Differential evolution method for workshop scheduling under worker constraint for electrolytic aluminum

Through feature-driven learning differential evolution optimization algorithm, flexible work workshop scheduling in the electrolytic aluminum production line is optimized under the constraints of workers, solving the problems of low production efficiency and high energy consumption, and achieving more efficient, flexible and user-friendly production scheduling.

CN118982163APending Publication Date: 2024-11-19LANZHOU UNIVERSITY OF TECHNOLOGY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410890881.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-04
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

In electrolytic aluminum production lines, it is difficult for the existing technology to effectively optimize the worker factors and resource allocation, resulting in low production efficiency, high energy consumption, and difficulty in responding to market changes and handling emergencies quickly.

Method used

Using a feature-driven learning differential evolution algorithm (MLDE), combining random initialization, differential evolution algorithm and decision-making neural network, flexible work workshop scheduling under worker constraints is optimized, and by adjusting the work Sequential and worker allocation, reducing maximum completion time and total energy consumption.

Benefits of technology

It achieves more efficient, flexible and user-friendly production scheduling, improves production efficiency, reduces energy consumption, and effectively responds to complex and dynamic production environment changes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118982163A_ABST
    Figure CN118982163A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of electrolytic aluminum production scheduling in the manufacturing industry, in particular to a differential evolution method for workshop scheduling of electrolytic aluminum under worker constraint, which is used for minimizing maximum completion time and total energy consumption by using a multi-feature driven learning type differential evolution optimization algorithm. According to the algorithm, the operation efficiency and performance of the flexible job shop scheduling system considering worker factors can be optimized; comprising the following steps: obtaining an initial scheduling solution by using a random initialization strategy, optimizing the scheduling solution by using a differential evolution operator, analyzing the scheduling solution by using a decision neural network and selecting a quadratic optimization operator, and completing quadratic optimization of the scheduling solution by using the quadratic optimization operator. The method has the advantages that a typical scheduling problem integer programming model in the aluminum machining process is defined, the flexible job shop scheduling problem of worker factors is considered, and an evaluation criterion combining energy consumption and completion time is provided.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of flexible job shop scheduling in manufacturing industry, and specifically is a differential evolution method for job shop scheduling under worker constraints for electrolytic aluminum. Background Art

[0002] In today's global market environment, the importance of scheduling issues, especially in electrolytic aluminum production lines, has become increasingly prominent. It is a key link in improving production efficiency, reducing operating costs, and ensuring product quality and delivery time. As a basic field of industrial manufacturing, the electrolytic aluminum industry has a complex production process and involves many processes, from ore processing, anode preparation, electrolysis and aluminum extraction and refining to the final finished product inspection. The efficient operation of each link is inseparable from accurate scheduling strategies. Optimizing these processes can not only increase production capacity and reduce resource waste, but also win valuable time and cost advantages for enterprises in the fierce market competition.

[0003] For China's manufacturing industry, the importance of the flexible job shop scheduling problem (FJSP) is particularly prominent. In the past few decades, China's manufacturing industry has experienced rapid development and growth, but in the face of rapid changes in global market demand and increasing personalized needs, the manufacturing industry urgently needs to improve its production flexibility and efficiency. Flexible job shops, with their ability to adapt to product diversification and small batch production, have become an effective way to achieve this goal. Through intelligent and flexible scheduling algorithms, it is possible to optimize resource allocation to the greatest extent, shorten production cycles, and improve response speed to market changes, which is an important support for promoting high-quality development of the manufacturing industry.

[0004] In addition, the importance of worker factors in workshop scheduling cannot be ignored. Every worker on the production line has unique skills, efficiency and work preferences. The comprehensive consideration of these factors has a decisive impact on improving production efficiency, ensuring product quality and enhancing worker satisfaction. Therefore, an advanced scheduling system should not only optimize the operation plan of machinery and equipment, but also reasonably plan human resources and realize human-machine collaboration to achieve optimal production efficiency.

[0005] Among many scheduling algorithms, the differential evolution algorithm has shown great application potential in the flexible job shop scheduling problem with its unique advantages. It is based on the idea of ​​population evolution. By performing differential and mutation operations between individuals in the population, it effectively explores the solution space while retaining individuals that are conducive to problem solving. This algorithm has a strong global search capability and the ability to quickly adapt to changing environments. This algorithm is not only suitable for static scheduling problems, but can also effectively respond to emergencies and dynamic changes in the production process, such as order changes, equipment failures, etc., showing excellent flexibility and robustness. Especially when considering complex worker factors, the parameter adjustment and adaptive strategy of the differential evolution algorithm can provide strong support for achieving more efficient and humane scheduling.

