Green dynamic scheduling method and system for distributed heterogeneous hybrid flow shop
Through rolling window technology and improved NSGA-II algorithm, combined with machine start-stop energy-saving strategy and workpiece status judgment, the dynamic disturbance problem in dynamic scheduling in distributed heterogeneous hybrid flow workshop is solved, production scheduling is optimized, and energy consumption is realized Dual optimization of time.
Patent Information
- Application Number
- CN202410949374.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-16
- Publication Date
- 2024-11-08
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing technology is difficult to effectively solve the problem of dynamic scheduling in distributed heterogeneous hybrid flow workshops, especially when facing uncertain factors such as machine failure and emergency orders. The existing research focuses on static scheduling, with few dynamic scheduling methods and no Fully consider energy consumption optimization.
The rolling window technology combined with the improved NSGA-II algorithm is adopted to realize the rescheduling of the scheduling workpieces through the machine start-stop energy-saving strategy and the workpiece status judgment rules, and optimize the minimum completion time, minimum energy consumption and minimum scheduling deviation.
It realizes production scheduling optimization in the face of dynamic disturbance factors, reduces energy consumption and completion time, and improves the robustness and efficiency of the production system.
Smart Images

Figure CN118915650A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of workshop scheduling, and in particular to a distributed heterogeneous hybrid flow shop green dynamic scheduling method and system based on rolling window technology and improved NSGA-II algorithm. Background Art
[0002] Distributed manufacturing, as a new production model, has become one of the important development directions of the manufacturing industry. Compared with traditional centralized production, distributed manufacturing can make full use of the resources of multiple factories or enterprises, and reduce production costs and inventory through the rational allocation of resources. The green transformation and sustainable development of the manufacturing industry has become a hot research topic. Distributed hybrid flow shop scheduling is to plan and coordinate the production resources of workshops distributed in different geographical locations to optimize the scheduling indicators of the entire production system. It is one of the key technologies for realizing intelligent green and low-carbon production. At the same time, attention is paid to the interference of uncertain factors such as machine failures and the arrival of emergency orders in the actual production process. The dynamic scheduling method of distributed heterogeneous hybrid flow shop is studied, which is of great significance to improving enterprise production efficiency, reducing energy consumption, and improving the stability of the production system.
[0003] The distributed hybrid flow shop scheduling problem is a typical NP-Hard problem, and the expansion problem is even more complicated. At present, most of the research in this field is based on homogeneous workshops, without considering the differences in the processing environment between distributed workshops, and most of the research focuses on the research of static scheduling methods, and there is less research on dynamic scheduling methods. In the production process, there are frequent situations where production plans cannot be carried out smoothly due to emergencies such as machine failures and the arrival of emergency orders. Therefore, the research on dynamic scheduling methods for distributed heterogeneous hybrid flow shops will be of great significance to enterprises in improving the robustness of production systems. In recent years, the concept of green and low-carbon production has been integrated into production scheduling, and the research on green production scheduling issues has become a research hotspot in the field of production scheduling. Most of the research on green scheduling of distributed hybrid flow shops only considers the energy consumption level in the processing process, and there is less research on no-load energy consumption and other energy consumption generated by the production process. Summary of the invention
[0004] In view of the shortcomings of the prior art, the present invention combines the rolling window technology to provide a green dynamic scheduling method for a distributed heterogeneous hybrid assembly line, and designs an improved NSGA-II algorithm to solve the problem based on the problem characteristics. The purpose is to update the information of the workpieces to be scheduled in the rolling window by identifying the rescheduling points, and use the improved NSGA-II algorithm to reschedule the workpieces in the scheduling window to achieve dynamic scheduling with minimum completion time, minimum energy consumption and minimum scheduling deviation.
[0005] Terminology explanation:
[0006] Traditional GLR initialization methods include global initialization, local initialization and random initialization. Global initialization and local initialization are mainly aimed at machine allocation and workshop allocation problems to balance the machine load of each workshop. Random initialization mainly considers distributing the initial solution as much as possible in the solution space. The combination of the three improves the quality of the initial solution. The EGLR1 initialization method considers the impact of energy consumption on the initial solution. It is extended on the basis of GLR and divides global initialization into time-based global initialization and energy-based global initialization; local initialization is divided into time-based local initialization and energy-based local initialization.
[0007] In order to achieve the above object, the technical solution adopted by the present invention is as follows:
[0008] A green dynamic scheduling method for a distributed heterogeneous hybrid flow shop, comprising:
[0009] S1: Propose energy-saving strategies for starting and stopping machines according to the machine processing status;
[0010] S2: According to the requirements of green dynamic scheduling of distributed heterogeneous hybrid flow workshops, a green dynamic scheduling planning model for distributed heterogeneous hybrid flow workshops is established;
[0011] S3: Construct artifact status judgment rules;
[0012] S4: The distributed heterogeneous hybrid flow shop green dynamic programming model is solved through the NSGA-II algorithm to reschedule the workpieces to be scheduled in the rolling window.
[0013] Preferably, according to the present invention, in step S1, according to the machine processing state, the energy consumption is divided into processing energy consumption, idle energy consumption and start-stop energy consumption, and a machine start-stop energy-saving strategy is proposed;
[0014] The machine start-stop energy-saving strategy means that when the idle waiting time between the processing of the two workpieces required by the machine is greater than the time it takes for the machine to perform a power on and off operation, and when the idle energy consumption between the processing of the two workpieces required by the machine is higher than the energy consumption of the machine performing a power on and off operation, the machine will be shut down.
[0015] Preferably, according to the present invention, in step S2, the green dynamic scheduling planning model of the distributed heterogeneous hybrid flow shop includes an optimization objective and a series of constraints;
[0016] The optimization objectives include: minimizing the maximum completion time, the minimum total energy consumption, and the minimum scheduling deviation; the maximum completion time refers to the latest completion time of the last process of all workpieces; the total energy consumption refers to the energy consumed after all workpieces are processed, which is divided into processing energy consumption, idle energy consumption, and start-stop energy consumption; the scheduling deviation refers to the difference between the start processing time of the jth process of workpiece i in the rescheduling and the start processing time of the jth process of workpiece i in the initial scheduling;
[0017] The set of constraints includes:
[0018] (1) Any workpiece can only be assigned to one workshop and can only be processed by one available machine at a time;
[0019] (2) The processing of any workpiece at any stage must start after the previous process is completed;
[0020] (3) In any workshop, two adjacent workpieces of the same machine must be processed after the previous workpiece is completed;
[0021] (4) Any machine can only process one workpiece at a time;
[0022] (5) The machine will only perform the start-stop operation when the start-stop time and start-stop energy consumption requirements are met at the same time;
[0023] (6) The start processing time and completion processing time of the completed workpiece in the rescheduling plan are the same as those in the previous scheduling plan;
[0024] (7) The start processing time and completion time of the undisturbed workpieces in the set of workpieces being processed in the rescheduling plan are the same as those in the previous scheduling plan;
[0025] (8) The workshop allocation, machine allocation and processing sequence of the workpieces that have been processed and the workpieces being processed are the same in the rescheduling as in the previous scheduling plan.
