each coal type and its proportion in the scheme, the key quality indicators of the final coke produced by the scheme (such as the coke post-reaction strength CSR), and the detailed physicochemical properties of the raw coal entering the plant (such as volatile matter Vdaf, caking index G, plastic layer thickness Y, etc.).

[0100] S12. Data filtering; specifically including the following steps:

[0101] S121. Integrity Screening: To ensure the accuracy of the analysis, the proportions of all coal types in each coal blending scheme are first summed. Only scheme records with a total proportion between 99% and 101% are retained, and invalid data with incomplete proportions due to data recording errors are removed.

[0102] S122. Success Screening: The core of this invention is "learning from successful experiences." Therefore, the formulas are screened based on coke quality indicators. Taking the post-reaction strength of coke as an example, a threshold is set (e.g., coke_CSR >= 62), and only "successful formula" records that meet this condition are retained.

[0103] S123. Representative Screening: In industrial production, the same coal blending scheme may correspond to multiple quality inspections. To avoid data redundancy and introduce unnecessary noise, the inspection results of the same scheme with multiple inspection records are averaged.

[0104] S13. Data Reconstruction: The cleaned and filtered data is reconstructed into a coal type structure data frame using pivot table operations. Each row of this data frame represents a unique and successful coal blending scheme, and each column represents a coal type, with the cell value indicating the proportion of that coal type in the scheme. This data is then further summarized to form a monthly structural trend.

[0105] In one embodiment of the present invention, the coal type structure data frame obtained by data reconstruction is shown in Table 1.

[0106] Table 1:

[0107]

[0108] S2, Breadth Exploration: Automatic Discovery of Implicit Coal Blending Strategies Based on Unsupervised Learning.

[0109] This step is unbiased and objectively, automatically and comprehensively, identifies all potential and valuable replacement opportunities from multi-dimensional data, and extracts effective information and hypotheses. Specific steps include:

[0110] S21. Correlation analysis of coal type combinations: Based on basic correlation analysis, the correlation and synergistic or substitution relationships between different coal type proportions are discovered.

[0111] Specifically, such as Figure 2 As shown, correlation statistics are performed on all preprocessed historical blending data for coal type proportions to obtain a correlation matrix, thereby quickly identifying significant "one-to-one" or "simple many-to-one" linear substitution relationships in the historical data. Coal type pairs exhibiting strong negative correlations (e.g., correlation coefficients below -0.5) intuitively indicate a historical "substitution" relationship where one coal type outperforms the other. Coal type pairs exhibiting strong positive correlations indicate a "synergistic" relationship, suggesting they tend to be bundled together for use.

[0112] S22. Data Standardization: Since the commonly used blending ratios for different coal types can vary significantly (e.g., 1 / 3 coking coal and gas coal often have high blending ratios, while some other coal types have low ratios), to avoid the clustering algorithm being dominated by coals with high blending ratios, standardization is first applied to the coal type structure data frame using Z-score standardization to obtain...

[0113] ;

[0114] in, This is the original formula. This is the average value obtained by statistically analyzing all the original formulas. This represents the standard deviation obtained by statistically analyzing all original formulations.

[0115] S23. Implicit Strategy Clustering: Using unsupervised clustering algorithms (such as K-Means), cluster analysis is performed on the standardized data to automatically divide all successful recipes into K clusters.

[0116] Specifically, N formula samples (each of the samples) (A vector containing the proportions of M types of coal) is divided into K clusters. In this context, the number of clusters K can be preset according to business needs, or determined through methods such as the "elbow rule" or silhouette coefficient.

[0117] The objective is to minimize the sum of squares within the cluster, i.e.:

[0118] ;

[0119] To obtain the optimal clustering scheme, and each cluster in it. The centroids (mean vectors, i.e., the average proportions of all formulas within the cluster) are then denormalized to obtain the proportioning schemes within the original proportioning space. . In this invention, it is defined as a strategy paradigm, which means: a "hidden coal blending strategy paradigm" that has been repeatedly proven successful in history.

[0120] By comparing the proportion differences among K strategy paradigms (preferably, comparing the differences in the central points of these K paradigms), the most important and highest frequency replacement patterns in the data are identified.

[0121] In one embodiment of the present invention, the resulting strategy paradigm is shown in Table 2.

