A raw material yard pre-blending real-time optimization and closed-loop control method, device, equipment and storage medium

Abstract

This invention relates to the field of raw material yard pre-mixing technology, providing a method, apparatus, equipment, and storage medium for real-time optimization and closed-loop control of raw material yard pre-mixing. By constructing a three-dimensional gridded digital twin model of the stockpile quality, real-time component data sampled at the batching and feeding device is precisely bound to the spatial location of the stockpile, achieving transparency in stockpile quality. Based on the updated digital twin model, with the stability of key chemical components of the blended ore as a constraint and the minimum total cost as the objective, linear programming is used to solve for the optimal feed quantity sequence, generating a pre-mixing operation plan. Simultaneously, based on deviation statistics, adaptive tuning optimizes the model parameters, achieving self-evolving closed-loop control. This invention replaces manual experience-based decision-making, transforming blended ore quality control from post-event verification to pre-event prediction and in-event control, significantly improving component stability and reducing batching costs. It can be widely applied to solid raw material batching scenarios in process industries.

Description

Technical Field

[0001] This invention relates to the field of raw material pre-mixing technology, and in particular to a method, apparatus, equipment and storage medium for real-time optimization and closed-loop control of raw material pre-mixing. Background Technology

[0002] In many process industries such as steel, non-ferrous metallurgy, building materials, power, and chemicals, the precise proportioning of various solid raw materials is a crucial process affecting quality, cost, and energy consumption. Currently, in the steel smelting sector, pre-mixing at the raw material yard is used to receive and store large quantities of iron ore of different types and grades. Through pre-mixing, a stable (especially TFe and SiO2) blended ore is prepared to provide high-quality raw materials for the downstream sintering process. This process is large-scale and difficult to control, and existing technologies face the following long-standing unresolved problems: (1) The quality of the stockpile is a “black box”: the actual composition of the ore in each stockpile and layer in the huge stockpile deviates from the initial data and is affected by weather and material handling operations. Traditional management relies on initial verification forms, resulting in serious data inaccuracies.

[0003] (2) Delayed verification of the mixing effect: The quality of the mixed ore can only be known after the batching and feeding device (such as the batching belt scale in the pre-batching room) completes a stack or even after sintering. If it is not qualified, the whole stack will be scrapped or the quality of the sintered ore will fluctuate, resulting in huge economic losses.

[0004] (3) Pre-mixing decision based on human experience: How to calculate a pre-mixing scheme that can output the target component of the mixed ore based on the uncertain quality status of each stockpile is extremely dependent on the personal experience of experts, and cannot achieve scientific and quantitative optimal decision-making.

[0005] Chinese patent CN114516553B designed a multi-variety iron ore blending production system in the port and used online detection and dynamic adjustment technology. However, its detection parameters are still not perfect and it needs to rely on offline test data input, which cannot realize the real-time perception of the quality of the stockpile. The literature "Analysis of Smart Steel Stockyard" (Metallurgical Equipment, No. 293, 2025 Special Issue (1)) shows that the construction of smart stockyard has significantly improved management efficiency, but it still needs to combine the material demand information of the raw material plant on-site online maintenance to form a material demand table, which lacks scientific real-time optimization methods.

[0006] In view of this, it is necessary to propose a method, device, equipment and storage medium for real-time optimization and closed-loop control of pre-mixed raw materials in the raw material yard to solve or at least alleviate the above-mentioned defects. Summary of the Invention

[0007] The main objective of this invention is to provide a method, apparatus, equipment, and storage medium for real-time optimization and closed-loop control of pre-mixed raw material yards, in order to solve the technical problems of unknown stockpile quality status, reliance on experience for batching decisions, and open-loop lag in process control in pre-mixed raw material yards.

[0008] To achieve the above objectives, the present invention provides a real-time optimization and closed-loop control method for pre-mixing raw materials in a raw material yard, comprising the following steps: S1, acquire first analytical data uploaded from a manual or online sampling device installed on the batching and feeding device; wherein, the first analytical data includes key chemical components and moisture data of the material at the current stockpile; S2, associate the first analysis data with the corresponding stockpile identification information, and update the real-time quality data of the corresponding stockpile material in the pre-constructed stockpile quality digital twin model according to the association relationship; wherein, the real-time quality data includes key chemical components and moisture data; S3, based on the updated stockpile quality digital twin model, obtain the key chemical component data from the real-time quality data of all available stockpile materials; then, with the optimization objective of stabilizing the key chemical components of the output blended ore within the target range within a future operating cycle, call the pre-configured multi-objective optimization model for processing; wherein, the input data of the processing includes the key chemical component data from the real-time quality data, the preset target components of the blended ore, and the cost data of each stockpile material; S4, Generate a pre-batching operation plan and control the operation of the batching and feeding device to form a mixed ore; wherein, the pre-batching operation plan includes the optimal material taking sequence from each available stockpile in the future operation cycle, obtained by solving the multi-objective optimization model; Preferably, the method further includes a feedback correction step S5, which involves acquiring second analysis data of the blended ore, feeding back the second analysis data to correct the stockpile quality digital twin model and / or the multi-objective optimization model, and initiating a new round of optimization control process based on the correction results.