[0026] Further preferably, the total energy consumption refers to the energy consumed after all workpieces are processed. According to the machine state, the energy consumption is divided into processing energy consumption, idle energy consumption and start-stop energy consumption, which is expressed as:
[0027] minTEC=PEC+WEC+OEC(I)
[0028] Where TEC is the total energy consumption; PEC is the processing energy consumption; WEC is the idle energy consumption; OEC is the start-stop energy consumption.
[0029] Among them, the processing energy consumption is expressed as:
[0030]
[0031] In the formula, Indicates processing O in workshop f i,j Processing power on machine m; O i,j It refers to the i-th process of workpiece j; Indicates workshop f in O i,j Processing time on machine m; The decision variable is 1 when the i-th process of workpiece j is assigned to the m-th machine in workshop f, otherwise it is 0; I represents the set of processes for workpiece processing, i is the process number, i∈I; J represents the set of workpieces to be processed, j is the workpiece number, j∈J; Indicates processing O in workshop f i,j Available machines; F is the number of workshops, f is the workshop number, f = {1,2,...,F};
[0032] No-load energy consumption is expressed as:
[0033]
[0034] Where, WP f,m is the no-load power of machine m in workshop f; For workshop f i,j Start processing time on machine m;
[0035] The start-stop energy consumption is expressed as:
[0036]
[0037] In the formula, h f,m The number of starts and stops of machine m in workshop f; OEC f,m is the energy consumption required for starting and stopping machine m in workshop f; f,m,a is a decision variable. It is 1 when machine m in workshop f is turned on or off between positions a and a+1, and 0 otherwise.
[0038] The total processing energy consumption TEC is expressed as:
[0039]
[0040] Further preferably, the start processing time and the completion processing time of the workpieces in the completed workpiece set in the rescheduling scheme are the same as those in the static scheduling scheme; as shown below:
[0041]
[0042] In the formula, For rescheduling workshop f in O i,j Start processing time on machine m; For rescheduling workshop f in O i,j Completion time on machine m; J finishis the set of completed workpieces. For workshop f i,j Start processing time on machine m, For workshop f i,j+1 The completion time on machine m.
[0043] Further preferably, the workshop allocation, machine allocation and processing sequence of the workpieces that have been processed and the workpieces being processed are the same in the rescheduling scheme as in the static scheduling scheme; as shown below:
[0044]
[0045] In the formula, is a decision variable, which is 1 when the i-th process of job j is assigned to the m-th machine in workshop f in the rescheduling plan, otherwise it is 0; is a decision variable, which is 1 if job j is processed in machine m in shop f before job j' in process i in the initial schedule, and 0 otherwise; is a decision variable. If job j is processed on machine m in workshop f before job j' in process i, it is 1, otherwise it is 0. f,j is a decision variable, which is 1 when job j is assigned to workshop f during rescheduling, otherwise it is 0; now is the set of workpieces being processed; J finish is a set of completed workpieces; Decision variable, 1 if the i-th process of workpiece j is assigned to the m-th machine in workshop f, otherwise 0; y f,j is a decision variable, which is 1 when job j is assigned to workshop f, otherwise it is 0;
[0046] Preferably, according to the present invention, in step S3, a workpiece state judgment rule is constructed; comprising:
[0047] When using the rolling window technology for dynamic scheduling, define the rolling window, that is, the workpiece window, and determine which workpieces belong to the workpiece window. According to the workpiece processing status, they are divided into five workpiece sets: the completed workpiece set, the processing workpiece set, the unprocessed workpiece set, the to-be-processed workpiece set, and the to-be-scheduled workpiece set;
[0048] The workpiece status judgment rule is based on the start processing time and completion processing time of the jth process of workpiece i in the uninitialized scheduling; when the completion time of the jth process of workpiece i is less than or equal to the rescheduling time, O i,j Belongs to the set of completed workpieces; when O i,j When the start processing time is less than the rescheduling time and the completion time is greater than the rescheduling time, O i,j Belongs to the set of workpieces being processed; when O i,j When the start processing time of is greater than the rescheduling time, Oi,j It belongs to the set of unprocessed workpieces; when different disturbance factors occur, the sets of workpieces that need to be rescheduled are different; when the disturbance factor is of the workpiece type, the workpieces in the unprocessed workpiece set and the workpieces in the changed processed workpiece set are rescheduled; when the disturbance factor is of the machine failure type, if the workpiece being processed is processed on a non-faulty machine, it will continue to be processed and will not be rescheduled; if the workpiece being processed is processed on a faulty machine, it is a disturbed workpiece.
[0049] Preferably, in step S4, an improved NSGA-II algorithm is designed to solve the green dynamic programming model of a distributed heterogeneous hybrid flow shop, and rescheduling the workpieces to be scheduled in the rolling window; including:
[0050] 1) Encoding and decoding;
[0051] A three-layer coding method is adopted. The first layer of coding represents the workshop allocation layer coding, the second layer of coding represents the workpiece layer coding to be processed, and the third layer of coding represents the machine selection layer coding, which respectively represent the workshop allocation of the workpiece to be processed, the processing order of the first process of the workpiece, and the processing machine selected by the workpiece in each process;
[0052] Adopt adaptive bidirectional decoding strategy for decoding;
[0053] By performing forward and reverse decoding on a single workshop respectively, the better decoding result is used as the decoding method of the workshop; wherein, forward decoding is to decode a given code from the first process to the last process, and reverse decoding is to decode a given code from the last process to the first process;
[0054] 2) Initialize the population;
[0055] Adopt EGLR1, including time-based global initialization, energy-based global initialization, time-based local initialization and energy-based local initialization;
[0056] The time-based global initialization is expressed as follows: initialize an array whose length is the total number of machines in each workshop, each position corresponds to the machine sequence in the workshop, and the value is the initial processing time of the machine; generate a set of workpieces to be processed according to the processing sequence of the workpieces; for each process in the set of workpieces to be processed, calculate the processing time of each available machine and add it to the corresponding array position; select the machine with the shortest processing time as the processing machine for the current process and update the array; repeat the operation until the processing machines are selected for the processes of all workpieces;
[0057] The global initialization based on energy consumption is expressed as follows: initialize an array with a length equal to the total number of machines in each workshop, each position corresponding to the machine sequence in the workshop, and the value being the processing energy consumption of the machine; generate a set of workpieces to be processed according to the processing sequence of the workpieces; for each process in the set of workpieces to be processed, calculate the processing energy consumption of each available machine and add it to the corresponding array position; select the machine with the minimum energy consumption as the processing machine for the current process and update the array; repeat the operation until all processes of all workpieces have selected processing machines;
[0058] Time-based local initialization;
[0059] Local initialization based on energy consumption;
[0060] 3) Genetic operator design;
[0061] The crossover operator crossover mode is: PBX crossover is used for sorting of workpieces to be processed, and MPX multi-point crossover is used for workshop allocation and machine selection;
[0062] Different mutation operators for shop assignment, machine selection, and the sequence of workpieces to be processed;
[0063] Mutation operators for the sequence of workpieces to be processed, including exchange, insertion and inverse transformation;
[0064] The adaptive mutation operators for workshop allocation are based on single-point insertion between key workshops and two-point exchange between key workshops;
[0065] 4) Elite individual selection strategy;
[0066] The number of excellent individuals retained in each level of non-dominated front end is selected by introducing a distribution function, so as to limit the number of elite solutions of each level of non-dominated front end entering the next generation of parents. The distribution function is expressed as:
[0067] v i =|V i |*s i s i =Rand(0.8,1) (XI)
[0068] In the formula, |V i | represents the total number of individuals in the i-th level non-dominated front end; v i represents the total number of individual selections from the i-th level non-dominated front end, r i is a random real number between 0.8 and 1.
