Coking coal blending method

Abstract

The invention relates to the technical field of coal blending, and discloses a coking coal blending method. The method comprises the steps that target functions are determined according to the wet basis ratio, the moisture, the ash content, the volatile component, the wet basis unit transportation cost and the coke dry basis volatile component of each single-variety coal in multiple single-variety coals, and the target functions comprise raw material and storage cost functions; determining a first constraint condition according to the wet basis ratio and moisture of each single-variety coal and a quality function relationship between the single-variety coal and the mixed coal for each single-variety coal in the first influence parameter set; determining a second constraint condition according to the wet basis ratio and moisture of each single-variety coal and a quality function relationship between the single-variety coal and coke for each single-variety coal in a second influence parameter set; and under a plurality of constraint conditions, a meta-heuristic algorithm is adopted to optimize the target function so as to determine the wet basis ratio of each single-variety coal, and the plurality of constraint conditions comprise the first constraint condition and the second constraint condition. According to the method, the coking coal blending ratio can be efficiently and accurately determined.

Description

Technical Field

[0001] The invention relates to the technical field of coal blending, and in particular to a coking coal blending method. Background Art

[0002] At present, the coking coal blending in coking plants is mainly based on empirical calculation or forward calculation. According to experience, single types of coal such as lean coal, PCI coal, fat coal, gas coal, 1 / 3 coal, main coking coal, lean coal, etc. from different manufacturers are selected, the raw material ratio is adjusted through experience, and the target composition of mixed coal and coke is calculated using an excel sheet with associated formulas. This method relies on the experience of the coal blender, is time-consuming and labor-intensive, and has low stability.

[0003] There are some public methods for calculating the coal blending of coke. In the practice of reducing the cost of coal blending in a coking plant, the influence of various types of single coal on the composition, performance and cost of coke was comprehensively evaluated through actual production quality data. The development and application of the coking optimization coal blending system, using linear programming algorithms, proposed factors and conditions related to coking coal blending, proposed a simple cost target optimization function, and used planning solutions, linear fitting and other algorithms to find the coal blending solution that meets the coke quality requirements and has the lowest cost.

[0004] However, the above methods all have the following problems: the coal blending models all use the polynomial of the sum of the amount of each raw material (ratio) multiplied by the corresponding raw material unit price as the objective function, the model is simplified and not accurate enough; all use dry basis ratio or wet basis ratio as input, which deviates from the actual situation, the model is not accurately established, resulting in inaccurate prediction of the blending results. Summary of the invention

[0005] The object of the present invention is to solve at least part or all of the above-mentioned defects and to provide a method for coking coal blending, which comprehensively considers various factors such as wet basis ratio, moisture, ash emission, volatile matter, wet basis unit transportation cost and coke dry basis volatile matter, establishes a coal blending optimization model including raw material cost and storage cost, and uses a meta-heuristic algorithm to ensure the stability and rapidity of the entire coking coal blending process, so that the coking coal blending ratio can be determined efficiently and accurately.

[0006] In order to achieve the above-mentioned purpose, the first aspect of the present invention provides a method for coking coal blending, which comprises: determining an objective function according to the wet basis ratio, moisture, ash, volatile matter, wet basis unit transportation cost and dry basis volatile matter of coke of each single variety of coal in a plurality of single varieties of coal, wherein the objective function comprises a raw material cost function and a storage cost function; determining a first constraint condition for each of the first influencing parameter set according to the wet basis ratio and moisture of each single variety of coal and the mass function relationship between the single variety coal and the mixed coal for each of the first influencing parameter set; determining a second constraint condition for each of the second influencing parameter set according to the wet basis ratio and moisture of each single variety of coal and the mass function relationship between the single variety coal and the coke for each of the second influencing parameter set; and optimizing the objective function under multiple constraints by using a metaheuristic algorithm to determine the wet basis ratio of each single variety of coal, wherein the multiple constraints comprise a first constraint condition for each of the first influencing parameter set and a second constraint condition for each of the second influencing parameter set.

[0007] Through the above technical scheme, the present invention creatively and comprehensively considers multiple factors such as wet basis ratio, moisture, ash, volatile matter, wet basis unit transportation cost and coke dry basis volatile matter, establishes a coal blending optimization model including raw material cost and storage cost, and uses a meta-heuristic algorithm to ensure the stability and rapidity of the entire coking coal blending process, so that the coking coal blending ratio can be determined efficiently and accurately.

[0008] A second aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, and the computer program implements the coking coal blending method when executed by a processor.

[0009] A third aspect of the present invention provides a computer program product, comprising a computer program, which implements the coking coal blending method when executed by a processor.

[0010] Other features and advantages of the present invention will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0012] Figure 1 is a flow chart of a coking coal blending method provided by one embodiment of the present invention;

[0013] Figure 2is a flow chart of a mixed coal quality prediction method provided by an embodiment of the present invention;

[0014] Figure 3 is a flow chart of a coke quality prediction method provided by one embodiment of the present invention;

[0015] Figure 4 is a flow chart of a coke quality prediction method provided by one embodiment of the present invention;

[0016] Figure 5 is a flow chart of a fireworks algorithm for solving a ratio scheme provided by an embodiment of the present invention; and

[0017] Figure 6 It is a flowchart of an overall algorithm provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.

[0019] The existing coal blending models all use the polynomial of the sum of the raw material usage (ratio) multiplied by the corresponding raw material unit price as the objective function, without considering the impact of inventory and transportation on the objective optimization function, and the model establishment is not accurate enough; the impact of the existing raw material inventory maintenance cost on the coal blending cost is not considered; all use dry basis ratio or wet basis ratio as input, and do not consider the multiple conversion relationships between actual dry and wet basis and other important factors such as actual test component items and combustion volume according to actual conditions.

[0020] In view of the defects of the existing technology, it is urgent to propose a more comprehensive coking coal blending model and its optimization algorithm, and based on this, develop a coking coal blending application system with a more accurate model, wider adaptability, and more in line with production and procurement decision-making requirements.

[0021] The present invention is an optimized coke coal blending control method for multiple processes of coke production based on a meta-heuristic comprehensive algorithm. In order to achieve accurate target optimal comprehensive cost optimization control, the following problems need to be solved: First, how to establish an accurate mathematical model, while considering the cost of coal use and the constraints of mixed coal and coke target components, to establish the relationship between the inventory of single-variety raw coal and the overall plan comprehensive cost optimization target; Second, refine the coal blending mathematical model as much as possible, consider the quality control constraints of the two-step production process of mixed coal and coke, and regress the quality relationship between single-variety coal and mixed coal and coke production; Third, the target optimization function and constraint conditions are all nonlinear function relationships, how to solve the problem efficiently and accurately. Fourth, the direct price of raw materials, unit transportation costs, etc. are directly related to the wet basis ratio, and the material balance, coking rate, etc. are directly related to the dry basis ratio. It is necessary to establish a logical relationship for each step of conversion.

