Steel production process real-time multi-objective optimization method and system based on genetic algorithm

Abstract

The application discloses a steel production process real-time multi-target optimization method and system based on a genetic algorithm, which comprises the following steps: step S1: initializing a production plan to obtain an initial production plan; step S2: deducing a production process based on the initial production plan to obtain a real-time production state and a deviation monitoring result; step S3: when the deviation monitoring result indicates that the deviation between the actual completion time and the planned time of any process exceeds a preset threshold, triggering a genetic algorithm optimizer to generate a rescheduling optimization scheme; step S4: comparing the rescheduling optimization scheme with the initial production plan to generate a scheduling adjustment suggestion containing process time adjustment or execution sequence adjustment; and step S5: repeating the steps S2 to S4 until all heats of production are completed. The application can realize real-time sensing of random deviations in the production process, quickly generate a rescheduling scheme that takes into account the total production time and the minimum plan change, and realize adaptive dynamic optimization of the production process.

Description

Technical Field

[0001] This invention relates to the field of steel production process optimization and intelligent scheduling technology, specifically to a real-time multi-objective optimization method and system for steel production processes based on genetic algorithms. Background Technology

[0002] Modern steel production processes are characterized by multiple steps, multiple pieces of equipment, high coupling, and strong constraints, making them typical complex process industries. Traditional production scheduling typically employs fixed-sequence or experience-based rule-based methods, which struggle to cope with uncertainties and dynamic disturbances in the production process, such as equipment failures, process fluctuations, and raw material changes. These factors often lead to process delays, equipment idleness, and production rhythm disruptions, thereby reducing production efficiency, increasing energy consumption, and raising production costs.

[0003] With the development of intelligent manufacturing and industry, the use of artificial intelligence and optimization algorithms to achieve real-time scheduling and dynamic optimization of production processes has become a research hotspot. Genetic algorithms, as a global optimization algorithm that simulates the natural evolutionary process, have advantages such as high stability and applicability to complex nonlinear problems, and have been applied in many scheduling problems. However, in production environments with high real-time requirements and multiple parallel processes, such as short-process steelmaking, how to combine genetic algorithms with real-time production status to achieve dynamic rescheduling at the minute or even second level remains a technical challenge.

[0004] Currently, most steel production scheduling systems remain at the offline planning or limited adjustment stage, lacking the ability to quickly respond to random deviations in the production process. This leads to a disconnect between planning and actual production, hindering further improvements in production efficiency. Therefore, developing an intelligent scheduling system capable of real-time perception of production deviations, rapid generation of optimized adjustment plans, and dynamic execution is of great significance for improving the production efficiency of steel enterprises, reducing energy consumption, and ensuring continuous and stable production. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, one of the objectives of this invention is to provide a real-time multi-objective optimization method for steel production processes based on genetic algorithms. This method can detect random deviations in the production process in real time, quickly generate a rescheduling scheme that takes into account both total production time and minimal changes in the plan, and achieve adaptive dynamic optimization of the production process.

[0006] The second objective of this invention is to provide a real-time multi-objective optimization system for steel production processes based on genetic algorithms. This system can detect random deviations in the production process in real time, quickly generate a rescheduling scheme that takes into account both total production time and minimal changes in the plan, and achieve adaptive dynamic optimization of the production process.

[0007] To achieve one of the objectives of this invention, the following solution is adopted: A real-time multi-objective optimization method for steel production processes based on genetic algorithms includes the following steps: Step S1: Initialize the production plan to obtain an initial production plan that includes multiple furnace batches, each furnace batch contains multiple sequentially executed processes, and the planned duration of each process; Step S2: Start the parallel simulator, deduce the production process based on the initial production plan, introduce random fluctuations in the actual time consumption of each process during the deduction process to simulate the uncertainty of real production, and monitor the deviation between the actual completion time and the planned time of each process in real time to obtain the real-time production status and deviation monitoring results. Step S3: When the deviation monitoring result indicates that the deviation between the actual completion time and the planned time of any process exceeds a preset threshold, the genetic algorithm optimizer is triggered. The optimizer takes the set of all unfinished processes at the current moment as the optimization object, minimizes the total production time of each casting and minimizes the plan change as the optimization objective, and runs the genetic algorithm to perform rescheduling optimization based on the current production status to generate a rescheduling optimization scheme. Step S4: Compare the rescheduling optimization scheme with the initial production plan, generate scheduling adjustment suggestions that include process time adjustments or execution sequence adjustments, and apply the suggestions to the current production plan to obtain the updated production plan; Step S5: Control the parallel simulator to continue production based on the updated production plan, and repeat steps S2 to S4 until all furnace batches are completed.