[0069] Further preferably, in encoding and decoding, the iterative sampling probability is expressed as:
[0070]
[0071] Where P d (t+1) is the sampling probability of the decoding direction in the current iteration; P d (t) is the sampling probability of the decoding direction in the last iteration; β is the learning factor; ET num Indicates the number of elite solutions; DEC num Indicates the number of times the current decoding direction has been selected.
[0072] Among them, β is expressed as:
[0073]
[0074] Where RT now Indicates the current machine running time; RT max represents the maximum machine running time; γ represents the learning efficiency modulation factor.
[0075] Further preferably, in the genetic operator design, the probability calculation formula for iterative selection is expressed as:
[0076]
[0077] Where P m (t+1) and P m (t) are the selection probabilities of mutation operations in the current iteration and the previous iteration respectively; β represents the learning factor; improve num The number of solutions increased by mutation operation; use num The number of times the mutation operation was used.
[0078] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of a green dynamic scheduling method for a distributed heterogeneous hybrid assembly line when executing the computer program.
[0079] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a green dynamic scheduling method for a distributed heterogeneous hybrid assembly line.
[0080] A distributed heterogeneous hybrid flow shop green dynamic scheduling system, comprising:
[0081] The machine start-stop energy-saving strategy obtaining module is configured to: propose a machine start-stop energy-saving strategy according to the machine processing state;
[0082] The distributed heterogeneous hybrid flow shop green dynamic scheduling planning model establishment module is configured to: establish a distributed heterogeneous hybrid flow shop green dynamic scheduling planning model according to the distributed heterogeneous hybrid flow shop green dynamic scheduling requirements;
[0083] The artifact status judgment rule building module is configured to: build an artifact status judgment rule;
[0084] The rescheduling implementation module is configured to solve the green dynamic programming model of the distributed heterogeneous hybrid flow shop through the NSGA-II algorithm, and reschedule the workpieces to be scheduled in the rolling window.
[0085] Compared with the prior art, the present invention has the following beneficial effects:
[0086] (1) This paper proposes a mathematical model for green dynamic scheduling of distributed heterogeneous hybrid flow shop, which addresses some deficiencies in the scheduling field and makes a practical expansion in the research field of hybrid flow shop scheduling.
[0087] (2) The present invention divides the machine status into processing, idle and start-stop, and correspondingly divides the energy consumption into processing energy consumption, idle energy consumption and start-stop energy consumption, proposes a machine start-stop energy-saving strategy, reduces the machine idle energy consumption, and provides a method for related enterprises to reduce energy consumption and realize green production transformation and upgrading;
[0088] (3) The present invention utilizes rolling window technology to construct workpiece status judgment rules to determine the workpiece status when dynamic disturbance factors occur, perform rolling window-band scheduled workpiece correction, and reschedule it to achieve the effect of dynamic disturbance events on production with minimal completion time, energy consumption and workpiece transfer.
[0089] (4) The present invention adopts an improved NSGA-II algorithm to solve the green dynamic scheduling model of distributed heterogeneous hybrid assembly lines. The adaptive two-way decoding can fully obtain global information. The population initialization strategy of EGLR1 based on time-based global initialization, energy consumption-based global initialization, time-based local initialization, and energy consumption-based local initialization fully considers the impact of energy consumption on the initial solution; adaptive mutation operation is used to produce a new population, and the distribution function is introduced to improve the elite selection strategy, thereby improving the efficiency and quality of the calculation. BRIEF DESCRIPTION OF THE DRAWINGS
[0090] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the figures required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0091] Figure 1 A schematic diagram of a distributed heterogeneous hybrid flow shop model provided by an embodiment of the present invention;
[0092] Figure 2A flow chart of a green dynamic scheduling method for a distributed heterogeneous hybrid flow shop provided by an embodiment of the present invention;
[0093] Figure 3 A corresponding relationship diagram between a rolling window and a workpiece set provided by an embodiment of the present invention;
[0094] Figure 4 Flow chart of the improved NSGA-II algorithm provided by the embodiment of the present invention;
[0095] Figure 5(a) is a schematic diagram of the comparison of the initial scheduling completion time of the NSGA-II algorithm and the improved NSGA-II algorithm;
[0096] Figure 5(b) is a schematic diagram of the comparison of the total energy consumption of the initial scheduling of the NSGA-II algorithm and the improved NSGA-II algorithm;
[0097] Figure 6(a) is a schematic diagram of the initial scheduling plan for workshop 1;
[0098] Figure 6(b) is a schematic diagram of the initial scheduling plan for workshop 2;
[0099] Figure 7(a) is a schematic diagram of the workshop 1-rescheduling scheme under emergency order disturbance;
[0100] Figure 7(b) is a schematic diagram of the dual scheduling scheme of the workshop under the emergency order disturbance;
[0101] Figure 8 Schematic diagram of coding for three-layer coding method;
[0102] FIG9( a ) is a schematic diagram of a swap operation;
[0103] FIG9( b ) is a schematic diagram of an insertion operation;
[0104] FIG9( c ) is a schematic diagram of the inverse transform operation. DETAILED DESCRIPTION
[0105] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0106] Example 1
[0107] The purpose of the present invention is to provide a green dynamic scheduling method for a distributed heterogeneous hybrid flow shop. Considering the different energy consumption of machines in different states, the energy consumption is divided into processing energy consumption, idle energy consumption and start-stop energy consumption. In order to reduce idle energy consumption, a machine start-stop energy-saving strategy is proposed. Considering the existence of multiple uncertain disturbance factors in the production process, a dynamic scheduling method based on rolling window technology is proposed. The maximum completion time, the minimum total energy consumption and the minimum scheduling deviation are minimized as the goal. Considering the heterogeneity of distributed workshops, the irrelevance between parallel machines and other constraints, a distributed heterogeneous hybrid flow shop green dynamic scheduling model is established. Based on the rolling window technology, a workpiece set judgment criterion is constructed, the traditional NSGA-II algorithm is improved, and an adaptive two-way coding strategy is adopted. The population initialization strategy of EGLR1 based on time-based global initialization, energy consumption-based global initialization, time-based local initialization and energy consumption-based local initialization is adopted. Two crossover operators are adopted, PBX crossover is adopted for sorting the workpieces to be processed, MPX multi-point crossover is adopted for workshop allocation and machine selection, and adaptive mutation operation is adopted to produce a new population. The elite selection strategy is improved by introducing a distribution function to achieve scientific, accurate and timely production scheduling.
[0108] The distributed heterogeneous hybrid flow shop scheduling problem can be described as follows: there are J workpieces to be processed that need to be assigned to F non-identical hybrid flow shops for processing. The fth shop contains I processes, f = 1, 2, ..., F. Among the I processes, at least one process has z (z ≥ 2) unrelated parallel machines. Each workpiece only needs to be processed on one machine in each process and must be processed according to the processing path from the first process to the end of the I process. All processing of the same workpiece needs to be completed in the same shop. Its form is as follows: Figure 1 shown.