[0022] Figure 1 1 is a flow chart of a method for coking coal blending provided by one embodiment of the present invention. Figure 1 As shown, the coking coal blending method includes: step S101, determining the objective function according to the wet basis ratio, moisture, ash, volatile matter, wet basis unit transportation cost and dry basis volatile matter of coke of each single variety of coal in multiple single varieties of coal, wherein the objective function includes a raw material cost function and a storage cost function; step S102, determining the first constraint condition for each of the first influencing parameter set according to the wet basis ratio and moisture of each single variety of coal and the mass function relationship between the single variety coal and the mixed coal for each of the first influencing parameter set; step S103, determining the second constraint condition for each of the second influencing parameter set according to the wet basis ratio and moisture of each single variety of coal and the mass function relationship between the single variety coal and the coke for each of the second influencing parameter set; and step S104, optimizing the objective function under multiple constraints by using a metaheuristic algorithm to determine the wet basis ratio of each single variety of coal, wherein the multiple constraints include the first constraint condition for each of the first influencing parameter set and the second constraint condition for each of the second influencing parameter set.

[0023] The coal blending model of this embodiment comprehensively considers the coal blending consumption cost, storage cost, raw material cost and coke formation rate, takes the Ad, Vdaf, Stad, G, X, Y of the mixed coal formed by a single type of coal and the Ad, Vdaf, Stad, M10, M40, CRI, CSR of the coke as constraints, takes the optimal comprehensive cost as the goal, and is more in line with the coal blending cost control requirements of the coking plant.

[0024] The above steps S101 to S104 are explained and illustrated below respectively.

[0025] Step S101, determining the objective function according to the wet basis proportion, moisture, ash content, volatile matter, wet basis unit transportation cost and dry basis volatile matter of coke of each single type of coal among the multiple single types of coal.

[0026] Wherein, the objective function includes a raw material cost function and a storage cost function.

[0027] For step S101, the establishment of the target optimization function includes: determining the dry basis ratio of each single variety of coal according to the wet basis ratio and moisture of each single variety of coal; determining the coking rate according to the dry basis ratio, ash and volatile matter of each single variety of coal and the dry basis volatile matter of the coke; determining the wet basis demand of each single variety of coal according to the wet basis ratio and moisture of each single variety of coal, the coking rate and the coke demand; determining the raw material cost function according to the wet basis demand, wet basis unit transportation cost and wet basis ex-factory unit price of each single variety of coal; determining the storage cost function according to the inventory quantity, wet basis demand and unit inventory maintenance cost of each single variety of coal; and establishing the target optimization function according to the raw material cost function and the storage cost function.

[0028] The raw coal inventory of the coking plant will form storage costs, and the actual cost of raw coal is also related to the unit transportation cost. According to the research on the inventory cost structure of the raw material plant, the influence of common and uncertain cost factors such as warehousing and out-of-warehouse is removed, and the storage cost is mainly inventory maintenance cost, which is related to the raw material inventory.

[0029] Let the number of raw coal types (differentiated by manufacturer) in the coal blending plan be n (i.e. the number of single-variety coals is n), and the wet basis ratio of the i-th single-variety coal is x i , water content is h i First, the dry basis ratio of the i-th single coal is determined by the following formula (1): for:

[0030]

[0031] Next, the coking rate k is determined by the following formula (2): a :

[0032]

[0033] Among them, a Vdaf,i represents the Vdaf (volatile matter) value of the i-th raw coal, a Ad,i represents the value of Ad of the i-th raw coal, V dj It represents the value of the volatile matter on a dry basis of coke, and Q is a correction factor (e.g., 1%).

[0034] Then, the wet basis demand q for the i-th raw coal is determined by the following formula (3): i (x i ):

[0035]

[0036] Here, q0 is the demand for coke (the unit is, for example, tons).

[0037] Further, the determining of the raw material cost function includes: according to the wet basis demand q of the i-th single type of coal i (x i ), wet basis unit transportation cost C T,i (The unit is, for example, yuan / ton) and the wet base ex-factory unit price C i (unit is, for example, yuan / ton) and the following formula (4) to determine the raw material cost function

[0038]

[0039] Where n is the number of single type of coal.

[0040] Furthermore, the storage cost function is determined by: according to the inventory quantity s of the i-th single type of coal i and wet basis requirement q i (x i ), unit inventory maintenance cost c0 and the following formula to determine the storage cost function

[0041]

[0042] Where n is the number of single-type coals. ∈(x) is the Relu function, that is, ∈(x) = max(x, 0).

[0043] Finally, the objective function expressed by the following formula (6) can be established:

[0044]

[0045] Of course, in other embodiments, the two cost functions mentioned above may be weighted and summed, and the specific weights may be reasonably set according to actual needs, but are generally close to 1.

[0046] Before executing step S101, in the first type of embodiment, the coking coal blending method also includes: using the RANSAC algorithm to determine the quality function relationship between the single type coal and the mixed coal for each of the first influencing parameter set; and using the RANSAC algorithm to determine the quality function relationship between the single type coal and the coke for each of the second influencing parameter set.

[0047] Among them, the first influencing parameter set includes: Ad (ash content), Vdaf (volatile matter), Stad (sulfur content), G (bonding index), X (final shrinkage of the colloid layer), and Y (maximum thickness of the colloid layer); the second influencing parameter set includes: Ad (ash content), Vdaf (volatile matter), Stad (sulfur content), M10 (wear resistance), M40 (crushing strength), CRI (reactivity) and CSR (strength after reaction).

[0048] Specifically, the RANSAC algorithm is used to determine the quality function relationship between the single-variety coal and the mixed coal for each of the first influencing parameter set, including: S1: normalizing a plurality of small coke oven test data groups, wherein each of the plurality of small coke oven test data groups includes a plurality of single-variety coal ratios, six-dimensional data corresponding to each single-variety coal under each single-variety coal ratio, and the experimental results of each dimensional data of the mixed coal under each single-variety coal ratio, and the six-dimensional data includes ash content, volatile matter, sulfur content, bonding index, final shrinkage of the gelatinous layer and maximum thickness of the gelatinous layer; S2: according to Randomly select at least 3 test data from a test data group from multiple small coke oven test data groups to determine the corresponding plane; S3: according to the corresponding plane, traverse each test data in the selected test data group respectively, and judge whether the distance from each test data to the corresponding plane is less than a threshold value. If so, determine the corresponding test data as an inlier and count the number of inliers; and S4, repeat S2-S3 until the set number of experiments is reached, select the test data group containing the most inliers, and fit the quality function relationship between the single-variety coal and the mixed coal for each of the first influencing parameter set.

[0049] Among them, the quality function relationship between the single-variety coal and the mixed coal for each of the first influencing parameter sets is a linear relationship, such as shown in formula (7).

[0050] Specifically, the RANSAC algorithm is used to determine the quality function relationship between the single-variety coal and the coke for each of the second influencing parameter sets, including: S1: normalizing a plurality of small coke oven test data groups, wherein each of the plurality of small coke oven test data groups includes a plurality of single-variety coal ratios, six-dimensional data corresponding to each single-variety coal under each single-variety coal ratio, and experimental results of each dimensional data of the mixed coal under each single-variety coal ratio, and the six-dimensional data includes ash content, volatile matter, sulfur content, bonding index, final shrinkage of the gelatinous layer and maximum thickness of the gelatinous layer; S2: according to Randomly select at least 3 test data from a test data group from multiple small coke oven test data groups to determine the corresponding plane; S3: according to the corresponding plane, traverse each test data in the selected test data group respectively, and judge whether the distance from each test data to the corresponding plane is less than a threshold value. If so, determine the corresponding test data as an inlier and count the number of inliers; and S4, repeat S2-S3 until the set number of experiments is reached, select the test data group containing the most inliers, and fit the mass function relationship between single-variety coal and coke for each of the second influencing parameter set.

