An intelligent production scheduling optimization method and system based on large models

Abstract

The present invention discloses an intelligent production scheduling optimization method based on a large model, comprising the following steps: S1, data collection and preprocessing; collecting equipment operation data, material supply data, production progress data, and production task information; cleaning, normalizing, and outlier processing the collected data to generate structured data; S2, determining the optimization objectives of the scheduling scheme and the constraint conditions of the production workshop, constructing a scheduling model, and generating a preliminary scheduling scheme; S3, using a scheduling scheme evaluation algorithm to evaluate the preliminary scheduling scheme; S4, outputting the optimal production scheduling scheme and applying this scheme to the production system. The present invention adopts the entropy weight multi-objective decision-making method combined with the multi-objective dragonfly optimization algorithm, which can effectively evaluate and optimize the production scheduling scheme. Compared with traditional mathematical programming methods and heuristic algorithms, the method of the present invention can significantly improve the scheduling efficiency while ensuring the quality of the solution.

Description

Technical Field

[0001] The present invention relates to the technical field of production scheduling control, and particularly relates to an intelligent production scheduling optimization method and system based on a large model. Background Art

[0002] With the advent of the Industrial 4.0 era, intelligent manufacturing has become an important direction for the development of the manufacturing industry. In intelligent manufacturing, production scheduling optimization is a key link to improve production efficiency, reduce costs, and enhance product quality. Production scheduling optimization involves multiple objectives, such as minimizing production costs, minimizing production time, maximizing equipment utilization rate, etc. These objectives often conflict with each other, making production scheduling a complex multi-objective optimization problem. Currently, production scheduling optimization methods mainly include traditional mathematical programming methods, heuristic algorithms, and intelligent optimization algorithms. Traditional mathematical programming methods have a large amount of calculation when dealing with large-scale problems and are difficult to meet real-time requirements. Although heuristic algorithms have a fast calculation speed, they often cannot guarantee the optimality of the solution. Intelligent optimization algorithms, such as genetic algorithms, particle swarm optimization algorithms, etc., although have solved the above problems to a certain extent, still have problems such as slow convergence speed, easy to fall into local optimum, and insufficient diversity of solutions when dealing with multi-objective optimization problems.

[0003] In addition, when the existing technology deals with multi-objective optimization problems, it often lacks an effective method for determining objective weights, resulting in difficulty in balancing the importance of different objectives in practical applications. At the same time, the existing algorithms also have deficiencies in maintaining the diversity of solutions, and are prone to generating overly concentrated solution sets, thus affecting the flexibility and adaptability of the scheduling scheme. Summary of the Invention

[0004] In view of the above problems, the present invention proposes an intelligent production scheduling optimization method based on a large model. This method combines the entropy weight multi-objective decision-making method and the multi-objective dragonfly optimization algorithm, and realizes the efficient optimization of production scheduling problems through steps such as data collection and preprocessing, scheduling model construction, scheduling scheme evaluation, and optimal scheme output. The present invention determines the weights of each optimization objective through the entropy weight method, enabling the algorithm to more reasonably balance the importance of different objectives; at the same time, introducing the multi-objective dragonfly optimization algorithm improves the convergence speed and diversity of the solutions of the algorithm, thereby obtaining a better production scheduling scheme.

[0005] The technical solution adopted by the present invention to solve its technical problems is: First, provide an intelligent production scheduling optimization method based on a large model, including the following steps:

[0006] S1. Data collection and preprocessing; collect equipment operation data, material supply data, production progress data, and production task information; clean, normalize, and process outliers for the collected data to generate structured data;

[0007] S2. Determine the optimization objectives of the scheduling plan and the constraint conditions of the production workshop. Extract features and learn patterns from historical production data through a large model, construct a scheduling model, and generate a preliminary scheduling plan;

[0008] S3. Use the scheduling plan evaluation algorithm to evaluate the preliminary scheduling plan based on the dynamic parameter adjustment ability of the large model;

[0009] S4. Output the optimal production scheduling plan and apply this plan to the production system;

[0010] Among them, the scheduling plan evaluation algorithm in S3 uses the entropy weight multi-objective decision-making method combined with the multi-objective dragonfly optimization algorithm to evaluate the scheduling plan.

