Machine learning based multi-cycle rolling batch optimization method and system for mix sintering
By employing machine learning and nonlinear programming optimization methods, the problems of inaccurate quality prediction and unreasonable inventory management in sintering batching optimization were solved, achieving accurate quality prediction and inventory balance, and improving the feasibility and economy of the batching scheme.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI BAOSIGHT SOFTWARE CO LTD
- Filing Date
- 2026-03-18
- Publication Date
- 2026-07-14
AI Technical Summary
Existing sintering batching optimization methods suffer from problems such as insufficient accuracy in quality prediction, unreasonable inventory management, inability of static optimization models to adapt to dynamic factors, difficulty in cost control, and insufficient balance of multi-variety inventory, resulting in substandard product quality, large inventory fluctuations, and high risk of production interruption.
A machine learning-based approach was adopted, which involved data preprocessing, prediction of sinter quality indicators, dynamic inventory management, and multi-cycle rolling optimization to establish a nonlinear programming model. This model was then solved using heuristic and precise algorithms to optimize the batching scheme for the sintering process.
It has improved the accuracy of sinter quality prediction, enabled scientific inventory management, optimized both cost and quality objectives, enhanced the feasibility and adaptability of batching schemes, and reduced production risks.
Smart Images
Figure CN122390263A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of ironmaking batching technology, specifically to a machine learning-based method and system for optimizing multi-cycle rolling batching in the mixing and sintering process. Background Technology
[0002] In the steel plant production process, the mixing and sintering process is a key link connecting iron ore raw materials and blast furnace smelting. The rationality of its batching scheme directly determines the quality of sinter, production cost and production continuity, involving multiple complex factors such as iron ore inventory, resource delivery plan, ore chemical composition, storage capacity limitations, and cost control.
[0003] Existing sintering batching optimization methods are mainly based on static linear programming or manual empirical rules, which have the following specific technical problems, all of which can be solved by the technical solution of this invention: First, the accuracy of quality prediction is insufficient. Without a standardized machine learning model, it is impossible to accurately predict sinter quality indicators based on iron ore proportions and chemical composition, easily leading to substandard product quality. Second, inventory management lacks a time difference matching mechanism. It does not consider the production time difference between sintering materials and blending materials, and has not established corresponding time difference constraints and dynamic inventory accounting rules, resulting in large inventory fluctuations and a tendency for stockpiling or shortages. Third, the static optimization model cannot adjust the batching scheme according to dynamic factors such as changes in ore arrival and inventory deviations, resulting in poor executability. Fourth, a scientific cost-quality dual-objective optimization framework has not been constructed, making it difficult to achieve cost control while ensuring quality, or to ensure quality while controlling costs. Fifth, a sound multi-variety, multi-time period inventory balance constraint has not been established, making it impossible to achieve an optimal balance between safety stock and cost, increasing the risk of capital occupation and production interruption.
[0004] A search of patent literature revealed that invention patent application number CN202510935690.8 discloses a sintering batching optimization scheme, which uses a random forest surrogate model to predict the performance of the finished ore, and uses NSGA... Algorithm II searches for the Pareto optimal solution set, calculates the index weights using the entropy weight method, and then uses the TOPSIS method to screen the optimal solution, achieving a multi-objective trade-off to improve sinter quality and reduce costs. However, this patent does not consider inventory tracking and feeding schemes during the multi-cycle batching process of mixing and sintering, leading to unreasonable inventory adjustments in practical applications. Patent CN118692594A discloses a sinter batching optimization method and device based on reinforcement learning evolutionary multi-objectives. It inputs batching plans, raw material information, etc., establishes a cost objective function, and solves it using an intelligent optimization algorithm to obtain the optimal solution. However, this patent does not combine with multi-cycle optimization, failing to achieve inventory balance and a reasonable feeding scheme.
[0005] In summary, given the problems of the existing technologies, researching a machine learning-based multi-cycle rolling batching optimization method and system for the mixing and sintering process has become a critical task that urgently needs to be addressed. Summary of the Invention
[0006] To address the shortcomings of existing technologies, the purpose of this invention is to provide a multi-cycle rolling batching optimization method and system for the mixing and sintering process based on machine learning.
