Search

CN-121998366-A - Multi-target energy-saving distributed flow shop group scheduling method and system

CN121998366ACN 121998366 ACN121998366 ACN 121998366ACN-121998366-A

Abstract

The invention discloses a multi-target energy-saving distributed flow shop scheduling method and a system, which relate to the technical field of intelligent production scheduling and optimization, wherein the method comprises the steps of obtaining operation parameters of a distributed flow shop scheduling scene, and constructing a multi-target mixed integer linear programming model taking the minimum maximum finishing time and the total energy consumption as optimization targets; the method comprises the steps of generating an initial population containing a plurality of feasible solutions by adopting a plurality of initialization strategies, dividing the initial population into two sub-populations by adopting a sorting mechanism based on pareto fronts, constructing a double Q learning mechanism, establishing two independent Q tables to respectively optimize two targets, generating candidate solutions in each sub-population by adopting a self-adaptive iterative greedy algorithm, a Q learning algorithm and a mixed algorithm of the two algorithms, screening the optimal solutions by adopting non-dominant sorting and crowding distance sorting to update the sub-populations, iterating and updating until termination conditions are met, and combining the solutions of the two sub-populations to form a pareto fronts solution set to obtain an optimal scheduling scheme with better effect.

Inventors

  • LIU BEIBEI
  • DU YU

Assignees

  • 济南大学

Dates

Publication Date
20260508
Application Date
20260129

Claims (10)