[0051] Among them, the quality function relationship between a single type of coal and coke for each of the second influencing parameter sets is a linear relationship, such as shown in formula (8).

[0052] The RANSAC algorithm (Random Sample Consensus) is an iterative method for estimating the parameters of a mathematical model from an observed data set that contains outliers (or outliers). The algorithm randomly selects a portion of the data points, fits a model based on these points, and then statistically analyzes the deviations between the model and other data points, and finally selects data points that meet a certain threshold to estimate the model parameters. The RANSAC algorithm is highly robust to noisy data and outliers, and can effectively estimate the model parameters.

[0053] In order to avoid noise caused by inaccurate individual experimental results, improve the rationality of sample data, and improve the robustness of the mixed coal quality prediction method, this embodiment takes a whole set of experimental results as a sample. Considering that there is a certain correlation between the main quality inspection items Ad, Vdaf, Stad, G, X, and Y of single-variety coal and mixed coal, rather than being independent of each other, this method is not limited to the independent relationship mapping of quality inspection items, but takes the overall experimental results into consideration as a whole. The n quality inspection items form an n+1-dimensional data verification space. The RANSAC algorithm is used to perform interior point selection and function relationship fitting in the n+1-dimensional space. When one or more quality inspection items are abnormal beyond a reasonable range, the sample is considered to be abnormal, and the sample is not considered as a basis for prediction. Therefore, it is considered that this group of samples is abnormal, thereby improving the accuracy of the overall experimental data.

[0054] Figure 2 The specific process of determining the blended coal quality prediction model is demonstrated.

[0055] According to the small coke oven test data, the mixed coal quality model is predicted by the RANSAC algorithm. Because the numerical dimensions of each quality inspection item are quite different, the data can be normalized by the Z-score standardization method. The normalization method of the quality inspection results of each type of coal (expressed by item, including Ad, Vdaf, Stad, G, X, Y, etc.) of the mixed coal prediction model is as follows:

[0056]

[0057] in is the normalized single-variety coal item quality inspection result data, a item,i is the quality inspection result of the item quality inspection item of the i-th single variety of coal in the group (i.e., the quality inspection result of the item quality inspection item of the i-th single variety of coal on a batching list), μ item is the expected quality inspection result of the i-th single-variety coal in each group of item quality inspection items, σitem It is the standard deviation of the quality inspection results of the i-th single variety of coal in each group of item quality inspection items.

[0058]

[0059] in is the normalized quality inspection result data of mixed coal item, a coal,item is the quality inspection result of the mixed coal quality inspection item item (i.e. the experimental result of the mixed coal under any ratio for each dimension data), μ coal,item is the expected quality inspection result of each group of item, σ item is the standard deviation of the quality inspection items of each group of items.

[0060] Since the regression relationships between the quality inspection items of single-variety coal and mixed coal in each set of data are independent of each other, a set of data can determine the quality regression relationship between single-variety coal and mixed coal. For example, the algorithm removes noise data with an accuracy rate of z = 99.95% or more as the goal, and the noise data accounts for 1-p = 20%. The number of algorithm iterations and the number of iterations of the mixed coal regression relationship are initialized respectively:

[0061]

[0062] Randomly select a set of experimental data (i.e. coal blending data), calculate the distance between its position in n+1 dimensional space and the position of other experimental results in space, which is less than or equal to the threshold d stad The points are recorded as interior points. Through iterative calculation, a set of experimental data with the most interior points in each quality inspection item is selected to perform linear regression fitting to solve δ coal,item .

[0063] The experimental data of small coke ovens were collected, and the quality of single-variety coal and mixed coal was regressed using the Ransac algorithm. Because the mixing process of single-variety coal mainly undergoes physical changes, the main index relationships between mixed coal and each single-variety coal at any ratio are as follows:

[0064]

[0065] where a coal,Ad , a coal,Vdaf , a coal,S , a coal,G ,a coal,X ,a coal,Y are Ad, Vdaf, Stad, G, X, and Y of the mixed coal respectively (which may be normalized values ​​or unnormalized values); a Ad,i , a Vdaf,i , a S,i , a G,i , a X,i , aY,i are Ad, Vdaf, Stad, G, X, and Y of the i-th single coal variety (which can be normalized values ​​or unnormalized values); δ coal,Ad , δ coal,Vdaf , δ coal,S , δ coal,G , δ coal,X , δ coal,Y are respectively the predicted correction values ​​of mixed coal quality, is the dry basis ratio of the i-th single coal variety.

[0066] Figure 3 The specific process of determining the coke quality prediction model is demonstrated.

[0067] The normalization method of the quality inspection results of each type of coal in the coke prediction model (expressed by item, including Ad, Vdaf, Stad, M10, M40, CRI, CSR, etc.) is as follows:

[0068]

[0069] in is the normalized single-variety coal item quality inspection result data, a item,i is the quality inspection result of the item of the i-th single-variety coal in this group, μ item is the expected quality inspection result of the i-th single-variety coal in each group of item quality inspection items, σ item It is the standard deviation of the quality inspection results of the i-th single variety of coal in each group of item quality inspection items.

[0070]

[0071] in is the normalized coke item quality inspection result data, a coal,item is the quality inspection result of the coke quality inspection item item (i.e. the experimental result of coke under any ratio for each dimension data), μ coal,item is the expected quality inspection results of each group of coke items, σ item is the standard deviation of the quality inspection items of each group of coke items.

[0072] Since the regression relationships between the quality inspection items of single-variety coal and coke in each set of data are independent of each other, the two sets of data can determine the quality regression relationship between single-variety coal and coke. For example, the algorithm removes noise data with an accuracy rate of z = 99.95% or more as the goal, and the noise data accounts for 1-p = 20%. The number of algorithm iterations and the number of iterations of the coke regression relationship are initialized respectively:

[0073]

[0074] Randomly select two sets of experimental data, calculate the distance between the straight line formed by them in the n+1-dimensional space and the point positions of various other experimental results in the space, which is less than or equal to the threshold d stad The points are recorded as interior points. Through iterative calculation, a set of experimental data with the most interior points in each quality inspection item is selected to solve ρ coke,item , δ coke,item .