[0006] In summary, the scheduling problem of electrolytic aluminum production line is of great significance to improving the competitiveness of enterprises. The flexible job shop scheduling problem plays a key role in promoting the high-quality development of China's manufacturing industry, and the reasonable consideration of worker factors is an important part of achieving effective scheduling. The differential evolution algorithm, with its excellent search ability and adaptability, has shown great potential and advantages in solving this series of complex scheduling problems, and has provided the possibility of achieving more efficient, flexible and humane production scheduling. Summary of the invention

[0007] The purpose of the present invention is to address the problems existing in the prior art and provide a differential evolution method for workshop scheduling under worker constraints for electrolytic aluminum. A feature-driven learning differential evolution optimization algorithm (Multi-feature Based Learning Differential Evolution Algorithm, MLDE) is used to minimize the maximum make span (makespan, CMAX) and total energy consumption (total energy consumption, TEC). The algorithm can optimize the operating efficiency and performance of the flexible job shop scheduling system taking worker factors into consideration.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A differential evolution method for workshop scheduling under worker constraints for electrolytic aluminum, characterized by comprising the following steps:

[0010] Step 1: Use a random initialization strategy to obtain an initial scheduling solution. The random initialization strategy means that at the beginning of the solution process, one or more groups of initial solutions are randomly generated. These initial solutions serve as the starting point of the algorithm search, aiming to explore the solution space and provide a springboard for finding the optimal or approximately optimal solution.

[0011] Step 2: Differential evolution algorithm optimizes the scheduling solution. In the flexible job shop scheduling problem considering worker factors, the differential evolution algorithm first randomly generates an initial population to represent different scheduling schemes, where each individual includes not only the processing order of the job, but also the workers assigned to a specific task. Then, the algorithm updates the population through an iterative process, using differential mutation, crossover, and selection operations in each generation. The whole process is repeated until the stopping condition is met, such as reaching a predetermined number of iterations or the quality of the solution meets the requirements.

[0012] In the differential mutation step, candidate solutions are generated based on other individuals in the population; in the crossover step, the diversity of solutions is increased; in the selection step, individuals with better performance are selected to enter the next generation based on the fitness function, usually minimizing the total completion time or maximizing production efficiency.

[0013] Step 3: The decision neural network analyzes the scheduling solution and selects the quadratic optimization operator; in the flexible job shop scheduling problem considering worker factors, the application of the quadratic optimization operator aims to further improve the optimization level of the existing scheduling scheme. This process is usually carried out after the preliminary scheduling scheme is formed. The specific steps include evaluating each job arrangement in the current scheduling scheme and its allocation to specific workers, and then identifying potential optimization space based on the worker's skill level and the skill requirements of the job; the quadratic optimization operator tries to reduce the total completion time, reduce idle time or improve production efficiency by adjusting the job sequence, reallocating workers or adjusting the match between jobs and workers. This process involves complex decision models and heuristic strategies to ensure the overall optimization of the scheduling scheme without violating worker restrictions.

[0014] Preferably, the five stages of ore processing, anode preparation, electrolysis, and aluminum extraction and refining in the electrolytic aluminum process are modeled as a flexible job shop scheduling model with worker factors.

[0015] Preferably, in step 1, by introducing randomness, this method can increase the diversity of solutions and prevent the algorithm from falling into the local optimum too early, thereby improving the efficiency and quality of the solution. In the application of flexible job shop scheduling, the random initialization method also helps to adapt to complex and changeable production tasks and process routes, and provides the possibility for efficient and flexible production scheduling;

[0016] The specific constraints are: The MILP formulas of OCRA-FJSP with maximum completion time and minimum TEC are (1) and (2), where f 1 is the maximum completion time of the scheduling scheme, f 2is the TEC of the scheduling scheme, PE is the working energy consumption of the machine, IE is the idle energy consumption of the machine, and AE is the auxiliary energy consumption of the factory, including lighting and air conditioning; the operation priority constraint is introduced to ensure that each operation in each job is completed in sequence, and the constraint formula is shown in (3); the constraint (4) is introduced to require that each operation can only be completed on one machine; the constraint (5) is introduced to ensure that each operation is performed by only one worker; the machine constraint (6) is introduced to ensure that any machine only processes one operation at a time; the worker constraint (7) is introduced to ensure that any worker only processes one operation at a time; the constraints (8) and (9) are introduced to require that each operation has a corresponding worker and machine;