[0122] Table 2:

[0123]

[0124] Figure 3 This image presents N strategy paradigms discovered by the clustering algorithm in the form of a heatmap. Each row represents a strategy paradigm, and each column represents a type of coal. The color intensity of the cells indicates the average proportion of that coal type within that strategy. This image allows for the rapid identification of the core differences in coal composition among different strategies.

[0125] Furthermore, Figure 4 Several core coal blending strategy paradigms are compared using radar charts. The chart provides a unique "shape fingerprint" for each paradigm, with each axis representing a specific coal type. By comparing the shapes of different strategies (polygons of different colors), the tendency of each strategy to use a particular coal type becomes very intuitive. For example, if a strategy is unusually prominent along a specific coal type axis, it indicates a high dependence on that coal type.

[0126] S3. Feasibility Verification and Attribute Balance Analysis: This step uses domain knowledge to verify the "reasonableness" of data discovery. Based on coal attribute balance analysis, after identifying multiple data-driven strategy paradigms, the "feasibility verification" stage connects the data patterns with physical reality. Specific steps include:

[0127] S31. Construct a coal type attribute database: Group coal types by type, calculate the average physicochemical properties (such as Vdaf, G, Y) of each type of coal, and form a "coal type attribute quick reference database" indexed by coal type.

[0128] In one embodiment of the present invention, the obtained coal type attribute database is shown in Table 3.

[0129] Table 3:

[0130]

[0131] S32. Perform attribute balance verification: Iterate through all transformation combinations between strategy paradigms (e.g., from strategy paradigm 2 to strategy paradigm 3) and perform the following steps:

[0132] S321. Determine the "Addition / Decrease Package": Calculate the difference vector Δ between the center points of the two strategies. In vector Δ, coal types with a value less than -0.01 (threshold) constitute the "Decrease Package"; coal types with a value greater than +0.01 constitute the "Addition Package".

[0133] S322. Calculate the weighted attributes: For the "reduction package," the weighted attribute is determined by the reduction in the proportion of each coal type within it. Similarly, calculate the increase in the proportion of each coal type within the "increase package." Then, calculate the equivalent Vdaf, G, Y values ​​for the "reduction package" and the "increase package" as a whole.

[0134] In a computational process of transitioning from a high-cost strategy A to a low-cost strategy B, the two strategy paradigms are first calculated. and Difference vector The negative components in this vector are defined as "reduced packets". This refers to a coal type combination where strategy A has a high proportion and strategy B has a low proportion; the positive component is defined as "additional package". This refers to a coal type combination where the proportion is low in Strategy A and high in Strategy B. Subsequently, the system queries a pre-built coal type attribute database (such as volatile matter Vdaf, caking index G, and plastic layer thickness Y) to calculate the weighted sum of attributes for the "reduction package". The weighted sum property of "add package" :

[0135] ;

[0136] in, It refers to the coal quality properties of coal type j; It means reducing or increasing the vector component corresponding to coal type j in the package.

[0137] S323. Generate conclusions: Comparison and The similarity is used to determine whether the strategy transition is "balanced", thereby effectively filtering out "spurious correlation" replacement paths and ensuring that subsequent analysis focuses on truly valuable directions.

[0138] Specifically, the physicochemical feasibility of the strategy conversion is determined by calculating the attribute differences and matching degree between "reduced package" and "increased package" ((1 - abs(difference / coal properties)) * 100%). Only conversions with an average matching degree higher than a preset threshold (e.g., 95%) are considered "highly feasible" high-potential replacement paths.

[0139] This step yields an attribute balance verification table for the strategy transition, which displays a comparison of the weighted average values ​​of the "reduction package" and the "increase package" output by the system on several key physicochemical properties (such as Vdaf, G, Y) when verifying the transition from "strategy A to strategy B". The table includes both difference values ​​and matching degrees, providing direct quantitative evidence for determining the feasibility of the replacement scheme.

[0140] In one embodiment of the present invention, during the process of converting from strategy paradigm 0 to strategy paradigm 1, an attribute balance verification table for strategy conversion is obtained by calculation, as shown in Table 4.

[0141] Table 4:

[0142]

[0143] Based on the above results, the conversion is deemed highly feasible. The average attribute matching rate is 97.6%, indicating a high degree of balance in the properties of the blended coal.

[0144] The balancing results obtained during the transition from strategy paradigm 0 to strategy paradigm 3 are shown in Table 5.

[0145] Table 5:

[0146]

[0147] Based on the above results, the conversion is deemed highly feasible. The average property matching rate is 97.5%, and the replacement is highly balanced in terms of the properties of the blended coal.