[0109] A green dynamic scheduling method for distributed heterogeneous hybrid flow shop, such as Figure 2 As shown, including:
[0110] S1: Propose energy-saving strategies for starting and stopping machines according to the machine processing status;
[0111] S2: According to the requirements of green dynamic scheduling of distributed heterogeneous hybrid flow workshops, a green dynamic scheduling planning model for distributed heterogeneous hybrid flow workshops is established;
[0112] S3: Construct artifact status judgment rules;
[0113] S4: The distributed heterogeneous hybrid flow shop green dynamic programming model is solved through the NSGA-II algorithm to reschedule the workpieces to be scheduled in the rolling window.
[0114] Example 2
[0115] The green dynamic scheduling method for a distributed heterogeneous hybrid assembly line shop described in Example 1 is different in that:
[0116] In step S1, according to the machine processing state, the energy consumption is divided into processing energy consumption, idle energy consumption and start-stop energy consumption, and a machine start-stop energy-saving strategy is proposed;
[0117] The machine start-stop energy-saving strategy means that when the idle waiting time between the processing of the two workpieces required by the machine is greater than the time it takes for the machine to perform a power on and off operation, and when the idle energy consumption between the processing of the two workpieces required by the machine is higher than the energy consumption of the machine performing a power on and off operation, the machine will be shut down.
[0118] When the idle waiting time between the processing of two workpieces required by the machine is longer than the time it takes for the machine to perform a power on and off operation and the energy consumed by the machine during the waiting time is higher than the energy consumed by starting and stopping, the machine is shut down to save idle energy.
[0119] In the production scheduling process, the start-stop energy consumption is positively correlated with the number of machine starts and stops. The energy consumption required for machine m to perform a restart operation, including the energy consumption of the machine starting, preparing, shutting down, etc. until the machine can be processed. The shortest time for machine m to perform start-stop operations satisfy:
[0120]
[0121] Where, t m WP is the time for machine m to start and stop operations; m is the no-load power of machine m;
[0122] Then the energy consumption WEC between the workpiece processing positions a-1 to a on machine m is m As shown below:
[0123]
[0124] In the formula, ST m,a is the start time of processing the ath workpiece on machine m, FT m,a-1 It is the finishing time of the workpiece at the a-1th serial position on machine m.
[0125] The total start-stop energy consumption of machine m is OEC m for:
[0126]
[0127] In the formula, h m is the number of starts and stops of machine m.
[0128] In step S2, the green dynamic scheduling planning model of the distributed heterogeneous hybrid flow shop includes an optimization objective and a series of constraints;
[0129] The optimization objectives include: minimizing the maximum completion time, the minimum total energy consumption, and the minimum scheduling deviation; the maximum completion time refers to the latest completion time of the last process of all workpieces; the total energy consumption refers to the energy consumed after all workpieces are processed, which is divided into processing energy consumption, idle energy consumption, and start-stop energy consumption; the scheduling deviation refers to the difference between the start processing time of the jth process of workpiece i in the rescheduling and the start processing time of the jth process of workpiece i in the initial scheduling;
[0130] The set of constraints includes:
[0131] (1) Any workpiece can only be assigned to one workshop and can only be processed by one available machine at the same time; it can be expressed as:
[0132]
[0133] In the formula, y f,j is a decision variable, which is 1 when job j is assigned to workshop f and 0 otherwise.
[0134] Workshop f i,j The calculation formula of processing time is expressed as:
[0135]
[0136] In the formula, To represent the O in workshop f i,j processing time.
[0137] (2) The processing of any workpiece at any stage must start after the previous process is completed; it can be expressed as:
[0138]
[0139] (3) In any workshop, two adjacent workpieces of the same machine must be processed after the previous workpiece is processed; it can be expressed as:
[0140]
[0141] In the formula, For workshop f i,j+1 The completion time on machine m.
[0142] (4) Any machine can only process one workpiece at a time; expressed as:
[0143]
[0144] (5) The machine will only perform the start-stop operation when the start-stop time and start-stop energy consumption requirements are met at the same time;
[0145] (6) The start processing time and completion processing time of the completed workpiece in the rescheduling plan are the same as those in the previous scheduling plan;
[0146] (7) The start processing time and completion time of the undisturbed workpieces in the set of workpieces being processed in the rescheduling plan are the same as those in the previous scheduling plan;
[0147] (8) The workshop allocation, machine allocation and processing sequence of the workpieces that have been processed and the workpieces being processed are the same in the rescheduling as in the previous scheduling plan.
[0148] Optimization goal:
[0149] Minimize the maximum completion time. The maximum completion time refers to the minimum latest completion time of the last process of all workpieces, expressed as:
[0150] minC max =min{max{C 1,max ,C 2,max ,...,C F,max}}
[0151] Where minC max The total maximum completion time is the minimum; C F,max is the maximum completion time in workshop F.
[0152] Total energy consumption refers to the energy consumed after all workpieces are processed. According to the machine status, the energy consumption is divided into processing energy consumption, idle energy consumption and start-stop energy consumption, expressed as:
[0153] minTEC=PEC+WEC+OEC(I)
[0154] Where TEC is the total energy consumption; PEC is the processing energy consumption; WEC is the idle energy consumption; OEC is the start-stop energy consumption.
[0155] Among them, the processing energy consumption is expressed as:
[0156]
[0157] In the formula, Indicates processing O in workshop f i,j Processing power on machine m; O i,j It refers to the i-th process of workpiece j; Indicates workshop f in O i,j Processing time on machine m; The decision variable is 1 when the i-th process of workpiece j is assigned to the m-th machine in workshop f, otherwise it is 0; I represents the set of processes for workpiece processing, i is the process number, i∈I; J represents the set of workpieces to be processed, j is the workpiece number, j∈J; Indicates processing O in workshop f i,j Available machines; F is the number of workshops, f is the workshop number, f = {1,2,...,F};
[0158] No-load energy consumption is expressed as:
[0159]
[0160] Where, WP f,m is the no-load power of machine m in workshop f; For workshop f i,j Start processing time on machine m;
[0161] The start-stop energy consumption is expressed as:
[0162]
[0163] In the formula, h f,m The number of starts and stops of machine m in workshop f; OEC f,m is the energy consumption required for starting and stopping machine m in workshop f; f,m,a is a decision variable. It is 1 when machine m in workshop f is turned on or off between positions a and a+1, and 0 otherwise.
[0164] The total processing energy consumption TEC is expressed as:
[0165]
[0166] Minimum scheduling deviation. Use rescheduling i,j The start processing time and the initial scheduling i,j The difference in the start processing time represents the scheduling deviation, which is expressed as:
[0167]
[0168] In the formula, DEV represents the scheduling deviation; For rescheduling workshop f in O i,j Start processing time on machine m.
[0169] The start processing time and completion processing time of the workpieces in the completed workpiece set in the rescheduling scheme are the same as those in the static scheduling scheme; as shown below:
[0170]
[0171] In the formula, For rescheduling workshop f in O i,j Start processing time on machine m; For rescheduling workshop f in O i,j Completion time on machine m; J finish is the set of completed workpieces. For workshop f i,j Start processing time on machine m, For workshop f i,j+1 The completion time on machine m.