[0008] Furthermore, the genetic algorithm in step S3 specifically includes: Population initialization: Based on the current production state, an initial population containing multiple individuals is generated by randomly fine-tuning the duration of unfinished processes, where each individual represents a scheduling scheme. Fitness assessment: For each individual in the initial population, calculate the estimated total production time and the planned deviation for each batch, and calculate the fitness score of each individual according to the preset multi-objective weighted rule; Selection operation: Based on the fitness score of each individual, the roulette wheel selection method is used to select multiple individuals with higher fitness scores from the current population as excellent individuals; Cross-operation: For the aforementioned outstanding individuals, the process time settings of different individuals are merged through cross-operation to generate offspring individuals; Mutation operation: For the offspring individual, the duration of some of the processes is randomly adjusted to obtain the mutated offspring individual; Elite retention: The mutated offspring individuals are merged with the individuals in the current generation population, and a number of individuals with the highest fitness scores are retained and directly enter the next generation population as at least a part of the next generation population.

[0009] Furthermore, the multi-objective weighted rule in the fitness assessment is as follows: by combining the estimated total production time and the planned deviation of each batch, weights are assigned to the total production time target and the planned deviation target respectively, and a weighted score is calculated as the fitness score of the individual; wherein, the shorter the total production time and the smaller the sum of planned deviations, the higher the fitness score.

[0010] Furthermore, minimizing the plan variation in step S3 specifically means minimizing the sum of the absolute values ​​of the time deviations between the actual start time of each process in each furnace after rescheduling and the original planned start time.

[0011] Furthermore, in step S3, when the genetic algorithm optimizer performs rescheduling optimization on incomplete processes, it sets the production plans of completed processes or processes with special production constraints to be prohibited from adjustment.

[0012] Furthermore, the processes that are prohibited from adjustment include completed processes, processes that are currently in operation, smelting processes, and continuous casting processes; refining processes and transportation processes are processes that are allowed to be adjusted.

[0013] Furthermore, the preset threshold is configured to be 0.01 minutes, that is, when the actual completion time of any process deviates from the planned time by more than 0.01 minutes, the genetic algorithm optimizer is triggered.

[0014] Furthermore, the parallel simulator adopts an event-driven mechanism to advance the simulation time, specifically including: in each simulation loop, traversing all incomplete furnaces, checking whether the next process of each furnace meets the execution conditions, the execution conditions including the previous process has been completed and the equipment required for the current process is idle; taking all processes that meet the execution conditions as executable tasks, selecting the earliest task that can start from them for execution, and updating the equipment occupancy time and the current simulation time.

[0015] Furthermore, the scheduling adjustment suggestions include process delay, process advancement, or process time compression; the scheduling adjustment suggestions can be configured to be automatically applied to the current production plan, or provided to the scheduler for confirmation via a human-machine interface before application.