[0075] The experimental data of small coke ovens were collected, and the quality of single coal and coke was regressed by Ransac algorithm. The nonlinear relationship between the Ad, Vdaf, Stad and other components of mixed coal and coke was initially used as the data screening condition to establish the corresponding correlation between the Ad, Vdaf, Stad and other components of single coal and the Ad, Vdaf, Stad and other components of coke. Therefore, the main index relationships between coke and each single coal under any ratio are as follows:

[0076]

[0077] where a coke,Ad , a coke,Vdaf , a coke,S , a coke,M10 , a coke,M40 , a coke,CRI , a coke,CSR Respectively represent Ad, Vdaf, Stad, M10, M40, CRI, CSR of coke (which can be normalized values ​​or unnormalized values); a Ad,i , a Vdaf,i , a S,i , a M10,i , a M40,i , a CRI,i 、a CSR,i are Ad, Vdaf, Stad, M10, M40, CRI, and CSR of the i-th single coal variety (which can be normalized values ​​or unnormalized values); ρ coke,Ad , ρ coke,Vdaf , ρ coke,S , ρ coke,M10 , ρ coke,M40 , a coke,CRI , a coke,CSR Respectively represent the predicted proportion correction coefficients of Ad, Vdaf, Stad, M10, M40, CRI, and CSR of coke quality; δ coke,Ad , δ coke,Vdaf , δ coke,S , δ coke,M10 , δ coke,M40 , δ coke,CRI , δ coke,CSR Respectively represent the predicted correction values ​​of Ad, Vdaf, Stad, M10, M40, CRI, and CSR of coke quality. is the dry basis ratio of the i-th single-variety coal, k a Indicates the coking rate.

[0078] Before executing step S101, in the second class of embodiments, when the second influencing parameter set includes a first subset and a second subset, the first subset includes ash, volatile matter and sulfur content, and the second subset includes wear resistance, crushing strength, reactivity and post-reaction strength, the coking coal blending method also includes: using the RANSAC algorithm to determine the mass function relationship between single-variety coal and mixed coal for each of the first influencing parameter set; and using the RANSAC algorithm to determine the mass function relationship between single-variety coal and coke for each of the first subset and using a machine learning method or a statistical regression method to determine the mass function of single-variety coal and coke for each of the second subset.

[0079] The specific process of using the RANSAC algorithm to determine the quality function relationship between the single-variety coal and the mixed coal for each of the first influencing parameter set can be found in the relevant description in the previous embodiment, which will not be repeated here.

[0080] Wherein, the machine learning method may include a BP neural network.

[0081] Specifically, in the case where the second influencing parameter set includes the first subset and the second subset, first, the RANSAC algorithm is used to determine the mass function relationship between the single-variety coal and the coke for each of the first subsets, including: S1: normalizing a plurality of small coke oven test data groups, wherein each test data group in the plurality of small coke oven test data groups includes a plurality of single-variety coal ratios, three-dimensional data corresponding to each single-variety coal under each single-variety coal ratio, and experimental results of each dimensional data of the mixed coal under each single-variety coal ratio, and the three-dimensional data includes ash content, volatile matter, and sulfur content; S2: Determine a corresponding plane based on at least 3 test data in a test data group randomly selected from multiple small coke oven test data groups; S3: According to the corresponding plane, traverse each test data in the selected test data group respectively, and judge whether the distance from each test data to the corresponding plane is less than a threshold value. If so, determine the corresponding test data as an inlier and count the number of inliers; and S4, repeat S2-S3 until the set number of experiments is reached, select the test data group containing the most inliers, and fit the quality function relationship between the single-variety coal and the mixed coal for each of the first subset.

[0082] Secondly, a BP neural network is used to determine the quality function of a single type of coal and coke for each of the second subsets, including: training the BP neural network through a training set to determine the quality function of a single type of coal and coke for each of the second subsets, wherein the training set includes multiple training groups, each training group including: a theoretical value of each of a first influencing parameter set of a single type of coal under any ratio, an actual value of each of a first influencing parameter set of a mixed coal formed by a batch small coke oven test using any of the ratios, an actual value of each of the first subset of coke formed by the batch small coke oven test using any of the ratios, and an actual value of each of the coke formed by actual production using any of the ratios for the second subset.

[0083] Among them, the quality function relationship between a single type of coal and coke for each of the first subsets in the second influencing parameter set is a linear relationship, such as shown in formula (9); the quality function relationship between a single type of coal and coke for each of the second subsets in the second influencing parameter set is a nonlinear relationship, such as shown in formula (10).

[0084] The coke prediction model is mainly divided into two parts due to the different relationships between quality inspection items. Considering that there is a certain correlation between the main quality inspection items Ad, Vdaf, and Stad of single-variety coal and coke, rather than being independent of each other, this embodiment is not limited to the independent relationship mapping of quality inspection items, but takes the overall experimental results into consideration as a whole. The n quality inspection items form an n+1-dimensional data verification space, and the RANSAC algorithm is used to perform interior point selection and function relationship fitting in the n+1-dimensional space.

[0085] Figure 4 The specific process of determining the coke quality prediction model is demonstrated.

[0086] The normalization method of the quality inspection results of each type of coal in the coke prediction model (expressed by item, including Ad, Vdaf, Stad) is as follows:

[0087]

[0088] in is the normalized single-variety coal item quality inspection result data, a item,i is the quality inspection result of the item of the i-th single-variety coal in this group, μ item is the expected quality inspection result of the i-th single-variety coal in each group of item quality inspection items, σ item It is the standard deviation of the quality inspection results of the i-th single variety of coal in each group of item quality inspection items.

[0089]

[0090] in is the normalized coke item quality inspection result data, a coal,item is the quality inspection result of the coke quality inspection item item (i.e. the experimental result of coke under any ratio for each dimension data), μ coal,item is the expected quality inspection results of each group of coke items, σ item is the standard deviation of the quality inspection items of each group of coke items.

[0091] Since the regression relationships between the quality inspection items of single-variety coal and coke in each set of data are independent of each other, the two sets of data can determine the quality regression relationship between single-variety coal and coke. For example, the algorithm removes noise data with an accuracy rate of z = 99.95% or more as the goal, and the noise data accounts for 1-p = 20%. The number of algorithm iterations and the number of iterations of the coke regression relationship are initialized respectively:

[0092]

[0093] Randomly select two sets of experimental data, calculate the distance between the straight line formed by them in n+1 dimensional space and the point positions of various other experimental results in space, which is less than or equal to the threshold d stad The points are recorded as interior points. Through iterative calculation, a set of experimental data with the most interior points in each quality inspection item is selected to solve ρ cok,item , δ coke,item .

[0094] The prediction of coke M10, M40, CRI, and CSR is based on the data set corresponding to the internal point of the best fitting data selected in the above steps (including the mixed coal composition corresponding to the batch of small coke oven experiments, and the quality data of the coke actually produced). The BP neural network is trained with the data set as a sample, and the nonlinear functional relationship of coke M10, M40, CRI, and CSR is fitted with the BP neural network. The Ad, Vdaf, Stad, G, X, and Y of the mixed coal in the batch of mixed coal and the Ad, Vdaf, Stad, G, X, and Y of the corresponding batch of small coke oven experiments, as well as the Ad, Vdaf, and Stad experimental results of the coke in the batch of small coke oven experiments corresponding to the batch of mixed coal in ...

[0095] Initially, the nonlinear relationship between the Ad, Vdaf, Stad and other components of mixed coal and coke is used as the data screening condition to establish the correlation between the Ad, Vdaf, Stad of single-variety coal and the Ad, Vdaf, Stad of coke under any ratio:

[0096]

[0097] Among them, acoke,Ad , a coke,Vdaf , a coke,S Respectively represent Ad, Vdaf, Stad of coke (which can be normalized values ​​or unnormalized values); a Ad,i , a Vdaf,i , a S,i are Ad, Vdaf, Stad of the ith single coal variety (which can be normalized values ​​or unnormalized values); ρ coke,Ad , ρ coke,Vdaf , ρ coke,S Respectively represent the Ad, Vdaf, Stad prediction ratio correction coefficients of coke quality; δ coke,Ad , δ coke,Vdaf , δ coke,S They represent the predicted correction values ​​of Ad, Vdaf and Stad of coke quality respectively; k is the dry basis ratio of the i-th single coal variety; a Indicates the coking rate.

