Flexible job shop lean production self-organization optimization method of peak clipping and valley filling strategy

Abstract

The invention belongs to the technical field of industrial scheduling, and discloses a flexible job shop lean production self-organization optimization method for a peak load shifting strategy, and the method comprises the steps: building a problem model, and determining an optimization target and a constraint condition; encoding and decoding chromosomes, and initializing a population; sorting the individuals, executing selection, crossover and variation, and forming a primary filial generation population by using an equipment efficiency cooperative operator and an energy efficiency cooperative operator; combining the parent population and the filial population, performing sorting, selection, crossover and variation again, forming a new filial population by using an equipment efficiency cooperative operator and an energy efficiency cooperative operator, and taking the previous filial population as a new parent population; and repeatedly executing the iteration step, and outputting an optimal scheduling scheme after a termination condition is reached. According to the method, the productivity and the energy efficiency can be effectively coordinated, the actual technical support is provided for the flexible job shop to realize peak clipping and valley filling and reduce the peak-valley electricity proportion, and the method system of the shop scheduling problem is enriched.

Description

Technical Field

[0001] The invention relates to the technical field of industrial scheduling, and in particular to a self-organizing optimization method for lean production in a flexible job shop with a peak-cutting and valley-filling strategy. Background Art

[0002] Energy issues have become the central issue of political economy, environmental protection and other issues in my country and even the world. Although with the continuous development of science and technology, the way of energy utilization and the utilization rate of energy are increasing, the increasing depletion of energy is still an important problem faced by mankind. Therefore, promoting the sustainable development of the manufacturing industry is an inevitable choice to deal with the problem of energy shortage, and it is also the key to enhance the core competitiveness of the manufacturing industry.

[0003] The sustainable development of manufacturing industry requires intelligent manufacturing to empower it, and the intelligent workshop is the main carrier of intelligent manufacturing, which has gradually become the production system direction pursued by manufacturers. Therefore, the intelligent workshop has a higher degree of flexibility and is more suitable for the current widespread needs of multi-variety, small batch, customization, batch production and mixed line production. In addition, the production flexibility of the intelligent workshop can also play an important role in improving the comprehensive energy efficiency of workshop production. Workshop production depends on energy, and the production activities of the workshop will inevitably have an impact on energy consumption and production energy efficiency. Similarly, the restrictions and constraints on energy use in the workshop will also affect the arrangement of production activities. The intelligent workshop can arrange a more reasonable production plan through the flexible adjustment of the production line and the rapid conversion of the production process, thereby creating space for the efficient use of energy. At the same time, the automated energy monitoring technology and networked communication means of the intelligent workshop are conducive to the construction of an interconnected energy-consuming equipment management system, collecting, integrating and analyzing the energy consumption data of the equipment layer, production line layer and workshop layer, promoting the refined control of energy-consuming equipment, and guiding more energy-saving and efficient production activities. In summary, in the context of intelligent workshops, the research on the optimization method of lean production collaborative scheduling is an important topic that meets actual production conditions and helps to improve the sustainable development capabilities of manufacturing enterprises. Flexible adjustment of production activities is one of the important measures to achieve energy conservation in the workshop.

[0004] As a world-class manufacturing center, my country has become a fast-growing power consumer. The power development model that used to track the growth of power load has reached a point where it must change. With the rapid development of smart grids and the acceleration of power market reform, power demand response (DR) as an important interactive technology for smart power consumption has received more and more attention. Power demand response refers to the behavior of users responding to price signals or incentive mechanisms when the price of the wholesale electricity market rises or the stability of the system is threatened, and by changing their own habitual electricity consumption methods, thereby suppressing the rise in electricity prices and maintaining the stability of the power grid. Today, industry has become the largest power consumption industry in my country, and industrial electricity consumption accounts for more than 70% of the total social power generation. Compared with other industries, industrial users have the advantages of large power consumption, strong stability, and high degree of automation. They are very important demand response resources. Among them, most of the equipment in the machinery industry does not require high power reliability, and the demand response resources in terms of peak shifting have great potential for regulation. Moreover, the rapid development of smart workshops provides a good foundation for the implementation of demand response.