[0016] To achieve the second objective of this invention, the following solution is adopted: A real-time multi-objective optimization system for steel production processes based on genetic algorithms includes: The planning initialization module is used to initialize the production plan and obtain an initial production plan that includes multiple furnace batches, each furnace batch contains multiple sequentially executed processes, and the planned duration of each process. The parallel simulation and monitoring module is connected to the planning initialization module and is used to extrapolate the production process based on the initial production plan. During the extrapolation process, random fluctuations are introduced into the actual time consumption of each process to simulate the uncertainty of real production, and the deviation between the actual completion time and the planned time of each process is monitored in real time to obtain the real-time production status and deviation monitoring results. The genetic algorithm optimization trigger module is connected to the parallel simulation and monitoring module. When the deviation monitoring result indicates that the deviation between the actual completion time and the planned time of any process exceeds a preset threshold, the genetic algorithm optimizer is triggered. The optimizer takes the set of all unfinished processes at the current moment as the optimization object, minimizes the total production time of each casting and minimizes the plan change as the optimization objective, runs the genetic algorithm to perform rescheduling optimization based on the current production status, and generates a rescheduling optimization scheme. The scheduling adjustment suggestion generation and application module is connected to the genetic algorithm optimization triggering module. It is used to compare the rescheduling optimization scheme with the initial production plan, generate scheduling adjustment suggestions including process time adjustment or execution order adjustment, and apply the suggestions to the current production plan to obtain the updated production plan. The loop control module, connected to the scheduling adjustment suggestion generation and application module and the parallel simulation and monitoring module, is used to control the parallel simulation and monitoring module to continue production based on the updated production plan, and to repeatedly execute the monitoring, optimization and adjustment steps until all furnace batches are completed.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention possesses real-time response and dynamic optimization capabilities. By monitoring the deviation between the actual and planned times of each process in real time and setting deviation thresholds to trigger a genetic algorithm optimizer, this invention achieves minute-level or even second-level rapid response to random fluctuations in the production process (such as equipment failure, process fluctuations, raw material changes, etc.). Compared to traditional scheduling methods that use fixed time sequences or empirical rules and are difficult to cope with dynamic disturbances, this invention can detect production deviations online and trigger rescheduling in a timely manner, effectively reducing the cumulative impact of deviations and avoiding process delays and equipment idleness.

[0018] 2. This invention enables global multi-objective optimization. The invention aims to minimize the total production time per casting cycle and minimize plan variations, employing a genetic algorithm for rescheduling optimization. Genetic algorithms possess strong global search capabilities and are suitable for complex nonlinear problems. They can comprehensively consider the coupling constraints between multiple processes and equipment from a global perspective, avoiding local optima, while simultaneously balancing the conflicting objectives of production efficiency and plan stability, thus obtaining a superior overall scheduling scheme.

[0019] 3. This invention establishes an adaptive closed-loop scheduling mechanism. The invention compares the optimized plan with the original plan to generate adjustment suggestions, which are then applied to the current production plan. The simulator is then controlled to continue advancing based on the updated plan, cyclically executing the monitoring-optimization-adjustment process, thus forming a closed-loop adaptive scheduling. This mechanism solves the problems of traditional offline planning being disconnected from actual production and lacking rapid response capabilities, enabling the production plan to dynamically evolve with the actual execution status, significantly improving the continuity and stability of the production process.

[0020] 4. This invention can reduce energy consumption and production costs. Through real-time dynamic optimization, this invention reduces additional waiting time and energy waste caused by process delays, equipment idleness, and production rhythm disruptions. Using this invention, the total production time for the post-casting cycle is shortened by approximately 25 minutes compared to the unoptimized scenario, directly reducing energy consumption and production costs per unit product.

[0021] 5. This invention possesses high flexibility and configurability. The system of this invention adopts a modular architecture design, and the production model, optimization objective, and genetic algorithm parameters can all be flexibly configured according to the actual production line process characteristics and production needs. Therefore, this invention can quickly adapt to the process characteristics and scheduling requirements of different steel mills and different production lines, exhibiting good portability and versatility.

[0022] 6. This invention supports a human-machine collaborative scheduling mode. It can either execute optimization suggestions fully automatically or provide them to schedulers via a human-machine interface for review and confirmation before application. This human-machine collaborative mode balances the efficiency of automated scheduling with the flexibility of human experience. Schedulers can review, adjust, or reject optimization suggestions based on actual production conditions, thus ensuring optimization effectiveness while retaining human intervention capabilities, enhancing the system's acceptability and practicality in real-world production environments. Attached Figure Description