[0098] The relationship between coke performance and mixed coal performance, in order to establish a more accurate regression relationship, the coal blending model is refined, and the nonlinear functional relationship of coke M10, M40, CRI, and CSR is fitted by BP neural network. The Ad, Vdaf, Stad, G, X, and Y of the mixed coal in the batch list and the Ad, Vdaf, Stad, G, X, and Y of the corresponding batch of small coke oven experiments, as well as the Ad, Vdaf, and Stad experimental results of the coke in the batch of small coke oven experiments corresponding to the batch list are used as network inputs, and the M10, M40, CRI, and CSR of the coke produced in the actual coke oven are used as network outputs. The network input and output under any ratio are equivalent to the same training group.

[0099] The functional relationship can be expressed as

[0100]

[0101] Among them, y coke,M10 ,y coke,M40 ,y coke,CRI ,y coke,CSR Respectively represent the M10, M40, CRI, and CSR of the coke actually produced by the coke oven; x coal,Ad ,x coal,Vdaf ,x coal,S ,x coal,G ,x coal,X ,x coal,Y They represent the theoretical prediction results of Ad, Vdaf, Stad, G, X, and Y of the single-batch mixed coal respectively; Indicates the experimental results of Ad, Vdaf, Stad, G, X, and Y of the mixed coal corresponding to the batch of small coke oven experiments in the batching list; It indicates the experimental results of coke Ad, Vdaf and Stad of the small coke oven experiment of the corresponding batch in the batching list.

[0102] The actual coking production data and the corresponding small coke oven experimental data are collected as samples for training.

[0103] In this embodiment, the quality relationship prediction is mainly divided into mixed coal quality prediction and coke quality prediction. The coke quality prediction is further divided into Ad, Vdaf, Stad prediction and M10, M40, CRI, CSR prediction according to the objective characteristics of its main quality inspection items. The small coke oven experimental sample data is considered as a whole to form a high-dimensional space, and the RANSAC algorithm is used to reduce noise and fit the prediction model of the mixed coal quality based on the quality of a single coal variety; the small coke oven experimental sample data is considered as a whole to form a high-dimensional space, and the RANSAC algorithm is used to reduce noise and fit the prediction model of the mixed coal quality based on the quality of a single coal variety. af, Stad for the prediction model of coke Ad, Vdaf, Stad, and then use the data of the interior points selected by the RANSAC algorithm, select the corresponding small coke oven experimental data, actual coking production data, and ingredient list data as samples, combine the prediction model of coke Ad, Vdaf, Stad, and use BP neural network to fit the M10, M40, CRI, CSR prediction model of coke. After convergence, combine the prediction model of coke Ad, Vdaf, Stad to form a complete prediction model for coke Ad, Vdaf, Stad.

[0104] Step S102, determining the first constraint condition for each of the first influencing parameter set according to the wet basis ratio and moisture content of each single type of coal and the mass function relationship between the single type of coal and the mixed coal for each of the first influencing parameter set.

[0105] For step S102, determining the first constraint condition for each of the first influencing parameter set includes: determining the intercept in the quality function relationship between the single-variety coal and the mixed coal for each of the first influencing parameter set as the mixed coal quality prediction correction value for each of the first influencing parameter set; and determining the first constraint condition for each of the first influencing parameter set based on the wet basis ratio and moisture content of each single-variety coal and the mixed coal quality prediction correction value for each of the first influencing parameter set.

[0106] Specifically, the following first constraint is established:

[0107]

[0108] Among them l lb,coal,item is the lower limit of the quality inspection item of the target mixed coal item, l ub,coal,item is the upper limit of the quality inspection item of the target mixed coal item; coal,itemis the intercept in the quality function relationship between single-variety coal and mixed coal for item; a item,i is the quality inspection result of item quality inspection of the i-th single type coal in the group (it can be a normalized value or an unnormalized value); x i 、h i are the wet basis ratio and moisture content of the i-th single variety coal respectively.

[0109] Step S103, determining the second constraint condition for each of the second influencing parameter set according to the wet basis ratio and moisture content of each single type of coal and the mass function relationship between the single type of coal and coke for each of the second influencing parameter set.

[0110] In the first type of embodiment, step S103 is performed on the basis of using the RANSAC algorithm to determine the quality function relationship between the single type of coal and coke for each of the second influencing parameter set. Specifically, the second constraint condition for each of the second influencing parameter set is determined as follows: the intercept in the quality function relationship between the single type of coal and coke for each of the second influencing parameter set is determined as the predicted coke quality correction value for each of the second influencing parameter set; and the proportional correction coefficient for each of the second influencing parameter set is determined according to the slope in the quality function relationship between the single type of coal and coke for each of the second influencing parameter set; and the second constraint condition for each of the second influencing parameter set is determined according to the wet basis ratio and moisture content, coking rate of each single type of coal, and the predicted coke quality correction value and proportional correction coefficient for each of the second influencing parameter set.

[0111] In the first type of embodiment, regardless of the first set of influencing parameters or the second set of influencing parameters, for each influencing parameter, the process of determining the quality function relationship between a single type of coal and mixed coal / coke is the same (for example, both use the RANSAC algorithm), and the process of determining the corresponding constraint conditions is also similar.

[0112] Specifically, the following second constraint is established:

[0113]

[0114] Among them, l lb,coke,item is the lower limit of the quality inspection item of the target coke item, l ub,coke,item is the upper limit of the quality inspection item of the target coke item; ρ coke,item is the slope of the mass function relationship between single-variety coal and coke for item (i.e., proportional correction coefficient); coke,item is the intercept in the quality function relationship between single-variety coal and coke for item (i.e., the predicted correction value of coke quality); aitem,i is the quality inspection result of item quality inspection of the i-th single type coal in the group (it can be a normalized value or an unnormalized value); x i 、h i are the wet basis ratio and moisture content of the i-th single coal variety; k a Indicates the coking rate.

[0115] In the second type of embodiment, step S103 is performed on the basis of using the RANSAC algorithm to determine the quality function relationship between a single type of coal and coke for each of the first subset in the second influencing parameter set and using the BP neural network to determine the quality function relationship between a single type of coal and coke for each of the second subset in the second influencing parameter set. Specifically, the determination of the second constraint condition for each of the second influencing parameter set includes: determining the second constraint condition for each of the first subset in the following manner: determining the second constraint condition for each of the second subset in the following manner.

[0116] Specifically, the second constraint condition for each of the first subset is determined in the following manner, including: determining the intercept in the mass function relationship between the single type of coal and coke for each of the first subset as the predicted coke quality correction value for each of the first subset; and determining the proportional correction coefficient for each of the first subset based on the slope in the mass function relationship between the single type of coal and coke for each of the first subset; and determining the second constraint condition for each of the first subset based on the wet basis ratio and moisture content, coking rate of each single type of coal, and the predicted coke quality correction value and proportional correction coefficient for each of the first subset.