[0005] In summary, industrial users have great potential for demand response, and the production flexibility of smart workshops can help industrial users give full play to their demand response potential. Therefore, it is of great significance to carry out research on the optimization method of lean production collaborative scheduling of smart workshops under the industrial demand response mode. It is a key issue that needs to be solved to further rationalize the electricity consumption behavior of smart workshops to help industrial users improve energy efficiency, deeply promote the application of demand response to help power companies improve power supply reliability, and ultimately achieve the comprehensive optimization of the benefits of power companies and industrial users. Summary of the invention

[0006] In view of the problems in the related technology, the present invention proposes a self-organizing optimization method for lean production in a flexible job shop with a peak shaving and valley filling strategy to overcome the above-mentioned technical problems existing in the existing related technology.

[0007] To this end, the specific technical solution adopted by the present invention is as follows:

[0008] A method for self-organizing optimization of lean production in a flexible workshop with a peak-cutting and valley-filling strategy comprises the following steps:

[0009] S1. Based on the flexible job shop scheduling problem under peak shaving and valley filling, establish the problem model and determine the optimization objectives and constraints;

[0010] S2. Encode and decode the chromosomes using a genetic algorithm based on improved global sorting, and initialize the population, which is used as the first-generation parent population;

[0011] S3. Sort the individuals in the initial parent population based on the improved global sorting method, perform selection, crossover and mutation according to the sorting results, and use the equipment efficiency synergy operator and the energy efficiency synergy operator to form the initial child population;

[0012] S4, merge the parent and child populations, perform sorting, selection, crossover and mutation again, and use the equipment efficiency synergy operator and the energy efficiency synergy operator to form a new child population, with the previous child population as the new parent population;

[0013] S5. Repeat the iterative steps of S4 and output the optimal scheduling plan after reaching the termination condition.

[0014] Furthermore, the optimization goal of the problem model is to smooth the load curve of production activities, reduce the peak-valley electricity ratio and reduce electricity costs;

[0015] In the problem model, the objective function established is:

[0016] f=min(C max , TEC, PVER)

[0017]

[0018] TEC=WEC+REC

[0019]

[0020] PVER=PEC / VEC

[0021]

[0022] The constraints are:

[0023]

[0024]

[0025] In the formula, C max represents the maximum completion time; TEC represents the total energy cost; PVER represents the peak-valley electricity ratio; C i represents the completion time of workpiece i; n and m represent the number of processed workpieces and the number of processing machines respectively; WEC, REC, and TEC represent the total processing energy consumption cost, the total standby energy consumption cost, and the total energy consumption cost respectively; i, j, and k represent the workpiece number, the process number of the workpiece, and the machine number respectively; n i Indicates the number of processes included in workpiece i; m i,j represents the total number of available machines for the jth process of job i; L represents the total number of electricity price intervals; t i，j，k，lW represents the processing time of the jth process of workpiece i on machine k in the lth electricity price interval; i,j,k represents the processing power of the jth process of machine k processing workpiece i; EP l represents the electricity price in the lth electricity price interval; t k，l represents the idle time of machine k in the lth electricity price interval; W k represents the no-load power of machine k; PEC and VEC represent the peak energy consumption and valley energy consumption respectively; p l Indicates whether the lth electricity price interval is a peak electricity price interval; v l Indicates whether the lth electricity price interval is a valley electricity price interval; S i,j represents the starting time of the jth process of workpiece i; p i,j,k represents the processing time of the jth process of workpiece i on machine k; x i,j,k Indicates whether the j-th process of workpiece i is processed on machine k; J represents the set of workpieces; represents the start time of the rth process of machine k; y i，j，k，r Indicates whether the j-th process of workpiece i is the r-th process of machine k; r represents the processing priority of the machine; u k represents the total number of processes completed by machine k; Q represents an infinite number; c i represents the completion time of workpiece i; The nth i The start time of the process; The nth i The processing time of the process on machine k; The nth i Whether the process is processed on machine k; T l Indicates the duration of the lth electricity price interval.

[0026] Furthermore, the genetic algorithm based on improved global sorting is used to encode and decode chromosomes, and the population is initialized, and the initialized population is used as the first-generation parent population, including the following steps:

[0027] S21, using a three-layer coding method of process sorting, machine selection and the type of electricity price in the first process of the workpiece to encode the chromosome;

[0028] S22, actively decoding the encoded chromosome using an insertion method combining a genetic algorithm with improved global sorting and an insertion greedy decoding algorithm;

[0029] S23, using a hybrid heuristic method to form a preset number of populations, the number of individuals in each population is 2N, and N is the expected number of individuals in the population;

[0030] S24, traverse each population, traverse each individual in each population, and calculate the crowding degree of the individual;

[0031] S25. Arrange the individuals in each population in descending order of crowding degree, and select the first N individuals to form several new populations;

[0032] S26. Calculate the entropy of each new population, and arrange the new populations in order from large to small, select the new population with the largest entropy, and use the new population as the initial population for evolution.