[0023] Figure 1 This is a flowchart of a real-time multi-objective optimization method for steel production process based on genetic algorithm in an embodiment of the present invention; Figure 2 This is a principle block diagram of the multi-objective optimization scheduling architecture in an embodiment of the present invention; Figure 3 This is a flowchart illustrating the real-time multi-objective optimization method for steel production process based on genetic algorithm in an embodiment of the present invention. Figure 4 This is a flowchart illustrating the workflow of the genetic algorithm optimizer in an embodiment of the present invention. Figure 5 This is a schematic diagram of the operating interface of the production simulator in an embodiment of the present invention; Figure 6 This is the output interface for scheduling and adjustment suggestions for four batches of production in this embodiment of the invention; Figure 7 This is a Gantt chart of the four production processes in an embodiment of the present invention; Figure 8 This is an output interface for scheduling adjustment suggestions in another production scenario in this embodiment of the invention; Figure 9 This is a Gantt chart of the production process in another production scenario according to an embodiment of the present invention; Figure 10 This is a block diagram of a real-time multi-objective optimization system for steel production process based on genetic algorithms, as described in an embodiment of the present invention. Detailed Implementation

[0024] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments. It should be noted that, without conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0025] Example 1 like Figure 1-9 As shown, this embodiment of the invention provides a real-time multi-objective optimization method for steel production processes based on genetic algorithms, including the following steps: Step S1: Initialize the production plan to obtain an initial production plan that includes multiple furnace batches, each furnace batch contains multiple sequentially executed processes, and the planned duration of each process.

[0026] Step S2: Start the parallel simulator and simulate the production process based on the initial production plan. During the simulation, random fluctuations are introduced into the actual time consumption of each process to simulate the uncertainty of real production. The deviation between the actual completion time and the planned time of each process is monitored in real time to obtain the real-time production status and deviation monitoring results.

[0027] Step S3: When the deviation monitoring result indicates that the deviation between the actual completion time and the planned time of any process exceeds a preset threshold, the genetic algorithm optimizer is triggered. The optimizer takes the set of all unfinished processes at the current moment as the optimization object, minimizes the total production time of each casting and minimizes the plan change as the optimization objective, and runs the genetic algorithm to perform rescheduling optimization based on the current production status to generate a rescheduling optimization scheme.

[0028] Step S4: Compare the rescheduling optimization scheme with the initial production plan, generate scheduling adjustment suggestions that include process time adjustments or execution sequence adjustments, and apply the suggestions to the current production plan to obtain the updated production plan.

[0029] Step S5: Control the parallel simulator to continue production based on the updated production plan, and repeat steps S2 to S4 until all furnace batches are completed.

[0030] Furthermore, the genetic algorithm in step S3 specifically includes: Population initialization: Based on the current production state, an initial population containing multiple individuals is generated by randomly fine-tuning the duration of unfinished processes, where each individual represents a scheduling scheme. Fitness assessment: For each individual in the initial population, calculate the estimated total production time and the planned deviation for each batch, and calculate the fitness score of each individual according to the preset multi-objective weighted rule; Selection operation: Based on the fitness score of each individual, the roulette wheel selection method is used to select multiple individuals with higher fitness scores from the current population as excellent individuals; Cross-operation: For the aforementioned outstanding individuals, the process time settings of different individuals are merged through cross-operation to generate offspring individuals; Mutation operation: For the offspring individual, the duration of some of the processes is randomly adjusted to obtain the mutated offspring individual; Elite retention: The mutated offspring individuals are merged with the individuals in the current generation population, and a number of individuals with the highest fitness scores are retained and directly enter the next generation population as at least a part of the next generation population.

[0031] Furthermore, the multi-objective weighted rule in the fitness assessment is as follows: by combining the estimated total production time and the planned deviation of each batch, weights are assigned to the total production time target and the planned deviation target respectively, and a weighted score is calculated as the fitness score of the individual; wherein, the shorter the total production time and the smaller the sum of planned deviations, the higher the fitness score.

[0032] Furthermore, minimizing the plan variation in step S3 specifically means minimizing the sum of the absolute values ​​of the time deviations between the actual start time of each process in each furnace after rescheduling and the original planned start time.

[0033] Furthermore, in step S3, when the genetic algorithm optimizer performs rescheduling optimization on incomplete processes, it sets the production plans of completed processes or processes with special production constraints to be prohibited from adjustment.

[0034] Furthermore, the processes that are prohibited from adjustment include completed processes, processes that are currently in operation, smelting processes, and continuous casting processes; refining processes and transportation processes are processes that are allowed to be adjusted.

