Iron ore smelting cost analysis method and system based on multi-dimensional data fusion

Abstract

The invention provides an iron ore smelting cost analysis method and system based on multi-dimensional data fusion. The method comprises the following steps: collecting cost factor data in each iron ore smelting process; performing importance analysis of each cost factor and preliminary prediction of unit molten iron cost by using a random forest algorithm to obtain a first prediction result; on the basis of the importance analysis result, a multiple regression model of the key cost factors is constructed, parameters of the multiple regression model are optimized through an ant colony algorithm, the unit molten iron cost is predicted again, and a second prediction result is obtained; and the first prediction result and the second prediction result are fused, the final unit molten iron cost of each iron ore is obtained, and cost performance comparison is carried out. According to the method, various complex cost factors of iron ore smelting are comprehensively analyzed, rapid and accurate prediction of the unit molten iron cost is achieved, the smelting cost performance of each iron ore can be rapidly and accurately obtained, and enterprises can rapidly adjust purchase production strategies and cost control measures.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of iron ore smelting cost analysis, and particularly relates to an iron ore smelting cost analysis method and system based on multi-dimensional data fusion. BACKGROUND

[0002] As a key link of steel production, the cost composition of iron ore smelting is complex, and is affected by many factors such as raw material price fluctuation, energy consumption, process parameters, international market, etc., and the cost level is directly related to the economic benefit and market competitiveness of enterprises.

[0003] At present, the traditional cost accounting method mainly relies on historical data and periodic statistics, and the accounting result has obvious lag. In today's environment where market situation changes rapidly, key cost factors such as raw material price and energy cost frequently fluctuate, and the traditional cost accounting method cannot timely reflect these changes, so that enterprises cannot quickly adjust production strategy and cost control measures. When the market appears adverse changes, enterprises often fall into a passive situation due to the inability to obtain accurate cost information in time, and it is difficult to effectively control cost risk, and even may cause economic loss;

[0004] In addition, the traditional cost accounting method usually only focuses on the overall cost, lacks detailed decomposition and analysis of each cost factor, and cannot clearly understand the specific influence degree of each cost factor on the cost, so that it is difficult for enterprises to identify the key factors affecting the cost, and it is more difficult to develop effective cost control measures, so as to realize the optimization of production process and effective reduction of cost. SUMMARY

[0005] Therefore, the purpose of the present application is to provide an iron ore smelting cost analysis method and system based on multi-dimensional data fusion to solve the above-mentioned problems.

[0006] The iron ore smelting cost analysis method based on multi-dimensional data fusion provided by the present application comprises:

[0007] Collecting each cost factor data related to unit molten iron cost in each iron ore smelting process;

[0008] Using a random forest algorithm to analyze the importance of each cost factor and preliminarily predict the unit molten iron cost to obtain a first prediction result;

[0009] Based on the importance analysis result of the cost factor, a multiple regression model of the key cost factor is constructed, and the parameters of the multiple regression model are optimized by an ant colony algorithm, and the optimized multiple regression model is used to predict the unit molten iron cost again to obtain a second prediction result;

[0010] fuse the first prediction result and the second prediction result of the unit hot metal cost of each iron ore to obtain a final unit hot metal cost of each iron ore;

[0011] compare the smelting cost performance of various iron ores according to the final unit hot metal cost of each iron ore.

[0012] Further, the step of using the random forest algorithm to analyze the importance of each cost factor and predict the unit hot metal cost to obtain the first prediction result comprises:

[0013] For the target type of iron ore, the pre-processed historical multiple smelting cycle data of each cost factor is taken as input and is denoted as a feature matrix X, each row of the feature matrix X represents a smelting cycle, and each column represents a cost factor;

[0014] The unit hot metal cost of the target type of iron ore is taken as the output target and is denoted as a vector Y, each element of which corresponds to the unit hot metal cost of a smelting cycle;

[0015] Based on the feature matrix X and the output target Y, a random forest model is constructed, and the random forest is composed of multiple decision trees, and each tree is constructed by randomly sampling features and data points;

[0016] For each tree, the data not involved in tree construction is used for prediction, and the prediction error is calculated;

[0017] Randomly select all values of a cost factor x i in the data not involved in tree construction, predict again, and calculate the new prediction error;

[0018] Calculate the change in prediction error before and after the values of the cost factor x i are disturbed to obtain the importance score of the cost factor x i , and repeat the above operation for each cost factor until the importance scores of all cost factors are obtained;

[0019] Record the cost factor data of the current smelting cycle of the target type of iron ore as a vector x current , and use the trained random forest model to predict the unit hot metal cost of the current smelting cycle to obtain the first prediction result y 1，pre of the unit hot metal cost of the target type of iron ore.

[0020] Further, the step of constructing a multiple regression model of key cost factors based on the importance analysis result of the cost factors comprises:

[0021] From the cost factors, select the top n cost factors in the importance score ranking as key cost factors;

[0022] Data values of each key cost factor are taken as input variables, and unit hot metal cost of target iron ore is taken as output target, a multiple regression model of key cost factors is constructed, and the model expression is:

[0023] y 2，pre = β0+ β1x1+ β2x2+ … + β n x n + ∈,

[0024] wherein y 2，pre is the second prediction result of unit hot metal cost, x u is the data value of the u-th key cost factor, β u is a regression coefficient, and ∈ is an error term.

[0025] Further, the step of optimizing the parameters of the multiple regression model by using the ant colony algorithm and predicting the unit hot metal cost by using the optimized multiple regression model to obtain the second prediction result comprises the following steps.

[0026] Different weights are given to initial pheromone values of each regression coefficient;

[0027] Importance scores of key cost factors are taken as heuristic information of corresponding regression coefficients;

[0028] When searching, each ant selects adjustment direction and amplitude of each regression coefficient according to current pheromone concentration and heuristic information of each regression coefficient;

[0029] After multiple iterations, an optimal regression coefficient combination {β0, β1, β2, …, β n} is obtained, and an optimized multiple regression model is constructed;

[0030] Based on data of each key cost factor of the target type of iron ore in the current smelting period, the optimized multiple regression model is used to predict and calculate the unit hot metal cost of the current smelting period, and the second prediction result y 2，pre of the optimized unit hot metal cost of the target type of iron ore is obtained.

[0031] Further, before the step of each ant selecting adjustment direction and amplitude of each regression coefficient according to current pheromone concentration and heuristic information of each regression coefficient, the following step is further included.

[0032] The pheromone is updated according to the prediction accuracy of the multiple regression model, and the formula is:

[0033]

[0034] wherein, is the updated u-th regression coefficient β uwhere ρ is the pheromone evaporation coefficient, Q is a constant indicating the strength of pheromone enhancement, MSE v is the mean square error of the multiple regression model after the vth ant adjusts the regression coefficients, reflecting the prediction accuracy of the multiple regression model.