[0172] The start processing time and completion time of the undisturbed workpieces in the rescheduling scheme are the same as those in the initial scheduling scheme, which can be expressed as:
[0173]
[0174] In the formula, J now is the set of workpieces being processed; J' is the set of workpieces being disturbed.
[0175] The workshop allocation, machine allocation, and processing sequence of completed workpieces and workpieces being processed are the same in the rescheduling and static scheduling schemes; as shown below:
[0176]
[0177] In the formula, is a decision variable, which is 1 when the i-th process of job j is assigned to the m-th machine in workshop f in the rescheduling plan, otherwise it is 0; is a decision variable, which is 1 if job j is processed in machine m in shop f before job j' in process i in the initial schedule, and 0 otherwise; is a decision variable. If job j is processed on machine m in workshop f before job j' in process i, it is 1, otherwise it is 0. f,j is a decision variable, which is 1 when job j is assigned to workshop f during rescheduling, otherwise it is 0; now is the set of workpieces being processed; J finish is a set of completed workpieces; Decision variable, 1 if the i-th process of workpiece j is assigned to the m-th machine in workshop f, otherwise 0; y f,j is a decision variable, which is 1 when job j is assigned to workshop f, otherwise it is 0;
[0178] In step S3, a workpiece status judgment rule is constructed, including:
[0179] Based on the rolling window technology, the dynamic scheduling problem of distributed heterogeneous hybrid flow shop is studied. It is crucial to transform the dynamic scheduling into static scheduling and determine which type of workpiece set the workpiece belongs to at the time of rescheduling. The judgment is based on the static scheduling scheme. and Make a judgment. i,j Completion time When it is less than or equal to the rescheduling time t', O i,j Belongs to the processed workpiece set J finish ; when O i,j Start processing time Less than the rescheduling time t' and the completion time When it is greater than the rescheduling time t', O i,j Belongs to the processing workpiece set J now ; when O i,j Start processing time When it is greater than the rescheduling time t', O i,j Belongs to the unprocessed workpiece set J unfinish , which can be expressed as:
[0180]
[0181] The corresponding relationship between the rolling window and the artifact set is as follows Figure 3 As shown. The finished workpiece, that is, the completed workpiece set J finish The completed workpieces in the processing workpiece set are stored in the completion window; the unfinished workpieces, that is, the unprocessed workpiece set J unfinish , the set of workpieces to be processed J wait The unprocessed parts of the workpieces in the processing workpiece set are stored in the waiting-to-be-processed window. wait , unprocessed workpiece set J unfinish It is stored in the scheduling window and needs to be rescheduled during rescheduling.
[0182] In step S4, an improved NSGA-II algorithm is designed to solve the green dynamic programming model of the distributed heterogeneous hybrid flow shop and reschedule the workpieces to be scheduled in the rolling window; including:
[0183] Combined with the characteristics of green dynamic scheduling problem of distributed heterogeneous hybrid flow shop, the traditional NSGA-II algorithm is improved, including using adaptive two-way decoding to fully obtain global information and avoid falling into local optimum;
[0184] The population initialization strategy of EGLR1, which includes global initialization based on time, global initialization based on energy consumption, local initialization based on time, and local initialization based on energy consumption, is adopted to fully consider the impact of energy consumption on the initial solution and improve the quality of the initial solution.
[0185] Two crossover operators are used: PBX crossover is used for sorting workpieces, and MPX multi-point crossover is used for workshop allocation and machine selection. Adaptive mutation operation is used to produce new populations, which can reduce time complexity, improve convergence speed, and ensure the diversity of the gene pool.
[0186] By introducing the distribution function to improve the elite selection strategy, the number of elite solutions of each level of non-dominated front end entering the next generation of parents is limited, which can effectively prevent the population from converging prematurely and improve the diversity of the population.
[0187] 1) Encoding and decoding;
[0188] The dynamic scheduling problem of distributed heterogeneous hybrid flow shop includes three problems: shop allocation, machine selection and workpiece processing sequence. In order to represent the shop allocation of workpieces and the workpiece processing sequence in the shop, a three-layer coding method is adopted. The first layer of coding represents the shop allocation layer coding, the second layer of coding represents the workpiece layer coding to be processed, and the third layer of coding represents the machine selection layer coding, which respectively represent the shop allocation of workpieces to be processed, the processing sequence of the first process of workpieces and the processing machines selected for workpieces in each process; in order to clearly introduce the three-layer coding method, examples are given for explanation.
[0189] like Figure 8 As shown, workpieces 2, 4, and 5 are assigned to production workshop 1, and workpiece 2 is processed first, workpiece 4 is processed second, and workpiece 3 is processed last. Workpieces 1 and 3 are assigned to production workshop 2, workpiece 1 is processed first, and workpiece 3 is processed last. Workpiece 2 is processed on parallel machine No. 2 in the first process of workshop 1.
[0190] Adopt adaptive bidirectional decoding strategy for decoding;
[0191] Since the solution space of the distributed heterogeneous hybrid flow shop scheduling problem is large, the complete encoding of the solution space will lead to low search efficiency of the metaheuristic algorithm. At the same time, considering the different processing capabilities of parallel machines, the machine selection rules of related parallel machines cannot be directly used for the selection of unrelated parallel machines. Therefore, when decoding, based on the arrangement method, the workpiece sequence of the first process in the process sorting is based on the given coding sequence, and the subsequent processes are carried out according to the principle of first come first served, that is, the completion time of the workpieces of the previous process is sorted in descending order, and then the workpiece sequence of the subsequent process is obtained. The machine selection is carried out in the way of the earliest completion time rule, that is, the machines are arranged according to the given sequence of workpieces. If several workpieces have the same completion time, the workpiece is randomly selected for priority processing. Secondly, in order to fully mine the elite solution information, an adaptive two-way decoding strategy is designed. By performing forward and reverse decoding on a single workshop respectively, the better result of the two decodings is used as the decoding method of the workshop; among them, forward decoding is to decode the given code from the first process to the last process, and reverse decoding is to decode the given code from the last process to the first process;
[0192] 2) Initialize the population;
[0193] Considering the influence of energy consumption on the initial solution, EGLR1 is adopted, including time-based global initialization, energy-based global initialization, time-based local initialization and energy-based local initialization;
[0194] The time-based global initialization is expressed as follows: initialize an array whose length is the total number of machines in each workshop, each position corresponds to the machine sequence in the workshop, and the value is the initial processing time of the machine; generate a set of workpieces to be processed according to the processing sequence of the workpieces; for each process in the set of workpieces to be processed, calculate the processing time of each available machine and add it to the corresponding array position; select the machine with the shortest processing time as the processing machine for the current process and update the array; repeat the operation until the processing machines are selected for the processes of all workpieces;
[0195] The global initialization based on energy consumption is expressed as follows: initialize an array whose length is the total number of machines in each workshop, each position corresponds to the machine order in the workshop, and the value is the processing energy consumption of the machine; generate a set of workpieces to be processed according to the processing order of the workpieces; for each process in the set of workpieces to be processed, calculate the processing energy consumption of each available machine and add it to the corresponding array position; select the machine with the minimum energy consumption as the processing machine for the current process and update the array; repeat the operation until the processing machines are selected for all processes of all workpieces; the principle of global initialization based on energy consumption is the same as that based on time, except that the value of each bit in the array represents the processing energy consumption on the corresponding machine.