[0035] Furthermore, the preset threshold is configured to be 0.01 minutes, that is, when the actual completion time of any process deviates from the planned time by more than 0.01 minutes, the genetic algorithm optimizer is triggered.

[0036] Furthermore, the parallel simulator adopts an event-driven mechanism to advance the simulation time, specifically including: in each simulation loop, traversing all incomplete furnaces, checking whether the next process of each furnace meets the execution conditions, the execution conditions including the previous process has been completed and the equipment required for the current process is idle; taking all processes that meet the execution conditions as executable tasks, selecting the earliest task that can start from them for execution, and updating the equipment occupancy time and the current simulation time.

[0037] Furthermore, the scheduling adjustment suggestions include process delay, process advancement, or process time compression; the scheduling adjustment suggestions can be configured to be automatically applied to the current production plan, or provided to the scheduler for confirmation via a human-machine interface before application.

[0038] The following is a more detailed description of the real-time multi-objective optimization method for steel production process based on genetic algorithm according to an embodiment of the present invention.

[0039] I. Data Structure Design of the Production Model: This invention first performs structured modeling of the production plan. The entire production plan is defined as a casting, each casting contains multiple heats, and each heat goes through multiple process steps in sequence, such as smelting, refining, transportation, and continuous casting.

[0040] Therefore, three core data classes are defined: Process Steps: Record the process name, planned duration, actual duration, actual start time, actual end time, current status (planned, running, completed, etc.), and original planned start time.

[0041] Furnace batch class: Contains furnace batch number and a process dictionary, where the process name is the key and the corresponding process step object is the value. The furnace batch class also provides methods to obtain the actual total duration, the planned total duration, and information about a specified process.

[0042] Casting sequence category: includes casting sequence number, furnace sequence list, casting sequence start time, casting sequence end time, and casting sequence status, and provides a method to obtain the corresponding furnace sequence based on the furnace sequence number.

[0043] II. Event-driven scheduling mechanism of parallel simulators: The simulator uses an event-driven approach to advance time, rather than continuous time steps, thereby improving simulation efficiency. Its core scheduling logic is as follows: Each piece of equipment (smelting, refining, transportation, continuous casting) can only handle one heat at a time; each process in each heat must be executed in sequence, but the same process in different heats can be performed in parallel; the simulator maintains the busy / idle schedule for each piece of equipment, as well as the process index of each heat.

[0044] The main loop repeatedly performs the following operations: iterates through all incomplete furnace runs, checks whether the next process meets the execution conditions, i.e., the preceding process is completed and the required equipment is currently idle; treats all processes that meet the conditions as executable tasks and calculates their earliest start time (taking the maximum value of the current simulation time, equipment idle time, and the end time of the preceding process); selects the task with the earliest start time from all executable tasks for execution; during execution, adds a random fluctuation (e.g., evenly distributed within ±5 minutes) to the actual time of the process based on the planned time to simulate the uncertainty in real production; updates the actual start time, actual end time, and status of the process, and updates the busy / idle time of the corresponding equipment; if the absolute value of the deviation between the actual time and the planned time is detected to exceed a preset threshold (e.g., 0.01 minutes), the real-time optimizer is triggered; advances the simulation time to the next event occurrence time (if there are no tasks to be executed, advances to the next equipment availability time or advances in fixed steps); loops until all processes of all furnace runs are completed.

[0045] III. Design of the Genetic Algorithm Optimizer: When the simulator detects that the process time deviation exceeds the threshold, it calls the genetic algorithm optimizer for rescheduling optimization. The optimizer takes all incomplete processes at the current moment as the optimization object, and minimizes the total production time of each casting and minimizes the plan variation as the optimization objectives.

[0046] 1. Individual chromosome representation: Each individual represents a complete rescheduling scheme, which is encapsulated as a casting object and records the fitness score, estimated total production time and total plan deviation of the scheme.

[0047] 2. Fitness Assessment: For each individual, the simulator's unbiased planning time calculation function is called to obtain the total production time under the given scheme and the sum of the absolute values ​​of the planning deviations for all batches. Then, the fitness score is calculated: a baseline optimal time is set (e.g., 210 minutes). If the total time is less than or equal to the baseline time, the time score is 100 minus the saved time; if it is greater than the baseline time, the time score is 100 minus twice the excess time (minimum 0). The deviation score is 100 minus the total deviation (minimum 0). The final fitness is a weighted sum of the time score and the deviation score, each accounting for 50%.