[0035] Further, the step of selecting the adjustment direction and amplitude of each regression coefficient by each ant according to the current pheromone concentration and heuristic information of each regression coefficient comprises:

[0036] For each ant v and each regression coefficient β u , the probability P u selected by the ant v to adjust the regression coefficient β vu is calculated according to the pheromone concentration and heuristic information, and is:

[0037]

[0038] where P vu is the probability of the ant v to adjust the regression coefficient β u , τ vu and τ vg are the pheromone concentrations of the vth ant on the regression coefficients β u and β g , α1 is a parameter controlling the relative importance of pheromone, α2 is a parameter controlling the relative importance of heuristic information, H u and H g are the heuristic information of the regression coefficients β u and β g , and n is the total number of regression coefficients, i.e., the number of key cost factors.

[0039] A first random number r1 between 0 and 1 is generated, and the cumulative probability CP is calculated. v(u-1) Starting from u = 1, the cumulative probability CP vu is calculated in sequence until u satisfying CP u is found, then the ant v selects to adjust the regression coefficient β u currently found, and r1 is the first random number between 0 and 1.

[0040] The value of the selected regression coefficient β u is updated according to the random adjustment direction and the dynamically set amplitude.

[0041] Further, the step of updating the value of the selected regression coefficient β u according to the random adjustment direction and the dynamically set amplitude comprises:

[0042] A second random number r2 between 0 and 1 is generated, and if r2 < 0.5, the selected regression coefficient βu if r2≥0.5, then the value of the selected adjusted regression coefficient β u is reduced

[0043] a third random number r3 between 0 and 1 is generated, and the adjustment range Δβ u of the selected adjusted regression coefficient β u is calculated u = min_step + r3·(max_step-min_step), [min_step, max_step] is the step range

[0044] According to the determined adjustment direction and range, the value of the regression coefficient β u is updated

[0045] if the value of the regression coefficient β u is increased, the updated value is

[0046] if the value of the regression coefficient β u is reduced, the updated value is wherein, is the value of the regression coefficient β u before updating.

[0047] Further, the step of fusing the first prediction result and the second prediction result of the unit hot metal cost of each iron ore to obtain the final unit hot metal cost of each iron ore comprises:

[0048] the first prediction result y 1，pre and the second prediction result y 2，pre of the unit hot metal cost of each iron ore are weighted and fused to obtain the final unit hot metal cost y final，pre pre of each iron ore = λ·y 1，pre +(1-λ)·y 2， pre, wherein λ is a weighting coefficient.

[0049] Further, in the step of collecting each cost factor data related to the unit hot metal cost in the smelting process of each iron ore:

[0050] The cost factor data includes iron ore index, coke index, ore consumption, fuel consumption, flux cost, wind water electricity cost, worker wages, manufacturing cost, depreciation, and current RMB exchange rate data.

[0051] The application further proposes an iron ore smelting cost analysis system based on multi-dimensional data fusion, which is used to realize the above-mentioned iron ore smelting cost analysis method based on multi-dimensional data fusion, and the system comprises:

[0052] A data collection module is configured to collect data of each cost factor related to the unit molten iron cost in the smelting process of each iron ore;

[0053] A first prediction module is configured to analyze the importance of each cost factor and preliminarily predict the unit molten iron cost by using a random forest algorithm, and obtain a first prediction result;

[0054] A second prediction module is configured to construct a multiple regression model of key cost factors based on the importance analysis result of the cost factors, optimize the parameters of the multiple regression model by using an ant colony algorithm, and predict the unit molten iron cost again by using the optimized multiple regression model, and obtain a second prediction result;

[0055] A final molten iron cost module is configured to fuse the first prediction result and the second prediction result of the unit molten iron cost of each iron ore, and obtain the final unit molten iron cost of each iron ore.

[0056] In summary, the smelting cost analysis method for iron ore based on multi-dimensional data fusion provided by the present application firstly analyzes the importance of each cost factor by using a random forest algorithm, obtains the influence degree of each cost factor on the unit molten iron cost, and preliminarily predicts the unit molten iron cost based on the model to obtain a preliminary prediction result; then, the key cost factors are screened based on the importance analysis result, a multiple regression model of the key cost factors is constructed to predict the unit molten iron cost again, which greatly reduces the complexity of modeling, and the parameters of the regression model are globally optimized by using an ant colony algorithm to find the optimal regression coefficient combination, which not only makes the prediction performance of the multiple regression model globally optimal, ensures the stability and prediction accuracy of the model in a complex smelting data environment, but also simplifies the model structure, and is suitable for non-convex optimization problems in complex smelting scenarios, can well handle the non-linear relationship in the multiple regression model, and balance the calculation efficiency and cost prediction accuracy; then, the two prediction results are fused to obtain the final unit molten iron cost of each iron ore, which can significantly reduce the calculation bias and variance of a single model, enhance the prediction stability and accuracy, and improve the adaptability of the overall analysis scheme to complex iron ore smelting cost analysis scenarios; finally, the smelting cost-effectiveness of various iron ores is compared according to the final unit molten iron cost of each iron ore, so as to understand which iron ore is more economical in the smelting process, and make a more intelligent purchasing and smelting production decision.

[0057] The present application can comprehensively and intelligently analyze each complex cost factor in the smelting process of each iron ore, realize real-time, rapid and accurate prediction and analysis of the unit molten iron cost, quickly and accurately obtain the smelting cost-effectiveness of each iron ore, enable the enterprise to quickly adjust the purchasing strategy, production strategy and cost control measures, effectively cope with market risks, and improve the market response ability of the enterprise.

[0058] Additional aspects and advantages of the application will be set forth in part in the description which follows, and in part will become apparent to those skilled in the art upon examination of the following and the accompanying drawings or can be learned by practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0059] The foregoing and / or additional aspects and advantages of the application are achieved by providing what is described below and / or claimed by the application as can be learned by perusal of the following description and the accompanying drawings.

[0060] Figure 1 A flow chart of the iron ore smelting cost analysis method based on multi-dimensional data fusion according to an embodiment of the present application;

[0061] Figure 2 A system block diagram of the iron ore smelting cost analysis system based on multi-dimensional data fusion according to an embodiment of the present application. DETAILED DESCRIPTION

[0062] In order to facilitate the understanding of the present application, a more complete and comprehensive description of the present application will be provided below with reference to the relevant drawings. The drawings show several embodiments of the present application. However, the present application can be realized in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive.

[0063] It should be noted that when an element is referred to as being "fixed to" another element, it can be directly on the other element or there can be a middle element. When an element is referred to as being "connected" to another element, it can be directly connected to the other element or there can be a middle element. The terms "vertical", "horizontal", "left", "right", and similar expressions used herein are for illustrative purposes only.