[0011] Furthermore, the entropy weight multi-objective decision-making method combined with the multi-objective dragonfly optimization algorithm in S3 includes the following steps:

[0012] A1. Extract multiple production samples from the preprocessed data, use the large model to generate simulated samples of future production scenarios, and standardize the production samples to unify each optimization objective in the same direction. Each sample represents the actual production situation within a time period;

[0013] A2. Calculate the entropy value of each optimization index according to the extracted sample data E j , and calculate the weight of each optimization index based on the entropy value ω j ;

[0014] A3. Initialize the dragonfly population, generate a candidate solution set of the initial scheduling plan through the large model, and assign an initial position and speed to each dragonfly. Each dragonfly represents a scheduling plan;

[0015] A4. Calculate the fitness of each dragonfly. According to the optimization objective weight ω j and the values of each optimization objective f j (X i ), calculate the comprehensive fitness; The formula for calculating the comprehensive fitness is:

[0016] ;

[0017] Among them, f j ( X l ) is the value of the i th scheduling plan on the j th optimization objective; ω j is the weight of each optimization objective; n is the total number of optimization objectives;

[0018] A5. Use the Pareto dominance relationship to find the current non-dominated solution set as the current optimal solution;

[0019] A6. For each dragonfly, calculate the contributions of separation, alignment, cohesion, food attraction, and predator avoidance behaviors based on the information of its neighbors, and update the speed and position of the dragonfly by integrating these behaviors;

[0020] A7. Re-evaluate the fitness of all dragonflies and update the Pareto optimal solution set; if the fitness of the newly generated scheduling scheme is better than that of some scheduling schemes in the current non-dominated solution set S then add it to the non-dominated solution set S and remove the dominated scheduling schemes;

[0021] A8. When the predetermined number of iterations G is reached, the algorithm terminates; otherwise, repeat steps A4 to A7;

[0022] A9. Select the optimal production scheduling scheme from the Pareto optimal solution set.

[0023] Furthermore, the formula for A2 to calculate the entropy value of each optimization objective is:

[0024]

[0025] where p jl is the proportion of the l th sample under the jth optimization objective, and m is the number of samples.

[0026] Furthermore, the formula for A2 to calculate the weight of each optimization index is:

[0027] ;

[0028] where ω j is the weight of the jth optimization objective, n is the total number of optimization objectives, E j is the entropy value of each optimization index.

[0029] Furthermore, A3 initializes the dragonfly population including:

[0030] Set the size N of the dragonfly population, that is, the number of initial scheduling schemes; the size N of the dragonfly population is dynamically adjusted by the large model according to the complexity of the production task and the learning results of historical scheduling data;

[0031] Randomly generate an initial position for each dragonfly X i , X i∈ lb , ub where lb and ub are the lower and upper bounds of the decision variables, respectively;

[0032] Generate an initial velocity for each dragonfly V i , V i ∈ [- V max , V max ;

[0033] where V max is the maximum velocity, set according to the scale and complexity of the production scheduling problem;

[0034] Set the maximum number of iterations G of the algorithm;

[0035] Set the neighborhood radius r to determine the neighbors of the dragonflies.

[0036] Furthermore, A5 uses the Pareto dominance relationship to find the current non-dominated solution set, which includes the current optimal solutions:

[0037] A51. Initialize the non-dominated solution set; use all scheduling plans in the population as the initial candidate solution set;

[0038] A52. Screen the non-dominated solutions;

[0039] Traverse each scheduling plan in the initial candidate solution set i , and compare it pairwise with other scheduling plans in the candidate solution set g ;

[0040] If the fitness of scheduling plan i is greater than that of scheduling plan g , and scheduling plan i is significantly better than scheduling plan g in at least one key optimization objective, then scheduling plan i dominates scheduling plan g ;

[0041] If scheduling plan i dominates scheduling plan g , then remove scheduling plan g from the candidate solution set;

[0042] If scheduling plan i is not dominated by any other scheduling plan, then add scheduling plan i to the non-dominated solution set;

[0043] A53. Maintain the diversity of solutions; according to the distance, retain the scheduling schemes with sparse distribution;

[0044] A54. Termination condition judgment; when the preset number of iterations G is reached, stop the iteration;

[0045] A55. Take the final non-dominated solution set as the Pareto optimal solution set.

[0046] Further, in step A52, the key optimization objectives include the following objectives: minimizing production cost, minimizing production time, or maximizing equipment utilization.