[0007] The present invention provides a machine learning-based multi-cycle rolling batching optimization method for homogenization sintering, comprising the following steps:
[0008] Data input steps: Acquire and manage the initial data of the blended sintered ore. The initial data includes initial inventory data, resource arrival plan, resource succession plan, batching constraint configuration and basic data of ore properties; Machine learning prediction steps: After preprocessing the initial data, a pre-trained machine learning model is used to predict the main quality indicators of sinter. Inventory management steps: Perform dynamic inventory management within a set time period; Multi-period rolling optimization modeling steps: Establish a nonlinear programming model containing multiple types of decision variables, set multiple constraints, and construct an objective function that minimizes cost or maximizes quality; Decision-making steps: Use a nonlinear programming algorithm to solve the nonlinear programming model and output the material optimization scheme for the mixing and sintering process.
[0009] Preferably, in the data input step, the basic data of ore properties includes material name, material category, chemical composition, price and inventory; the resource arrival plan includes arrival time and maximum receiving quantity; and the chemical composition includes TFe, SiO2, CaO and MgO.
[0010] Preferably, in the machine learning prediction step, the preprocessing includes data standardization, feature selection, and feature dimensionality reduction; the machine learning model includes linear regression, random forest, and neural network; and the main quality indicators of sinter include TFe, R2, drum index, and basicity.
[0011] Preferably, the inventory management steps include: The time difference between sintering materials and mixing materials is set according to the pile-building mode and the non-pile-building mode. The time difference is 7-15 days in the pile-building mode and 0 days in the non-pile-building mode.
[0012] Preferably, the dynamic inventory change is calculated, with a time series of t = 1 to T days, and the calculation formula is as follows: Under the non-stacking mode, the daily consumption formula for the i-th raw material is as follows: consumption [t][i]= sinter_daily_demand [t]* sinter_ratio [t][blend] / 100*blend_ratio[t][i]; In the stacking mode, the daily consumption of the i-th raw material is calculated using the following formula: consumption [t][i]= sinter_daily_demand [t+delay] * sinter_ratio [t+delay][blend] / 100*blend_ratio [t][i]; Where delay is the time difference between sintering material and blending material, sinter_daily_demand is the daily demand for sintering, sinter_ratio is the proportion of sintering mixture, and blend_ratio is the proportion of ore raw materials in the blend.
[0013] The formula for calculating the daily inventory of the i-th raw material is: inventory[t][i]= inventory[t-1][i]+arrival[t][i]- consumption[t][i]; Where, inventory[t][i] represents the inventory of type i ore on day t, inventory[t-1][i] represents the inventory of type i ore on day t-1, arrival[t][i] represents the arrival quantity of type i ore on day t, and consumption[t][i] represents the consumption quantity of type i ore on day t.
[0014] Preferably, in the multi-cycle rolling optimization modeling step, the multiple decision variables include inventory variables, proportion variables, feed variables and time difference adjustment variables; proportion variable x[i] = model.addVar(lb=lower_bound, ub=upper_bound), and inventory variable inventory[t][i] = model.addVar(lb=0).
[0015] Preferably, in the multi-period rolling optimization modeling step, the multiple constraints include inventory balance constraints, ingredient allocation constraints, capacity constraints, quality constraints, and time difference constraints. The ingredient allocation constraints include the total proportion constraint, and the formula is: quicksum (x_blend [i] for i in materials) == 100, Where x_blend[i] is the ratio variable of each ore raw material, and materials represents the total number of ore raw materials.
[0016] Inventory balance constraints include daily inventory constraints for the i-th type of raw material, as shown in the formula: inventory [t][i]>= safety_stock; Here, safety_stock represents the safety stock level of ore raw materials.
[0017] Preferably, the decision-solving steps include: integrating heuristic algorithms and exact algorithms to form multiple solution strategies, calling a nonlinear programming solver to solve the nonlinear programming model using the solution strategies, extracting the decision variable values corresponding to the optimal solution of the model, and outputting the batching scheme, sinter quality prediction results, and batching cost analysis results for the homogenizing and sintering process.