[0148] S4. In-depth exploration and quantitative guidance: non-predictive and refined quantification of high-potential paths.

[0149] For high-potential replacement paths that have passed the attribute balance verification in step S3 (e.g., replacing SM with {PS, 1 / 3JM}), this step performs in-depth, refined quantitative analysis to provide specific parameters for small coke oven experiments. This process does not rely on any predictive models. Specifically, it includes the following steps:

[0150] S41. Data Focus: From the coal type structure data frame, filter out all historically successful formulations related to this replacement path.

[0151] Preferably, the formulas that simultaneously use all key coal types in both the "reduction package" and the "increase package" of a certain replacement path are selected.

[0152] S42. Conditional Relationship Analysis: Capturing replacement dynamics through historical replacement rates.

[0153] Based on the historical successful formula obtained in step S41, the formulas are stratified according to the amount of the replaced coal type, and conditional statistics are used to compare the corresponding changes in the amount of "increased package" coal type in different strata.

[0154] As shown in Table 6, in one embodiment of the present invention, the coal types SM being replaced are divided into three groups based on their usage: high, medium, and low (represented by usages greater than 15%, usages between 10% and 15%, and usages less than 10%). Each group contains the same amount of data. For each group, the average proportion of the replacement combination and the replaced coal type is calculated. By comparing the average proportions between different groups, the replacement relationship can be clearly observed: when the usage of SM transitions from the "high usage group" to the "low usage group," if the average proportion of another coal type shows a systematic increase, it indicates a potential replacement relationship between it and SM, i.e., the "historical replacement rate."

[0155] Preferably, in the process of capturing historical replacement rates, by comparing the corresponding changes in the amount of "addition package" in different strata, a stable quantitative relationship between the replacement combination and the replaced coal type is statistically determined. :

[0156] The stable quantitative relationship between the replacement combination and the replaced coal type was statistically determined. :

[0157] ;

[0158] in and These represent the average proportions of coal type j in the lower and higher usage strata, respectively.

[0159] Table 6:

[0160]

[0161] S43. Internal Structure Analysis: Replace the golden ratio of the combination; calculate the distribution of each coal type in the "addition package" (e.g., {PS, 1 / 3JM}) in all relevant cases, analyze its mean and stability, and obtain the "golden ratio" inside the combination package.

[0162] Specifically, for each historically successful formulation obtained from step S41, the relative proportion of each coal type within the "addition package" (replacement combination) is calculated (e.g., ratio_PS / (ratio_PS + ratio_1 / 3JM)). Statistical analysis is then performed on the calculation results for each formulation, calculating its mean, standard deviation, 25%-75% quantile, and coefficient of variation (cv) (i.e., standard deviation / mean). If the coefficient of variation is less than a set threshold (exemplarily set to 0.15), and the 25%-75% quantile is narrower than a set threshold (exemplarily set to 10%), it demonstrates that the internal proportions ("golden ratio") of the replacement combination are very stable and are key parameters that must be strictly adhered to.

[0163] In one embodiment of the present invention, the internal proportion analysis results of an additive package {PS, 1 / 3JM} are shown in Table 7.

[0164] Table 7:

[0165]

[0166] S44. Safety Dosage Boundary Statistics: Statistically analyze the maximum ratio in which this "additional pack" has been successfully used in history to provide a safety reference for experimental design.

[0167] S5. Output Report: Through the progressive analysis in steps S2-S4, the messy historical data is systematically transformed into a series of high-potential coal type replacement schemes that have been verified for feasibility and come with clear quantitative operation guidelines, providing a reliable basis for subsequent physical experiments and production practices.

[0168] The results obtained from the above steps are integrated into a quantitative recommendation report containing information such as historical replacement rate, golden ratio, and safety margin, and output to guide the experimental design of small coke ovens.

[0169] In one embodiment of the present invention, the output quantitative recommendation report is as follows:

[0170] Historical replacement rate reference: According to the statistics of step S42, for every 11.31% reduction of SM, approximately 10.56% of the [PS, 1 / 3JM] combination needs to be added. That is, for every 1% reduction of SM, approximately 0.93% of the [PS, 1 / 3JM] combination needs to be added, with a replacement ratio of nearly 1:1.

[0171] Core internal ratio: Based on the analysis in step S43, it is recommended that the internal ratio of the replacement pack be strictly controlled around the following golden ratio in the experiment: PS: 45.5% and 1 / 3JM: 54.5%.