[0047] Further, A53 retaining the scheduling schemes with sparse distribution includes:

[0048] Calculate the distance between each scheduling scheme in the non-dominated solution set;

[0049] Calculate any two scheduling schemes in the solution set X i and X g The Euclidean distance d( X i , X g ) between them, and the formula is:

[0050] ;

[0051] where f k ( X i ) and f k ( X g ) are the fitness values of the scheduling schemes X i and X g on the k th optimization objective respectively, and n is the total number of optimization objectives;

[0052] Calculate the density X i of each scheduling scheme ρ ( X i ), and the formula is:

[0053] ;

[0054] where S is the current non-dominated solution set, ρ ( X i) Represents the scheduling plan X i of the density;

[0055] According to the density value ρ ( X i ) Sort all the scheduling plans in the non-dominated solution set S , and select the top h scheduling plans with the smallest density values as the reserved solution set, where h is the preset number of reserved solutions.

[0056] Further, the formula for A6 to update the position of the dragonfly is:

[0057] ;

[0058] Among them, represents the new position of the i-th dragonfly at time step t + 1; represents the position of the i-th dragonfly at the current time step t; α is the influence coefficient that controls the separation force D i on the overall position update; D i , A i , C i , F i , R i and E enemy respectively represent the separation force, alignment force, cohesion force, food attraction force, natural enemy avoidance force, and additional displacement amount for avoiding natural enemies, and Δt is the time step size.

[0059] An intelligent production scheduling optimization system based on a large model using the above optimization method, including:

[0060] A data collection module for collecting equipment operation data, material supply data, production progress data, and production task information;

[0061] A preprocessing module for cleaning, normalizing, and handling outliers of the collected data to generate structured data;

[0062] A model construction module that integrates a pre-trained large model and analyzes the non-linear relationships in production data through its attention mechanism to automatically generate the optimization objectives of the scheduling model and the constraint conditions of the production workshop;

[0063] A scheduling plan evaluation module that, based on the dynamic parameter adjustment ability of the large model, optimizes the weight assignment and neighborhood radius parameters in the multi-objective dragonfly optimization algorithm in real time, and evaluates the preliminary scheduling plan in combination with the entropy weight multi-objective decision-making method;

[0064] An output module, configured to select an optimal production scheduling plan from the Pareto optimal solution set and apply it to the production system.

[0065] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0066] (1) The present invention adopts an entropy weight multi-objective decision-making method combined with a multi-objective dragonfly optimization algorithm, which can effectively evaluate and optimize the production scheduling plan. Compared with traditional mathematical programming methods and heuristic algorithms, the method of the present invention significantly improves the scheduling efficiency while ensuring the quality of the solution.

[0067] (2) The present invention considers the diversity of solutions in the algorithm design. By retaining the scheduling plans with sparse distribution, the problem of over-concentration of the solution set is avoided. This helps to find more comprehensive and adaptable scheduling plans, and improves the flexibility and robustness of the scheduling system.

[0068] (3) The present invention dynamically determines the weights of each optimization objective through an entropy value calculation method, enabling the algorithm to reasonably balance the importance of different objectives according to the actual production situation and business requirements. This dynamic adjustment mechanism improves the practicality and adaptability of the scheduling plan. Description of the Drawings

[0069] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings;

[0070] Figure 1 It is a flowchart of an intelligent production scheduling optimization method based on a large model of the present invention;

[0071] Figure 2 It is a flowchart of a scheduling plan evaluation algorithm of an intelligent production scheduling optimization method based on a large model of the present invention;

[0072] Figure 3 It is a flowchart of finding the current non-dominated solution set by using the Pareto dominance relationship in an intelligent production scheduling optimization method based on a large model of the present invention;

[0073] Figure 4 It is a structural diagram of an intelligent production scheduling optimization system based on a large model of the present invention. Detailed Embodiments

[0074] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0075] This application discloses an intelligent production scheduling optimization method based on a large model, as Figure 1 shown, including the following steps:

[0076] S1. Data collection and preprocessing: Collect equipment operation data, material supply data, production progress data, and production task information; clean, normalize, and handle outliers for the collected data to generate structured data;

[0077] S2. Determine the optimization objectives of the scheduling plan and the constraint conditions of the production workshop, extract features and learn patterns from historical production data through a large model, construct a scheduling model, and generate a preliminary scheduling plan;

[0078] S3. Use a scheduling plan evaluation algorithm to evaluate the preliminary scheduling plan based on the dynamic parameter adjustment ability of the large model;

[0079] S4. Output the optimal production scheduling plan and apply this plan to the production system;

[0080] Among them, the scheduling plan evaluation algorithm in S3 uses the entropy weight multi-objective decision-making method combined with the multi-objective dragonfly optimization algorithm to evaluate the scheduling plan.