[0018] This invention also provides a machine learning-based multi-cycle rolling batching optimization system for homogenized sintering, employing the aforementioned machine learning-based multi-cycle rolling batching optimization method for homogenized sintering, comprising: Data input module: Acquires and manages initial data for blended sintered ore, including initial inventory data, resource arrival plan, resource succession plan, batching constraint configuration, and basic data on ore properties; Machine learning prediction module: After preprocessing the initial data, it uses a pre-trained machine learning model to predict the main quality indicators of sinter. Inventory Management Module: Performs dynamic inventory management within a set time period; Multi-period rolling optimization modeling module: Establish a nonlinear programming model containing multiple types of decision variables, set multiple constraints, and construct an objective function that minimizes cost or maximizes quality; Decision Solving Module: Uses nonlinear programming algorithm to solve nonlinear programming model and outputs optimized material batching scheme for the mixing and sintering process.
[0019] Preferably, in the data input module, the basic data of ore properties includes material name, material category, chemical composition, price and inventory; the resource arrival plan includes arrival time and maximum receiving quantity; the chemical composition includes TFe, SiO2, CaO and MgO.
[0020] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention standardizes, filters features, and performs dimensionality reduction preprocessing on initial data. It then uses pre-trained machine learning models such as linear regression, random forest, or neural networks to accurately predict quality indicators of sinter such as TFe, R2, drum index, and basicity based on iron ore proportions and chemical composition, thus solving the problem of insufficient accuracy in quality prediction by traditional methods.
[0021] 2. By distinguishing between stockpiling mode and non-stockpiling mode, and combining dynamic inventory recursion formula, this invention can calculate the daily inventory balance of each iron ore within a preset period, set safety stock constraints to avoid stockpiling or shortage, and realize scientific inventory management of the mixing and sintering process.
[0022] 3. This invention establishes a nonlinear programming model that includes inventory variables, proportioning variables, material input variables, and time difference adjustment variables, and configures inventory balance constraints, material input constraints, capacity constraints, quality constraints, and time difference constraints, thereby ensuring the feasibility and executability of the material input scheme under multi-dimensional constraints.
[0023] 4. This invention utilizes a nonlinear programming solution strategy that integrates heuristic and exact algorithms to construct an objective function that minimizes cost or maximizes quality while satisfying quality constraints. This allows for obtaining an optimal ingredient ratio that balances quality and cost within a finite time, thus achieving dual-objective optimization of cost and quality.
[0024] 5. This invention, through a multi-cycle rolling optimization mechanism, can update data and re-solve based on dynamic factors such as changes in resource arrival at the plant and inventory deviations, thus solving the defect that traditional static batching schemes cannot adapt to dynamic changes in production and improving the continuity and adaptability of batching optimization. Attached Figure Description
[0025] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart illustrating a machine learning-based multi-cycle rolling batching optimization method for homogenization and sintering, as described in an embodiment of the present invention. Detailed Implementation
[0026] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.
[0027] This invention discloses a machine learning-based multi-cycle rolling batching optimization method and system for homogenized sintering. The method acquires initial data of homogenized sintering ore, preprocesses the data, and uses a pre-trained machine learning model to predict the main quality indicators of the sinter. Within a set time period, it distinguishes between stockpiling and non-stockpiling modes for dynamic inventory management, establishes a nonlinear programming model containing multiple decision variables, sets multiple constraints, and constructs an objective function for minimizing cost or maximizing quality. A nonlinear programming algorithm integrating heuristic and exact algorithms is used to solve the model, outputting an optimized batching scheme for the homogenized sintering process. This invention can minimize batching costs while meeting the quality requirements of the sinter, taking into account the time difference between sintering and homogenizing materials and iron ore inventory balance management, thus improving the feasibility and economy of the batching scheme.
[0028] Example 1: Figure 1 This is a flowchart illustrating a machine learning-based multi-cycle rolling batching optimization method for homogenization and sintering, as described in an embodiment of the present invention.
[0029] like Figure 1 As shown, this embodiment provides a multi-cycle rolling batching optimization method for the mixing and sintering process based on machine learning, including the following steps: Data input steps: Acquire and manage initial data for blended sintered ore. Initial data includes initial inventory data, resource delivery plan, resource succession plan, batching constraint configuration, and basic data on ore properties. Specifically, basic data on ore properties includes material name, material category, chemical composition such as TFe, SiO2, CaO, and MgO, price, and inventory. Resource delivery plan includes key information such as delivery time and maximum receiving quantity.