[0009] Preferably, step S2 includes the following steps: S21, associate the first analysis data with the corresponding stockpile location identification information; wherein, the stockpile location identification information is generated by the batching and feeding device according to its operating position; the stockpile quality digital twin model maps the physical stockpile in the form of a three-dimensional grid, and each grid cell is associated with a unique stockpile location identification and stores the time-series quality data of the corresponding material; S22, Based on the pile location identification information, extract multiple historical key chemical component data, historical moisture data and historical particle size data that have been stored in the corresponding grid cell in the digital twin model of the pile quality in a preset historical period to form a historical quality data sequence; S23, compare the key chemical component data and moisture data in the first analysis data with the corresponding category data in the historical quality data sequence according to time; S24. For each type of data in the first analysis data, calculate the statistical mean and standard deviation of the historical quality data sequence, and set the reasonableness condition that the data value of that type in the first analysis data falls within a preset interval. If the reasonableness condition is met, then execute step S25. S25, perform data fusion, where for each type of data, calculate and generate updated real-time quality data; S26, the updated real-time quality data is stored as a new time-series data point in the grid cell corresponding to the stack location identifier in the digital twin model of the stack quality.

[0010] Preferably, step S3 includes the following steps: S31, Based on the updated digital twin model of the stockpile quality, obtain real-time key chemical composition data of all available stockpile materials, unit cost data of each stockpile material, and current available stock of each stockpile material. S32, establish a linear programming model with minimizing total cost as the optimization direction, set the planned material extraction quantity of each stockpile as the decision variable, and set constraints; wherein, the constraints include: the predicted blending ore composition calculated based on the real-time key chemical composition data of the material at each stockpile and the decision variable must fall within the target range, thus serving as a hard constraint to achieve multi-objective synergistic optimization of quality and cost; and the planned material extraction quantity of each stockpile does not exceed the corresponding current available inventory and is a non-negative number; S33, Solve the linear programming model, and output the optimal decision variable values ​​obtained as the optimal material taking sequence from each available stack location in the next working cycle.

[0011] Preferably, step S4 includes the following steps: S41, the optimal material taking sequence is decomposed into multiple consecutive operation periods in chronological order, and a corresponding operation task is generated for each operation period; wherein, each operation task includes the planned material taking amount within the operation period; S42, the time period operation task is converted into a control command for the batching and feeding device to be executed, and sent to the corresponding batching and feeding device to form a mixed ore.

[0012] Preferably, step S5 includes the following steps: S51, Obtain second analytical data by online sampling and analysis of the homogenized ore formed in step S4; wherein, the second analytical data includes the measured values ​​of the key chemical components of the homogenized ore; S52, compare the measured values ​​of key chemical components in the second analysis data with the predicted values ​​calculated based on the real-time quality data of the corresponding stockpile material in the digital twin model of stockpile quality and the actual material handling data that has been executed, to obtain the comparison deviation, and perform reverse correction on the real-time quality data of the corresponding stockpile material in the digital twin model of stockpile quality based on the comparison deviation. S53, Based on the statistical analysis of the deviation between the second analysis data and the predicted value, the model parameters used to calculate the predicted composition of the mixed ore in the multi-objective optimization model are adjusted; S54, determine whether the conditions for triggering a new round of optimization are met; wherein, the triggering conditions include at least one of the following: reaching the preset fixed operation cycle end point, the actual material taking amount accumulated since the last optimization reaching a threshold, or the comparison deviation exceeding the preset tolerance. S55, if the optimization triggering condition is met, then based on the corrected digital twin model of the stockpile quality, a new round of optimization control process starting from step S1 is initiated.

[0013] Preferably, the reverse correction of the real-time quality data of the corresponding stockpile material in the digital twin model of stockpile quality based on the comparison deviation in step S52 includes the following steps: The contribution weight of each stockpile to the comparison deviation is calculated based on the proportion of the actual material taken from each stockpile to the total material taken in the operation cycle of forming the mixed ore. For each stockpile participating in this operation in the stockpile quality digital twin model, the correction amount for the real-time key chemical composition data of the material at that stockpile is calculated in reverse based on the comparison deviation, the contribution weight, and the actual material handling data that has been executed. Obtain the historical reliability index of the real-time quality data of the material at the stockpile location in the digital twin model of stockpile quality; adjust the correction amount by weighting according to the historical reliability index to obtain the final correction value; The final correction value is applied to the real-time key chemical composition data of the corresponding stockpile material in the digital twin model of stockpile quality to complete the reverse correction.

[0014] The present invention also provides a real-time optimization and closed-loop control device for pre-mixing raw material yards, used to execute the real-time optimization and closed-loop control method for pre-mixing raw material yards as described above, comprising: The sampling and analysis unit is used to acquire first analysis data uploaded from a manual or online sampling device installed on the batching and feeding device; wherein, the first analysis data includes key chemical components and moisture data of the material at the current stockpile. An update unit is used to associate the first analysis data with the corresponding stockpile identification information, and update the real-time quality data of the corresponding stockpile material in the pre-constructed stockpile quality digital twin model according to the association relationship; wherein, the real-time quality data includes key chemical components and moisture data; The optimization calculation unit is used to obtain key chemical component data from the real-time quality data of all available stockpile materials based on the updated stockpile quality digital twin model; then, with the optimization objective of stabilizing the key chemical components of the output blended ore within the target range in the next operating cycle, it calls a pre-configured multi-objective optimization model for processing; wherein, the input data of the processing includes the key chemical component data in the real-time quality data, the preset target components of the blended ore, and the cost data of each stockpile material; A control unit is used to generate a pre-batching operation plan and control the operation of the batching and feeding device to form a mixed ore; wherein, the pre-batching operation plan includes the optimal material extraction sequence from each available stockpile within a future operation cycle, obtained by solving the multi-objective optimization model; The feedback correction unit is used to acquire the second analysis data of the blended ore, feed the second analysis data back to correct the stockpile quality digital twin model and the multi-objective optimization model, and then return to step S1 to start a new round of optimization control process.