[0196] The principles of time-based local initialization and energy-based local initialization are the same as those of time-based global initialization and energy-based global initialization, but when a workpiece is selected, the machine array needs to be reset to 0.
[0197] Time-based local initialization; initialize an array whose length is the total number of machines in each workshop, each position corresponds to the machine sequence in the workshop, and the value is the initial processing time of the machine; generate a set of workpieces to be processed according to the processing sequence of the workpieces; for each process in the set of workpieces to be processed, calculate the processing time of each available machine and add it to the corresponding array position; select the machine with the shortest processing time as the processing machine for the current process, and set the array corresponding to the selected machine to 0; repeat the operation until the processing machines are selected for the processes of all workpieces;
[0198] Local initialization based on energy consumption; initialize an array with the length of the total number of machines in each workshop, each position corresponds to the order of machines in the workshop, and the value is the processing energy consumption of the machine; generate a set of workpieces to be processed according to the processing order of the workpieces; for each process in the set of workpieces to be processed, calculate the processing energy consumption of each available machine and add it to the corresponding array position; select the machine with the minimum energy consumption as the processing machine for the current process, and set the array corresponding to the selected machine to 0; repeat the operation until the processing machine is selected for the processes of all workpieces;
[0199] 3) Genetic operator design;
[0200] The crossover operator crossover mode is: PBX crossover is used for sorting of workpieces to be processed, and MPX multi-point crossover is used for workshop allocation and machine selection;
[0201] Regarding the adaptive mutation operator, the present invention proposes different mutation operators for workshop allocation, machine selection and the sequence of workpieces to be processed;
[0202] Mutation operators for the sequence of workpieces to be processed, including exchange, insertion and inverse transformation;
[0203] Exchange: Randomly select two sites from the job sequence vector and exchange the corresponding elements;
[0204] Insertion: randomly select a site from the job sequence vector and move the corresponding element to another previous site;
[0205] Inverse transformation: randomly select two sites from the job sequence vector and rearrange the elements between them;
[0206] FIG9(a) is a schematic diagram of a swap operation; FIG9(b) is a schematic diagram of an insertion operation; and FIG9(c) is a schematic diagram of an inverse transformation operation.
[0207] The adaptive mutation operators for workshop allocation are based on single-point insertion between key workshops and two-point exchange between key workshops;
[0208] Single-point insertion between key workshops: Randomly select two key workshops from the current solution. If there is only one key workshop, randomly select a non-key workshop for the other one. Randomly select a workpiece from the key workshop to delete it, and insert the workpiece into a random position of another workshop.
[0209] Two-point exchange between key workshops: Randomly select two key workshops from the current solution. If there is only one key workshop, the other one is randomly selected from a non-key workshop; Randomly select two workpieces at any two positions in the two workshops for exchange;
[0210] The adaptive mutation operator for machine selection randomly selects from the available parallel machines for a certain process in a specific process, and replaces the processing machine until a machine with better processing time and processing energy consumption is selected;
[0211] 4) Elite individual selection strategy;
[0212] The number of excellent individuals retained in each level of non-dominated front end is selected by introducing a distribution function, so as to limit the number of elite solutions of each level of non-dominated front end entering the next generation of parents. The distribution function is expressed as:
[0213] v i =|V i |*s i s i =Rand(0.8,1) (XI)
[0214] In the formula, |V i | represents the total number of individuals in the i-th level non-dominated front end; v i represents the total number of individual selections from the i-th level non-dominated front end, r i is a random real number between 0.8 and 1.
[0215] The workpieces to be processed are included in the rolling window, and the improved NSGA-II algorithm is used to schedule the workpieces in the set of workpieces to be scheduled in the rolling window to generate an initial scheduling plan. When a disturbance factor occurs, the state of each workpiece is judged according to the workpiece state judgment rule, and each workpiece set is updated. The improved NSGA-II algorithm is used to reschedule the workpieces to be scheduled in the updated rolling window to generate a new rescheduling plan.
[0216] In encoding and decoding, the iterative sampling probability is expressed as:
[0217]
[0218] Where Pd (t+1) is the sampling probability of the decoding direction in the current iteration; P d (t) is the sampling probability of the decoding direction in the last iteration; β is the learning factor; ET num Indicates the number of elite solutions; DEC num Indicates the number of times the current decoding direction has been selected.
[0219] Among them, β is expressed as:
[0220]
[0221] Where RT now Indicates the current machine running time; RT max represents the maximum machine running time; γ represents the learning efficiency modulation factor.
[0222] In genetic operator design, the probability calculation formula for iterative selection is expressed as:
[0223]
[0224] Where P m (t+1) and P m (t) are the selection probabilities of mutation operations in the current iteration and the previous iteration respectively; β represents the learning factor; improve num The number of solutions increased by mutation operation; use num The number of times the mutation operation was used.
[0225] Improve the NSGA-II algorithm process as follows Figure 4 .
[0226] In a specific embodiment, taking the processing information of 10 workpieces and 13 available machine parameters of 5 processes in a production workshop as an example, the 10 workpieces are numbered J1 to J10; the 13 machines are numbered M1 to M13, the workpiece information of the production workshop and the data of the available machine parameters of each process are sorted out, Table 1 is the processing information of the workpieces in workshop 1 and workshop 2, Table 2 is the table of available machine parameters for each process in workshop 1 and workshop 2, and at production time 1056, a new urgent order arrives, which needs to be expedited production. The order includes three workpieces, and Table 3 is its production information.
[0227] Table 1
[0228]
[0229] Table 2
[0230]
[0231] Table 3
[0232]
[0233]
[0234] In the table, the processing time of the workpiece on the corresponding machine is in min; P and WP represent the machine processing power and no-load power respectively, in KW; OT represents the time required to turn the machine on and off once, in min; OEC represents the energy consumption consumed when turning the machine on and off once, in KWiH.
[0235] The improved NSGA-II algorithm program is encoded, and the workpiece processing information and machine information are imported into the program. After multiple program runs, the simulation results are good. Figure 5(a) is a schematic diagram of the comparison of the initial scheduling completion time between the NSGA-II algorithm and the improved NSGA-II algorithm; Figure 5(b) is a schematic diagram of the comparison of the initial scheduling total energy consumption between the NSGA-II algorithm and the improved NSGA-II algorithm.
[0236] Figure 6(a) is a schematic diagram of the initial scheduling plan for workshop 1; Figure 6(b) is a schematic diagram of the initial scheduling plan for workshop 2; Figure 7(a) is a schematic diagram of the rescheduling plan for workshop 1 under the disturbance of urgent orders; Figure 7(b) is a schematic diagram of the rescheduling plan for workshop 2 under the disturbance of urgent orders; The results of the dynamic scheduling strategy based on rolling window technology and the general right-shift rescheduling results are shown in Table 4.