[0064] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs. The terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0065] Embodiment One

[0066] Referring to Figure 1 The present application proposes an iron ore smelting cost analysis method based on multi-dimensional data fusion, which comprises steps S101 to S105:

[0067] S101, collecting each cost factor data related to the cost of unit molten iron in the smelting process of each iron ore.

[0068] It is understood that first, the historical data of each cost factor related to the cost per ton of molten iron in the smelting process of each iron ore needs to be collected in order to subsequently analyze the smelting cost and cost performance of each iron ore. The cost factor data includes but is not limited to iron ore index (reflecting the index of iron ore market price fluctuation), coke index (reflecting the index of coke market price change, coke is one of the main fuels for blast furnace smelting, and its price fluctuation directly affects the smelting cost), ore consumption (referring to the amount of iron ore consumed per ton of molten iron), fuel consumption (including the consumption of coke, coal powder and other fuels, which directly affects the energy cost), flux cost (such as limestone, dolomite cost), water, electricity and gas cost (the cost of water, wind, electricity, gas and other energy and public utilities required in the smelting process), worker wages (worker wages and welfare costs involved in smelting production), manufacturing cost (including equipment maintenance, repair, spare parts replacement and other indirect production costs), depreciation (depreciation of production equipment, plant and other fixed assets), and current RMB exchange rate (since iron ore and other raw materials may involve import, exchange rate fluctuation will affect the import cost) and other data.

[0069] The collected data is preprocessed, including data cleaning, data standardization, data normalization, feature engineering, etc., to remove outliers, missing values, duplicate data, etc. in the data, eliminate the influence of dimension difference on model training, and scale the data to a unified range.

[0070] In S102, the importance of each cost factor is analyzed and the cost per ton of molten iron is preliminarily predicted using the random forest algorithm, and a first prediction result is obtained.

[0071] Further optionally, the step of analyzing the importance of each cost factor and predicting the cost per ton of molten iron using the random forest algorithm to obtain the first prediction result comprises:

[0072] For the target type of iron ore, the preprocessed historical multiple smelting period data of each cost factor of the target type of iron ore is taken as input and is denoted as a feature matrix X, each row of the feature matrix X represents a smelting period, and each column represents a cost factor;

[0073] The cost per ton of molten iron of the target type of iron ore is taken as the output target and is denoted as a vector Y, each element of which corresponds to the cost per ton of molten iron of a smelting period;

[0074] A random forest model is constructed based on the feature matrix X and the output target Y, the random forest is composed of multiple decision trees, and each tree is constructed by randomly sampling features and data points;

[0075] For each tree, the data not involved in tree construction is used for prediction, and the prediction error is calculated;

[0076] A cost factor x is randomly selected from the data not involved in tree constructioni all values of the cost factor x, again make a prediction and calculate a new prediction error;

[0077] calculate the disturbance cost factor x i the change in prediction error before and after the value of the cost factor x i importance score, and repeat the above operation for each cost factor until the importance score of all cost factors is obtained;

[0078] The data of each cost factor of the target type of iron ore in the current smelting period is recorded as a vector x current , using the trained random forest model to predict the unit molten iron cost of the current smelting period, obtaining the first prediction result y 1，pre .

[0079] It can be understood that the embodiment quantitatively analyzes the importance of each cost factor in the smelting period of iron ore based on out-of-bag data (data not participating in tree construction). Specifically, by randomly disturbing the value of a certain cost factor and calculating the change in prediction error, the influence degree of the factor on the unit molten iron cost can be intuitively reflected.

[0080] The multi-tree ensemble prediction mechanism of the random forest model can effectively solve the problem of easy overfitting of a single decision tree. In the construction process, through the voting or averaging strategy of multiple decision trees, the complex nonlinear relationship between the cost factor and the unit molten iron cost can be accurately captured. At the same time, using out-of-bag data error calculation for model optimization can further enhance the stability and reliability of the prediction results.

[0081] The cost factor data of the target type of iron ore in the current smelting period is input into the trained random forest model to dynamically predict the unit molten iron cost in real time, and the first prediction result of the unit molten iron cost of the target type of iron ore is obtained.

[0082] S103, based on the importance analysis result of the cost factor, a multiple regression model of the key cost factor is constructed, and the parameters of the multiple regression model are optimized by an ant colony algorithm, and the second prediction result is obtained by using the optimized multiple regression model for re-prediction of the unit molten iron cost.

[0083] Further optionally, the step of constructing a multiple regression model of the key cost factor based on the importance analysis result of the cost factor comprises:

[0084] From each cost factor, the top n cost factors in the importance score ranking are selected as the key cost factors;

[0085] The data values of the key cost factors are taken as input variables, and the unit hot metal cost of the target iron ore is taken as an output target to build a multiple regression model of the key cost factors, and the model expression is:

[0086] y 2，pre =β0+β1x1+β2x2+…+β n x n +∈,

[0087] Wherein, y 2，pre is the second prediction result of the unit hot metal cost, x u is the data value of the u-th key cost factor, β u is the regression coefficient, and ∈ is the error term.

[0088] It can be understood that by screening the n cost factors with high importance scores, the key cost factors that have the most significant impact on the unit hot metal cost can be accurately locked, the complexity of cost analysis is effectively simplified, and the focus of cost analysis can be more focused on core factors and problems. And screening the key cost factors can narrow the analysis range and build a more targeted multiple regression model to ensure that the model can more accurately capture the key driving factors of cost changes.

[0089] The multiple regression model can quantify the specific comprehensive influence of each key cost factor of the target type of iron ore on the unit hot metal cost. Through the model expression y 2，pre =β0+β1x1+β2x2+…+β n x n +∈, it can be known that the relationship between each key cost factor x u and the unit hot metal cost y 2，pre , and the influence represented by the regression coefficient β u . Enterprises can adjust the data values of key cost factors according to the size and direction of the regression coefficient, that is, control each key cost factor to effectively reduce the cost.

[0090] Further, the step of optimizing the parameters of the multiple regression model by the ant colony algorithm and using the optimized multiple regression model to predict the unit hot metal cost to obtain the second prediction result comprises:

[0091] Different weights are given to the initial pheromone values of each regression coefficient;

[0092] The importance score of the key cost factor is taken as the heuristic information of the corresponding regression coefficient;

[0093] When the ants search, each ant selects the adjustment direction and amplitude of each regression coefficient according to the current pheromone concentration and heuristic information of each regression coefficient.

[0094] After multiple iterations, the optimal regression coefficient combination {β0, β1, β2, …, β n} is obtained, and an optimized multiple regression model is constructed.