[0015] The present invention also provides a real-time optimization and closed-loop control device for pre-mixing raw materials, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the real-time optimization and closed-loop control method for pre-mixing raw materials as described above.

[0016] The present invention also provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for real-time optimization and closed-loop control of raw material pre-mixing.

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention utilizes manual or online sampling devices on the batching and feeding equipment to perform real-time analysis of the stockpiled materials during operations, obtaining first-hand, accurate composition data. Furthermore, by constructing a digital twin model of the stockpile quality using a three-dimensional grid as the unit, it precisely binds discrete detection data to the spatial location of the stockpile, achieving a complete mapping from the physical stockpile to a digital mirror. This completely changes the traditional blind management of raw material yards, providing a reliable data foundation for precise decision-making. Based on the updated digital twin model, this invention invokes a multi-objective optimization model centered on linear programming, with the stability of the blended ore composition as a hard constraint and the minimum total batching cost as the optimization objective, automatically solving for the optimal solution for future operation cycles. By optimizing the feed rate sequence, the quality and cost are synergistically optimized. This transforms the quality control of blended ore from post-event verification to pre-event prediction and in-event control, significantly reducing the standard deviation of blended ore and creating stable conditions for downstream sintering production. The invention utilizes an online analyzer on the blended ore output conveyor to obtain the measured composition of the finished product. Through a reverse correction algorithm, the deviation between the measured and predicted values ​​is traced back to each participating stockpile. The digital twin model is then differentiated and corrected based on historical reliability indicators. Simultaneously, the optimization model parameters are adaptively adjusted based on the statistical characteristics of the deviation, achieving self-evolution and eliminating the long-term dependence on offline testing and manual parameter adjustment. This reduces the cost per ton of blended ore and improves stockpile operation efficiency. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the downstream sintering process; Figure 2 This is a schematic diagram showing the connection relationship of the pre-mixing chamber process equipment in one embodiment of the present invention; Figure 3 This is a schematic flowchart of one embodiment of the present invention.

[0020] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0023] In this invention, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. Furthermore, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. If the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0024] Example 1: Please refer to Figures 1 to 3 The present invention provides a real-time optimization and closed-loop control method for pre-mixing raw materials in a raw material yard, comprising the following steps: S1, acquire the first analytical data uploaded from the manual or online sampling device installed on the batching and feeding device; wherein, the first analytical data includes the key chemical components and moisture data of the material at the current stockpile; as an example, after the material at the batching and feeding device of the raw material yard (such as the batching belt scale in the pre-batching room) is manually or online sampled, a dedicated online X-ray fluorescence analyzer or LIBS laser component analyzer is installed, which can obtain the total iron grade and silica grade in real time, and can also obtain other chemical components when necessary; a microwave moisture meter, infrared moisture meter or microwave rapid drying moisture meter is installed, which can obtain the moisture content in real time.

[0025] S2, associate the first analysis data with the corresponding stockpile identification information, and update the real-time quality data of the corresponding stockpile material in the pre-constructed stockpile quality digital twin model according to the association relationship; wherein, the real-time quality data includes key chemical components and moisture data; in another preferred example, the real-time quality data may include key chemical components and moisture data, and particle size data, where particle size is an optional optimization parameter, and particle size distribution data can be obtained in real time by installing a visual particle size analyzer and a laser particle size analyzer.

[0026] S3, based on the updated stockpile quality digital twin model, obtain the key chemical component data from the real-time quality data of all available stockpile materials; then, with the optimization objective of stabilizing the key chemical components of the output blended ore within the target range within a future operating cycle, call the pre-configured multi-objective optimization model for processing; wherein, the input data of the processing includes the key chemical component data from the real-time quality data, the preset target components of the blended ore, and the cost data of each stockpile material; S4, Generate a pre-batching operation plan and control the operation of the batching and feeding device to form a mixed ore; wherein, the pre-batching operation plan includes the optimal material taking sequence from each available stockpile in the future operation cycle, obtained by solving the multi-objective optimization model; In a preferred embodiment, the method further includes a feedback correction step S5, which involves acquiring second analysis data of the blended ore, feeding back the second analysis data to correct the stockpile quality digital twin model and / or the multi-objective optimization model, and initiating a new round of optimization control process based on the correction results.

[0027] This invention utilizes manual or online sampling devices on the batching and feeding equipment to perform real-time analysis of the stockpiled materials during operations, obtaining first-hand, accurate composition data. Furthermore, by constructing a digital twin model of the stockpile quality using a three-dimensional grid as the unit, it precisely binds discrete detection data to the spatial location of the stockpile, achieving a complete mapping from the physical stockpile to a digital mirror. This completely changes the traditional blind management of raw material yards, providing a reliable data foundation for precise decision-making. Based on the updated digital twin model, this invention invokes a multi-objective optimization model centered on linear programming, with the stability of the blended ore composition as a hard constraint and the minimum total batching cost as the optimization objective, automatically solving for the optimal solution for future operation cycles. By optimizing the feed rate sequence, the quality and cost are synergistically optimized. This transforms the quality control of blended ore from post-event verification to pre-event prediction and in-event control, significantly reducing the standard deviation of blended ore and creating stable conditions for downstream sintering production. The invention utilizes an online analyzer on the blended ore output conveyor to obtain the measured composition of the finished product. Through a reverse correction algorithm, the deviation between the measured and predicted values ​​is traced back to each participating stockpile. The digital twin model is then differentiated and corrected based on historical reliability indicators. Simultaneously, the optimization model parameters are adaptively adjusted based on the statistical characteristics of the deviation, achieving self-evolution and eliminating the long-term dependence on offline testing and manual parameter adjustment. This reduces the cost per ton of blended ore and improves stockpile operation efficiency.