[0017]

[0018] Preferably, in step 2, the powerful global search capability and good adaptability of the differential evolution algorithm can effectively handle the complexity and multi-objective optimization problems of FJSP. Specifically, the differential evolution operator is an optimization method based on SHADE. Through the differences between individuals in the population, the differential evolution operator is used to complete the mutation of the operation sequence. Each time the differential evolution operator is executed, the machine sequence and the worker sequence are randomly generated under the constraints. The solution generated by the differential evolution operator is a continuous solution corresponding to a unique operation sequence, and an operation sequence corresponds to countless continuous solutions. The mutation formula of the differential evolution operator in SHADE is current-to-best. The current-to-best formula is shown in (10), where F is an adaptive parameter, X_rand1 and X_rand2 are randomly selected from the current population, V is the mutated individual, and X is the original individual. In OCRAFJSP, the Pareto front individuals are defined as the best individuals, X_best is randomly selected from the Pareto front individuals, and all non-Pareto front individuals tend to be close to the Pareto front individuals. The Pareto front individuals are not only optimized by the global optimization operator, but also use random methods to complete secondary optimization in the local optimization operator. The Pareto front individuals that have completed two optimizations are conducive to maintaining the diversity of the population.

[0019] V=X+F·((X best -X)+(X rand1 -X rand2 )) (10)

[0020] Because the operation sequence is generated by the differential evolution operator, the differential evolution operator determines the quality of the final solution. Compared with randomly generated new individuals, individuals that do not meet the constraints contain more useful information. The mutation of the global optimization operator depends on the differences between different individuals, and the addition of new individuals slows down the convergence of the population.

[0021] Preferably, in step three, the advantage of the quadratic optimization operator is that it can improve the accuracy and convergence speed of the search process, especially when dealing with optimization problems with complex search spaces, it can more effectively refine the solution space and accelerate the search for the optimal or approximately optimal solution; specifically, the quadratic optimization operator consists of 4 independent search operators. Compared with the global search operator, the local search operator is only used to optimize the Pareto frontier individuals. The local search operator not only improves the diversity of the population, but also directly improves the quality of the final solution. Four different local search operators are designed to further optimize the scheduling solution, namely operator 1, operator 2, operator 3 and operator 4. The four local search operators are shown in Table 3;

[0022] Table 3 Quadratic optimization operators

[0023]

[0024] A computer-readable storage medium contains a computer program, which can implement the method steps provided above when processed by a CPU.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] (1) The present invention defines an integer programming model for a typical type of scheduling problem in the electrolytic aluminum process: the flexible job shop scheduling problem considering worker factors, and proposes an evaluation criterion that combines energy consumption and completion time.

[0027] (2) A high-quality random initialization method is constructed using a new population generation strategy, which enables the algorithm to quickly search for better areas.

[0028] (3) Use the differential evolution operator to perform population optimization and obtain a preliminary optimized scheduling solution.

[0029] (4) During the population evolution process, the operation of the neural network decision model is used to select the secondary optimization operator, and the characteristic information of each operation is summarized as knowledge to guide the algorithm to perform the secondary optimization operation and improve the success rate of the operation.

[0030] (5) The present invention has simple logic, is easy to implement and expand, and can expand the optimizer to meet most scheduling problems in the current intelligent manufacturing production field. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Schematic diagram of the learning differential evolution algorithm solving the scheduling problem in the aluminum industry;

[0032] Figure 2 This is a diagram of the electrolytic aluminum process;

[0033] Figure 3This is the process diagram of electrolytic aluminum;

[0034] Figure 4 Schematic diagram of differential evolution operator;

[0035] Figure 5 It is a schematic diagram of the neural network decision model;

[0036] Figure 6 Schematic diagram of the quadratic optimization operator; Figure (a) is the Gantt chart of the original scheduling solution, and Figure (b) is the Gantt chart of the scheduling solution after quadratic optimization. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0038] Example 1

[0039] A differential evolution method for workshop scheduling under worker constraints for electrolytic aluminum includes the following steps:

[0040] Step 1: Use a random initialization strategy to obtain an initial scheduling solution. The random initialization strategy means that at the beginning of the solution process, one or more groups of initial solutions are randomly generated. These initial solutions serve as the starting point of the algorithm search, aiming to explore the solution space and provide a springboard for finding the optimal or approximately optimal solution.