[0081] As a specific implementation manner, the entropy weight multi-objective decision-making method combined with the multi-objective dragonfly optimization algorithm in S3 includes the following steps, as Figure 2 shown:

[0082] A1. Extract multiple production samples from the preprocessed data, use the large model to generate simulated samples of future production scenarios, and standardize the production samples. Each sample represents the actual production situation within a time period;

[0083] A2. According to the extracted sample data, calculate the entropy value of each optimization index E j , and calculate the weight of each optimization index based on the entropy value ω j ;

[0084] A3. Initialize the dragonfly population, generate a candidate solution set of the initial scheduling plan through the large model, and assign an initial position and speed to each dragonfly. Each dragonfly is a scheduling plan;

[0085] A4. Calculate the fitness of each dragonfly according to the optimization objective weights ω j and the values of each optimization objective f j (X i ), and calculate the comprehensive fitness; the formula for calculating the comprehensive fitness is:

[0086] ;

[0087] where f j ( X l ) is the value of the i th scheduling scheme on the j th optimization objective; ω j is the weight of each optimization objective; n is the total number of optimization objectives;

[0088] A5. Use the Pareto dominance relationship to find the current non-dominated solution set as the current optimal solution;

[0089] A6. For each dragonfly, calculate the contributions of separation, alignment, cohesion, food attraction, and predator avoidance behaviors based on the information of its neighbors, and update the speed and position of the dragonfly by integrating these behaviors;

[0090] A7. Re-evaluate the fitness of all dragonflies and update the Pareto optimal solution set; if the fitness of the newly generated scheduling scheme is better than some scheduling schemes in the current non-dominated solution set S , then add it to the non-dominated solution set S , and remove the dominated scheduling schemes;

[0091] A8. When the predetermined number of iterations G is reached, the algorithm terminates; otherwise, repeat steps A4 to A7;

[0092] A9. Select the optimal production scheduling scheme from the Pareto optimal solution set.

[0093] As a specific implementation, the formula for A2 to calculate the entropy value of each optimization objective is:

[0094]

[0095] where p jl is the proportion of the l th sample under the

[0096] As a specific implementation, the formula for A2 to calculate the weight of each optimization metric is:

[0097] ;

[0098] where ω j is the weight of the j-th optimization objective, n is the total number of optimization objectives, E j is the entropy value of each optimization metric.

[0099] The large model is a pre-trained Transformer architecture model, which interacts with the scheduling optimization process in the following ways: In the data preprocessing stage, the embedding layer of the large model is used to perform semantic parsing on unstructured production logs; in the model construction stage, the attention mechanism of the large model is used to identify the relevance between key production constraints and objectives; in the solution evaluation stage, the reinforcement learning module of the large model is called to adaptively adjust the convergence speed of the dragonfly optimization algorithm.

[0100] Example:

[0101] Suppose we have collected the following sample data, where each sample represents the actual production situation within a week, as shown in Table 1:

[0102] Table 1 shows the initial sample data collected.

[0103]

[0104] Using the min-max normalization method, each metric is transformed into the range of 0 to 1:

[0105] Normalized value = (Original value - Minimum value) / (Maximum value - Minimum value)

[0106] For production cost (the lower the cost, the better), production time, and equipment utilization rate (the higher the better), we perform the following normalization:

[0107] Production cost (Minimum value = 90, Maximum value = 130):

[0108] Sample 1: (100 - 90) / (130 - 90) = 0.25

[0109] Sample 2: (90 - 90) / (130 - 90) = 0

[0110] Sample 3: (120 - 90) / (130 - 90) = 0.75

[0111] Sample 4: (110 - 90) / (130 - 90) = 0.5

[0112] Sample 5: (130 - 90) / (130 - 90) = 1

[0113] Production time (minimum = 8 hours, maximum = 12 hours):

[0114] Sample 1: 1 - (10 - 8) / (12 - 8) = 0.5

[0115] Sample 2: 1 - (12 - 8) / (12 - 8) = 0

[0116] Sample 3: 1 - (8 - 8) / (12 - 8) = 1

[0117] Sample 4: 1 - (9 - 8) / (12 - 8) = 0.75

[0118] Sample 5: 1 - (11 - 8) / (12 - 8) = 0.25

[0119] Equipment utilization rate (minimum = 70%, maximum = 90%):