[0028] In a preferred embodiment, step S2 includes the following steps: S21, associate the first analysis data with the corresponding stockpile location identification information; wherein, the stockpile location identification information is generated by the batching and feeding device according to its operating position; the stockpile quality digital twin model maps the physical stockpile in the form of a three-dimensional grid, and each grid cell is associated with a unique stockpile location identification and stores the time-series quality data of the corresponding material; in this step, the stockpile location identification information is generated in real time by the vehicle-mounted positioning system of the batching and feeding device according to the coordinate position of its current operating machine, and the stockpile quality digital twin model is constructed using three-dimensional structured grid technology, which is a well-known and mature technology in the fields of computer-aided design and geographic information systems.

[0029] Explanation of the digital twin model of stockpile quality: The digital twin model of stockpile quality described in this invention is a dynamic virtual model that maps in real time to the physical stockpile. Its construction method is as follows: (1) Three-dimensional mesh mapping: The physical storage field is divided into several hexahedral mesh units along the length (L), width (W), and height (H) directions according to the preset unit size. Each mesh unit represents a minimum spatial storage unit and has a unique spatial coordinate range.

[0030] (2) Unique identifier: Each grid cell is assigned a globally unique heap identifier ID as its identity identifier in the digital twin model.

[0031] (3) Data storage structure: Each grid cell is associated with an independent storage area in a relational database or time-series database, which is used to store the material quality data corresponding to the cell in the order of timestamps. The quality data includes at least key chemical components (such as TFe, SiO2) and moisture content, and preferably also particle size distribution data.

[0032] (4) Dynamic update mechanism: When the first analysis data is obtained from the batching and feeding device, the analysis data and historical data are fused together according to the stacking position ID corresponding to the current material taking position, and then stored as a new time series data point in the storage area of ​​the corresponding grid cell to realize the real-time update of the digital twin model.

[0033] (5) Two-way mapping function: By using the stack location identifier ID, you can query the current and historical quality data of any physical stack location in the forward direction; or you can locate the physical stack location that meets the specific quality requirements based on the quality data characteristics in the reverse direction.

[0034] Through the above construction method, the digital twin model of stockpile quality of the present invention realizes a complete mapping from physical stockpile to digital image, providing a real-time and accurate data foundation for subsequent optimization decisions.

[0035] S22, based on the stockpile identification information, extract multiple historical key chemical component data, historical moisture data, and historical particle size data stored in the corresponding grid cells of the stockpile quality digital twin model within a preset historical period to form a historical quality data sequence. In this step, the preset historical period is a parameter that can be configured according to the stockpile operation frequency and material change rate. For example, for a frequently operated blending ore main stockpile, the preset historical period can be set to "the past 72 hours". By extracting all the key chemical component (TFe, SiO2) and moisture data of the material in this stockpile within the preset historical period and arranging them in order from oldest to newest sampling time, the historical quality data sequence is formed.

[0036] S23, compare the key chemical component data and moisture data in the first analysis data with the corresponding categories of data in the historical quality data sequence by time. Specifically, extract the TFe value (%), SiO2 value (%), moisture value (%), and particle size value (%) from the first analysis data (for sintering mixed ore, this may include 0-3mm particle size, 3-5mm, and 5-8mm, specifically determined by the on-site process requirements or specific demand rules). For each category of data, compare it with the corresponding category of data in the historical quality data sequence. The comparison operations include, but are not limited to: calculating the difference between the current value and the latest historical value to detect short-term changes in material quality; calculating the deviation of the current value from the mean of the historical quality data sequence to assess the position of the current value within the historical fluctuation range; and determining whether the current value falls within the maximum and minimum value intervals of the historical sequence to quickly screen for obvious outliers.

[0037] S24: For each type of data in the first analysis data, calculate the statistical mean and standard deviation of the historical quality data sequence, and set the reasonableness condition that the data value of that type in the first analysis data falls within a preset interval. If the reasonableness condition is met, proceed to step S25. Specifically, for each type of data (TFe, SiO2, moisture, particle size), calculate the arithmetic mean μ and sample standard deviation σ of the historical quality data sequence. The reasonableness condition is: the current detection value satisfies [μ-kσ, μ+kσ], where k is a preset constant (usually 2 or 3). If the condition is met, the data is determined to be valid, and proceed to S25; otherwise, it is determined to be abnormal data, and the system can perform preset processing such as discarding, downgrading, or alarming.

[0038] S25, perform data fusion, whereby for each type of data, updated real-time quality data is calculated and generated; specifically, the most basic conventional algorithm in the field of signal processing and data fusion, the weighted moving average method, can be used for data fusion.

[0039] S26, the updated real-time quality data is stored as a new time-series data point in the grid cell corresponding to the stack location identifier in the digital twin model of the stack quality.

[0040] In a preferred embodiment, step S3 includes the following steps: S31, based on the updated digital twin model of the stockpile quality, obtain real-time key chemical composition data of materials in all currently available stockpile locations, unit cost data of materials in each stockpile location, and current available inventory of materials in each stockpile location. This step extracts all the basic data required for this optimization decision from the latest updated digital twin model of the stockpile quality in one complete step, ensuring the real-time, synchronous, and comprehensive nature of the decision information. Through standard database query operations, filter stockpile locations with available status, such as those that are not blocked, not emptied, or whose quality status is known. For each available stockpile location, extract the total iron (TFe) value and silicon dioxide (SiO2) value from the real-time quality data corresponding to the latest timestamp from the digital twin model of the stockpile quality, extract the unit cost (yuan / ton) of that stockpile location pre-stored in the material master data table, and extract the current available inventory (tons) of that stockpile location.