[0095] Based on the data of each key cost factor of the target type of iron ore in the current smelting period, the optimized multiple regression model is used to predict and calculate the unit molten iron cost of the target type of iron ore in the current smelting period, and the second prediction result y 2，pre of the unit molten iron cost of the target type of iron ore is obtained.

[0096] It can be understood that the ant colony algorithm is used to optimize the parameters of the multiple regression model, and the optimized multiple regression model is used to predict the unit molten iron cost of the target type of iron ore, which further improves the accuracy and reliability of the cost prediction. Specifically, different weights are assigned to the initial pheromone values of each regression coefficient as the starting point of the ant colony algorithm optimization. Through the differentiated weight setting, the algorithm has a preliminary distinction of the importance of each regression coefficient in the initial stage, providing a reasonable guide for the subsequent optimization process.

[0097] Then, the importance score of the key cost factor is used as the heuristic information of the corresponding regression coefficient, and the importance score reflects the influence degree of each key cost factor on the unit molten iron cost. By integrating it into the algorithm as heuristic information, the algorithm can be more inclined to adjust the regression coefficients that have a greater impact on the cost during the search process. This heuristic guidance enhances the optimization direction of the algorithm, enabling the algorithm to quickly find a regression coefficient combination close to the optimal solution, thereby greatly improving the accuracy and efficiency of the optimization.

[0098] During the ant search process, each ant selects the adjustment direction and amplitude according to the pheromone concentration and heuristic information of each regression coefficient. This mechanism simulates the behavior of ants searching for food in nature, and through the positive feedback of pheromones and the guidance of heuristic information, the algorithm can dynamically adjust the regression coefficients and gradually approach the optimal solution.

[0099] After multiple iterations, the algorithm can obtain the optimal regression coefficient combination {β0, β1, β2, …, β n}. This combination is based on a large amount of historical data and current heuristic information, and can best fit the actual cost data to improve the accuracy of the prediction. Through the iterative optimization process, the algorithm continuously adjusts the regression coefficients and gradually approaches the global optimal solution, which also makes the prediction performance of the multiple regression model optimal, thereby ensuring the reliability of the optimization result and the accuracy of the second prediction result.

[0100] Based on the optimized combination of regression coefficients, an optimized multiple regression model is constructed. The model retains the advantages of simplicity and easy interpretation of the multiple regression model, while further improving the accuracy of prediction through the optimization of the ant colony algorithm. The optimized multiple regression model is used to predict the unit molten iron cost of the target type of iron ore in the current smelting cycle, and a more accurate second prediction result y of the unit molten iron cost of the target type of iron ore is obtained 2，pre .

[0101] Further optionally, before the step of each ant selecting the adjustment direction and amplitude of each regression coefficient according to the current pheromone concentration and heuristic information of each regression coefficient, the method further comprises the steps of:

[0102] The pheromone is updated according to the prediction accuracy of the multiple regression model, and the formula is:

[0103]

[0104] wherein, is the updated pheromone concentration of the vth ant on the u th regression coefficient β u , ρ is the pheromone evaporation coefficient, which is 0.1-0.3, used to control the evaporation speed of the pheromone, Q is a constant, representing the strength of the pheromone enhancement, and MSE v is the mean square error of the multiple regression model after the vth ant adjusts the regression coefficient, reflecting the prediction accuracy of the multiple regression model.

[0105] It can be understood that updating the pheromone according to the prediction accuracy of the multiple regression model is the key in the process of optimizing the multiple regression model parameters by the ant colony algorithm. Specifically, the pheromone concentration can be updated by the quantitative index of mean square error MSE, which can closely combine the model prediction performance with the dynamic change of pheromone. The mean square error MSE can intuitively reflect the fitting degree of the multiple regression model to the actual data, and the higher the prediction accuracy, the smaller the MSE value. Using this index to update the pheromone concentration makes the pheromone concentration on the path of the ants with good performance (i.e. the regression coefficient combination adjusted by them makes the model prediction accuracy higher) increase, and vice versa. This dynamic adjustment mechanism ensures that the pheromone can truly reflect the advantages and disadvantages of each regression coefficient combination, and provides effective guidance for the search of subsequent ants.

[0106] The whole pheromone updating process constitutes a feedback loop mechanism: ants choose the adjustment direction and amplitude of the regression coefficient according to the current pheromone concentration and heuristic information, which in turn affects the prediction accuracy of the multiple regression model; the prediction accuracy is fed back to the pheromone updating formula through the mean square error (MSE), which determines the pheromone concentration on each path in the next iteration; the change of pheromone concentration guides ants to conduct new search. This loop mechanism continuously optimizes the parameters of the multiple regression model, gradually improves the prediction accuracy of the multiple regression model, and ultimately obtains a more accurate and reliable multiple regression model, achieving accurate prediction of the unit molten iron cost of each type of iron ore smelting again.

[0107] Further optionally, the step of each ant selecting the adjustment direction and amplitude of each regression coefficient according to the current pheromone concentration and heuristic information comprises:

[0108] For each ant v and each regression coefficient β u , according to the pheromone concentration and heuristic information, the probability P u of ant v selecting to adjust the regression coefficient β vu is calculated as:

[0109]

[0110] Where P vu is the probability of ant v selecting to adjust the regression coefficient β u , τ vu and τ vg are the pheromone concentrations of the vth ant on the regression coefficients β u and β g , respectively. The pheromone concentration reflects the likelihood of the regression coefficient value combination being proven to produce better prediction accuracy in history. The higher the concentration, the greater the probability of being selected. α1 is a parameter that controls the relative importance of pheromone, which determines the weight of pheromone concentration in the calculation of selection probability. The larger α1, the greater the influence of pheromone on selection probability. α2 is a parameter that controls the relative importance of heuristic information, which determines the weight of heuristic information in the calculation of selection probability. The larger α2, the greater the influence of heuristic information on selection probability. H u and H g are the heuristic information of the regression coefficients β u and β g , respectively. n is the total number of regression coefficients, i.e., the number of key cost factors.

[0111] A first random number r1 between 0 and 1 is generated, and the cumulative probability CP is calculated as: v(u-1) Starting from u = 1, calculate in turn until the cumulative probability CP vuu, then ant v chooses to adjust the currently found regression coefficient β u , r1 is the first random number between 0 and 1;

[0112] Update the selected adjusted regression coefficient β according to the random adjustment direction and dynamically set amplitude u value.