[0120] Sample 1: 1 - (80 - 70) / (90 - 70) = 0.5

[0121] Sample 2: 1 - (75 - 70) / (90 - 70) = 0.75

[0122] Sample 3: 1 - (90 - 70) / (90 - 70) = 0

[0123] Sample 4: 1 - (85 - 70) / (90 - 70) = 0.25

[0124] Sample 5: 1 - (70 - 70) / (90 - 70) = 1

[0125] Step 1: Calculate the proportion p jl

[0126] For each optimization objective, calculate its proportion p among all samples jl :

[0127] Production cost: p11 = 0.1818, p21 = 0.1636, p31 = 0.2182, p41 = 0.2, p51 = 0.2364;

[0128] Production time: p12 = 0.2, p22 = 0.24, p32 = 0.16, p42 = 0.18, p52 = 0.22;

[0129] Equipment utilization rate: p13 = 0.2, p23 = 0.1875, p33 = 0.225, p43 = 0.2125, p53 = 0.175;

[0130] Step 2: Calculate the entropy value E j

[0131] Use the formula to calculate the entropy value of each optimization objective:

[0132] =0.9948

[0133] =0.9938

[0134] =0.9976

[0135] Step 3: Calculate the weights ω j

[0136] Use the formula to calculate the weights of each optimization objective:

[0137] =0.3768

[0138] =0.4493

[0139] =0.1739

[0140] Step 4: Use the formula to calculate the comprehensive fitness:

[0141] =0.4058

[0142] =0.1304

[0143] =0.7319

[0144] =0.5689

[0145] =0.6630

[0146] Furthermore, the initial dragonfly population of A3 includes:

[0147] Set the size N of the dragonfly population, that is, the number of initial scheduling plans; the size N of the dragonfly population is dynamically adjusted by the large model according to the complexity of the production task and the learning results of historical scheduling data;

[0148] Randomly generate an initial position for each dragonfly X i , X i ∈lb , ub wherein, lb and ub are the lower bound and upper bound of the decision variable respectively;

[0149] Generate an initial velocity for each dragonfly V i , V i ∈ [- V max , V max ;

[0150] wherein, V max is the maximum velocity, which is set according to the scale and complexity of the production scheduling problem;

[0151] Set the maximum number of iterations G of the algorithm;

[0152] Set the neighborhood radius r to determine the neighbors of the dragonflies.

[0153] Embodiment:

[0154] This embodiment describes how to use the dragonfly algorithm in the intelligent production scheduling optimization method based on the large model to initialize the dragonfly population and generate an initial solution set for the production scheduling problem.

[0155] Determine the decision variables in the production scheduling problem, such as machine allocation, process sequence, etc.

[0156] Determine the upper and lower bounds of each decision variable, namely: lb and ub.

[0157] According to the scale and complexity of the problem, set the size of the dragonfly population: N, which represents the number of initial scheduling schemes.

[0158] For each dragonfly i (i = 1, 2,..., N), randomly generate an initial position: X i . The position: X i : is a vector, each component of which represents the value of a decision variable, and the value of each component is within the corresponding upper and lower bounds: [lb, ub].

[0159] For example, if the decision variable is the working time of the machine, its upper and lower bounds may be [0, 24] (hours).

[0160] For each dragonfly i, randomly generate an initial velocity V i . Each component of the velocity vector represents the change amount of the decision variable value, and the value of each component is within: [- Vmax , V max within. V max is the maximum speed set according to the scale and complexity of the production scheduling problem.

[0161] Set the maximum number of iterations of the algorithm: G, which determines the total number of times the algorithm runs.

[0162] Set the neighborhood radius: r, which is used to determine the neighbors of the dragonflies. The neighborhood radius determines the range of other dragonflies that a dragonfly considers when updating its position.

[0163] Apply the initialized dragonfly population to the production scheduling optimization problem. Each dragonfly represents a possible scheduling solution. Use the dragonfly algorithm to iteratively optimize these initial solutions to find the optimal or approximately optimal production scheduling solution.

[0164] As a specific implementation manner, A5 uses the Pareto dominance relationship to find the current non-dominated solution set, as Figure 3 shown, the current optimal solutions include:

[0165] A51. Initialize the non-dominated solution set; use all the scheduling solutions in the population as the initial candidate solution set;

[0166] A52. Screen the non-dominated solutions;

[0167] Traverse each scheduling solution in the initial candidate solution set i , and compare it pairwise with other scheduling solutions in the candidate solution set g ;

[0168] If the fitness of scheduling solution i is greater than that of scheduling solution g , and scheduling solution i is significantly better than scheduling solution g in at least one key optimization objective, then scheduling solution i dominates scheduling solution g ;

[0169] If scheduling solution i dominates scheduling solution g , then remove scheduling solution g from the candidate solution set;

[0170] If scheduling solution i is not dominated by any other scheduling solution, then add scheduling solution i to the non-dominated solution set S ;

[0171] A53. Maintain the diversity of solutions; according to the said distance, retain the scheduling schemes with sparse distribution;

[0172] A54. Termination condition judgment; when the preset number of iterations G is reached, stop the iteration;

[0173] A55. Take the final non-dominated solution set as the Pareto optimal solution set.