[0030] In this embodiment, initial data is read, including initial inventory {'KSK001': 20000, 'KSK003': 18000, 'KSK015': 16000}, delivery schedule {day_1: {'KSK001': {'max_receive': 10000}}}, and material information in the batching constraint configuration [{'materialID': 'KSK001', 'Category': 'Iron ore powder, hematite', 'Price': 850.50, 'TFe': 62.5}].
[0031] Machine learning prediction steps: After preprocessing the basic data such as initial inventory data, resource arrival plan, resource succession plan, batching constraint configuration, and ore properties, a pre-trained machine learning model is used to predict the main quality indicators of sinter.
[0032] Specifically, the machine learning models include linear regression, random forest, and neural networks. Key quality metrics include TFe, R2, drum index, and alkalinity. Preprocessing includes data standardization, feature selection, and feature dimensionality reduction, resulting in a format suitable for machine learning models.
[0033] In this embodiment, a pre-trained model is loaded, including a normalizer scaler_model['TFe'] = MinMaxScaler(), a dimensionality reduction model DR_model['TFe'] = DR_MODEL (n_components=6), and a predictor ['TFe'] = RandomForestRegressor(). The input features are then normalized, such as feature_normalized = (feature_value - data_min) / (data_max - data_min), and feature filtering and dimensionality reduction are performed sequentially to obtain the model input feature set.
[0034] Inventory management steps: Perform dynamic inventory management within a set time period.
[0035] Considering the time difference between the sintering process and the mixing process, corresponding time differences are set for the stacking mode and the non-stacking mode. The time difference in the stacking mode is 7-15 days, and the time difference in the non-stacking mode is 0 days.
[0036] Based on the usage and stacking status of the mixing piles over a period of time, the specific time difference between the use and stacking of each mixing pile is determined. By adjusting the inventory variables, the time of mixing output and sintering input is matched, thereby reducing inventory costs.
[0037] Specifically, to calculate dynamic inventory changes, for a set of mixing and sintering processes, a time series of t=1~T days is set.
[0038] Under the non-stacking mode, the daily consumption formula for the i-th raw material is as follows: consumption [t][i]= sinter_daily_demand [t]* sinter_ratio [t][blend] / 100*blend_ratio[t][i]; In the stacking mode, the daily consumption of the i-th raw material is calculated using the following formula: consumption [t][i]= sinter_daily_demand [t+delay] * sinter_ratio [t+delay][blend] / 100*blend_ratio [t][i]; Where delay is the time difference between sintering material and blending material, sinter_daily_demand is the daily demand for sintering, sinter_ratio is the proportion of sintering mixture, and blend_ratio is the proportion of ore raw materials in the blend.
[0039] The formula for calculating the daily inventory of the i-th raw material is: inventory[t][i]= inventory[t-1][i]+arrival[t][i]- consumption[t][i]; Where, inventory[t][i] represents the inventory of type i ore on day t, inventory[t-1][i] represents the inventory of type i ore on day t-1, arrival[t][i] represents the arrival quantity of type i ore on day t, and consumption[t][i] represents the consumption quantity of type i ore on day t.
[0040] Multi-period rolling optimization modeling steps: Establish a nonlinear programming model containing multiple types of decision variables, set multiple constraints, construct an objective function to minimize cost or maximize quality, and realize multi-period rolling optimization of the mixed sintering batch.
[0041] Specifically, the decision variables include inventory variables, allocation variables, material input variables, and time difference adjustment variables. The allocation variable x[i] = model.addVar(lb=lower_bound, ub=upper_bound), and the inventory variable inventory[t][i] = model.addVar(lb=0).
[0042] Multiple constraints include inventory balance constraints, ingredient allocation constraints, capacity constraints, quality constraints, and time difference constraints. Ingredient allocation constraints include a total proportion constraint, the formula of which is: quicksum (x_blend [i] for i in materials) == 100, Where x_blend[i] is the ratio variable of each ore raw material, and materials represents the total number of ore raw materials.
[0043] Inventory balance constraints include daily inventory constraints for the i-th type of raw material, as shown in the formula: inventory [t][i]>= safety_stock; Here, safety_stock represents the safety stock level of ore raw materials.
[0044] Decision-making steps: The nonlinear programming model is solved by using a nonlinear programming algorithm. Heuristic and exact algorithms are integrated to provide multiple solution strategies, ensuring that a high-quality multi-cycle rolling batching scheme for the mixing and sintering process is obtained within a limited time.