[0237] Table 4
[0238]
[0239] Analysis shows that the improved NSGA-II algorithm converges faster than the traditional NSGA-II algorithm. The optimized scheduling scheme shortens the completion time by 456 minutes, reduces production energy consumption by 11093.57 Kwh, and increases the on-time delivery rate of orders by 60%. At the same time, the optimized scheduling deviation is reduced by 7675 minutes compared with the original scheduling scheme, making the transfer of workpieces smaller, effectively reducing the transfer cost of workpieces and improving production efficiency. Among the 12 Pareto frontier solutions of rescheduling based on rolling window technology, 9 completion times are better than right-shift rescheduling, with an efficiency of 75%, and 11 energy consumption are better than right-shift rescheduling, with an efficiency of 91.66%. Although the scheduling deviation of the rescheduling scheme generated by the rescheduling strategy of the rolling window is significantly inferior to that of the rescheduling scheme generated by the right-shift rescheduling, its completion time and total energy consumption are significantly lower than those of the right-shift rescheduling, which proves that the rolling window dynamic scheduling strategy proposed in this invention is effective for emergency order insertion disturbances.
[0240] In summary, the green dynamic scheduling method of distributed heterogeneous hybrid flow shops can better respond to market demand and production resource status in real time, and improve the flexibility and adaptability of the production process.
[0241] Example 3
[0242] A computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the distributed heterogeneous hybrid assembly line shop green dynamic scheduling method described in Example 1 or 2 are implemented.
[0243] Example 4
[0244] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the green dynamic scheduling method for a distributed heterogeneous hybrid assembly line shop described in Example 1 or 2.
[0245] Example 5
[0246] A distributed heterogeneous hybrid flow shop green dynamic scheduling system, comprising:
[0247] The machine start-stop energy-saving strategy obtaining module is configured to: propose a machine start-stop energy-saving strategy according to the machine processing state;
[0248] The distributed heterogeneous hybrid flow shop green dynamic scheduling planning model establishment module is configured to: establish a distributed heterogeneous hybrid flow shop green dynamic scheduling planning model according to the distributed heterogeneous hybrid flow shop green dynamic scheduling requirements;
[0249] The artifact status judgment rule building module is configured to: build an artifact status judgment rule;
[0250] The rescheduling implementation module is configured to solve the green dynamic programming model of the distributed heterogeneous hybrid flow shop through the NSGA-II algorithm, and reschedule the workpieces to be scheduled in the rolling window.
Claims
1. A green dynamic scheduling method for a distributed heterogeneous hybrid flow shop, characterized in that: include: S1: Propose energy-saving strategies for starting and stopping machines according to the machine processing status; S2: According to the requirements of green dynamic scheduling of distributed heterogeneous hybrid flow workshops, a green dynamic scheduling planning model for distributed heterogeneous hybrid flow workshops is established; S3: Construct artifact status judgment rules; S4: The distributed heterogeneous hybrid flow shop green dynamic programming model is solved through the NSGA-II algorithm to reschedule the workpieces to be scheduled in the rolling window.
2. A distributed heterogeneous hybrid flow shop green dynamic scheduling method according to claim 1, characterized in that: In step S1, according to the machine processing state, the energy consumption is divided into processing energy consumption, idle energy consumption and start-stop energy consumption, and a machine start-stop energy-saving strategy is proposed; The machine start-stop energy-saving strategy means that when the idle waiting time between the processing of the two workpieces required by the machine is greater than the time it takes for the machine to perform a power on and off operation, and when the idle energy consumption between the processing of the two workpieces required by the machine is higher than the energy consumption of the machine performing a power on and off operation, the machine will be shut down.
3. A distributed heterogeneous hybrid flow shop green dynamic scheduling method according to claim 1, characterized in that: In step S2, the green dynamic scheduling planning model of the distributed heterogeneous hybrid flow shop includes an optimization objective and a series of constraints; The optimization objectives include: minimizing the maximum completion time, the minimum total energy consumption, and the minimum scheduling deviation; the maximum completion time refers to the latest completion time of the last process of all workpieces; the total energy consumption refers to the energy consumed after all workpieces are processed, which is divided into processing energy consumption, idle energy consumption, and start-stop energy consumption; the scheduling deviation refers to the difference between the start processing time of the jth process of workpiece i in the rescheduling and the start processing time of the jth process of workpiece i in the initial scheduling; The set of constraints includes: (1) Any workpiece can only be assigned to one workshop and can only be processed by one available machine at a time; (2) The processing of any workpiece at any stage must start after the previous process is completed; (3) In any workshop, two adjacent workpieces of the same machine must be processed after the previous workpiece is completed; (4) Any machine can only process one workpiece at a time; (5) The machine will only perform the start-stop operation when the start-stop time and start-stop energy consumption requirements are met at the same time; (6) The start processing time and completion processing time of the completed workpiece in the rescheduling plan are the same as those in the previous scheduling plan; (7) The start processing time and completion time of the undisturbed workpieces in the set of workpieces being processed in the rescheduling plan are the same as those in the previous scheduling plan; (8) The workshop allocation, machine allocation and processing sequence of the workpieces that have been processed and the workpieces being processed are the same in the rescheduling as in the previous scheduling plan.
4. A distributed heterogeneous hybrid flow shop green dynamic scheduling method according to claim 3, characterized in that: Total energy consumption refers to the energy consumed after all workpieces are processed. According to the machine status, the energy consumption is divided into processing energy consumption, idle energy consumption and start-stop energy consumption, expressed as: minTEC=PEC+WEC+OEC(I) In the formula, TEC is the total energy consumption; PEC is the processing energy consumption; WEC is the idle energy consumption; OEC is the start-stop energy consumption; Among them, the processing energy consumption is expressed as: In the formula, Indicates processing O in workshop f i,j Processing power on machine m; O i,j It refers to the i-th process of workpiece j; Indicates workshop f in O i,j Processing time on machine m; The decision variable is 1 when the i-th process of workpiece j is assigned to the m-th machine in workshop f, otherwise it is 0; I represents the set of processes for workpiece processing, i is the process number, i∈I; J represents the set of workpieces to be processed, j is the workpiece number, j∈J; Indicates processing O in workshop f i,j Available machines; F is the number of workshops, f is the workshop number, f = {1,2,...,F}; No-load energy consumption is expressed as: Where, WP f,m is the no-load power of machine m in workshop f; For workshop f i,j Start processing time on machine m; The start-stop energy consumption is expressed as: In the formula, h f,m The number of starts and stops of machine m in workshop f; OEC f,m is the energy consumption required for starting and stopping machine m in workshop f; f,m,a is a decision variable, which is 1 when machine m in workshop f is switched on and off between positions a and a+1, and 0 otherwise; The total processing energy consumption TEC is expressed as: Further preferably, the start processing time and the completion processing time of the workpieces in the completed workpiece set in the rescheduling scheme are the same as those in the static scheduling scheme; as shown below: In the formula, For rescheduling workshop f in O i,j Start processing time on machine m; For rescheduling workshop f in O i,j Completion time on machine m; J finish is the set of completed workpieces. For workshop f i,j Start processing time on machine m, For workshop f i,j+1 Completion time on machine m; Further preferably, the workshop allocation, machine allocation and processing sequence of the workpieces that have been processed and the workpieces being processed are the same in the rescheduling scheme as in the static scheduling scheme; as shown below: In the formula, is a decision variable, which is 1 when the i-th process of job j is assigned to the m-th machine in workshop f in the rescheduling plan, otherwise it is 0; is a decision variable, which is 1 if job j is processed in machine m in shop f before job j' in process i in the initial schedule, and 0 otherwise; is a decision variable. If job j is processed on machine m in workshop f before job j' in process i, it is 1, otherwise it is 0. f,j is a decision variable, which is 1 when job j is assigned to workshop f during rescheduling, otherwise it is 0; now is the set of workpieces being processed; J finish is a set of completed workpieces; Decision variable, 1 if the i-th process of workpiece j is assigned to the m-th machine in workshop f, otherwise 0; y f,j is a decision variable, which is 1 when job j is assigned to workshop f and 0 otherwise.