[0174] As a specific implementation manner, in step A52, the key optimization objectives include but are not limited to the following objectives: minimizing production costs, minimizing production time, or maximizing equipment utilization.

[0175] As a specific implementation manner, A53 retaining the scheduling schemes with sparse distribution includes:

[0176] Calculate the distances between the scheduling schemes in the non-dominated solution set;

[0177] Calculate any two scheduling schemes in the solution set X i and X g the Euclidean distance d( X i , X g ), and the formula is:

[0178] ;

[0179] where f k ( X i ) and f k ( X g ) are the fitness values of the scheduling schemes X i and X g on the k-th optimization objective respectively, and n is the total number of optimization objectives;

[0180] Calculate the density X i of each scheduling scheme ρ ( X i ), and the formula is:

[0181] ;

[0182] where S is the current non-dominated solution set, ρ ( X i ) represents the scheduling schemeX i The density;

[0183] According to the density value ρ ( X i ) sort all the scheduling schemes in the non-dominated solution set S and select the top h scheduling schemes with the smallest density values as the retained solution set, where h is the number of retained solutions preset.

[0184] Example:

[0185] 1. Initialize the non-dominated solution set:

[0186] Suppose we have the following 5 scheduling schemes (each scheme has three key optimization indicators: equipment utilization rate, production cost, production time), as shown in Table 2:

[0187] Table 2 is the initial scheduling scheme collected.

[0188]

[0189] 2. Screen the non-dominated solution set:

[0190] First, we need to determine which schemes are non-dominated. A scheme is non-dominated if it is not inferior to any other scheme in all optimization indicators and is superior to another scheme in at least one indicator.

[0191] By comparing the above schemes, it can be concluded that the non-dominated solution set includes Scheme 2, Scheme 3, and Scheme 4.

[0192] 3. Calculate the distance between every two scheduling schemes

[0193] d(2, 3) ≈ 1.458; d(2, 4) ≈ 1.031; d(3, 4) ≈ 0.433.

[0194] Calculate the density of each scheduling scheme ρ ( X i ).

[0195] ≈ 1.6558;

[0196] ≈ 2.9953;

[0197] ≈ 3.2794;

[0198] Sort the scheduling schemes according to the density values, and select the top h scheduling schemes with the smallest density values as the retained solution set. Set h to 2.

[0199] Result:

[0200] Suppose that after calculation and sorting, the top 2 scheduling schemes with the smallest density values are selected as the retained solution set. Retain Scheme 3 and Scheme 4.

[0201] Through this embodiment, we demonstrated how to use the A45 algorithm to screen out the non-dominated solution set from a group of scheduling schemes with three key optimization metrics (production cost, production time, equipment utilization rate), and retain the scheduling schemes with a relatively sparse distribution according to the distance between these solutions. This method helps to find the best solution that balances different requirements in multi-objective optimization problems. Note that in practical applications, it may be necessary to standardize the data, and the number of retained schemes can be adjusted according to specific circumstances.

[0202] As a specific implementation, the formula for updating the position of the dragonfly by A6 is:

[0203] ;

[0204] Among them, represents the new position of the i-th dragonfly at time step t + 1; represents the position of the i-th dragonfly at the current time step t; α is the influence coefficient that controls the separation force D i on the overall position update; D i , A i , C i , F i and R i represent the separation force, alignment force, cohesion force, food attraction force, and natural enemy avoidance force respectively, and Δt is the time step. Calculate the new position of each dragonfly according to the above formula. Update the positions of all dragonflies to complete one iteration.

[0205] Separation force D i : Calculate the distance between each dragonfly and the surrounding individuals according to the information of the neighbors to ensure that they are not too close.

[0206] Alignment force A i : Adjust the speed to match the speed of the neighbors to maintain group consistency.

[0207] Cohesion force C i : Move towards the center of the group to enhance group cohesion.

[0208] Food attraction force F i : Move towards the direction of the food source to simulate foraging behavior.

[0209] Natural enemy avoidance force R i: Stay away from potential threats and simulate behaviors to avoid natural enemies.