[0033] Furthermore, the three-layer encoding method of process sorting, machine selection and the type of electricity price in the first process processing interval of the workpiece is used to encode the chromosome, including:

[0034] For the coding of process chains, the workpiece number is used as the gene for identification. The total number of genes corresponds to the accumulation of the number of processes of all workpieces. The number of times a workpiece number appears in the coding represents the number of processes contained in the workpiece. Specifically, the jth occurrence of workpiece number i represents workpiece j i The jth process O i,j The coding of the machine chain adopts the serial number of the processing machine specified by the process in the set of optional machines in the current process, rather than the direct machine code. The specific coding method of the third layer is: determine the interval electricity price type of the first process of the workpiece. If it is the peak electricity price, the first digit of the code is 0, if it is the valley electricity price, it is 1, and if it is the flat electricity price, the first digit of the code is 2.

[0035] Furthermore, the plug-in method combining the improved global sorting genetic algorithm with the plug-in greedy decoding algorithm to actively decode the encoded chromosome includes:

[0036] When the completion time of the j-th process of workpiece i is earlier than the end time of the idle state of machine k, and the time difference between the two times is greater than the processing time of the j+1-th process of workpiece i, the j+1-th process of workpiece i is scheduled to the idle time period, while ensuring that the start time of the j+1-th process is not earlier than the completion time of the j-th process. When there are multiple such idle time periods, the idle time period that minimizes the energy consumption cost of the processing machine is selected for insertion. If the conditions are not met, the processes will be processed on the corresponding machines in a pre-set order.

[0037] Further, forming a population of a preset number using a hybrid heuristic method includes:

[0038] Generate the first layer of coding: 40% of the population is generated using the strategy with the most remaining processes, that is, all workpieces are sorted according to the remaining processes, and the workpieces with the most remaining processes are selected for processing first; 30% of the population is generated using the strategy with the minimum average processing time, that is, the average processing time of each process is calculated, and the process with the shortest average processing time is selected for processing first; the rest are generated randomly.

[0039] Generate the second layer of code: 40% of the population is generated using the global selection strategy, that is, the machine with the smallest total load in the current process machine set is selected; 30% of the population is generated using the local selection strategy, that is, the machine with the smallest processing power is selected in the current process machine set; the rest is generated randomly.

[0040] Randomly generate the third layer code.

[0041] Furthermore, the crowding degree D of individual i i The calculation method is as follows: for each objective function, all individuals are arranged from large to small according to the current objective function value, and the ratio of the smaller value to the larger value of the distance to the individual i adjacent to the individual is calculated, and the sum is calculated after multiple objectives are calculated; the calculation formula of the crowding degree is:

[0042]

[0043] The formula for calculating the entropy of a population is:

[0044]

[0045] Where D ci represents the crowding degree of individual ci; md represents the target dimension, m∈[1,M]; M represents the total number of target dimensions; f md (ci) represents the function value of individual ci on target md; f md (ci+1) represents the function value of individual ci+1 on the target md; f md (ci-1) represents the function value of individual ci-1 on the target md; H represents the entropy of the population; LC represents the length of the chromosome; s represents the size of the symbol set D, |D| = s. D is a symbol set, that is, the set of symbols that may appear at each locus in the chromosome, that is, s represents the total number of symbols that may appear at each locus in the chromosome; P gk represents the probability that the kth symbol in the symbol set D appears at the locus g, where h∈{1,2,…,L}.