[0113] As can be understood, each ant selects the direction and magnitude of its adjustment based on the pheromone concentration and heuristic information for each regression coefficient. Specifically, the probability of the ant choosing to adjust each regression coefficient is calculated by comprehensively considering pheromone concentration and heuristic information. Pheromone concentration reflects the likelihood that a combination of regression coefficient values ​​has historically produced good predictive accuracy. A higher concentration indicates a combination with better past performance and a greater probability of selection. Heuristic information, based on the importance scores of key cost factors, provides additional guidance for the ant's current search direction. α1 and α2 control the relative importance of pheromone and heuristic information, respectively. The algorithm can flexibly adjust the weight of each in the probability calculation to suit different optimization requirements. This integrated calculation method enables ants to more scientifically and rationally select the direction of regression coefficient adjustment, improving the accuracy and efficiency of multivariate regression model optimization.

[0114] By calculating cumulative probabilities to determine the regression coefficients that ants choose to adjust, the algorithm achieves a balance between exploration and exploitation. By calculating cumulative probabilities, the algorithm can determine the relative probability of each regression coefficient being selected based on pheromone concentration and heuristic information, ensuring that excellent regression coefficient combinations have a higher chance of being selected. Furthermore, the random number r1 introduces randomness, allowing ants to explore regression coefficients that may have potential optimization value despite low current pheromone concentrations.

[0115] The selected regression coefficients are updated based on the random adjustment direction and dynamically set amplitude, allowing them to be continuously optimized with each algorithm iteration. The random adjustment direction provides more possibilities, avoiding a fixed adjustment pattern and helping to discover more optimal regression coefficient combinations. The dynamically set amplitude can be adjusted based on actual conditions, making the adjustment process more flexible and efficient. By continuously updating the regression coefficients, the algorithm can gradually approach the optimal regression coefficient combination, achieving optimal prediction accuracy for the multivariate regression model.

[0116] Further optionally, the regression coefficient β selected for adjustment is updated according to the random adjustment direction and the dynamically set amplitude. u The steps for finding the value include:

[0117] Generate a second random number r2 between 0 and 1. If r2<0.5, increase the selected adjusted regression coefficient β uIf r2≥0.5, reduce the selected adjusted regression coefficient β u The value of

[0118] Generate a third random number r3 between 0 and 1 and calculate the selected adjusted regression coefficient β u The adjustment range Δβ u =min_step+r3·(max_step-min_step), [min_step, max_step] is the step size range;

[0119] According to the determined adjustment direction and amplitude, update the regression coefficient β u Value:

[0120] If we increase the regression coefficient β u The updated value is:

[0121] If the regression coefficient β is reduced u The updated value is: in, is the regression coefficient β before updating u value.

[0122] Understandably, the regression coefficient β selected for adjustment is updated according to the random adjustment direction and the dynamically set amplitude u In the process of adjusting the value of , first, randomness is introduced to determine the adjustment direction. Specifically, a second random number r2 between 0 and 1 is generated, and the second random number r2 is used to determine whether to increase or decrease the regression coefficient β selected for adjustment. u The value of r² is used to inject randomness. During the optimization process, adjusting the regression coefficients based solely on fixed rules or considering only pheromones and heuristic information can easily lead to the algorithm falling into a local optimum. Increasing the regression coefficient when r² < 0.5 and decreasing it when r² ≥ 0.5 allows the algorithm to try different adjustment paths, potentially discovering more optimal regression coefficient combinations that would otherwise be difficult to achieve under deterministic rules, thereby increasing the likelihood of finding the global optimal regression coefficient combination.

[0123] Then, the adjustment range is dynamically set to balance optimization accuracy and efficiency. Specifically, a second random number r3 between 0 and 1 is generated, and the regression coefficient β selected for adjustment is calculated based on it. u The adjustment range Δβ u= min_step + r3 · (max_step - min_step) to balance the accuracy and efficiency of optimization. On the one hand, the setting of the step range [min_step, max_step] limits the maximum amplitude of each adjustment, avoiding the dramatic fluctuation of model prediction accuracy caused by excessive adjustment amplitude, and ensuring the stability of the optimization process. On the other hand, by randomly selecting a value within the step range through the random number r3, the adjustment amplitude is dynamic.

[0124] updating the regression coefficients according to the determined direction and amplitude, updating the values of the regression coefficients β u , gradually adjusting the parameters of the multiple regression model, so that the model better fits the actual data. With the increase of the number of iterations, the regression coefficients gradually approach the optimal value, and the prediction accuracy of the multiple regression model is also continuously improved. Finally, after multiple iterations of optimization, a multiple regression model that can accurately predict the unit hot metal cost of iron ore smelting is obtained.

[0125] S104, fusing the first prediction result and the second prediction result of the unit hot metal cost of each kind of iron ore to obtain the final unit hot metal cost of each kind of iron ore.

[0126] Understandably, the unit hot metal cost refers to the total cost required for producing each ton of hot metal (i.e. liquid iron) when smelting with a specific iron ore. This cost indicator needs to consider various direct and indirect costs such as the procurement cost of iron ore, energy consumption during the smelting process, auxiliary material consumption, equipment depreciation, labor cost, and possible waste disposal cost, etc. The cost composition includes but is not limited to the purchase price, transportation cost, customs duty, etc. during the procurement process of iron ore, the energy cost of a large amount of coke, electricity, natural gas, etc. required during the smelting process, the flux cost of limestone, dolomite, etc. used to remove impurities in iron ore, the purchase cost, depreciation cost, and daily maintenance and repair cost of smelting equipment, the labor cost required during the smelting process, the disposal cost of waste slag, waste gas, etc. generated during the smelting process, etc.

[0127] Further optionally, the step of fusing the first prediction result and the second prediction result of the unit hot metal cost of each kind of iron ore to obtain the final unit hot metal cost of each kind of iron ore comprises:

[0128] weighting and fusing the first prediction result y 1，pre and the second prediction result y 2，pre of the unit hot metal cost of each kind of iron ore to obtain the final unit hot metal cost y final，pre of each kind of iron ore = λ · y 1，pre + (1 - λ) · y 2，pre , wherein λ is a weighting coefficient.

[0129] It can be understood that for each kind of iron ore, the first prediction result y of its unit hot metal cost 1，pre The second prediction result y 2，pre The final unit hot metal cost y of each kind of iron ore is obtained by weighted fusion of the first prediction result and the second prediction result final，pre In order to comprehensively consider the advantages of different prediction methods and obtain accurate and stable prediction results, the first prediction result is generated by a random forest algorithm, which can process complex data relationships and give a prediction value of unit hot metal cost, but may not consider some special cases; the second prediction result is obtained by a multiple regression model based on key cost factors and optimized parameters by an ant colony algorithm, which has advantages in considering the interaction of multiple factors and optimizing parameters, but also has limitations. Therefore, by weighted fusion, the advantages of the two prediction methods can be comprehensively utilized to make up for the shortcomings of single prediction. In addition, in actual iron ore smelting cost analysis, due to the influence of many factors such as raw material quality fluctuation, production process change, market price instability, etc., the prediction result often has uncertainty, and a single prediction model is difficult to fully consider these complex factors, which is easy to lead to large deviation. The unit hot metal cost calculated by weighted fusion can offset each other's errors, reduce the uncertainty of the prediction result, make the final prediction value more stable, provide an accurate reference for the subsequent performance-price ratio judgment of various kinds of iron ore, and help smelting enterprises better control the cost.