  1. 1. A multi-target energy-saving distributed flow shop group scheduling method is characterized by comprising the following steps: acquiring operation parameters of a distributed flow shop group scheduling scene, and constructing a DFGSP multi-objective mixed integer linear programming model taking the minimum maximum finishing time and the total energy consumption as optimization targets; Generating an initial population containing a plurality of feasible solutions by adopting a plurality of initialization strategies, and dividing the initial population into two sub-populations respectively optimizing the maximum finishing time and the total energy consumption by adopting a sorting mechanism based on the pareto front; Constructing a dual Q learning mechanism, establishing two independent Q tables to respectively optimize two targets, generating candidate solutions in each sub-population by adopting a self-adaptive iterative greedy algorithm, a Q learning algorithm and a mixed algorithm of the two, and then adopting non-dominant sorting and crowding distance sorting to screen optimal solutions to update the sub-population, and continuously carrying out iterative updating until the termination condition is met; and combining the solutions of the two sub-populations to form a pareto front solution set, thereby obtaining an optimal scheduling scheme.
  2. 2. The multi-target energy efficient distributed flow shop scheduling method according to claim 1, wherein the operating parameters include a number of plants, a number of machines configured in each plant, a number of work pieces in each work piece; in the multi-objective mixed integer linear programming model, the optimization objective comprises maximum finishing time and total energy consumption, the total energy consumption comprises processing energy consumption, preparation energy consumption and idle energy consumption, the decision variables comprise a group sorting sequence allocated by each factory, a workpiece sorting sequence in each group and a processing speed of each workpiece on a machine, and the constraint conditions comprise factory allocation constraint, intra-group sorting constraint, machine processing sequence constraint, speed selection constraint, time connection constraint and energy consumption calculation constraint.
  3. 3. The multi-objective energy-saving distributed flow shop scheduling method according to claim 1, wherein the population initialization and sub-population division processes are as follows: generating feasible solutions by adopting various initialization strategies, respectively focusing on the initial population with minimized maximum finishing time, focusing on the initial population with minimized total energy consumption and random initialization, and combining various initial populations; And (3) evaluating the initial population by adopting a sorting mechanism based on the pareto front through non-dominant sorting and crowding distance sorting, extracting an elite solution set of the first pareto front, dividing the elite solution set into two types of sub-populations according to the preference of each elite solution on two optimization targets, and distributing the rest non-elite solutions by adopting a direct target deflection method so as to balance the scales of the two sub-populations.
  4. 4. The multi-target energy-efficient distributed flow shop group scheduling method of claim 1, wherein generating the candidate solution using an adaptive iterative greedy algorithm comprises: Calculating the average group number of each factory based on the operation parameters in the distributed flow shop group scheduling scene, determining the scale of the factory, and dynamically adjusting the damage intensity coefficient d according to the scale of the factory; Randomly removing d workpiece groups in a factory corresponding to the current solution, temporarily storing the d workpiece groups to a temporary set, inserting the temporarily stored d workpiece groups into all feasible positions one by one, selecting an insertion scheme for optimizing two optimization targets, and allocating and sequencing a reconstruction group; The method comprises the steps of taking a factory with the longest finishing time in all factories as a key factory, identifying a work group with the largest number of workpieces in the key factory, randomly removing half of the total number of workpieces in the current group, temporarily storing the workpieces, inserting temporary storage workpieces into the optimal positions in the group after the sequence of the rest workpieces in the group is disturbed, and reconstructing the distribution and the sequencing of the workpieces in the group; And reconstructing the new solution, verifying whether the new solution meets all scheduling constraints, and outputting a candidate solution.
  5. 5. The multi-objective energy-efficient distributed flow shop scheduling method of claim 4, wherein generating the candidate solution using a Q-learning algorithm comprises: Taking any random solution in the population as the current solution and carrying out random disturbance action, calculating the time-varying quantity and the energy consumption varying quantity of completion of the solutions before and after disturbance, matching the time-varying quantity and the energy consumption varying quantity with 9 system states of two-dimensional representation, and determining the current state; Inquiring the optimal action corresponding to the current state based on the pre-training Q table corresponding to the optimization target, and disturbing the current solution by the optimal action to generate a new solution; Checking whether the generated new solution meets the scheduling constraint, so that the solution meeting the constraint is a candidate solution and is output; The pre-training process of the Q table is as follows: Taking any random solution in the population as the current solution and carrying out random disturbance action, calculating the time-varying quantity and the energy consumption varying quantity of completion of the solutions before and after disturbance, matching the time-varying quantity and the energy consumption varying quantity with 9 system states of two-dimensional representation, and determining the current state; Recording the current state-action pair, calculating the rewarding value of the action according to a rewarding formula to quantify the target optimization degree, correcting and updating the Q value of the current state-action pair in the Q table by using the rewarding value, and calculating the estimation error of the Q value; and (5) continuously iterating training until the Q value estimation error tends to be stable and approaches to 0, finishing the Q-Learning training, and storing a final Q table to obtain a pre-training Q table.
  6. 6. The multi-target energy-efficient distributed flow shop group scheduling method of claim 5, wherein generating the candidate solution using a hybrid algorithm of an adaptive iterative greedy algorithm and a Q-learning algorithm comprises: Taking any random solution in the population as a current solution, adopting a self-adaptive iterative greedy algorithm to optimize the current solution, and generating an intermediate solution through the steps of judging the scale of a factory, determining the damage intensity coefficient, destroying and reconstructing at the group level and optimizing at the workpiece level; Optimizing the intermediate solution by adopting a Q learning algorithm, calculating the time change and the energy consumption change of completion between the intermediate solution and the current solution, determining the state of the intermediate solution, inquiring the optimal action based on a Q table of a corresponding target, and finely adjusting the intermediate solution to generate a new solution; and checking whether the new solution meets the scheduling constraint, and outputting a candidate solution fusing global search and local optimization.
  7. 7. A multi-objective energy-efficient distributed flow shop group scheduling system, comprising: The data acquisition and model construction module is used for acquiring the operation parameters of the distributed flow shop group scheduling scene and constructing a DFGSP multi-target mixed integer linear programming model taking the minimum completion time and the total energy consumption as optimization targets; The population initialization module is used for generating an initial population containing a plurality of feasible solutions by adopting a plurality of initialization strategies, and dividing the initial population into two sub-populations respectively optimizing the maximum finishing time and the total energy consumption by adopting a sorting mechanism based on the pareto front; The iterative optimization module is used for constructing a dual Q learning mechanism, establishing two independent Q tables to respectively optimize two targets, generating candidate solutions in each sub-population by adopting a self-adaptive iterative greedy algorithm, a Q learning algorithm and a mixed algorithm of the two, and then screening the optimal solutions by adopting non-dominant sorting and crowding distance sorting to update the sub-population, and continuously carrying out iterative updating until the termination condition is met; And the scheduling scheme generating module is used for combining the solutions of the two sub-populations to form a pareto front solution set so as to obtain an optimal scheduling scheme.
  8. 8. An electronic device, comprising: The multi-objective energy-saving distributed flow shop scheduling method according to any one of claims 1 to 6 when the executable instructions stored in the memory are executed.
  9. 9. A computer readable storage medium having stored thereon executable instructions for causing a processor to execute the executable instructions, implementing the multi-objective energy-efficient distributed flow shop floor scheduling method of any one of claims 1-6.
  10. 10. A computer program product, the computer program product comprising executable instructions stored on a computer readable storage medium; the multi-objective energy-efficient distributed flow shop group scheduling method of any one of claims 1-6 is implemented when a processor of an electronic device reads the executable instructions from the computer-readable storage medium and executes the executable instructions.