[0046] Furthermore, the process of sorting individuals in the current population based on the improved global sorting method, performing selection, crossover and mutation according to the sorting results, and forming an initial generation population using the equipment efficiency synergy operator and the energy efficiency synergy operator includes the following steps:

[0047] S31, normalize the function value data of the individual multidimensional target, and based on the normalized results, calculate the global ranking value GR(i) and the individual aggregation degree A of each individual in the population P. i And global fitness fitness(i), and arrange the individuals in the population in descending order according to the global fitness;

[0048] S32, calculating the entropy of the parent population and the global fitness of the individuals in the population, and determining whether the entropy of the population satisfies the requirement of continuous G-generation decrease;

[0049] If so, Boltzmann selection is applied, and the individual's selection probability P(i) is:

[0050]

[0051] Among them, f i represents individual fitness, b is a scaling factor. When b<0, the selection process is conducive to jumping out of the local optimum, so b<0 is taken. Initially, b=[-1,0] is randomly selected, and the b value is periodically randomly changed every G generations. This method is used to generate a population of N individuals.

[0052] If not or the entropy of the population has not reached G generations, then a binary tournament selection is applied. Two individuals are randomly selected from the parent population. The global ranking values ​​of the two individuals are compared. The individual with the larger global ranking value is selected. If there is no difference in the global ranking values, the individual with the smaller individual aggregation degree is selected. This method is applied to generate a population of N individuals, where N is the size of the offspring population, and G is 0.1 times the number of iterations.

[0053] S33, randomly select k solutions from the parent population as the centroids of each cluster Where l = 1, 2, ..., k;

[0054] S34. Divide the population into k subclusters C l , traverse each individual, calculate the similarity between the individual and each centroid, and batch the individual to the cluster where the centroid is located. If the similarity with different centroids is equal, the individual is assigned to the cluster with less than the preset number of individuals in the cluster;

[0055] S35. Calculate the Euclidean distance between each individual and other individuals respectively, and take the individual with the smallest Euclidean distance variance as the new centroid of each sub-cluster

[0056] S36: Determine whether the conditions are met Or the maximum number of cycles is reached, where ε represents the acceptable similarity between the new and old centroids. If yes, the algorithm ends. If no, And return to S33;

[0057] S37, randomly pairing individuals from two different clusters for crossover; and selecting individuals for mutation at a preset mutation rate;

[0058] S38, calculating the number of times each machine in the machine selection segment is selected for the chromosome after crossover mutation, and determining the machine with the largest number of processes to be processed and the machine with the smallest number of processes to be processed; traversing each process of the machine with the largest number of processes to be processed, and if the current process can be processed by the machine with the smallest number of processes to be processed, replacing the processing machine of the process;

[0059] S39. Calculate and sort the power consumption of each process of each workpiece of the chromosome of the application equipment efficiency coordination operator, determine the process with the largest power consumption, and replace the selected machine with the currently unloaded minimum power machine from the optional machine set of the process to form an initial generation population.

[0060] Furthermore, the expression of the normalization process is:

[0061]

[0062] The calculation formula of the global ranking value GR(i) is:

[0063]

[0064] Individual aggregation degree A i The calculation formula is:

[0065]

[0066] When considering the objective function value and the diverse distribution, we hope that the global ranking value GR(i) is maximized and the individual aggregation degree A i Minimum, so use individual aggregation degree A i When solving the diversity of the set, it is constructed as The calculation formula of global fitness fitness(i) is:

[0067]

[0068] In the formula, f m (i) represents the function value of individual i at target m; f m ′(i) represents the normalized function value of individual i at target m; and Respectively represent the maximum and minimum values ​​of individual i in the current target; S ij represents the similarity between individuals i and j, where individual j is any individual in population P that is different from i; γ represents the similarity threshold, which is a constant and takes γ∈[0.9,1]; represents the number of identical genes between individuals i and j; N represents the number of individuals in the population; ω1 and ω2 represent the weight coefficients of the global ranking value and individual aggregation, respectively, and both are taken as 0.5.

[0069] Furthermore, the individual j and each centroid The calculation formula for the similarity between is:

[0070]

[0071] In the formula, Represents individual j and each centroid The similarity between Represents individual j and the centroid The same number of genes.

[0072] Furthermore, the calculation formula of the variance is:

[0073]

[0074] In the formula, Represents individual j and each centroid The similarity between Represents individual j and the centroid The number of identical genes; μ i represents the mean of the Euclidean distance; g represents the gene locus in the chromosome; x ig represents the g-th gene value in individual i; x jg represents the g-th gene value in individual j; (C l ) represents the subcluster C l The number of individuals in s i It represents the variance of the Euclidean distance between individual i and other individuals.