[0130] S105, comparing the performance-price ratio of smelting of various kinds of iron ore according to the final unit hot metal cost of each kind of iron ore.

[0131] It can be understood that due to the significant differences in chemical composition (such as iron content, impurity content, etc.) of different types of iron ore, these differences will directly affect the energy consumption, auxiliary material consumption and hot metal yield in the smelting process, thereby causing different unit hot metal costs. The physical properties of iron ore such as particle size, hardness and porosity also affect the smelting efficiency and cost, thereby affecting the unit hot metal cost. And for different types of iron ore, the smelting process parameters (such as temperature, pressure, atmosphere, etc.) will also differ, which will directly affect the smelting cost and hot metal yield, and thus affect the unit hot metal cost.

[0132] Therefore, when purchasing iron ore, the smelting plant needs to know the unit hot metal cost of each kind of iron ore in order to select the iron ore with the highest performance-price ratio. By calculating the unit hot metal cost of each kind of iron ore, the smelting plant can more accurately control the cost, identify the key points of cost control, and take corresponding optimization measures.

[0133] After the final unit hot metal cost of each iron ore is accurately calculated, the smelting cost-effectiveness of each iron ore is evaluated by comparing and ranking the final unit hot metal costs of each iron ore, so that it can be directly seen which iron ore has the highest cost-effectiveness and which has the lowest, providing a clear basis for smelting decisions.

[0134] Specifically, the specific evaluation criteria for the smelting cost-effectiveness of iron ore can be: the iron ore with the lowest unit hot metal cost has the highest cost-effectiveness, indicating that under the same smelting conditions, the cost of producing each ton of hot metal using this iron ore is the lowest, and the economic benefit is the best. The iron ore with the second lowest unit hot metal cost is the iron ore with the second highest cost-effectiveness, although the cost is slightly higher than that of the iron ore with the highest cost-effectiveness, but it may still have competitiveness under certain specific conditions (such as supply stability, transportation cost, etc.). The iron ore with a unit hot metal cost at an intermediate level is the iron ore with an average cost-effectiveness, and other factors (such as quality stability, procurement cost fluctuations, etc.) need to be considered to determine whether to use it. The iron ore with a higher unit hot metal cost is the iron ore with lower cost-effectiveness, which is relatively poor in economic benefit, unless there are special needs or advantages (such as high grade, low impurities, etc.), otherwise it is generally not recommended to use.

[0135] In some special cases, such as tight supply of iron ore, high transportation cost or special requirements for iron ore quality for certain steel products, the cost-effectiveness of iron ore may need to be evaluated by considering the unit hot metal cost and other factors.

[0136] In summary, the iron ore smelting cost analysis method based on multi-dimensional data fusion provided by the present application firstly uses a random forest algorithm to analyze the importance of each cost factor, obtains the influence degree of each cost factor on the unit molten iron cost, and performs preliminary prediction of the unit molten iron cost based on the model to obtain a preliminary prediction result; then, key cost factors are screened based on the importance analysis result, a multiple regression model of the key cost factors is constructed to perform secondary prediction of the unit molten iron cost, which greatly reduces the complexity of modeling, and the ant colony algorithm is used to globally optimize the regression model parameters to find the optimal regression coefficient combination, which not only enables the prediction performance of the multiple regression model to reach a global optimum, ensures the stability and prediction accuracy of the model in a complex smelting data environment, but also simplifies the model structure, is suitable for non-convex optimization problems in a complex smelting scene, can well handle the non-linear relationship in the multiple regression model, and balances the calculation efficiency and cost prediction accuracy; the two prediction results are fused to obtain the final unit molten iron cost of each kind of iron ore, which can significantly reduce the calculation deviation and variance of a single model, enhance the prediction stability and accuracy, and improve the adaptability of the overall analysis scheme to a complex iron ore smelting cost analysis scene; finally, the smelting cost-effectiveness of various iron ores is compared according to the final unit molten iron cost of each kind of iron ore, and it is understood which kind of iron ore is more economical in the smelting process, so that a more intelligent purchasing and smelting production decision can be made.

[0137] The present application can comprehensively and intelligently analyze various complex cost factors in the smelting process of each kind of iron ore, realize real-time, rapid and accurate prediction and analysis of the unit molten iron cost, quickly and accurately know the smelting cost-effectiveness of each kind of iron ore, enable enterprises to quickly adjust purchasing strategies, production strategies and cost control measures, effectively cope with market risks, and improve the market response ability of enterprises.

[0138] Embodiment two

[0139] Please refer to Figure 2 The iron ore smelting cost analysis system based on multi-dimensional data fusion provided by the present application comprises:

[0140] A data collection module is configured to collect various cost factor data related to the unit molten iron cost in the smelting process of each kind of iron ore.

[0141] A first prediction module is configured to use a random forest algorithm to analyze the importance of each cost factor and perform preliminary prediction of the unit molten iron cost to obtain a first prediction result.

[0142] A second prediction module is configured to construct a multiple regression model of key cost factors based on the importance analysis result of the cost factors, optimize the parameters of the multiple regression model through an ant colony algorithm, use the optimized multiple regression model to perform secondary prediction of the unit molten iron cost, and obtain a second prediction result.

[0143] Final hot metal cost module: used for fusing the first prediction result and the second prediction result of the unit hot metal cost of each iron ore to obtain the final unit hot metal cost of each iron ore.

[0144] Cost performance comparison module: used for comparing the smelting cost performance of various iron ores according to the final unit hot metal cost of each iron ore.

[0145] Further optionally, the first prediction module is further used for:

[0146] For the target kind of iron ore, the preprocessed historical cost factor data of multiple smelting cycles are taken as input and recorded as a feature matrix X, each row of the feature matrix X represents a smelting cycle, and each column represents a cost factor;

[0147] The unit hot metal cost of the target kind of iron ore is taken as output target and recorded as a vector Y, each element of which corresponds to the unit hot metal cost of a smelting cycle;

[0148] A random forest model is constructed based on the feature matrix X and the output target Y, the random forest is composed of multiple decision trees, and each tree is constructed by randomly sampling features and data points;

[0149] For each tree, the data not participating in tree construction is used for prediction, and the prediction error is calculated;

[0150] Randomly select all values of a cost factor x i in the data not participating in tree construction, make a prediction again, and calculate a new prediction error;

[0151] Calculate the change amount of the prediction error before and after the values of the cost factor x i are shuffled to obtain the importance score of the cost factor x i , and repeat the above operation for each cost factor until the importance scores of all cost factors are obtained;

[0152] The cost factor data of the current smelting cycle of the target kind of iron ore is recorded as a vector x current , and the trained random forest model is used to predict the unit hot metal cost of the current smelting cycle to obtain the first prediction result y 1，pre of the unit hot metal cost of the target kind of iron ore.