[0041] Step 2: Differential evolution algorithm optimizes the scheduling solution. In the flexible job shop scheduling problem considering worker factors, the differential evolution algorithm first randomly generates an initial population to represent different scheduling schemes, where each individual includes not only the processing order of the job, but also the workers assigned to a specific task. Then, the algorithm updates the population through an iterative process, using differential mutation, crossover, and selection operations in each generation. The whole process is repeated until the stopping condition is met, such as reaching a predetermined number of iterations or the quality of the solution meets the requirements.

[0042] In the differential mutation step, candidate solutions are generated based on other individuals in the population; in the crossover step, the diversity of solutions is increased; in the selection step, individuals with better performance are selected to enter the next generation based on the fitness function, usually minimizing the total completion time or maximizing production efficiency.

[0043] Step 3: The decision neural network analyzes the scheduling solution and selects the quadratic optimization operator; in the flexible job shop scheduling problem considering worker factors, the application of the quadratic optimization operator aims to further improve the optimization level of the existing scheduling scheme. This process is usually carried out after the preliminary scheduling scheme is formed. The specific steps include evaluating each job arrangement in the current scheduling scheme and its allocation to specific workers, and then identifying potential optimization space based on the worker's skill level and the skill requirements of the job; the quadratic optimization operator tries to reduce the total completion time, reduce idle time or improve production efficiency by adjusting the job sequence, reallocating workers or adjusting the match between jobs and workers. This process involves complex decision models and heuristic strategies to ensure the overall optimization of the scheduling scheme without violating worker restrictions.

[0044] Figure 1 This is a schematic diagram of the learning differential evolution algorithm solving the scheduling problem in the aluminum industry. The problem to be solved comes from the industrial electrolytic aluminum process. Problem characteristics refer to the use of prior knowledge of the problem to improve the algorithm optimization performance. Multi-objective optimization is to fit the actual optimization goal. Worker constraints are factors that cannot be ignored. Key operations are the key to guiding algorithm optimization. In the algorithm space, a learning optimization algorithm based on differential evolution is used, including the random initialization method, neural network decision model, differential evolution optimization operator and quadratic optimization operator, which together constitute the proposed algorithm.

[0045] Example 2

[0046] A differential evolution method for worker-constrained job shop scheduling for electrolytic aluminum is proposed. The proposed flexible job shop scheduling problem considering worker factors is described as follows:

[0047] Figure 2 This is a diagram of the electrolytic aluminum process. The main steps include mining, anode preparation, electrolysis, and aluminum extraction and refining. Figure 3 is the process diagram of aluminum electrolysis. The MILP parameters of OCRA-FJSP are shown in Table 1. The MILP formulas of OCRA-FJSP with maximum completion time and minimum TEC are (1) and (2), where f 1 is the maximum completion time of the scheduling scheme, f 2 is the TEC of the scheduling scheme, PE is the working energy consumption of the machine, IE is the idle energy consumption of the machine, and AE is the auxiliary energy consumption of the factory, including lighting and air conditioning.

[0048] Table 1 Parameter settings of OCRA-FJSP model

[0049]

[0050]

[0051] To illustrate this problem clearly, Table 2 shows a simple OCRA-FJSP example with three jobs, four machines, and four workers. It is worth noting that each operation can only be processed on one machine selected from the predefined machine set at the same time. The processing time on the same machine is related to the proficiency of the operator. i As an example, the first operation of M 11 ={M 1 ,M 3}Execute 11 , the basic processing times are 20 and 10 respectively. Assume that worker W is selected 1 and W 2 Operating Machine M 1 and M 3 ,W 1 Operation M 1 The skill proficiency of W is 0.9. 1 In M 1 Upper processing operation O 11 The actual processing time is 20×0.9=18.0. Similarly, if we choose W 2 To operate M 3 , Operation O 11 The actual processing time is 10×0.8=8.0.

[0052] The operation priority constraint is introduced to ensure that each operation in each job is completed in order. The constraint formula is shown in (3). Constraint (4) is introduced to require that each operation can only be completed on one machine. Constraint (5) ensures that each operation is performed by only one worker. Machine constraint (6) is introduced to ensure that any machine only processes one operation at a time. Worker constraint (7) ensures that any worker only processes one operation at a time. Constraints (8) and (9) are introduced to require that each operation has a corresponding worker and machine.