[0075] The beneficial effects of the present invention are:

[0076] 1) The present invention is applied to flexible operation workshops. Under the peak shaving and valley filling strategy, a problem model containing multiple objectives and constraints is established, aiming to minimize the maximum completion time, energy consumption cost and peak-to-valley electricity ratio, and smooth the production load curve. In the scheduling process, the equipment efficiency synergy operator and the energy efficiency synergy operator are combined to optimize machine selection and process scheduling, reduce power consumption, and thus reduce the overall energy consumption cost. Through the application of the equipment efficiency synergy operator and the energy efficiency synergy operator, production efficiency is further improved.

[0077] 2) The present invention designs a genetic algorithm based on improved global sorting. By combining the improved global sorting genetic algorithm with the inserted greedy decoding algorithm, the reasonable arrangement of the workpiece process is ensured, the conflict between the processes is avoided, and the rationality and effectiveness of the scheduling are guaranteed. In the process of population initialization and selection, the population diversity is maintained, premature convergence is avoided, and the global search ability of the algorithm is improved by calculating the entropy of the population and the crowding of the individuals. In the selection operator, by applying the binary tournament selection and Boltzmann selection methods, the selection strategy is adaptively adjusted to further improve the optimization effect. In addition, by dividing the population into multiple subclusters, recalculating the center of mass, and randomly pairing individuals from different clusters for crossover, the effectiveness of crossover and mutation is enhanced, and the diversity and optimization ability of the population are guaranteed.

[0078] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings 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.

[0080] Figure 1 This is a flow chart of a self-organizing optimization method for lean production in a flexible job shop under a peak-cutting and valley-filling strategy according to an embodiment of the present invention;

[0081] Figure 2 is a schematic diagram of chromosome encoding according to an embodiment of the present invention;

[0082] Figure 3 is a schematic diagram of a crossover method of process coding layers according to an embodiment of the present invention;

[0083] Figure 4 is a schematic diagram of a chromosome segment crossover selected by a machine according to an embodiment of the present invention;

[0084] Figure 5 is a schematic diagram of chromosome segment variation according to the process sequence in an embodiment of the present invention;

[0085] Figure 6 is a schematic diagram of machine selection of chromosome segment variation according to an embodiment of the present invention;

[0086] Figure 7 It is a Pareto frontier distribution diagram formed on three scheduling target dimensions of maximum completion time, total energy consumption cost, and peak-valley electricity ratio according to an embodiment of the present invention;

[0087] Figure 8 It is an optimal scheduling Gantt chart under the efficient optimal scheduling scheme according to an embodiment of the present invention;

[0088] Fig. 9 is an optimal scheduling Gantt chart under an economical optimal scheduling solution according to an embodiment of the present invention;

[0089] Fig.10 It is the optimal scheduling Gantt chart under the stable optimal scheduling scheme according to the embodiment of the present invention. DETAILED DESCRIPTION

[0090] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention and are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, ordinary technicians in the field should be able to understand other possible implementations and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0091] According to an embodiment of the present invention, a self-organizing optimization method for lean production in a flexible job shop with a peak shaving and valley filling strategy is provided.

[0092] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1-Figure 10 As shown, the lean production self-organization optimization method of a flexible job shop with a peak shaving and valley filling strategy according to an embodiment of the present invention includes the following steps:

[0093] S1. Based on the flexible job shop scheduling problem under peak shaving and valley filling, establish the problem model and determine the optimization objectives and constraints;

[0094] Specifically, the optimization goal of the problem model is to smooth the load curve of production activities, reduce the peak-valley electricity ratio, and reduce electricity costs.

[0095] In the problem model, the objective function established is:

[0096] f=min(C max ,TEC,PVER) (1)

[0097]

[0098] TEC=WEC+REC (3)

[0099]

[0100] PVER=PEC / VEC (6)

[0101]

[0102] The constraints of the established problem model are:

[0103]

[0104]

[0105]

[0106] Among them, formula (9) represents the process constraint of the same workpiece, that is, the j+1th process of the workpiece must be started after the jth process is completed; formula (10) means that the same machine must complete the processing of the current process before starting to process another process; formulas (11) and (12) mean that each process can only be processed after the previous process on its processing machine and all its previous processes are completed; formula (15) means that the completion time of the workpiece must be after the completion of its last process; formula (16) means that the processing time of the same machine in any electricity price range cannot exceed the interval length; formula (17) means that the sum of the processing time of the same process in each electricity price range is equal to its established processing time;