[0041] S32, establish a linear programming model with minimizing total cost as the optimization direction, setting the planned material extraction quantity for each stockpile as a decision variable and defining constraints. These constraints include: the predicted blending composition calculated based on real-time key chemical composition data of the material at each stockpile and the decision variables must fall within the target range, serving as a hard constraint to achieve multi-objective synergistic optimization of quality and cost; and the planned material extraction quantity for each stockpile must not exceed the corresponding current available inventory and must be non-negative. This step transforms the production decision problem of how to allocate material extraction quantities to each stockpile into a linear programming model with a clear mathematical structure. Specifically, a decision variable is assigned to each available stockpile, representing the planned weight (in tons) of material to be extracted from that stockpile in a future operating cycle. The values ​​of the decision variables will be determined by the solver in subsequent solution processes.

[0042] This embodiment optimizes for minimizing total material cost. The total cost is calculated by multiplying the planned material extraction quantity for each stockpile by the corresponding unit cost, and then summing the products for all stockpiles. The optimization solver automatically finds the combination of material extraction quantities that minimizes this sum. Based on the planned material extraction quantity and real-time grade of each stockpile, the system uses a weighted average method to calculate and predict the content of key components in the blended ore. Taking TFe as an example, the calculation steps are as follows: (1) Multiply the amount of material taken from each stockpile by its TFe grade to obtain the amount of iron contributed by that stockpile to the total iron content of the blended ore. (2) Add up the iron content of all piles to obtain the total iron content of the blended ore; (3) Add up the amount of material taken from all piles to get the total weight of the mixed ore; (4) Divide the total iron content by the total weight to obtain the predicted TFe grade of the blended ore. The weighted average calculation of the composition of the mixed materials is a fundamental principle of batching calculation in the process industry and is common knowledge for those skilled in the art. This predicted value falls within a pre-set target range (e.g., TFe between 62.0% and 62.5%). The planned material extraction quantity for each stockpile must not exceed the current available stockpile quantity and must not be negative.

[0043] S33. Solve the linear programming model and output the optimal decision variable values ​​as the optimal material retrieval sequence from each available stack location within the next work cycle. A mature optimization solver (commonly used solvers in industry include IBM CPLEX, Gurobi, COIN-OR CLP, etc.) can be called to quickly calculate the constructed linear programming model and convert the solution results into standard operating instructions that the production system can directly execute. This step involves a routine call to the solver and does not involve any modification to the solver's underlying algorithm. The optimal material retrieval quantities for each stack location are summarized by stack location identifier to generate a structured pre-batching work plan. This plan may include fields such as stack location number, planned material retrieval quantity, work priority, and estimated work duration, and is stored in a standard data format.

[0044] This embodiment transforms the traditional pre-batching decision-making process, which relies on human experience, into a quantifiable, solvable, and verifiable mathematical optimization problem through linear programming. This allows for the automatic minimization of the total batching cost under the hard constraint of ensuring the stability of the mixed mineral composition.

[0045] In a preferred embodiment, step S4 includes the following steps: S41, the optimal material taking sequence is decomposed into multiple consecutive operation periods in chronological order, and a corresponding operation task is generated for each operation period; wherein, each operation task includes the planned material taking quantity within the operation period; the standard duration or standard material taking quantity of a single operation period can be determined according to the rated material taking efficiency of the batching and feeding device (unit: tons / hour) and the instruction execution cycle of the automated control system. The decomposed unit tasks are sorted according to the operation priority rules to form an ordered sequence of operation tasks for each operation period.

[0046] S42, the time period operation task is converted into a control command for the batching and feeding device to be executed, and sent to the corresponding batching and feeding device to form a mixed ore.

[0047] In a preferred embodiment, step S5 includes the following steps: S51, Obtain second analytical data by online sampling and analysis of the blended ore formed in step S4; wherein, the second analytical data includes the measured values ​​of key chemical components of the blended ore; obtain real-time chemical component detection data of the finished blended ore as the basis for feedback correction, and provide a reference true value for subsequent deviation analysis and model correction. An online analyzer can be installed on the blended ore output conveyor belt to perform real-time sampling and analysis of the blended ore continuously passing through the belt to obtain the second analytical data.

[0048] S52, compare the measured values ​​of key chemical components in the second analysis data with the predicted values ​​calculated based on the real-time quality data of the corresponding stockpile material in the digital twin model of stockpile quality and the actual material handling data that has been executed, to obtain the comparison deviation, and perform reverse correction on the real-time quality data of the corresponding stockpile material in the digital twin model of stockpile quality based on the comparison deviation. Specifically, the real-time quality data of the stockpile material comes from the latest timestamp-corresponding real-time quality data stored in the stockpile quality digital twin model, which is part of the batch of blended ore. The actual material handling data is from the execution log of step S4, including the actual material handling volume (in tons) completed by each stockpile during the current operation cycle, the stockpile identifier, start and end time, and completion status of each material handling task. A weighted average algorithm can be used to calculate the predicted composition of the blended ore. The specific steps include: (1) From the execution log of step S4, summarize the actual material taken from each stack location during this operation cycle and calculate the total material taken during this cycle; (2) For each stockpile involved in the formation of this batch of blended ore, extract the real-time grade data of the latest timestamp of the stockpile from the stockpile quality digital twin model; (3) Multiply the actual amount of material taken from each stockpile by its corresponding real-time grade to obtain the component contribution of each stockpile; (4) Add up the component contributions of all heap positions to obtain the total component contribution; (5) Divide the total component contribution by the total material intake in this cycle to obtain the predicted components of the blended ore (i.e., the predicted value).