[0045] Specifically, the nonlinear programming solver is invoked to solve the model, and model.optimize() is executed to complete the solution calculation; the solution x_vals = {i: model.getVal(x[i])} is extracted, and the results of the batching scheme {'KSK001': 20.5%, 'KSK003': 18.2%}, quality prediction {'TFe': 58.3, 'R2': 1.85}, and cost analysis {'total_cost': 950.2} are output.
[0046] This invention aims to solve the technical problems in traditional sintering batching optimization methods, such as inaccurate quality prediction, unbalanced multi-cycle inventory management, poor cost control, and neglect of mixing and sintering time differences. Through the collaborative process of data input, machine learning prediction, inventory management, multi-cycle rolling optimization modeling, and decision solving, it improves the hit rate of sinter quality prediction, minimizes batching costs, and achieves dynamic inventory balance management, thereby enhancing the feasibility and economy of batching schemes in the mixing and sintering process.
[0047] Example 2: The present invention also provides a machine learning-based multi-cycle rolling batching optimization system for the mixing and sintering process. The machine learning-based multi-cycle rolling batching optimization system for the mixing and sintering process can be implemented by executing the process steps of the machine learning-based multi-cycle rolling batching optimization method for the mixing and sintering process. That is, those skilled in the art can understand the machine learning-based multi-cycle rolling batching optimization method for the mixing and sintering process as a preferred embodiment of the machine learning-based multi-cycle rolling batching optimization system for the mixing and sintering process.
[0048] Specifically, the machine learning-based multi-cycle rolling batching optimization system for the mixing and sintering process includes: Data Input Module: Acquires and manages initial data for blended sintered ore, including initial inventory data, resource arrival plan, resource succession plan, batching constraint configuration, and basic data on ore properties.
[0049] The basic data for ore properties includes material name, material category, chemical composition, price, and inventory; the resource delivery plan includes delivery time and maximum receiving quantity; the chemical composition includes TFe, SiO2, CaO, and MgO.
[0050] Machine learning prediction module: After preprocessing the initial data, it uses a pre-trained machine learning model to predict the main quality indicators of sinter. Inventory Management Module: Performs dynamic inventory management within a set time period; Multi-period rolling optimization modeling module: Establish a nonlinear programming model containing multiple types of decision variables, set multiple constraints, and construct an objective function that minimizes cost or maximizes quality; Decision Solving Module: Uses nonlinear programming algorithm to solve nonlinear programming model and outputs optimized material batching scheme for the mixing and sintering process.
[0051] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.
[0052] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.
Claims
1. A machine learning-based method for optimizing multi-cycle rolling batching in homogenization and sintering, characterized in that, Includes the following steps: Data input steps: Acquire and manage initial data for blended sintered ore, including initial inventory data, resource arrival plan, resource succession plan, batching constraint configuration, and basic data on ore properties; Machine learning prediction steps: After preprocessing the initial data, a pre-trained machine learning model is used to predict the main quality indicators of sinter. Inventory management steps: Perform dynamic inventory management within a set time period; Multi-period rolling optimization modeling steps: Establish a nonlinear programming model containing multiple types of decision variables, set multiple constraints, and construct an objective function that minimizes cost or maximizes quality; Decision-making steps: The nonlinear programming model is solved using a nonlinear programming algorithm, and the optimized batching scheme for the mixing and sintering process is output.
2. The machine learning-based multi-cycle rolling batching optimization method for homogenization and sintering according to claim 1, characterized in that, In the data input step, the basic data of the ore attributes includes material name, material category, chemical composition, price, and inventory; the resource delivery plan includes delivery time and maximum receiving quantity; the chemical composition includes TFe, SiO2, CaO, and MgO.
3. The machine learning-based multi-cycle rolling batching optimization method for homogenization and sintering according to claim 1, characterized in that, In the machine learning prediction step, the preprocessing includes data standardization, feature selection, and feature dimensionality reduction; the machine learning model includes linear regression, random forest, and neural network; and the main quality indicators of the sinter include TFe, R2, drum index, and alkalinity.