[0107] In the formula, C max represents the maximum completion time; TEC represents the total energy cost; PVER represents the peak-valley electricity ratio; C i represents the completion time of workpiece i; n and m represent the number of processed workpieces and the number of processing machines respectively; WEC, REC, and TEC represent the total processing energy consumption cost, the total standby energy consumption cost, and the total energy consumption cost respectively; i, j, and k represent the workpiece number, the process number of the workpiece, and the machine number respectively; n i Indicates the number of processes included in workpiece i; m i,j represents the total number of available machines for the jth process of job i; L represents the total number of electricity price intervals; t i，j，k，l W represents the processing time of the jth process of workpiece i on machine k in the lth electricity price interval; i,j,k represents the processing power of the jth process of machine k processing workpiece i; EP l represents the electricity price in the lth electricity price interval; t k，l represents the idle time of machine k in the lth electricity price interval; W k represents the no-load power of machine k; PEC and VEC represent the peak energy consumption and valley energy consumption respectively; p l Indicates whether the lth electricity price interval is a peak electricity price interval; v l Indicates whether the lth electricity price interval is a valley electricity price interval; S i,j represents the starting time of the jth process of workpiece i; p i,j,k represents the processing time of the jth process of workpiece i on machine k; x i,j,kIndicates whether the j-th process of workpiece i is processed on machine k; J represents the set of workpieces; represents the start time of the rth process of machine k; y i，j，k，r Indicates whether the j-th process of workpiece i is the r-th process of machine k; r represents the processing priority of the machine; u k represents the total number of processes completed by machine k; Q represents an infinite number; c i represents the completion time of workpiece i; The nth i The start time of the process; The nth i The processing time of the process on machine k; The nth i Whether the first process is processed on machine k; T l Indicates the duration of the lth electricity price interval.

[0108] S2. Encode and decode the chromosomes using a genetic algorithm based on improved global sorting, and initialize the population, which is used as the first-generation parent population;

[0109] The method of encoding and decoding chromosomes using a genetic algorithm based on improved global sorting and initializing a population, wherein the initialized population is used as a first-generation parent population, includes the following steps:

[0110] S21, using a three-layer coding method of process sorting, machine selection and the type of electricity price in the first process of the workpiece to encode the chromosome;

[0111] For the coding of process chains, the workpiece number is used as the gene for identification. The total number of genes corresponds to the accumulation of the number of processes of all workpieces. The number of times a workpiece number appears in the coding represents the number of processes contained in the workpiece. Specifically, the jth occurrence of workpiece number i represents workpiece j i The jth process O i,j The coding of the machine chain adopts the serial number of the processing machine specified by the process in the set of optional machines in the current process, rather than the direct machine code. The specific coding method of the third layer is: determine the interval electricity price type of the first process of the workpiece. If it is the peak electricity price, the first digit of the code is 0, if it is the valley electricity price, it is 1, and if it is the flat electricity price, the first digit of the code is 2.

[0112] S22, actively decoding the encoded chromosome using an insertion method combining a genetic algorithm with improved global sorting and an insertion greedy decoding algorithm;

[0113] The decoding method is an insertion-type active decoding that combines the genetic algorithm of improved global sorting with the insertion-type greedy decoding algorithm. When the completion time of the jth process of workpiece i is earlier than the end time of the idle state of machine k, and the time difference between the two times is greater than the processing time of the j+1th process of workpiece i, the j+1th process of workpiece i is arranged into the idle time period. At the same time, it is necessary to ensure that the start time of the j+1th process is not earlier than the completion time of the jth process. When there are multiple such idle time periods, the idle time period that minimizes the energy consumption cost of the processing machine is selected for insertion. If the conditions are not met, the processes will be processed on the corresponding machines in the previously specified order.

[0114] S23, using a hybrid heuristic method to form a preset number of populations, the number of individuals in each population is 2N, and N is the expected number of individuals in the population;

[0115] The method of forming a preset number of populations using a hybrid heuristic method includes:

[0116] Generate the first level code. 40% of the population is generated using the strategy of most remaining processes, that is, all workpieces are sorted according to the remaining processes, and the workpieces with the most remaining processes are selected for processing first; 30% of the population is generated using the strategy of minimum average processing time, that is, the average processing time of each process is calculated, and the process with the shortest average processing time is selected for processing first; the rest are generated randomly.