[0049] S53, based on the statistical deviation between the second analysis data and the predicted values, the model parameters used to calculate and predict the blended ore components in the multi-objective optimization model are adjusted; using the deviation between the measured and predicted values ​​of the blended ore finished product, the key model parameters used to predict the blended ore components in the multi-objective optimization model are adjusted, so that the model's predictive ability continuously improves with production operation. For example, when there is a persistent systematic deviation in the system (such as the average deviation significantly deviating from zero), adjusting this coefficient can eliminate systematic errors; when the prediction deviation of a certain component is large, its weight in the objective function is adjusted, so that the optimizer pays more attention to the control of that component.

[0050] S54, determine whether the conditions for triggering a new round of optimization are met; wherein, the triggering conditions include at least one of the following: reaching the end of a preset fixed operation cycle (a fixed time window can be set (e.g., every 4 hours, every 8 hours, every shift), and triggering a new round of optimization upon expiration), the actual material taking amount accumulated since the last optimization reaching a threshold (a cumulative material taking amount threshold can be set (e.g., every 5000 tons, every 10000 tons), and triggering when the cumulative material taking amount reaches the threshold since the last optimization), or the comparison deviation exceeding a preset tolerance (a deviation tolerance value can be set (e.g., TFe deviation exceeds ±0.5%), and triggering when the current batch comparison deviation exceeds the tolerance); S55, if the optimization triggering condition is met, a new round of optimization control process starting from step S1 is initiated based on the corrected digital twin model of the stockpile quality. When S54 determines that the triggering condition is met, the complete optimization control process starting from step S1 is restarted based on the latest version of the digital twin model corrected by S52, forming a closed-loop continuous optimization.

[0051] As a preferred embodiment, the reverse correction of the real-time quality data of the corresponding stockpile material in the digital twin model of stockpile quality based on the comparison deviation in step S52 includes the following steps: Based on the proportion of the actual material taken from each stockpile to the total material taken in the operation cycle of forming the blended ore, the contribution weight of each stockpile to the comparison deviation is calculated; the actual material taken from each stockpile (unit: tons) in the current operation cycle is queried and summarized from the material taking operation execution log. For each stockpile participating in the formation of this batch of blended ore, the proportion of its actual material taken to the total material taken in this cycle is calculated, and this proportion is used as the contribution weight of that stockpile to this blended ore.

[0052] For each stockpile participating in this operation within the digital twin model of the stockpile quality, the correction amount for the real-time key chemical composition data of the material at that stockpile is calculated in reverse based on the comparison deviation, the contribution weight, and the actual material extraction data already performed. The total comparison deviation is then proportionally allocated according to the contribution weight of each stockpile to obtain the deviation component that each stockpile should bear. That is, the deviation component borne by a stockpile is equal to the total deviation multiplied by the contribution weight of that stockpile. The correction amount is calculated based on the principle of inverse material balance. Since the predicted composition of the blended ore is the result of a weighted average of the grades of each stockpile according to its extraction volume, when there is a deviation between the predicted value and the measured value, this deviation should be evenly distributed across the grade data of each participating stockpile.

[0053] The historical reliability index of the real-time quality data of the material at the specified stockpile location in the digital twin model of stockpile quality is obtained. The correction amount is then weighted and adjusted based on the historical reliability index to obtain the final correction value. For each stockpile location, a quantitative reliability index (range 0-1) is assessed based on the statistical characteristics of its historical quality data sequence. The reliability index can employ statistical stability evaluation methods known in the art. For example, if the historical data of the stockpile location is stable over a long period and the adjustment range of each reverse correction is small, it indicates that the confidence level of the twin model data for that stockpile location is high, and the system assigns it a reliability level close to 1. If the historical data of the stockpile location fluctuates drastically, or each reverse correction requires a large adjustment, it indicates that the reliability of the twin model data for that stockpile location is poor, and the system assigns it a reliability level close to 0. Based on the principle that higher reliability corresponds to a smaller correction range, the initial correction amount is weighted and adjusted to obtain the final correction value of the real-time quality data of each participating stockpile location twin model. The final correction value = initial correction amount × (1 - reliability index).

[0054] The final correction value is applied to the real-time key chemical composition data of the corresponding stockpile material in the stockpile quality digital twin model to complete the reverse correction. For each stockpile participating in this correction, the real-time key chemical composition data (such as TFe grade) in the stockpile quality digital twin model is algebraically superimposed with the final correction value to obtain the updated real-time quality data, and finally, the stockpile quality digital twin model updated by reverse correction is obtained.

[0055] The present invention also provides a real-time optimization and closed-loop control device for pre-mixing raw material yards, used to execute the real-time optimization and closed-loop control method for pre-mixing raw material yards as described above, comprising: The sampling and analysis unit is used to acquire first analysis data uploaded from a manual or online sampling device installed on the batching and feeding device; wherein, the first analysis data includes key chemical components and moisture data of the material at the current stockpile. An update unit is used to associate the first analysis data with the corresponding stockpile identification information, and update the real-time quality data of the corresponding stockpile material in the pre-constructed stockpile quality digital twin model according to the association relationship; wherein, the real-time quality data includes key chemical components and moisture data; The optimization calculation unit is used to obtain key chemical component data from the real-time quality data of all available stockpile materials based on the updated stockpile quality digital twin model; then, with the optimization objective of stabilizing the key chemical components of the output blended ore within the target range in the next operating cycle, it calls a pre-configured multi-objective optimization model for processing; wherein, the input data of the processing includes the key chemical component data in the real-time quality data, the preset target components of the blended ore, and the cost data of each stockpile material; A control unit is used to generate a pre-batching operation plan and control the operation of the batching and feeding device to form a mixed ore; wherein, the pre-batching operation plan includes the optimal material extraction sequence from each available stockpile within a future operation cycle, obtained by solving the multi-objective optimization model; The feedback correction unit is used to acquire the second analysis data of the blended ore, feed the second analysis data back to correct the stockpile quality digital twin model and the multi-objective optimization model, and then return to step S1 to start a new round of optimization control process.