[0117] The second constraint condition for each of the second subset is determined in the following manner, including: determining the second constraint condition for each of the second subset based on the quality function of the single type coal and coke for each of the second subset and the upper and lower limits of each of the second subset.

[0118] Specifically, the following second constraint is established:

[0119]

[0120] Among them, l lb,coke,item is the lower limit of the quality inspection item of the target coke item, l ub,coke,item is the upper limit of the quality inspection item of the target coke item; ρ coke,item is the slope of the mass function relationship between single-variety coal and coke for item (i.e., proportional correction coefficient); coke,itemis the intercept in the quality function relationship between single-variety coal and coke for item (i.e., the predicted correction value of coke quality); a item,i is the quality inspection result of item quality inspection of the i-th single type coal in the group (it can be a normalized value or an unnormalized value); x i 、h i are the wet basis ratio and moisture content of the i-th single coal variety; k a Indicates the coking rate; is the regression function relationship of the coke quality M10, M40, CRI, CSR neural network; l lb,coke,M10 ,l lb,coke,M40 ,l lb,coke,CRI ,l lb,coke,CSR They are the lower limits of target coke M10, M40, CRI, and CSR quality inspection items respectively; ub,coke,M10 ,l ub,coke,M40 ,l ub,coke,CRI ,l ub,coke,CSR They are the upper limits of target coke M10, M40, CRI and CSR quality inspection items respectively.

[0121] In one embodiment, the plurality of constraints further include: a third constraint, and the third constraint is determined according to the coking rate.

[0122] Specifically, according to the coking rate k a Establish the following third constraint:

[0123] k a ≥k a,lb

[0124] Among them, k a,lb is the lower limit of the target coking rate.

[0125] For the first type of example:

[0126]

[0127] in:

[0128]

[0129] where a Vdaf,i represents Vdaf of the i-th raw coal, a Ad,i represents Ad, V of the i-th raw coal dj represents the volatile matter on dry basis of coke, C i is the wet-basis ex-factory price of a single coal variety (yuan / ton), q0 is the demand for coke (tons), C T,i is the unit transportation cost on a wet basis (yuan / ton), s irepresents the wet basis inventory of the i-th type of raw coal, c0 represents the unit inventory maintenance cost, ∈(x) is the Relu function, that is, ∈(x)=max(x,0).

[0130] Constraints:

[0131]

[0132] k a ≥k a,lb

[0133] Among them l lb,coal,item is the lower limit of the quality inspection item of the target mixed coal item, l ub,coal,item is the upper limit of the quality inspection item of the target mixed coal item, l lb,coke,item is the lower limit of the quality inspection item of the target coke item, l ub,coke,item is the upper limit of the quality inspection item of the target coke item, k a,lb is the lower limit of the target coking rate, δ coal,item , coke,item , δ coke,item To solve the predicted relationship coefficient value of the quality inspection items of mixed coal and coke item.

[0134] This embodiment establishes the relationship between factors such as raw fuel inventory and transportation and the comprehensive coal blending cost, establishes a mathematical model of the comprehensive coal blending optimization model, uses a strong robustness method to effectively regress the quality relationship between single-variety coal and mixed coal and coke, and solves the meta-heuristic comprehensive algorithm for the optimal control problem of the nonlinear comprehensive coal blending optimization model (see step S104 for details), while ensuring the stability and speed of the algorithm.

[0135] For the second type of embodiment:

[0136] The above constraints can be replaced with the following:

[0137]

[0138] The overall algorithm of this coal blending method first uses the experimental data of a small coke oven, and adopts the RANSAC algorithm to denoise and regress the quality relationship between single-variety coal and mixed coal, and regress the quality relationship between mixed coal and coke Ad, Vdaf, and Stad. The RANSAC algorithm is used to denoise and select the batching list corresponding to the internal point of the optimal fitting relationship, the actual production and small coke oven experimental data as samples, and the neural network is trained to regress the quality relationship between mixed coal and coke M10, M40, CRI, and CSR; combined with the minimum limit on the coking rate, the constraint conditions for coal blending calculation are formed. In the process of coal blending calculation, the quality prediction results of mixed coal and coke are calculated based on individual inputs, and the comprehensive cost objective function integrating coal blending consumption cost, storage cost, and raw material cost is established. The meta-heuristic target optimization algorithm is used to solve it and the current optimal comprehensive cost coal blending plan is obtained.

[0139] Step S104, under multiple constraints, a metaheuristic algorithm is used to optimize the objective function to determine the wet basis ratio of each single type of coal, wherein the multiple constraints include a first constraint for each of the first influencing parameter set and a second constraint for each of the second influencing parameter set.

[0140] Because the nonlinear calculation of the model is complex and difficult to solve with analytical algorithms, the metaheuristic algorithm combines the random algorithm with the local search algorithm, which can balance the local optimal solution and the global optimal solution, thereby effectively solving the combinatorial optimization problem. This coal blending method adopts the metaheuristic target optimization algorithm. According to the speed and applicability of the algorithm, the algorithm uses the non-biological swarm intelligence algorithm to solve the coal blending scheme. The Fireworks Algorithm (FWA) is a new type of swarm intelligence optimization algorithm. It was originally used to solve the global optimization problem of composite functions. It has strong global and local search capabilities, and has the characteristics of easy implementation, distributed computing mechanism, strong robustness, good scalability and randomness.

[0141] Among them, the meta-heuristic algorithms include: particle swarm, genetic algorithm, fireworks algorithm, simulated annealing algorithm, ant colony algorithm, marine predator algorithm and machine learning algorithm.

[0142] Take the fireworks algorithm as an example to solve the optimal coal blending solution, such as Figure 5 shown.

[0143] Randomly initialize N fireworks explosion positions x i ,i∈1,2,…,N(coal blending ratio of single coal The position of the point in n-dimensional space), x i Number of fireworks that explode at this location:

[0144]

[0145] in is the spark number limit, f max is the maximum fitness value of the current fireworks, f(x i ) is x i The fitness value of the position, ε is a minimum constant.

[0146] x i Location fireworks explosion amplitude:

[0147]

[0148] in is the explosion amplitude limit of fireworks, f minis the maximum fitness value of each firework in the current iteration, f(x i ) is x i The fitness value of the position, ε is a minimum constant.

[0149] Determine the location of sparks after a firework explosion by displacement:

[0150]

[0151] Δx i p is the position of the pth spark of the ith firework, let p∈1,2,…,S i , σ i,p =rand(0,A i ), i∈1,2…n, from 0 to A i Uniform random number.

[0152] Perform Gaussian mutation on the random dimension (jth dimension) of randomly selected fireworks:

[0153] x j i =x j i *σ g

[0154] σ g is a random number that follows N(1,1) (N(1,1) is a Gaussian distribution with a mean of 1 and a mean square deviation of 1).

[0155] If the mutation exceeds the bounds, the position is updated through the modulo operation mapping:

[0156] x j i =x j min +|x j i |%(x j max -x j min )

[0157] x j max , x j min for The upper and lower bounds in the j dimension, % represents the modulo operation.