[0172] Verified safety boundary: According to the statistics in step S44, the total usage of the [PS, 1 / 3JM] combination reached a maximum of 42.00%, and was successful. It is recommended that experiments be conducted within this range.

[0173] In a second aspect, the present invention provides an embodiment of a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in the first aspect.

[0174] Thirdly, the present invention provides an embodiment of a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0175] Fourthly, the present invention provides a computer program product embodiment, which, when executed by a processor, implements the steps of the method described in the first aspect.

Claims

1. A method for generating coal type replacement paths based on implicit coal blending strategy identification, characterized in that, Includes the following steps: S1. Data acquisition and preprocessing; including: S11. Data Acquisition: Collect historical coal blending data of the enterprise; S12. Data Filtering: Retain coal blending schemes with complete proportions and compliant coke, and remove duplicate data; S13. Data Reconstruction: Reconstruct the filtered data into a tabular data frame for coal type structure. S2. Breadth-based exploration: Uncovering hidden coal blending strategies from multi-dimensional data; including: S21. Through correlation analysis, the correlation, synergy, or substitution relationships between the proportions of different coal types were discovered. S22. Data Standardization: Perform Z-score standardization on the coal type structure data frame; S23. Implicit Strategy Clustering: For the data obtained in S22, an unsupervised clustering algorithm is used to cluster the most common coal blending schemes. The average scheme of each cluster is taken as the implicit coal blending strategy that has been verified in history, i.e., the strategy paradigm. By comparing the ratio differences between strategy paradigms, coal blending replacement schemes are identified. S3. Based on the balance analysis of the properties of blended coal, verify the feasibility of the coal replacement scheme; including: S31. Construct a coal type attribute database, including the average physicochemical properties of each type of coal; S32. For all strategy paradigm transformations and combinations, verify the property balance before and after the transformation; S4. Conduct a detailed and quantitative analysis of feasible coal blending replacement schemes; specifically, based on the historical data in the coal type structure data frame, analyze the historical replacement rate, the proportion distribution of the replaced coal types, and the historical safe usage of the coal blending replacement schemes. S5. Output Report: Output a quantitative coal type replacement report.

2. The method according to claim 1, characterized in that, In step S11, data is collected from the enterprise's MES or LIMS system. The data includes at least: coal blending scheme ID, the name of each type of coal and its proportion in the scheme, the key quality indicators of the final coke produced by the scheme, and the detailed physicochemical properties of the raw coal entering the plant.

3. The method according to claim 1, characterized in that, Step S12 specifically includes the following steps: S121. Completeness Screening: Summing up the proportions of all coal types in each coal blending scheme to obtain the total proportion, and eliminating invalid scheme data with incomplete proportions; S122. Success Screening: Screening coal blending schemes based on coke quality indicators; S123. Representative screening: For the same scheme with multiple test records, the test results are averaged.

4. The method according to claim 3, characterized in that, The specific screening method for step S121 is as follows: only retain the records of schemes with a total ratio between 99% and 101%.

5. The method according to claim 3, characterized in that, The specific screening method for step S122 is as follows: only coal blending schemes with coke strength not less than a set threshold after reaction are retained.

6. The method according to claim 1, characterized in that, The coal type structure data frame in step S13 consists of: each row representing a successful coal blending scheme, each column representing a type of coal, and the value in the cell being the proportion of that coal type in the scheme; summarizing the coal blending schemes to form a monthly structural trend.

7. The method according to claim 1, characterized in that, The specific steps of step S21 include: The coal blending data obtained in step S1 is used to obtain a correlation matrix for the proportion of different coal types; among which, There is a substitution relationship between coal types that exhibit a strong negative correlation; There is a synergistic relationship between coal types that show a strong positive correlation.

8. The method according to claim 1, characterized in that, The processing method for step S22 is as follows: ； in, It is a standardized formula. This is the original formula. This is the average value obtained by statistically analyzing all the original formulas. This represents the standard deviation obtained by statistically analyzing all original formulations.

9. The method according to claim 1, characterized in that, In step S23, the specific steps of the unsupervised clustering algorithm include: The standardized coal blending scheme data is divided into K clusters; The optimal clustering scheme and the centroids of the K clusters are obtained with the goal of minimizing the sum of squares within each cluster. By denormalizing the centroid, we obtain the proportioning schemes within the original proportioning space, namely K strategy paradigms.