[0048] 3. Population Initialization: Population size is set to 50. Each individual is generated by deep copying the original casting plan and randomly fine-tuning the planned durations of all unstarted refining and transport processes. The random adjustment range for refining processes is 27–37 minutes, and for transport processes it is 15–25 minutes.

[0049] 4. Selection Operation: A roulette wheel selection method is used. The probability of selection is calculated based on the fitness score of each individual in the current population. The higher the fitness score, the greater the probability of selection. In this way, superior individuals are selected to enter the next generation.

[0050] 5. Cross-operation: Single-point cross-operation is adopted. Two parent individuals are randomly selected, and with a 50% probability, the refining process planned time of some furnaces in the offspring individuals is copied from parent 2, while the rest retains the settings of parent 1.

[0051] 6. Mutation Operation: Perform mutation on individuals at a certain mutation rate (e.g., 0.3). Randomly select an incomplete refining or transportation process and add a random perturbation of -2 to 2 minutes (refining) or -3 to 3 minutes (transportation) to its current planned duration, ensuring that the adjusted time is still within a reasonable range.

[0052] 7. Elite Preservation and Iterative Evolution: In each generation, the two individuals with the highest fitness are retained and directly enter the next generation (elite preservation). The remaining individuals are generated through selection, crossover, and mutation. The population size remains constant at 50. The evolutionary cycle is 100 generations. The final output is the scheduling scheme corresponding to the individual with the highest fitness, and specific process time adjustment suggestions are generated based on the differences between this scheme and the original plan.

[0053] The following example illustrates the process of a short-process steelmaking plant where one casting cycle consists of four heats, with each heat going through four processes in sequence: smelting, refining, transportation, and continuous casting.

[0054] The system configuration is as follows: Planned processing times: smelting 32 minutes, refining 32 minutes, transportation 20 minutes, continuous casting 30 minutes.

[0055] Parallel equipment: Each process has independent equipment, which can process the same process in different batches in parallel.

[0056] Deviation trigger threshold: Optimization is triggered when the actual time of a process deviates from the planned time by more than 0.01 minutes.

[0057] Genetic algorithm parameters: population size 50, number of iterations 100, crossover rate 0.7, mutation rate 0.3.

[0058] The implementation process is as follows: The system initializes and generates a casting schedule containing four furnace batches, with each process arranged in the order of the scheduled time.

[0059] The simulator begins simulating production. When the smelting process of the first furnace is delayed by 2.5 minutes due to random factors, the system detects that the deviation exceeds the threshold.

[0060] The real-time optimizer is triggered. The genetic algorithm optimizer takes the current state (smelting delay, subsequent processes have not started) as input and reschedules all remaining processes. Due to special restrictions, the production plans of the smelting and continuous casting processes are prohibited from being adjusted. Only the refining and running processes can be adjusted.

[0061] After 100 generations of evolution, the genetic algorithm outputs an optimized solution to balance the production line load, and the system outputs optimization suggestions.

[0062] The simulator continues to advance, and if deviations occur again during subsequent production processes, the above optimization process is repeated.

[0063] Optimization Results: Under the original fixed plan, the total production time for each casting cycle was approximately 210 minutes. Introducing random deviations without dynamic optimization could extend the total time to over 230 minutes. After dynamic optimization using the system of this invention, the total production time was controlled at around 205.7 minutes, a reduction of approximately 25 minutes compared to the unoptimized scenario, and the total variation in the plan for each casting cycle was minimized.

[0064] The real-time multi-objective optimization method for steel production processes based on genetic algorithms in this invention employs an optimization scheduling framework. This framework includes core modules such as a simulator, optimizer, and production model, enabling flexible adaptation to different steelmaking processes, equipment configurations, and production objectives. This method can monitor the progress of each process in real time during production. When a process delay or advancement is detected, the genetic algorithm is used to quickly re-optimize the subsequent production plan. By dynamically adjusting process times and rearranging execution sequences, the total production time and plan deviation are minimized, achieving adaptive optimization of the production process.