5. A distributed heterogeneous hybrid flow shop green dynamic scheduling method according to claim 1, characterized in that: In step S3, a workpiece status judgment rule is constructed, including: When using the rolling window technology for dynamic scheduling, define the rolling window, that is, the workpiece window, and determine which workpieces belong to the workpiece window. According to the workpiece processing status, they are divided into five workpiece sets: the completed workpiece set, the processing workpiece set, the unprocessed workpiece set, the to-be-processed workpiece set, and the to-be-scheduled workpiece set; The workpiece status judgment rule is based on the start processing time and completion processing time of the jth process of workpiece i in the uninitialized scheduling; when the completion time of the jth process of workpiece i is less than or equal to the rescheduling time, O i,j Belongs to the set of completed workpieces; when O i,j When the start processing time is less than the rescheduling time and the completion time is greater than the rescheduling time, O i,j Belongs to the set of workpieces being processed; when O i,j When the start processing time of is greater than the rescheduling time, O i,j It belongs to the unprocessed workpiece set; when the disturbance factor is of the workpiece type, the workpieces in the unprocessed workpiece set and the workpieces in the changed processed workpiece set are rescheduled; when the disturbance factor is of the machine failure type, if the workpiece being processed is processed on a non-faulty machine, it will continue to be processed and will not be rescheduled; if the workpiece being processed is processed on a faulty machine, it is a disturbed workpiece.
6. A distributed heterogeneous hybrid flow shop green dynamic scheduling method according to claim 1, characterized in that: In step S4, an improved NSGA-II algorithm is designed to solve the green dynamic programming model of the distributed heterogeneous hybrid flow shop and reschedule the workpieces to be scheduled in the rolling window; including: 1) Encoding and decoding; A three-layer coding method is adopted. The first layer of coding represents the workshop allocation layer coding, the second layer of coding represents the workpiece layer coding to be processed, and the third layer of coding represents the machine selection layer coding, which respectively represent the workshop allocation of the workpiece to be processed, the processing order of the first process of the workpiece, and the processing machine selected by the workpiece in each process; Adopt adaptive bidirectional decoding strategy for decoding; By performing forward and reverse decoding on a single workshop respectively, the better decoding result is used as the decoding method of the workshop; wherein, forward decoding is to decode a given code from the first process to the last process, and reverse decoding is to decode a given code from the last process to the first process; 2) Initialize the population; Adopt EGLR1, including time-based global initialization, energy-based global initialization, time-based local initialization and energy-based local initialization; The time-based global initialization is expressed as follows: initialize an array whose length is the total number of machines in each workshop, each position corresponds to the machine sequence in the workshop, and the value is the initial processing time of the machine; generate a set of workpieces to be processed according to the processing sequence of the workpieces; for each process in the set of workpieces to be processed, calculate the processing time of each available machine and add it to the corresponding array position; select the machine with the shortest processing time as the processing machine for the current process and update the array; repeat the operation until the processing machines are selected for the processes of all workpieces; The global initialization based on energy consumption is expressed as follows: initialize an array with a length equal to the total number of machines in each workshop, each position corresponding to the machine sequence in the workshop, and the value being the processing energy consumption of the machine; generate a set of workpieces to be processed according to the processing sequence of the workpieces; for each process in the set of workpieces to be processed, calculate the processing energy consumption of each available machine and add it to the corresponding array position; select the machine with the minimum energy consumption as the processing machine for the current process and update the array; repeat the operation until all processes of all workpieces have selected processing machines; Time-based local initialization; Local initialization based on energy consumption; 3) Genetic operator design; The crossover operator crossover mode is: PBX crossover is used for sorting of workpieces to be processed, and MPX multi-point crossover is used for workshop allocation and machine selection; Different mutation operators for shop assignment, machine selection, and the sequence of workpieces to be processed; Mutation operators for the sequence of workpieces to be processed, including exchange, insertion and inverse transformation; The adaptive mutation operators for workshop allocation are based on single-point insertion between key workshops and two-point exchange between key workshops; 4) Elite individual selection strategy; The number of excellent individuals retained in each level of non-dominated front end is selected by introducing a distribution function, so as to limit the number of elite solutions of each level of non-dominated front end entering the next generation of parents. The distribution function is expressed as: v i =|V i |*s i s i =Rand(0.8,1) (XI) In the formula, |V i | represents the total number of individuals in the i-th level non-dominated front end; v i represents the total number of individual selections from the i-th level non-dominated front end, r i is a random real number between 0.8 and 1.
7. A distributed heterogeneous hybrid flow shop green dynamic scheduling method according to claim 6, characterized in that: In encoding and decoding, the iterative sampling probability is expressed as: Where P d (t+1) is the sampling probability of the decoding direction in the current iteration; P d (t) is the sampling probability of the decoding direction in the last iteration; β is the learning factor; ET num Indicates the number of elite solutions; DEC num Indicates the number of selections for the current decoding direction; Among them, β is expressed as: Where RT now Indicates the current machine running time; RT max represents the maximum machine running time; γ represents the learning efficiency modulation factor; Further preferably, in the genetic operator design, the probability calculation formula for iterative selection is expressed as: Where P m (t+1) and P m (t) are the selection probabilities of mutation operations in the current iteration and the previous iteration respectively; β represents the learning factor; improve num The number of solutions increased by mutation operation; use num The number of times the mutation operation was used.
8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the distributed heterogeneous hybrid assembly line green dynamic scheduling method described in any one of claims 1-7 are implemented.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the distributed heterogeneous hybrid assembly line green dynamic scheduling method described in any one of claims 1 to 7 are implemented.
10. A distributed heterogeneous hybrid flow shop green dynamic scheduling system, characterized in that: include: The machine start-stop energy-saving strategy obtaining module is configured to: propose a machine start-stop energy-saving strategy according to the machine processing state; The distributed heterogeneous hybrid flow shop green dynamic scheduling planning model establishment module is configured to: establish a distributed heterogeneous hybrid flow shop green dynamic scheduling planning model according to the distributed heterogeneous hybrid flow shop green dynamic scheduling requirements; The artifact status judgment rule building module is configured to: build an artifact status judgment rule; The rescheduling implementation module is configured to solve the green dynamic programming model of the distributed heterogeneous hybrid flow shop through the NSGA-II algorithm, and reschedule the workpieces to be scheduled in the rolling window.
Citation Information
Patent Citations
Energy-saving workshop scheduling system for distributed heterogeneous factories
CN113344383A
Distributed heterogeneous flow shop scheduling method based on improved hybrid memetic algorithm
CN117035364A
Optimization method for solving multi-objective flexible job shop scheduling problem
CN117148796A
Hybrid flow shop energy-saving scheduling method for parallel batch processors
CN117631624A
Multi-target flexible job shop scheduling method
CN117787600A
Cited By
Distributed heterogeneous assembly flexible job shop scheduling method with transfer
CN119882635A