4. The machine learning-based multi-cycle rolling batching optimization method for homogenization and sintering according to claim 1, characterized in that, The inventory management steps include: The time difference between sintering materials and mixing materials is set according to the pile-building mode and the non-pile-building mode. The time difference is 7-15 days in the pile-building mode and 0 days in the non-pile-building mode.
5. The machine learning-based multi-cycle rolling batching optimization method for homogenization and sintering according to claim 4, characterized in that, To calculate dynamic inventory changes, given a time series of t = 1 to T days, the calculation formula is as follows: Under the non-stacking mode, the daily consumption formula for the i-th raw material is as follows: consumption [t][i] = sinter_daily_demand [t]* sinter_ratio [t][blend] / 100*blend_ratio[t][i]; In the stacking mode, the daily consumption of the i-th raw material is calculated using the following formula: consumption [t][i] = sinter_daily_demand [t+delay] * sinter_ratio [t+delay][blend] / 100*blend_ratio [t][i]; Where delay is the time difference between sintering material and blending material, sinter_daily_demand is the daily demand for sintering, sinter_ratio is the proportion of sintering mixture, and blend_ratio is the proportion of ore raw materials in the blend. The formula for calculating the daily inventory of the i-th raw material is: inventory[t][i] = inventory[t-1][i]+arrival[t][i]- consumption[t][i]; Where, inventory[t][i] represents the inventory of type i ore on day t, inventory[t-1][i] represents the inventory of type i ore on day t-1, arrival[t][i] represents the arrival quantity of type i ore on day t, and consumption[t][i] represents the consumption quantity of type i ore on day t.
6. The machine learning-based multi-cycle rolling batching optimization method for homogenization and sintering according to claim 1, characterized in that, In the multi-cycle rolling optimization modeling step, the various decision variables include inventory variables, proportioning variables, material input variables, and time difference adjustment variables; The ratio variable x[i] = model.addVar(lb=lower_bound, ub=upper_bound), and the inventory variable inventory[t][i] = model.addVar(lb=0).
7. The machine learning-based multi-cycle rolling batching optimization method for homogenization and sintering according to claim 1, characterized in that, In the multi-period rolling optimization modeling step, the multiple constraints include inventory balance constraints, ingredient allocation constraints, capacity constraints, quality constraints, and time difference constraints. The ingredient allocation constraints include a total proportion constraint, and the formula is: quicksum (x_blend [i] for i in materials) == 100, Where x_blend[i] is the ratio variable of each ore raw material, and materials represents the total number of ore raw materials; The inventory balance constraint includes the daily inventory quantity constraint for the i-th type of raw material, as shown in the formula: inventory [t][i]>= safety_stock; Here, safety_stock represents the safety stock level of ore raw materials.
8. The machine learning-based multi-cycle rolling batching optimization method for homogenization and sintering according to claim 1, characterized in that, The decision-solving steps include: integrating heuristic algorithms and exact algorithms to form multiple solution strategies; calling a nonlinear programming solver to solve the nonlinear programming model using the solution strategies; extracting the decision variable values corresponding to the optimal solution of the model; and outputting the batching scheme, sinter quality prediction results, and batching cost analysis results for the homogenizing and sintering process.
9. A machine learning-based multi-cycle rolling batching optimization system for homogenizing sintering, employing the machine learning-based multi-cycle rolling batching optimization method for homogenizing sintering as described in any one of claims 1-8, characterized in that, include: Data input module: Acquires and manages initial data of blended sintered ore, including initial inventory data, resource arrival plan, resource succession plan, batching constraint configuration and basic data of ore properties; Machine learning prediction module: After preprocessing the initial data, a pre-trained machine learning model is used to predict the main quality indicators of sinter. Inventory Management Module: Performs dynamic inventory management within a set time period; Multi-period rolling optimization modeling module: Establish a nonlinear programming model containing multiple types of decision variables, set multiple constraints, and construct an objective function that minimizes cost or maximizes quality; Decision Solving Module: The nonlinear programming algorithm is used to solve the nonlinear programming model and output the material optimization scheme for the mixing and sintering process.
10. The machine learning-based multi-cycle rolling batching optimization system for homogenizing and sintering according to claim 9, characterized in that, In the data input module, the basic data of the ore attributes includes material name, material category, chemical composition, price, and inventory; the resource delivery plan includes delivery time and maximum receiving quantity; the chemical composition includes TFe, SiO2, CaO, and MgO.