[0056] The present invention also provides a real-time optimization and closed-loop control device for pre-mixing raw materials, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the real-time optimization and closed-loop control method for pre-mixing raw materials as described above.

[0057] Example 2 (Multi-component synergistic control): For sinter production scenarios with strict requirements on impurities such as Al2O3 and MgO, the key chemical components of this invention can be expanded to four items: TFe, SiO2, Al2O3, and MgO. In an application of this embodiment in a steel plant, the target ranges were set as follows: TFe 62.0%~62.5%, SiO2 4.8%~5.2%, Al2O3 ≤ 2.0%, and MgO ≤ 1.5%. When constructing the linear programming model in step S32, the constraints were correspondingly expanded to: all four components of the predicted blended ore must fall within their respective target ranges. The results showed that the standard deviation of Al2O3 decreased from 0.35% to 0.18%, and the standard deviation of MgO decreased from 0.28% to 0.15%, further stabilizing the sinter quality. The remaining steps were the same as in Example 1. This embodiment demonstrates that this invention can be extended to multi-component synergistic optimization scenarios.

[0058] Example 3 (Non-ferrous metallurgy scenario): In the pre-batching of copper smelting raw materials, the key chemical components can be defined as Cu, S, Fe, and SiO2. The real-time Cu grade and S content of copper concentrate from each stockpile are input into an optimization model to achieve precise batching of multi-metal raw materials, with the goal of stabilizing the Cu grade and S content of the copper concentrate fed into the furnace. This embodiment demonstrates the versatility of the invention in different metallurgical fields.

[0059] After applying the system of this invention to the raw material yard of a large steel enterprise, the operating data for three consecutive months showed that: the standard deviation of TFe in the blended ore decreased from 0.65% before application to 0.45%, a reduction of 26.2%; the standard deviation of SiO2 decreased from 0.32% to 0.22%, a reduction of 31%; at the same time, due to the reduction in the use of high-priced high-quality ore powder and the increase in the proportion of low-priced ore, the average cost of ore blending was reduced by 8.7 yuan / ton, resulting in annual economic benefits of over 20 million yuan.

[0060] The present invention also provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for real-time optimization and closed-loop control of raw material pre-mixing.

[0061] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0062] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for real-time optimization and closed-loop control of raw yard pre-blending, characterized in that, Includes the following steps: S1, acquire first analytical data uploaded from a manual or online sampling device installed on the batching and feeding device; wherein, the first analytical data includes key chemical components and moisture data of the material at the current stockpile; S2, associate the first analysis data with the corresponding stockpile identification information, and update the real-time quality data of the corresponding stockpile material in the pre-constructed stockpile quality digital twin model according to the association relationship; wherein, the real-time quality data includes key chemical components and moisture data; S3, based on the updated stockpile quality digital twin model, obtain the key chemical component data from the real-time quality data of all available stockpile materials; then, with the optimization objective of stabilizing the key chemical components of the output blended ore within the target range within a future operating cycle, call the pre-configured multi-objective optimization model for processing; wherein, the input data of the processing includes the key chemical component data from the real-time quality data, the preset target components of the blended ore, and the cost data of each stockpile material; S4, Generate a pre-batching operation plan and control the operation of the batching and feeding device to form a mixed ore; wherein, the pre-batching operation plan includes the optimal material taking sequence from each available stockpile within a future operation cycle, obtained by solving the multi-objective optimization model.

2. The method according to claim 1, characterized in that, It also includes a feedback correction step S5, which involves acquiring the second analysis data of the blended ore, feeding back the second analysis data to correct the stockpile quality digital twin model and / or the multi-objective optimization model, and initiating a new round of optimization control process based on the correction results.

3. The real-time optimization and closed-loop control method for pre-mixing raw materials according to claim 1, characterized in that, Step S2 includes the following steps: S21, associate the first analysis data with the corresponding stockpile location identification information; wherein, the stockpile location identification information is generated by the batching and feeding device according to its operating position; the stockpile quality digital twin model maps the physical stockpile in the form of a three-dimensional grid, and each grid cell is associated with a unique stockpile location identification and stores the time-series quality data of the corresponding material; S22, Based on the pile location identification information, extract multiple historical key chemical component data, historical moisture data and historical particle size data that have been stored in the corresponding grid cell in the digital twin model of the pile quality in a preset historical period to form a historical quality data sequence; S23, compare the key chemical component data and moisture data in the first analysis data with the corresponding category data in the historical quality data sequence according to time; S24. For each type of data in the first analysis data, calculate the statistical mean and standard deviation of the historical quality data sequence, and set the reasonableness condition that the data value of that type in the first analysis data falls within a preset interval. If the reasonableness condition is met, then execute step S25. S25, perform data fusion, where for each type of data, calculate and generate updated real-time quality data; S26, the updated real-time quality data is stored as a new time-series data point in the grid cell corresponding to the stack location identifier in the digital twin model of the stack quality.