[0158] Use the roulette method to select the spark as the explosion location of the next firework, the selection probability is:

[0159]

[0160] where x ip represents the position of the pth spark of the ith firework, R(x i p ) represents the Euclidean distance between the jth spark of the ith firework and the other sparks of the ith firework:

[0161]

[0162] The coal blending model of the present invention fully considers the coal blending raw material consumption cost, storage cost, unit transportation cost, and the role of dry basis ratio and wet basis ratio at different positions of the objective function. Ad (ash content), Vdaf (volatile matter), Stad (sulfur content), G, X, Y and other components or properties are used as constraints for mixed coal production. Ad (ash content), Stad (sulfur content), M40, M10, CSR and CRI of coke are used as constraints for coke production. The overall algorithm takes the optimal comprehensive cost as the goal. The overall algorithm first uses the experimental data of a small coke oven as a sample, and uses the RANSAC linear regression method to fit the Ad, Vdaf, Stad, G, X, and Y functional relationship between the quality of a single coal variety and a mixed coal. The RANSAC nonlinear regression method is used to fit the relationship between the quality of a single coal variety and a coke variety. The Ad, Vdaf, and Stad functional relationship of the quality is selected through RANSAC algorithm denoising. The batching list corresponding to the internal point of the optimal fitting relationship, the actual production and small coke oven experimental data are used as samples. The neural network is trained to regress the quality relationship between mixed coal and coke M10, M40, CRI, and CSR. According to the quality, unit price, unit transportation cost, inventory and other information of each single variety of coal, the target optimization function is established. According to the quality control requirements of mixed coal and coke and the functional relationship between the quality of single variety of coal and mixed coal and coke, the constraint conditions are established. The solution is solved by intelligent target optimization algorithm (including but not limited to particle swarm, genetic algorithm, fireworks algorithm, simulated annealing algorithm, ant colony algorithm, marine predator algorithm, machine learning algorithm, etc.), and the optimal coal blending solution is obtained, such as Figure 6 shown.

[0163] In the specific application and implementation of the coal blending center of the coking plant, the specific steps are as follows:

[0164] Establish a coking intelligent coal blending optimization system based on this comprehensive optimization algorithm, connect to the inspection and testing system to collect small coke oven experimental data, including single coal Ad, Vdaf, Stad, G, X, Y, M10, M40, CRI, CSR, mixed coal Ad, Vdaf, Stad, G, X, Y, coke Ad, Vdaf, Stad, M10, M40, CRI, CSR, connect to the MES system to collect raw coal inventory data, connect to the ERP system to collect raw coal price data, and maintain the current inventory maintenance cost coefficient and interest rate coefficient according to actual conditions. Coal blending personnel set target mixed coal, coke composition control constraints and minimum coking rate requirements.

[0165] The coking intelligent coal blending optimization system regresses the quality relationship between mixed coal and coke Ad, Vdaf, and Stad according to the input, selects the corresponding internal point of the optimal fitting relationship through RANSAC algorithm denoising, selects the corresponding ingredient list, actual production and small coke oven experimental data as samples, and trains the neural network to regress the quality relationship between mixed coal and coke M10, M40, CRI, and CSR to form a mixed coal. The coal blending plan with the optimal comprehensive cost is automatically calculated. The results are verified to meet the target control requirements and component constraints. Compared with traditional coal blending and existing coal blending calculation models, it can be seen that the intelligent coal blending system using this comprehensive optimization algorithm can effectively and significantly reduce the comprehensive coal blending cost of coking.

[0166] The present invention belongs to an optimization method for a coking comprehensive cost coal blending model, adopts a RANSAC algorithm to perform denoising and regress the quality relationship between a single type of coal and mixed coal, regresses the quality relationship between a single type of coal and coke, combines the minimum limit on the coking rate, thereby forming constraint conditions for coal blending calculation, establishes a comprehensive cost objective function integrating coal blending consumption cost, storage cost and transportation cost, adopts a meta-heuristic target optimization algorithm to solve, and applies the algorithm to be implemented in a computer software system, so as to help coking plants make decisions on coking coal blending and raw coal procurement and provide guidance.

[0167] In summary, the present invention creatively and comprehensively considers multiple factors such as wet basis ratio, moisture, ash, volatile matter, wet basis unit transportation cost and coke dry basis volatile matter, establishes a coal blending optimization model including raw material cost and storage cost, and uses a meta-heuristic algorithm to ensure the stability and rapidity of the entire coking coal blending process, so that the coking coal blending ratio can be determined efficiently and accurately.

[0168] An embodiment of the present invention provides a computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the coking coal blending method is implemented.

[0169] An embodiment of the present invention provides a computer program product, including a computer program, which implements the coking coal blending method when executed by a processor.

[0170] The preferred embodiments of the present invention are described in detail above in conjunction with the accompanying drawings. However, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, a variety of simple modifications can be made to the technical solution of the present invention, and these simple modifications all belong to the protection scope of the present invention.

[0171] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations.

[0172] Those skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a program, and the program is stored in a storage medium, including a number of instructions to enable a single-chip microcomputer, a chip or a processor to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0173] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0174] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0175] In addition, various embodiments of the present invention may be arbitrarily combined, and as long as they do not violate the concept of the present invention, they should also be regarded as the contents disclosed by the present invention.

Claims

1. A method for coking coal blending, characterized in that: The coking coal blending method comprises: Determine the objective function according to the wet basis proportion, moisture, ash, volatile matter, wet basis unit transportation cost and coke dry basis volatile matter of each single type of coal in the multiple single types of coal, wherein the objective function includes a raw material cost function and a storage cost function; Determine a first constraint condition for each of the first influencing parameter sets according to the wet basis ratio and moisture content of each single type of coal and the mass function relationship between the single type of coal and the mixed coal for each of the first influencing parameter sets; Determine a second constraint condition for each of the second influencing parameter sets according to the wet basis ratio and moisture content of each single type of coal and the mass function relationship between the single type of coal and coke for each of the second influencing parameter sets; and Under multiple constraints, a metaheuristic algorithm is used to optimize the objective function to determine the wet basis ratio of each single type of coal, wherein the multiple constraints include a first constraint for each of the first influencing parameter set and a second constraint for each of the second influencing parameter set.

2. The method for coking coal blending according to claim 1, characterized in that: The establishment of the target optimization function comprises: Determine the dry basis proportion of each single type of coal according to the wet basis proportion and moisture of each single type of coal; Determine the coking rate according to the dry basis ratio, ash and volatile matter of each single type of coal and the dry basis volatile matter of the coke; Determine the wet basis requirement of each single type of coal according to the wet basis ratio and moisture content of each single type of coal, the coking rate and the coke requirement; Determine the raw material cost function according to the wet basis demand, wet basis unit transportation cost and wet basis ex-factory unit price of each single type of coal; Determining the storage cost function according to the inventory quantity and wet basis demand of each single type of coal and the unit inventory maintenance cost; and The target optimization function is established according to the raw material cost function and the storage cost function.

3. The method for coking coal blending according to claim 2, characterized in that: Determining the raw material cost function includes: According to the wet basis demand q of the i-th single type of coal i (x i ), wet basis unit transportation cost C T,i With wet base ex-factory unit price C i And the following formula determines the raw material cost function Where n is the number of single type of coal.

4. The method for coking coal blending according to claim 2, characterized in that: Determining the storage cost function includes: According to the inventory s of the i-th single type of coal i and wet basis requirement q i (x i ), unit inventory maintenance cost c0 and the following formula to determine the storage cost function Where n is the number of single type of coal.