[0065] Example 2 like Figure 10 As shown, embodiments of the present invention also provide a real-time multi-objective optimization system for steel production processes based on genetic algorithms, including: The planning initialization module is used to initialize the production plan and obtain an initial production plan that includes multiple furnace batches, each furnace batch contains multiple sequentially executed processes, and the planned duration of each process. The parallel simulation and monitoring module is connected to the planning initialization module and is used to extrapolate the production process based on the initial production plan. During the extrapolation process, random fluctuations are introduced into the actual time consumption of each process to simulate the uncertainty of real production, and the deviation between the actual completion time and the planned time of each process is monitored in real time to obtain the real-time production status and deviation monitoring results. The genetic algorithm optimization trigger module is connected to the parallel simulation and monitoring module. When the deviation monitoring result indicates that the deviation between the actual completion time and the planned time of any process exceeds a preset threshold, the genetic algorithm optimizer is triggered. The optimizer takes the set of all unfinished processes at the current moment as the optimization object, minimizes the total production time of each casting and minimizes the plan change as the optimization objective, runs the genetic algorithm to perform rescheduling optimization based on the current production status, and generates a rescheduling optimization scheme. The scheduling adjustment suggestion generation and application module is connected to the genetic algorithm optimization triggering module. It is used to compare the rescheduling optimization scheme with the initial production plan, generate scheduling adjustment suggestions including process time adjustment or execution order adjustment, and apply the suggestions to the current production plan to obtain the updated production plan. The loop control module, connected to the scheduling adjustment suggestion generation and application module and the parallel simulation and monitoring module, is used to control the parallel simulation and monitoring module to continue production based on the updated production plan, and to repeatedly execute the monitoring, optimization and adjustment steps until all furnace batches are completed.

[0066] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A real-time multi-objective optimization method for steel production process based on genetic algorithm, characterized in that, Includes the following steps: Step S1: Initialize the production plan to obtain an initial production plan that includes multiple furnace batches, each furnace batch contains multiple sequentially executed processes, and the planned duration of each process; Step S2: Start the parallel simulator, deduce the production process based on the initial production plan, introduce random fluctuations in the actual time consumption of each process during the deduction process to simulate the uncertainty of real production, and monitor the deviation between the actual completion time and the planned time of each process in real time to obtain the real-time production status and deviation monitoring results. Step S3: When the deviation monitoring result indicates that the deviation between the actual completion time and the planned time of any process exceeds a preset threshold, the genetic algorithm optimizer is triggered. The optimizer takes the set of all unfinished processes at the current moment as the optimization object, minimizes the total production time of each casting and minimizes the plan change as the optimization objective, and runs the genetic algorithm to perform rescheduling optimization based on the current production status to generate a rescheduling optimization scheme. Step S4: Compare the rescheduling optimization scheme with the initial production plan, generate scheduling adjustment suggestions that include process time adjustments or execution sequence adjustments, and apply the suggestions to the current production plan to obtain the updated production plan; Step S5: Control the parallel simulator to continue production based on the updated production plan, and repeat steps S2 to S4 until all furnace batches are completed.

2. The real-time multi-objective optimization method for steel production process based on genetic algorithm according to claim 1, characterized in that, The genetic algorithm in step S3 specifically includes: Population initialization: Based on the current production state, an initial population containing multiple individuals is generated by randomly fine-tuning the duration of unfinished processes, where each individual represents a scheduling scheme. Fitness assessment: For each individual in the initial population, calculate the estimated total production time and the planned deviation for each batch, and calculate the fitness score of each individual according to the preset multi-objective weighted rule; Selection operation: Based on the fitness score of each individual, the roulette wheel selection method is used to select multiple individuals with higher fitness scores from the current population as excellent individuals; Cross-operation: For the aforementioned outstanding individuals, the process time settings of different individuals are merged through cross-operation to generate offspring individuals; Mutation operation: For the offspring individual, the duration of some of the processes is randomly adjusted to obtain the mutated offspring individual; Elite retention: The mutated offspring individuals are merged with the individuals in the current generation population, and a number of individuals with the highest fitness scores are retained and directly enter the next generation population as at least a part of the next generation population.