Description

Multi-target energy-saving distributed flow shop group scheduling method and system Technical Field The invention relates to the technical field of intelligent production scheduling and optimization, in particular to a multi-target energy-saving distributed flow shop group scheduling method and system. Background The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art. As manufacturing has transitioned from traditional centralized production to distributed production, and to intensive, distributed flow shop scheduling problems (DFSP) have become critical to improving production efficiency. In practical production, part of the workpieces share similar processing routes or equipment requirements, in order to fully utilize the similarity of workpiece processing, a grouping technology is introduced to divide the similar workpieces into groups, all the workpieces in the groups are continuously and uninterruptedly processed in a process factory, and the DFSP introduced into the grouping technology forms a distributed flow shop group scheduling problem (Distributed Flowshop Group Scheduling Problem, DFGSP), which is widely existing in the industries of label printing, metal stamping, printed circuit board manufacturing and the like. DFGSP not only needs to consider decision factors such as factory allocation, intra-group sequencing, inter-group sequencing and the like, but also needs to deal with complexity caused by sequence related preparation time (SDST) and machine speed selection caused by group conversion, namely, working procedures such as cleaning or tool replacement and the like need to be carried out when switching among groups, longer SDST delays finishing, waiting time is increased, construction period is prolonged, and higher machine speed can shorten processing time, but energy consumption is increased. Therefore, under low carbon manufacturing requirements, the total energy consumption and maximum finishing time become equally important optimization objectives in DFGSP. In the prior art, the DFGSP-oriented optimization method mainly comprises an iterative greedy algorithm (IG), a reinforcement learning algorithm (RL) and the like, however, the methods have certain defects that most of the existing researches only focus on single-objective optimization, and the existing researches have insufficient consideration on multi-objective collaborative optimization such as energy consumption, maximum finishing time and the like. The existing scheme for multi-objective collaborative optimization mostly adopts a single strategy method to solve the scheduling strategies, but the single strategy such as the fixed damage strength of an IG algorithm is easy to cause premature convergence or insufficient searching diversity, the reinforcement learning algorithm adopts a single Q table to optimize a plurality of targets, so that learning is unstable, and finally the generated scheduling scheme is poor in effect. Disclosure of Invention Aiming at the multi-objective and multi-constraint characteristics of the distributed flow shop scheduling problem (DFGSP), the invention provides a multi-objective energy-saving distributed flow shop scheduling method and system, which combine an improved iterative greedy strategy with double Q learning, optimize the maximum finishing time and the total energy consumption, realize the balance of search diversity and convergence stability, and effectively improve the effect of a generated scheduling scheme. In a first aspect, the present invention provides a multi-objective energy-saving distributed flow shop group scheduling method. A multi-target energy-saving distributed flow shop group scheduling method comprises the following steps: acquiring operation parameters of a distributed flow shop group scheduling scene, and constructing a DFGSP multi-objective mixed integer linear programming model taking the minimum maximum finishing time and the total energy consumption as optimization targets; Generating an initial population containing a plurality of feasible solutions by adopting a plurality of initialization strategies, and dividing the initial population into two sub-populations respectively optimizing the maximum finishing time and the total energy consumption by adopting a sorting mechanism based on the pareto front; Constructing a dual Q learning mechanism, establishing two independent Q tables to respectively optimize two targets, generating candidate solutions in each sub-population by adopting a self-adaptive iterative greedy algorithm, a Q learning algorithm and a mixed algorithm of the two, and then adopting non-dominant sorting and crowding distance sorting to screen optimal solutions to update the sub-population, and continuously carrying out iterative updating until the termination condition is met; and combining the solutions of the two sub-populations to form a pareto front so