4. The real-time optimization and closed-loop control method for pre-mixing raw materials according to claim 1, characterized in that, Step S3 includes the following steps: S31, Based on the updated digital twin model of the stockpile quality, obtain real-time key chemical composition data of all available stockpile materials, unit cost data of each stockpile material, and current available stock of each stockpile material. S32, establish a linear programming model with minimizing total cost as the optimization direction, set the planned material extraction quantity of each stockpile as the decision variable, and set constraints; wherein, the constraints include: the predicted blending ore composition calculated based on the real-time key chemical composition data of the material at each stockpile and the decision variable must fall within the target range, thus serving as a hard constraint to achieve multi-objective synergistic optimization of quality and cost; and the planned material extraction quantity of each stockpile does not exceed the corresponding current available inventory and is a non-negative number; S33, Solve the linear programming model, and output the optimal decision variable values ​​obtained as the optimal material taking sequence from each available stack location in the next working cycle.

5. The real-time optimization and closed-loop control method for pre-mixing raw materials according to claim 1, characterized in that, Step S4 includes the following steps: S41, the optimal material taking sequence is decomposed into multiple consecutive operation periods in chronological order, and a corresponding operation task is generated for each operation period; wherein, each operation task includes the planned material taking amount within the operation period; S42, the time period operation task is converted into a control command for the batching and feeding device to be executed, and sent to the corresponding batching and feeding device to form a mixed ore.

6. The real-time optimization and closed-loop control method for pre-mixing raw materials according to claim 2, characterized in that, Step S5 includes the following steps: S51, Obtain second analytical data by online sampling and analysis of the homogenized ore formed in step S4; wherein, the second analytical data includes the measured values ​​of the key chemical components of the homogenized ore; S52, compare the measured values ​​of key chemical components in the second analysis data with the predicted values ​​calculated based on the real-time quality data of the corresponding stockpile material in the digital twin model of stockpile quality and the actual material handling data that has been executed, to obtain the comparison deviation, and perform reverse correction on the real-time quality data of the corresponding stockpile material in the digital twin model of stockpile quality based on the comparison deviation. S53, Based on the statistical analysis of the deviation between the second analysis data and the predicted value, the model parameters used to calculate the predicted composition of the mixed ore in the multi-objective optimization model are adjusted; S54, determine whether the conditions for triggering a new round of optimization are met; wherein, the triggering conditions include at least one of the following: reaching the preset fixed operation cycle end point, the actual material taking amount accumulated since the last optimization reaching a threshold, or the comparison deviation exceeding the preset tolerance. S55, if the optimization triggering condition is met, then based on the corrected digital twin model of the stockpile quality, a new round of optimization control process starting from step S1 is initiated.

7. The real-time optimization and closed-loop control method for pre-mixing raw materials according to claim 6, characterized in that, The reverse correction of the real-time quality data of the corresponding stockpile material in the digital twin model of stockpile quality based on the comparison deviation in step S52 includes the following steps: The contribution weight of each stockpile to the comparison deviation is calculated based on the proportion of the actual material taken from each stockpile to the total material taken in the operation cycle of forming the mixed ore. For each stockpile participating in this operation in the stockpile quality digital twin model, the correction amount for the real-time key chemical composition data of the material at that stockpile is calculated in reverse based on the comparison deviation, the contribution weight, and the actual material handling data that has been executed. Obtain the historical reliability index of the real-time quality data of the material at the stockpile location in the digital twin model of stockpile quality; adjust the correction amount by weighting according to the historical reliability index to obtain the final correction value; The final correction value is applied to the real-time key chemical composition data of the corresponding stockpile material in the digital twin model of stockpile quality to complete the reverse correction.

8. A real-time optimization and closed-loop control device for pre-mixing raw material yards, used to execute the real-time optimization and closed-loop control method for pre-mixing raw material yards as described in any one of claims 1-7, characterized in that, include: The sampling and analysis unit is used to acquire first analysis data uploaded from a manual or online sampling device installed on the batching and feeding device; wherein, the first analysis data includes key chemical components and moisture data of the material at the current stockpile. An update unit is used to associate the first analysis data with the corresponding stockpile identification information, and update the real-time quality data of the corresponding stockpile material in the pre-constructed stockpile quality digital twin model according to the association relationship; wherein, the real-time quality data includes key chemical components and moisture data; The optimization calculation unit is used to obtain key chemical component data from the real-time quality data of all available stockpile materials based on the updated stockpile quality digital twin model; then, with the optimization objective of stabilizing the key chemical components of the output blended ore within the target range in the next operating cycle, it calls a pre-configured multi-objective optimization model for processing; wherein, the input data of the processing includes the key chemical component data in the real-time quality data, the preset target components of the blended ore, and the cost data of each stockpile material; A control unit is used to generate a pre-batching operation plan and control the operation of the batching and feeding device to form a mixed ore; wherein, the pre-batching operation plan includes the optimal material extraction sequence from each available stockpile within a future operation cycle, obtained by solving the multi-objective optimization model; The feedback correction unit is used to acquire the second analysis data of the blended ore, feed the second analysis data back to correct the stockpile quality digital twin model and the multi-objective optimization model, and then return to step S1 to start a new round of optimization control process.

9. A real-time optimization and closed-loop control device for pre-mixing raw materials at a raw material yard, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the real-time optimization and closed-loop control method for pre-mixing raw materials as described in any one of claims 1 to 7.

10. A storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the real-time optimization and closed-loop control method for pre-mixing raw materials as described in any one of claims 1 to 7.

Images