10. The method according to claim 9, characterized in that, The unsupervised clustering algorithm uses the K-Means algorithm, and the specific method includes: N formula samples Divided into K clusters In, each sample It is a vector containing the proportions of M types of coal; Cluster analysis is performed with the goal of minimizing the sum of squares within each cluster, i.e.: ； To obtain the optimal clustering scheme, and each cluster in it. The centroids are the average proportions of all formulations within the cluster. Each centroid is then denormalized to obtain K formulation schemes within the original formulation space. This refers to the strategy paradigm, which represents the implicit coal blending strategy paradigm that has been repeatedly proven successful throughout history.

11. The method according to claim 10, characterized in that, The number of clusters K can be preset according to business needs, or determined by the elbow rule or the profile coefficient method.

12. The method according to claim 1, characterized in that, In step S23, the method for identifying coal blending replacement schemes includes: identifying the main, high-frequency coal blending replacement schemes by comparing the differences of K paradigm center points.

13. The method according to claim 1, characterized in that, The specific steps of step S32 include: S321. Determine the reduction and addition packages for different coal types during the strategy paradigm shift; S322. Calculate the equivalent coal type attribute changes corresponding to the reduction and increase of packages. and ; S323, with and The feasibility of conversion is judged based on the similarity.

14. The method according to claim 13, characterized in that, The specific steps of step S321 include: Calculate the difference vector Δ between the center points of the two strategy paradigms involved in the transformation; In vector Δ, coal types with values ​​less than a set threshold constitute a reduction package, while coal types with values ​​greater than the set threshold constitute an increase package.

15. The method according to claim 13, characterized in that, The specific steps of step S322 include: Calculate the reduction of packages The reduction in the proportion of each type of coal in the internal blending. Calculate the increase in packages The increase in the proportion of each type of coal in the internal blending; Calculate the weighted sum property of reducing and increasing packages. and : ； in, It refers to the coal quality properties of coal type j; It means reducing or increasing the vector component corresponding to coal type j in the package.

16. The method according to claim 13, characterized in that, Step S323 specifically involves calculating... and The conversion matching degree of each attribute is calculated by the following steps: Computational attribute differences, i.e. and The difference in each attribute; Calculate the properties of the blended coal, that is, the value of the property in the overall coal blending scheme; Calculate the transformation match degree of this attribute using the following formula: ； When the average conversion matching degree of each attribute is higher than the preset threshold, the conversion is deemed feasible.

17. The method according to claim 1, characterized in that, The specific steps of step S4 include: S41. Data Focus: From the coal type structure data frame, filter out all coal blending schemes that simultaneously use all coal types in both the reduction package and the increase package; S42. Conditional Relationship Analysis: Based on the amount of the replaced coal type, the coal blending scheme obtained in step S41 is stratified. By performing conditional statistics on different strata, the quantitative relationship between the replacement combination and the replaced coal type is obtained, namely the historical replacement rate. S43. Internal Structure Analysis: For each coal blending scheme obtained in step S41, calculate the proportion distribution of each coal type in the current replacement path within the added package, analyze its mean and stability, and obtain the golden ratio within the added package as a key parameter that must be strictly followed. S44. Safety Usage Boundary Statistics: Statistically analyze the maximum ratio of this increased package that has been successfully used in history, as a reference for safety usage.

18. The method according to claim 17, characterized in that, In step S42, the method for conditional statistics includes: ； in, This is a quantitative relationship between the replacement combination and the coal type being replaced. To reduce the number of bags, To add packages; and These represent the average proportions of coal type j in the lower and higher usage strata, respectively.

19. The method according to claim 17, characterized in that, In step S43, the specific steps for analyzing stability include: Calculate the coefficient of variation, 25%, 50%, and 75% quantiles. The ratios with a coefficient of variation less than a set threshold and a 25%-75% quantile interval narrower than the set threshold are determined to be stable ratios, i.e., the golden ratio. The coefficient of variation is calculated as: standard deviation / mean.

20. The method according to claim 1, characterized in that, The report output in step S5 provides quantitative recommendations based on historical replacement rates, optimal ratios, and safety boundaries for coal type replacement paths, in conjunction with the conclusions of step S4.

21. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1-20.

22. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method described in any one of claims 1-20.

23. A computer program product, characterized in that, When executed by a processor, the computer program implements the steps of the method described in any one of claims 1-20.

Images