[0117] Generate the second level code. 40% of the population is generated using the global selection strategy, that is, the machine with the smallest total load in the current process machine set is selected; 30% of the population is generated using the local selection strategy, that is, the machine with the smallest processing power is selected in the current process machine set; the rest is generated randomly.

[0118] Generate the third layer code: randomly generate.

[0119] S24, traverse each population, traverse each individual in each population, and calculate the crowding degree of the individual;

[0120] Specifically, the method for calculating the crowding degree is as follows: for each objective function, all individuals are arranged from large to small according to the current objective function value, and the ratio of the smaller value to the larger value of the distance between the individuals adjacent to the individual i is calculated. After calculating multiple objectives, the sum is calculated. The calculation formula for the crowding degree is:

[0121]

[0122] The formula for calculating the entropy of a population is:

[0123]

[0124] Where D cirepresents the crowding degree of individual ci; md represents the target dimension, m∈[1,M]; M represents the total number of target dimensions; f md (ci) represents the function value of individual ci on target md; f md (ci+1) represents the function value of individual ci+1 on the target md; f md (ci-1) represents the function value of individual ci-1 on the target md; H represents the entropy of the population; LC represents the length of the chromosome; s represents the size of the symbol set D, |D| = s. D is a symbol set, that is, the set of symbols that may appear at each locus in the chromosome, that is, s represents the total number of symbols that may appear at each locus in the chromosome; P gk represents the probability that the kth symbol in the symbol set D appears at the locus g, where g∈{1,2,…,L}.

[0125] S25. Arrange the individuals in each population in descending order of crowding degree, and select the first N individuals to form several new populations;

[0126] S26. Calculate the entropy of each new population, and arrange the new populations in order from large to small, select the new population with the largest entropy, and use the new population as the initial population for evolution.

[0127] S3. Sort the individuals in the current population based on the improved global sorting method, perform selection, crossover and mutation according to the sorting results, and use the equipment efficiency synergy operator and the energy efficiency synergy operator to form the initial generation population;

[0128] The steps of sorting individuals in the current population based on the improved global sorting method, performing selection, crossover and mutation according to the sorting results, and forming an initial generation population using the equipment efficiency synergy operator and the energy efficiency synergy operator include the following steps:

[0129] S31, normalize the function value data of the individual multidimensional target, and based on the normalized results, calculate the global ranking value GR(i) and the individual aggregation degree A of each individual in the population P. i And global fitness fitness(i), and arrange the individuals in the population in descending order according to the global fitness;

[0130] The expression of the normalization process is:

[0131]

[0132] The calculation formula of the global ranking value GR(i) is:

[0133]

[0134] Individual aggregation degree Ai The calculation formula is:

[0135]

[0136] When considering the objective function value and the diverse distribution, we hope that the global ranking value GR(i) is maximized and the individual aggregation degree A i Minimum, so use individual aggregation degree A i When solving the diversity of the set, it is constructed as The calculation formula of global fitness fitness(i) is:

[0137]

[0138] In the formula, f m (i) represents the function value of individual i at target m; f m ′(i) represents the normalized function value of individual i at target m; and Respectively represent the maximum and minimum values ​​of individual i in the current target; S ij represents the similarity between individuals i and j, where individual j is any individual in population P that is different from i; γ represents the similarity threshold, which is a constant and takes γ∈[0.9,1]; represents the number of identical genes between individuals i and j; N represents the number of individuals in the population; ω1 and ω2 represent the weight coefficients of the global ranking value and individual aggregation, respectively, and both are taken as 0.5.

[0139] S32, calculating the entropy of the parent population and the global fitness of the individuals in the population, and determining whether the entropy of the population satisfies the requirement of continuous G-generation decrease;

[0140] If so, Boltzmann selection is applied, and the individual's selection probability P(i) is:

[0141]

[0142] Among them, f i represents individual fitness, b is a scaling factor. When b<0, the selection process is conducive to jumping out of the local optimum, so b<0 is taken. Initially, b=[-1,0] is randomly selected, and the b value is periodically randomly changed every G generations. This method is used to generate a population of N individuals.

[0143] If not or the entropy of the population has not reached G generations, then a binary tournament selection is applied. Two individuals are randomly selected from the parent population. The global ranking values ​​of the two individuals are compared. The individual with the larger global ranking value is selected. If there is no difference in the global ranking values, the individual with the smaller individual aggregation degree is selected. This method is applied to generate a population of N individuals, where N is the size of the offspring population, and G is 0.1 times the number of iterations.