[0210] By implementing an intelligent production scheduling optimization method based on a large model, an enterprise can find a scheduling plan that performs best in minimizing production costs, minimizing production time, and maximizing equipment utilization. This method ensures the diversity of solutions by retaining sparsely distributed scheduling plans, thereby improving the flexibility and adaptability of the scheduling plan. By applying the optimal scheduling plan, the enterprise has improved production efficiency, reduced costs, and increased the utilization rate of equipment.

[0211] An intelligent production scheduling optimization system based on a large model using the above optimization method, as Figure 4 shown, includes:

[0212] A data collection module for collecting equipment operation data, material supply data, production progress data, and production task information;

[0213] A preprocessing module for cleaning, normalizing, and handling outliers of the collected data to generate structured data;

[0214] A model construction module that integrates a pre-trained large model and analyzes the non-linear relationships in production data through its attention mechanism to automatically generate the optimization objectives of the scheduling model and the constraint conditions of the production workshop;

[0215] A scheduling plan evaluation module, based on the dynamic parameter adjustment ability of the large model, real-time optimizes the weight allocation and neighborhood radius parameters in the multi-objective dragonfly optimization algorithm, and evaluates the preliminary scheduling plan in combination with the entropy weight multi-objective decision-making method;

[0216] An output module for selecting the optimal production scheduling plan from the Pareto optimal solution set and applying it to the production system.

[0217] Obviously, the above embodiments are merely examples given for clear illustration and are not limitations on the implementation manners. For those of ordinary skill in the art, other different forms of changes or variations can be made based on the above description. It is not necessary and impossible to enumerate all the implementation manners here. And the obvious changes or variations derived therefrom are still within the protection scope of this invention.

Claims

1. An intelligent production scheduling optimization method based on a large model, characterized in that, It includes the following steps: S1. Data collection and preprocessing; Collect equipment operation data, material supply data, production progress data, and production task information; Clean, normalize, and handle outliers for the collected data to generate structured data; S2. Determine the optimization objectives of the scheduling plan and the constraint conditions of the production workshop. Extract features and learn patterns from historical production data through a large model, construct a scheduling model, and generate a preliminary scheduling plan; S3. Use the scheduling plan evaluation algorithm to evaluate the preliminary scheduling plan based on the dynamic parameter adjustment ability of the large model; S4. Output the optimal production scheduling plan and apply this plan to the production system; Among them, the scheduling plan evaluation algorithm in S3 uses the entropy weight multi-objective decision-making method combined with the multi-objective dragonfly optimization algorithm to evaluate the scheduling plan; the entropy weight multi-objective decision-making method combined with the multi-objective dragonfly optimization algorithm in S3 includes the following steps: A1. Extract multiple production samples from the preprocessed data and standardize the production samples. Each sample represents the actual production situation within a time period; A2. Calculate the entropy value of each optimization metric based on the extracted sample data E j , and calculate the weight of each optimization metric based on the entropy value ω j ; A3. Initialize the dragonfly population, generate a candidate solution set for the initial scheduling plan through the large model, and assign an initial position and speed to each dragonfly; A4. Calculate the fitness of each dragonfly according to the optimization objective weights ω j and the values of each optimization objective f j (X i ), calculate the comprehensive fitness; the formula for calculating the comprehensive fitness is: ； Among them, f j ( X i ) is the value of the i th scheduling scheme on the j th optimization objective; ω j is the weight of each optimization objective; n is the total number of optimization objectives; A5. Use the Pareto dominance relationship to find the current non-dominated solution set as the current optimal solution; A6. For each dragonfly, calculate the contributions of separation, alignment, cohesion, food attraction, and predator avoidance behaviors based on the information of its neighbors, and update the speed and position of the dragonfly by integrating these behaviors; A7. Re-evaluate the fitness of all dragonflies and update the Pareto optimal solution set; if the fitness of the newly generated scheduling scheme is better than that of some scheduling schemes in the current non-dominated solution set S then add it to the non-dominated solution set S and remove the dominated scheduling schemes; A8. When the predetermined number of iterations G is reached, the algorithm terminates; otherwise, repeat steps A4 to A7; A9. Select the optimal production scheduling plan from the Pareto optimal solution set; Among them, using the Pareto dominance relationship to find the current non-dominated solution set as the current optimal solution in A5 includes: A51. Initialize the non-dominated solution set; use all scheduling plans in the population as the initial candidate solution set; A52. Screen non-dominated solutions; Traverse each scheduling plan i in the initial candidate solution set and compare it pairwise with other scheduling plans g in the candidate solution set; If the fitness of scheduling plan i is greater than that of scheduling plan g, and scheduling plan i is superior to scheduling plan g in at least one key optimization objective, then scheduling plan i dominates scheduling plan g; If scheduling plan i dominates scheduling plan g, then remove scheduling plan g from the candidate solution set; If scheduling plan i is not dominated by any other scheduling plan, then add scheduling plan i to the non-dominated solution set S; A53. Maintain the diversity of solutions; retain the scheduling plans with sparse distribution according to the distance; A54. Judge the termination condition; stop the iteration when the preset number of iterations G is reached; A55. Use the final non-dominated solution set as the Pareto optimal solution set.