[0153] Further optionally, the second prediction module is further used for:

[0154] From the cost factors, the top n cost factors in the importance score ranking are selected as key cost factors;

[0155] Data values of each key cost factor are taken as input variables, and unit hot metal cost of target iron ore is taken as output target, a multiple regression model of key cost factors is constructed, and the model expression is:

[0156] y 2，pre = β0+ β1x1+ β2x2+ … + β n x n + ∈,

[0157] wherein y 2，pre is the second prediction result of unit hot metal cost, x u is the data value of the u-th key cost factor, β u is a regression coefficient, and ∈ is an error term.

[0158] Further optionally, the second prediction module is further used for:

[0159] assigning different weights to initial pheromone values of each regression coefficient;

[0160] taking the importance score of a key cost factor as heuristic information of the corresponding regression coefficient;

[0161] when the ants search, each ant selects the adjustment direction and amplitude of each regression coefficient according to the current pheromone concentration and heuristic information of each regression coefficient;

[0162] after multiple iterations, an optimal regression coefficient combination {β0, β1, β2, …, β n} is obtained, and an optimized multiple regression model is constructed;

[0163] based on the data of each key cost factor of the current smelting period of the target type of iron ore, the optimized multiple regression model is used to predict and calculate the unit hot metal cost of the current smelting period, and a second prediction result y 2，pre of the optimized unit hot metal cost of the target type of iron ore is obtained.

[0164] Further optionally, the second prediction module is further used for:

[0165] the pheromone is updated according to the prediction accuracy of the multiple regression model, and the formula is:

[0166]

[0167] wherein, is the pheromone concentration of the v-th ant on the u-th regression coefficient β u after updating, ρ is a pheromone evaporation coefficient, Q is a constant, indicating the strength of pheromone enhancement, and MSE vis the mean square error of the multiple regression model after adjusting the regression coefficient for the vth ant, which reflects the prediction accuracy of the multiple regression model.

[0168] Further optionally, the second prediction module is further configured to:

[0169] For each ant v and each regression coefficient β u , according to the pheromone concentration and heuristic information, calculate the ant v selection adjustment regression coefficient β u The probability P vu for:

[0170]

[0171] Among them, P vu Select the regression coefficient β for ant v u The probability of making an adjustment, τ vu , τ vg are the regression coefficients β of the vth ant respectively. u , β g The pheromone concentration on the , α1 is the parameter that controls the relative importance of pheromone, α2 is the parameter that controls the relative importance of heuristic information, H u 、H g The regression coefficients β u , β g Heuristic information of , n is the total number of regression coefficients, that is, the number of key cost factors;

[0172] Generate a first random number r1 between 0 and 1 and calculate the cumulative probability Start from u=1 and calculate until you find a v(u-1) <r1≤CP vu u, then ant v chooses to adjust the currently found regression coefficient β u , r1 is the first random number between 0 and 1;

[0173] Update the selected adjusted regression coefficient β according to the random adjustment direction and dynamically set amplitude u value.

[0174] Further optionally, the second prediction module is further configured to:

[0175] Generate a second random number r2 between 0 and 1. If r2<0.5, increase the selected adjusted regression coefficient β u If r2≥0.5, reduce the selected adjusted regression coefficient β u The value of

[0176] Generate a third random number r3 between 0 and 1 and calculate the selected adjusted regression coefficient β uthe adjustment range Δβ u = min_step + r3·(max_step-min_step), [min_step, max_step] is the step range;

[0177] According to the determined adjustment direction and range, the value of the regression coefficient β u is updated as follows:

[0178] If the value of the regression coefficient β u is increased, the updated value is:

[0179] If the value of the regression coefficient β u is decreased, the updated value is: wherein, is the value of the regression coefficient β u before the update.

[0180] Further optionally, the final molten iron cost module is further configured to:

[0181] weight and fuse the first prediction result y 1，pre and the second prediction result y 2，pre of the unit molten iron cost of each iron ore to obtain the final unit molten iron cost y final，pre = λ·y 1，pre +(1-λ)·y 2，pre of each iron ore, wherein λ is a weighting coefficient.

[0182] Further optionally, the data collection module is further configured to:

[0183] The cost factor data includes iron ore index, coke index, ore consumption, fuel consumption, flux cost, wind water electricity cost, worker salary, manufacturing cost, depreciation, and current RMB exchange rate data.

[0184] The above-described embodiments only express several embodiments of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the present application. It should be noted that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of the present application. Therefore, the scope of protection of the present application should be subject to the appended claims.

Claims

1. The iron ore smelting cost analysis method based on multidimensional data fusion is characterized by: The method comprises: Collect data on various cost factors related to unit molten iron cost in each iron ore smelting process; The random forest algorithm was used to analyze the importance of each cost factor and make a preliminary forecast of the unit molten iron cost, obtaining the first forecast result; Based on the results of the importance analysis of cost factors, a multivariate regression model of key cost factors was constructed, and the parameters of the multivariate regression model were optimized using the ant colony algorithm. The optimized multivariate regression model was used to re-predict the unit molten iron cost and obtain the second prediction result; Merging the first prediction result and the second prediction result of the unit molten iron cost of each iron ore to obtain the final unit molten iron cost of each iron ore; The smelting cost-effectiveness of various iron ores is compared based on the final unit molten iron cost of each iron ore.

2. The iron ore smelting cost analysis method based on multidimensional data fusion according to claim 1, characterized in that: The step of using the random forest algorithm to perform importance analysis of each cost factor and unit molten iron cost prediction to obtain a first prediction result includes: For the target type of iron ore, the pre-processed cost factor data of multiple historical smelting cycles are used as input and recorded as a feature matrix X. Each row in the feature matrix X represents a smelting cycle, and each column represents a cost factor. The unit hot metal cost of the target type of iron ore is taken as the output target, recorded as vector Y, where each element corresponds to the unit hot metal cost of a smelting cycle; Construct a random forest model based on the feature matrix X and the output target Y. The random forest consists of multiple decision trees, each of which is constructed by randomly sampling features and data points; For each tree, predictions are made using data that did not participate in tree construction, and the prediction error is calculated; Randomly select a cost factor x from the data that does not participate in tree construction i For all values ​​of , make predictions again and calculate the new prediction errors; Calculate the disruption cost factor x i The change in prediction error before and after the value of is obtained to obtain the cost factor x i The importance score of cost factors is obtained, and the above operation is repeated for each cost factor until the importance scores of all cost factors are obtained; The cost factor data of the current smelting cycle of the target type of iron ore is recorded as vector x current , use the trained random forest model to predict the unit molten iron cost of the current smelting cycle, and get the first prediction result y of the unit molten iron cost of the target type of iron ore 1，pre .