3. The real-time multi-objective optimization method for steel production process based on genetic algorithm according to claim 2, characterized in that, The multi-objective weighted rule in the fitness assessment is as follows: combining the estimated total production time and the planned deviation of each batch, weights are assigned to the total production time target and the planned deviation target respectively, and the weighted score is calculated as the fitness score of the individual; wherein, the shorter the total production time and the smaller the sum of planned deviations, the higher the fitness score.

4. The real-time multi-objective optimization method for steel production process based on genetic algorithm according to claim 1, characterized in that, The minimization of plan changes in step S3 specifically means minimizing the sum of the absolute values ​​of the time deviations between the actual start time of each furnace and each process after rescheduling and the original planned start time.

5. The real-time multi-objective optimization method for steel production process based on genetic algorithm according to claim 1, characterized in that, In step S3, the genetic algorithm optimizer sets the production plans for completed processes or processes with special production constraints to be prohibited from adjustment when rescheduling and optimizing incomplete processes.

6. The real-time multi-objective optimization method for steel production process based on genetic algorithm according to claim 5, characterized in that, The processes that are prohibited from adjustment include completed processes, processes that are currently in operation, smelting processes, and continuous casting processes; refining processes and transportation processes are processes that are allowed to be adjusted.

7. The real-time multi-objective optimization method for steel production process based on genetic algorithm according to claim 1, characterized in that, The preset threshold is configured to be 0.01 minutes, meaning that when the actual completion time of any process deviates from the planned time by more than 0.01 minutes, the genetic algorithm optimizer is triggered.

8. The real-time multi-objective optimization method for steel production process based on genetic algorithm according to claim 1, characterized in that, The parallel simulator uses an event-driven mechanism to advance the simulation time, specifically including: in each simulation loop, traversing all incomplete furnace batches, checking whether the next process of each furnace batch meets the execution conditions, the execution conditions including that the previous process has been completed and the equipment required for the current process is idle; taking all processes that meet the execution conditions as executable tasks, selecting the earliest task that can start from them for execution, and updating the equipment occupancy time and the current simulation time.

9. The real-time multi-objective optimization method for steel production process based on genetic algorithm according to claim 1, characterized in that, The scheduling adjustment suggestions include process delay, process advancement, or process time compression; the scheduling adjustment suggestions can be configured to be automatically applied to the current production plan, or provided to the scheduler for confirmation via a human-machine interface before application.

10. A real-time multi-objective optimization system for steel production process based on genetic algorithm, characterized in that, include: The planning initialization module is used to initialize the production plan and obtain an initial production plan that includes multiple furnace batches, each furnace batch contains multiple sequentially executed processes, and the planned duration of each process. The parallel simulation and monitoring module is connected to the planning initialization module and is used to extrapolate the production process based on the initial production plan. During the extrapolation process, random fluctuations are introduced into the actual time consumption of each process to simulate the uncertainty of real production, and the deviation between the actual completion time and the planned time of each process is monitored in real time to obtain the real-time production status and deviation monitoring results. The genetic algorithm optimization trigger module is connected to the parallel simulation and monitoring module. When the deviation monitoring result indicates that the deviation between the actual completion time and the planned time of any process exceeds a preset threshold, the genetic algorithm optimizer is triggered. The optimizer takes the set of all unfinished processes at the current moment as the optimization object, minimizes the total production time of each casting and minimizes the plan change as the optimization objective, runs the genetic algorithm to perform rescheduling optimization based on the current production status, and generates a rescheduling optimization scheme. The scheduling adjustment suggestion generation and application module is connected to the genetic algorithm optimization triggering module. It is used to compare the rescheduling optimization scheme with the initial production plan, generate scheduling adjustment suggestions including process time adjustment or execution order adjustment, and apply the suggestions to the current production plan to obtain the updated production plan. The loop control module, connected to the scheduling adjustment suggestion generation and application module and the parallel simulation and monitoring module, is used to control the parallel simulation and monitoring module to continue production based on the updated production plan, and to repeatedly execute the monitoring, optimization and adjustment steps until all furnace batches are completed.

Images