[0053] Table 2 OCRA-FJSP example

[0054]

[0055]

[0056] Example 3

[0057] The differential evolution operator is an optimization method based on SHADE. Differential evolution algorithms such as Figure 4As shown, the differential evolution operator is used to complete the mutation of the operation sequence through the differences between individuals in the population. Each time the differential evolution operator is executed, the machine sequence and the worker sequence are randomly generated under the constraints. The solution generated by the differential evolution operator is a continuous solution corresponding to a unique operation sequence. An operation sequence corresponds to countless continuous solutions. The mutation formula of the differential evolution operator in SHADE is current-to-best. The current-to-best formula is shown in (10), and F is an adaptive parameter. X_rand1 and X_rand2 are randomly selected from the current population. V is the mutated individual, and X is the original individual. In OCRAFJSP, the Pareto frontier individual is defined as the optimal individual. X_best is randomly selected from the Pareto frontier individuals. All non-Pareto frontier individuals tend to be close to the Pareto frontier individuals. The Pareto frontier individuals are not only optimized by the global optimization operator, but also use random methods to complete secondary optimization in the local optimization operator. The Pareto frontier individuals that have completed two optimizations are conducive to maintaining the diversity of the population.

[0058] V=X+F·((X best -X)+(X rand1 -X rand2 )) (10)

[0059] Because the operation sequence is generated by the differential evolution operator, the differential evolution operator determines the quality of the final solution. Compared with randomly generated new individuals, individuals that do not meet the constraints contain more useful information. The mutation of the global optimization operator depends on the differences between different individuals, and the addition of new individuals slows down the convergence of the population.

[0060] Figure 4 A material processing flow is shown, covering the various steps from raw materials to the final product. Initially, the raw materials (marked with diamond-shaped icons) are transformed into primary castings through melting and casting steps. These castings are then further processed in the processing steps to form regular spherical materials. Subsequently, these spherical materials are integrated into a larger component during the assembly process. Finally, this component is used for performance testing, including mechanical testing and surface treatment, to ensure that the product meets the expected quality and performance standards. The entire process not only shows the deformation and assembly of the material, but also includes quality control and testing to ensure the final quality of the product.

[0061] Example 4

[0062] The quadratic optimization operator consists of four independent search operators. Compared with the global search operator, the local search operator is only used to optimize the Pareto frontier individuals. The local search operator not only improves the diversity of the population, but also directly improves the quality of the final solution. The choice of local search is determined by the neural network. The structure of the neural network is as follows Figure 5 shown. Figure 5 It represents a neural network structure based on the kernel method, which is used to process and analyze the input vector x. The input layer contains m0 elements, each of which is connected to the Ns nodes in the hidden layer through a kernel function K. This kernel function calculates the inner product between the input vector x and the weight x1, x2, ..., xNs of each node to capture the nonlinear relationship in the data. The result of each kernel is multiplied by the corresponding weight u1, u2, ..., uNs, and all are summed up, plus a bias term b, and finally the result yy is output through a linear layer. This structure is particularly suitable for pattern recognition and classification tasks of high-dimensional data.

[0063] Four different local search operators are designed to further optimize the scheduling solution, namely Operator 1, Operator 2, Operator 3 and Operator 4. The four local search operators are shown in Table 3. A quadratic optimization example is Figure 6 shown.

[0064] Table 3 Quadratic optimization operators

[0065]