5. The method for coking coal blending according to claim 1, characterized in that: The coking coal blending method also includes: Using the RANSAC algorithm to determine the quality function relationship between the single-type coal and the mixed coal for each of the first influencing parameter sets; and The RANSAC algorithm is used to determine the quality function relationship between a single type of coal and coke for each of the second influencing parameter set, or, when the second influencing parameter set includes a first subset and a second subset, the first subset includes ash, volatile matter and sulfur content, and the second subset includes abrasion resistance, crushing strength, reactivity and post-reaction strength, the RANSAC algorithm is used to determine the quality function relationship between a single type of coal and coke for each of the first subset, and a machine learning method or a statistical regression method is used to determine the quality function of a single type of coal and coke for each of the second subset.

6. The method for coking coal blending according to claim 5, characterized in that: The RANSAC algorithm is used to determine the quality function relationship between the single-variety coal and the mixed coal for each of the first influencing parameter sets, including: S1: normalizing a plurality of small coke oven test data groups, wherein each of the plurality of small coke oven test data groups comprises a plurality of single-variety coal proportions, six-dimensional data corresponding to each single-variety coal under each single-variety coal proportion, and experimental results of each dimension of mixed coal under each single-variety coal proportion, and the six-dimensional data comprises ash content, volatile matter, sulfur content, bonding index, final shrinkage of the gelatinous layer, and maximum thickness of the gelatinous layer; S2: determining a corresponding plane according to at least three test data in a test data group randomly selected from a plurality of small coke oven test data groups; S3: according to the corresponding plane, traverse each test data in the selected test data group respectively, determine whether the distance between each test data and the corresponding plane is less than a threshold value, and if so, determine the corresponding test data as an inlier and count the number of inliers; and S4. Repeat S2-S3 until the set number of experiments is reached, select the test data group containing the most internal points, and fit the quality function relationship between the single-variety coal and the mixed coal for each of the first influencing parameter sets.

7. The method for coking coal blending according to any one of claims 1 to 6, characterized in that: The determining of a first constraint condition for each of the first set of influencing parameters comprises: determining an intercept in the quality function relationship between a single type of coal and mixed coal for each of the first influencing parameter sets as a mixed coal quality prediction correction value for each of the first influencing parameter sets; and The first constraint condition for each of the first influencing parameter set is determined according to the wet basis ratio and moisture content of each single type of coal and the predicted correction value of the mixed coal quality for each of the first influencing parameter set.

8. The method for coking coal blending according to claim 5, characterized in that: The RANSAC algorithm is used to determine the quality function relationship between the single-variety coal and the coke for each of the second influencing parameter sets, including: S1: normalizing a plurality of small coke oven test data groups, wherein each of the plurality of small coke oven test data groups comprises a plurality of single-variety coal proportions, six-dimensional data corresponding to each single-variety coal under each single-variety coal proportion, and experimental results of each dimension of mixed coal under each single-variety coal proportion, and the six-dimensional data comprises ash content, volatile matter, sulfur content, bonding index, final shrinkage of the gelatinous layer, and maximum thickness of the gelatinous layer; S2: determining a corresponding plane according to at least three test data in a test data group randomly selected from a plurality of small coke oven test data groups; S3: according to the corresponding plane, traverse each test data in the selected test data group respectively, determine whether the distance between each test data and the corresponding plane is less than a threshold value, and if so, determine the corresponding test data as an inlier and count the number of inliers; and S4. Repeat S2-S3 until the set number of experiments is reached, select the test data set containing the most internal points, and fit the mass function relationship between the single-variety coal and coke for each of the second influencing parameter sets.

9. The method for coking coal blending according to claim 8, characterized in that: The determining of a second constraint condition for each of the second set of influencing parameters comprises: determining an intercept in the quality function relationship between a single type of coal and coke for each of the second influencing parameter sets as a coke quality prediction correction value for each of the second influencing parameter sets; and Determining a proportional correction coefficient for each of the second influencing parameter sets according to a slope in a mass function relationship between a single type of coal and coke for each of the second influencing parameter sets; and The second constraint condition for each of the second influencing parameter set is determined based on the wet basis ratio and moisture content, coking rate of each single type of coal, and the coke quality prediction correction value and proportional correction coefficient for each of the second influencing parameter set.

10. The method for coking coal blending according to claim 5, characterized in that: In the case where the second influencing parameter set includes a first subset and a second subset, the RANSAC algorithm is used to determine the quality function relationship between a single type of coal and coke for each of the first subsets, including: S1: normalizing a plurality of small coke oven test data groups, wherein each of the plurality of small coke oven test data groups comprises a plurality of single-variety coal proportions, three-dimensional data corresponding to each single-variety coal under each single-variety coal proportion, and experimental results of each dimensional data of mixed coal under each single-variety coal proportion, and the three-dimensional data comprises ash content, volatile matter, and sulfur content; S2: determining a corresponding plane according to at least three test data in a test data group randomly selected from a plurality of small coke oven test data groups; S3: according to the corresponding plane, traverse each test data in the selected test data group respectively, determine whether the distance between each test data and the corresponding plane is less than a threshold value, and if so, determine the corresponding test data as an inlier and count the number of inliers; and S4, repeat S2-S3 until the set number of experiments is reached, select the test data group containing the most internal points, and fit the mass function relationship between the single-variety coal and the mixed coal for each of the first subsets, The BP neural network is used to determine the quality function of the single-type coal and coke for each of the second subsets, including: The BP neural network is trained by means of a training set to determine the quality function of the single type of coal and coke for each of the second subsets, wherein the training set comprises a plurality of training groups, each training group comprising: the theoretical value of each of the first influencing parameter set of the single type of coal under any ratio, the actual value of each of the first influencing parameter set of the mixed coal formed by batch small coke oven tests using any of the ratios, the actual value of each of the coke in the first subset formed by the batch small coke oven tests using any of the ratios, and the actual value of each of the coke in the second subset formed by actual production using any of the ratios.

11. The method for coking coal blending according to claim 10, characterized in that: The determining of a second constraint condition for each of the second set of influencing parameters comprises: A second constraint condition for each of the first subset is determined by: determining an intercept in the quality function relationship between a single type of coal and coke for each of the first subset as a coke quality prediction correction value for each of the first subset; and determining a proportional correction coefficient for each of the first subsets according to a slope in a mass function relationship between a single type of coal and coke for each of the first subsets; and Determine a second constraint condition for each of the first subsets according to the wet basis proportion and moisture content, coking rate of each single type of coal, and the predicted coke quality correction value and proportion correction coefficient for each of the first subsets, A second constraint condition for each of the second subset is determined by: A second constraint condition for each of the second subset is determined based on the quality function of the single-species coal and coke for each of the second subset and the upper limit and lower limit for each of the second subset.

12. The method for coking coal blending according to any one of claims 1 to 11, characterized in that: The plurality of constraints also include: a third constraint, The third constraint condition is determined according to the coking rate.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the method for coking coal blending according to any one of claims 1 to 12.

14. A computer program product, characterized in that It comprises a computer program, which, when executed by a processor, implements the method for coking coal blending according to any one of claims 1-12.

Images