3. The iron ore smelting cost analysis method based on multidimensional data fusion according to claim 2, characterized in that: The steps of constructing a multivariate regression model of key cost factors based on the importance analysis results of cost factors include: Select n cost factors with the highest importance scores from various cost factors as key cost factors; The data values ​​of each key cost factor are used as input variables, and the unit molten iron cost of the target type of iron ore is used as the output target to construct a multiple regression model of key cost factors. The model expression is: y 2，pre =β0+β1x1+β2x2+…+β n x n +∈, Among them, y 2，pre is the second prediction result of unit molten iron cost, x u is the data value of the u-th key cost factor, β u is the regression coefficient, and ∈ is the error term.

4. The iron ore smelting cost analysis method based on multidimensional data fusion according to claim 3, characterized in that: The step of optimizing the parameters of the multivariate regression model by using the ant colony algorithm and using the optimized multivariate regression model to predict the unit molten iron cost to obtain the second prediction result comprises: Assign different weights to the initial pheromone value of each regression coefficient; The importance scores of key cost factors are used as heuristic information for the corresponding regression coefficients; When ants search, each ant chooses the adjustment direction and magnitude of each regression coefficient based on the current pheromone concentration and heuristic information of each regression coefficient; After multiple rounds of iteration, the optimal regression coefficient combination {β0, β1, β2, ..., β n }, construct the optimized multiple regression model; Based on the key cost factor data of the current smelting cycle of the target type of iron ore, the optimized multiple regression model is used to predict the unit molten iron cost of the current smelting cycle, and the second prediction result y of the optimized unit molten iron cost of the target type of iron ore is obtained. 2，pre .

5. The iron ore smelting cost analysis method based on multidimensional data fusion according to claim 4, characterized in that: Before the step of each ant selecting the adjustment direction and amplitude of each regression coefficient according to the current pheromone concentration of each regression coefficient and the heuristic information, the method further includes: The pheromone is updated according to the prediction accuracy of the multivariate regression model. The formula is: in, is the regression coefficient β of the vth ant at the uth position after update u The pheromone concentration on the surface, ρ is the pheromone volatility coefficient, Q is a constant, indicating the intensity of pheromone enhancement, MSE v is the mean square error of the multiple regression model after adjusting the regression coefficient for the vth ant, which reflects the prediction accuracy of the multiple regression model.

6. The iron ore smelting cost analysis method based on multidimensional data fusion according to claim 4, characterized in that: The step of each ant selecting the adjustment direction and amplitude of each regression coefficient according to the current pheromone concentration of each regression coefficient and heuristic information includes: For each ant v and each regression coefficient β u , according to the pheromone concentration and heuristic information, calculate the ant v selection adjustment regression coefficient β u The probability P vu for: Among them, P vu Select the regression coefficient β for ant v u The probability of making an adjustment, τ vu , τ vg are the regression coefficients β of the vth ant respectively. u , β g The pheromone concentration on the , α1 is the parameter that controls the relative importance of pheromone, α2 is the parameter that controls the relative importance of heuristic information, H u 、H g The regression coefficients β u , β g Heuristic information of , n is the total number of regression coefficients, that is, the number of key cost factors; Generate a first random number r1 between 0 and 1 and calculate the cumulative probability Start from u=1 and calculate until you find a v(u-1) <r1≤CP vu u, then ant v chooses to adjust the currently found regression coefficient β u , r1 is the first random number between 0 and 1; Update the selected adjusted regression coefficient β according to the random adjustment direction and dynamically set amplitude u The value of .

7. The iron ore smelting cost analysis method based on multidimensional data fusion according to claim 6, characterized in that: The regression coefficient β selected for adjustment is updated according to the random adjustment direction and the dynamically set amplitude. u The steps for finding the value include: Generate a second random number r2 between 0 and 1. If r2<0.5, increase the selected adjusted regression coefficient β u If r2≥0.5, reduce the selected adjusted regression coefficient β u The value of Generate a third random number r3 between 0 and 1 and calculate the selected adjusted regression coefficient β u The adjustment range Δβ u =min_step+r3·(max_step-min_step), [min_step, max_step] is the step size range; According to the determined adjustment direction and amplitude, update the regression coefficient β u Value: If we increase the regression coefficient β u The updated value is: If we reduce the regression coefficient β u The updated value is: in, is the regression coefficient β before updating u The value of .

8. The iron ore smelting cost analysis method based on multidimensional data fusion according to claim 1, characterized in that: The step of fusing the first prediction result and the second prediction result of the unit molten iron cost of each type of iron ore to obtain the final unit molten iron cost of each type of iron ore includes: The first prediction result y of the unit molten iron cost of each iron ore 1，pre and the second prediction result y 2，pre Perform weighted fusion to obtain the final unit molten iron cost y for each type of iron ore final，pre =λ·y 1，pre +(1-λ)·y2,pre, where λ is the weighting coefficient.

9. The iron ore smelting cost analysis method based on multidimensional data fusion according to claim 1, characterized in that: In the step of collecting data on various cost factors related to unit molten iron cost in each iron ore smelting process: The cost factor data include iron ore index, coking coal index, ore consumption, fuel consumption, flux cost, fengshui, electricity and gas expenses, worker wages, manufacturing expenses, depreciation and current RMB exchange rate data.

10. An iron ore smelting cost analysis system based on multidimensional data fusion, used to implement the iron ore smelting cost analysis method based on multidimensional data fusion according to any one of claims 1 to 9, characterized in that: The system comprises: Data collection module: used to collect data on various cost factors related to unit molten iron cost in each iron ore smelting process; The first prediction module is used to use the random forest algorithm to analyze the importance of various cost factors and make a preliminary prediction of the unit molten iron cost to obtain the first prediction result; The second prediction module is used to construct a multiple regression model of key cost factors based on the results of the importance analysis of cost factors, optimize the parameters of the multiple regression model through the ant colony algorithm, and use the optimized multiple regression model to re-predict the unit molten iron cost to obtain a second prediction result; Final hot metal cost module: used to merge the first prediction result and the second prediction result of the unit hot metal cost of each iron ore to obtain the final unit hot metal cost of each iron ore; Cost-effectiveness comparison module: used to compare the smelting cost-effectiveness of various iron ores based on the final unit molten iron cost of each iron ore.

Images