[0066] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A differential evolution method for worker-constrained workshop scheduling of electrolytic aluminum, characterized in that: The steps include: Step 1: Use a random initialization strategy to obtain an initial scheduling solution. The random initialization strategy means that at the beginning of the solution process, one or more groups of initial solutions are randomly generated. These initial solutions serve as the starting point of the algorithm search, aiming to explore the solution space and provide a springboard for finding the optimal or approximately optimal solution. Step 2: Differential evolution algorithm optimizes the scheduling solution. In the flexible job shop scheduling problem considering worker factors, the differential evolution algorithm first randomly generates an initial population to represent different scheduling schemes, where each individual includes not only the processing order of the job, but also the workers assigned to a specific task. Then, the algorithm updates the population through an iterative process, using differential mutation, crossover, and selection operations in each generation. The whole process is repeated until the stopping condition is met, such as reaching a predetermined number of iterations or the quality of the solution meets the requirements. In the differential mutation step, candidate solutions are generated based on other individuals in the population; in the crossover step, the diversity of solutions is increased; in the selection step, individuals with better performance are selected to enter the next generation based on the fitness function, usually minimizing the total completion time or maximizing production efficiency. Step 3: The decision neural network analyzes the scheduling solution and selects the quadratic optimization operator; in the flexible job shop scheduling problem considering worker factors, the application of the quadratic optimization operator aims to further improve the optimization level of the existing scheduling scheme. This process is usually carried out after the preliminary scheduling scheme is formed. The specific steps include evaluating each job arrangement in the current scheduling scheme and its allocation to specific workers, and then identifying potential optimization space based on the worker's skill level and the skill requirements of the job; the quadratic optimization operator tries to reduce the total completion time, reduce idle time or improve production efficiency by adjusting the job sequence, reallocating workers or adjusting the match between jobs and workers. This process involves complex decision models and heuristic strategies to ensure the overall optimization of the scheduling scheme without violating worker restrictions.

2. The differential evolution method for worker-constrained workshop scheduling for electrolytic aluminum according to claim 1, characterized in that: The five stages of ore processing, anode preparation, electrolysis, and aluminum extraction and refining in the electrolytic aluminum process are modeled as a flexible job shop scheduling model with worker factors taken into account.

3. The differential evolution method for worker-constrained workshop scheduling for electrolytic aluminum according to claim 1, characterized in that: In step 1, by introducing randomness, this method can increase the diversity of solutions and prevent the algorithm from falling into the local optimum too early, thereby improving the efficiency and quality of the solution. The specific constraints are: The MILP formulas of OCRA-FJSP with the minimum maximum completion time and TEC are (1) and (2), where f1 is the maximum completion time of the scheduling scheme, f2 is the TEC of the scheduling scheme, PE is the working energy consumption of the machine, IE is the idle energy consumption of the machine, and AE is the auxiliary energy consumption of the factory, including lighting and air conditioning; the operation priority constraint is introduced to ensure that each operation in each job is completed in sequence, and the constraint formula is shown in (3); the constraint (4) is introduced to require that each operation can only be completed on one machine; the constraint (5) is introduced to ensure that each operation is performed by only one worker; the machine constraint (6) is introduced to ensure that any machine only processes one operation at a time; the worker constraint (7) is introduced to ensure that any worker only processes one operation at a time; the constraints (8) and (9) are introduced to require that each operation has a corresponding worker and machine; 4. The differential evolution method for worker-constrained workshop scheduling for electrolytic aluminum according to claim 1 is characterized in that: In step 2, the differential evolution algorithm has a strong global search capability and good adaptability, which can effectively handle the complexity and multi-objective optimization problems of FJSP. Specifically, the differential evolution operator is an optimization method based on SHADE. Through the differences between individuals in the population, the differential evolution operator is used to complete the mutation of the operation sequence. Each time the differential evolution operator is executed, the machine sequence and the worker sequence are randomly generated under the constraints. The solution generated by the differential evolution operator is a continuous solution corresponding to a unique operation sequence, and an operation sequence corresponds to countless continuous solutions. The mutation formula of the differential evolution operator in SHADE is current-to-best. The current-to-best formula is shown in (10), where F is an adaptive parameter, X_rand1 and X_rand2 are randomly selected from the current population, V is the mutated individual, and X is the original individual. V=X+F·((XbestX)+(Xrand1-Xrand2))(10).

5. The differential evolution method for worker-constrained workshop scheduling for electrolytic aluminum according to claim 1 is characterized in that: In step 3, the advantage of the quadratic optimization operator is that it can improve the accuracy and convergence speed of the search process. Especially when dealing with optimization problems with complex search spaces, it can more effectively refine the solution space and accelerate the search for the optimal or approximately optimal solution. Specifically, the quadratic optimization operator consists of four independent search operators, namely: Operator 1: Randomly select a key operation and randomly insert it into the previous operation. The selected operation corresponds to the worker and machine unchanged. Operator 2: Randomly select a key operation and randomly assign available machines and available workers; Operator 3: randomly selects an operation from the operation sequence and inserts it into a random position, and randomly assigns an available worker to the operation; Operator 4: Randomly select a worker from the worker sequence and insert it into a random position.

6. A computer-readable storage medium comprising a computer program, which, when processed by a CPU, can implement the method steps provided in claims 1-5.