2. The intelligent production scheduling optimization method based on a large model according to claim 1, characterized in that The formula for calculating the entropy value of each optimization objective in A2 is: ； Among them, p jl is the proportion of the l th sample under the jth optimization objective, and m is the number of samples.

3. The intelligent production scheduling optimization method based on a large model according to claim 2, wherein The formula for calculating the weight of each optimization index in A2 is: ； wherein, ω j is the weight of the j-th optimization objective, and n is the total number of optimization objectives, E j is the entropy value of each optimization metric.

4. An intelligent production scheduling optimization method based on a large model according to claim 1, characterized in that Initializing the dragonfly population in A3 includes: Set the size N of the dragonfly population, that is, the number of initial scheduling plans; the size N of the dragonfly population is dynamically adjusted by the large model according to the complexity of the production task and the learning results of historical scheduling data; Randomly generate an initial position for each dragonfly X i , X i ∈ lb , ub where lb and ub are the lower and upper bounds of the decision variables respectively; Randomly generate an initial velocity for each dragonfly V i , V i ∈[− V max , V max ; Among them, V max is the maximum speed, which is set according to the scale and complexity of the production scheduling problem; Set the maximum number of iterations G of the algorithm; Set the neighborhood radius r to determine the neighbors of the dragonflies.

5. The intelligent production scheduling optimization method based on a large model according to claim 1, characterized in that, In A52, the key optimization objectives include the following: minimizing production costs, minimizing production time, or maximizing equipment utilization.

6. The intelligent production scheduling optimization method based on a large model according to claim 1, wherein, The scheduling schemes with sparse retention distribution in A53 include: Calculate the distances between the scheduling schemes in the non-dominated solution set; Calculate the Euclidean distance d( X i and X g ) between any two scheduling solutions in the solution set. The formula is: X i , X g ) ； Among them, f k ( X i ) and f k ( X g ) are the fitness values of the scheduling schemes X i and X g on the k-th optimization objective, respectively, and n is the total number of optimization objectives; Calculate each scheduling plan X i of the density ρ ( X i ), the formula is: ； Among them, S is the current non-dominated solution set, ρ ( X i ) represents the scheduling scheme X i 's density; According to the density value ρ ( X i ) sort all the scheduling schemes in the non-dominated solution set S and select the top h scheduling schemes with the smallest density value as the reserved solution set, where h is the preset number of reserved solutions.

7. An intelligent production scheduling optimization method based on a large model according to claim 1, characterized in that The formula for updating the position of the dragonfly in A6 is: ； Among them, represents the new position of the i-th dragonfly at time step t + 1; represents the position of the i-th dragonfly at the current time step t; α is the influence coefficient of the separation force D i on the overall position update; D i , A i , C i , F i , R i and E enemy represent the separation force, alignment force, cohesion force, food attraction force, natural enemy avoidance force, and additional displacement for avoiding natural enemies respectively, and Δt is the time step.

8. An intelligent production scheduling optimization system based on a large model for an intelligent production scheduling optimization method according to any one of claims 1-7, characterized in that, Include: A data collection module for collecting equipment operation data, material supply data, production progress data, and production task information; A preprocessing module for cleaning, normalizing, and handling outliers of the collected data to generate structured data; A model construction module that integrates a pre-trained large model and analyzes the non-linear relationships in production data through its attention mechanism to generate the optimization objectives of the scheduling model and the constraints of the production workshop; A scheduling scheme evaluation module that, based on the dynamic parameter adjustment ability of the large model, optimizes the weight allocation and neighborhood radius parameters in the multi-objective dragonfly optimization algorithm in real time, and evaluates the preliminary scheduling scheme in combination with the entropy weight multi-objective decision-making method; An output module for selecting the optimal production scheduling scheme from the Pareto optimal solution set and applying it to the production system.

Images