[0144] S33, randomly select k solutions from the parent population as the centroids of each cluster Where l = 1, 2, ..., k;

[0145] S34. Divide the population into k subclusters C l , traverse each individual, calculate the similarity between the individual and each centroid, and batch the individual to the cluster where the centroid is located. If the similarity with different centroids is equal, the individual is assigned to the cluster with fewer individuals in the cluster;

[0146] Individual j and each centroid The calculation formula for the similarity between is:

[0147]

[0148] In the formula, Represents individual j and each centroid The similarity between Represents individual j and the centroid The same number of genes.

[0149] S35. Recalculate the centroid of each sub-cluster

[0150] Calculate the Euclidean distance between each individual and other individuals, and take the individual with the smallest Euclidean distance variance as the new centroid of each subcluster

[0151] The calculation formula of the variance is:

[0152]

[0153] In the formula, μ i represents the mean of the Euclidean distance; g represents the locus in the chromosome, g∈[1,L]; x ig represents the g-th gene value in individual i; x jg represents the g-th gene value in individual j; (C l ) represents the subcluster C l The number of individuals in s i It represents the variance of the Euclidean distance between individual i and other individuals.

[0154] S36: Determine whether the conditions are met Or the maximum number of cycles is reached, where ε represents the acceptable similarity between the new and old centroids. If yes, the algorithm ends. If no, And return to S33;

[0155] S37, randomly pairing individuals from two different clusters for crossover, and selecting individuals for mutation at a preset mutation rate;

[0156] S38, calculating the number of times each machine in the machine selection segment is selected for the chromosome after crossover mutation, and determining the machine with the largest number of processes to be processed and the machine with the smallest number of processes to be processed; traversing each process of the machine with the largest number of processes to be processed, and if the current process can be processed by the machine with the smallest number of processes to be processed, replacing the processing machine of the process;

[0157] S39. Calculate and sort the power consumption of each process of each workpiece of the chromosome of the application equipment efficiency coordination operator. The power consumption is the product of the selected processing machine and the processing time. Determine the process with the largest power consumption and replace the selected machine with the currently unloaded minimum power machine from the optional machine set of the process to form the initial generation population.

[0158] S4, merge the parent and child populations, perform sorting, selection, crossover and mutation again, and use the equipment efficiency synergy operator and the energy efficiency synergy operator to form a new child population, with the previous child population as the new parent population;

[0159] S5. Repeat the iterative steps of S4 and output the optimal scheduling plan after reaching the termination condition.

[0160] The above technical scheme describes the overall framework of the present invention through modeling and training processes. The following is an example of application verification using the ×× workshop of ×× enterprise under the time-of-use electricity price adopted by the industry in XX Province. The data is simplified into a 10×5 (10 workpieces, 5 machines) flexible workshop scheduling problem, and the algorithm is set as follows: population size N=100, population genetic generation G=100.

[0161] According to the 2023 "Notice of the Development and Reform Commission of XX Province on the Release of the Electricity Price List and the Electricity Purchase Price List of Power Grid Enterprises Acting as Agents for Industrial and Commercial Users in July (Actual Settlement) in our Province", the daily peak hours are 9:00-12:00, 17:00-22:00; the normal hours are 8:00-9:00, 12:00-17:00, 22:00-23:00; the low hours are 23:00-8:00 the next day. According to the specific electricity price, the time-of-use electricity price function (unit: yuan / (KWh)) within 24 hours a day can be expressed as:

[0162]

[0163] The energy cost is in yuan / (KWh). In order to consider the scheduling scheme across different electricity price periods, the time step of the scheduling example is extended to a duration of 15 minutes. The maximum processing time of each process in each scheduling example is t max , the minimum processing time is t min , the machine has different processing power when processing different processes, which is in the interval [3·t min, 4·t max ](kW), where p min is the minimum processing power of the machine, the standby power of the machine is constant, and its calculation formula is p′idle=0.4×p min Among them, p' idle This is the machine standby power.

[0164] After 100 iterations, the Pareto frontier formed by the flexible job shop scheduling under time-of-use electricity prices in the three scheduling target dimensions of maximum completion time, total energy consumption cost, and peak-valley electricity ratio is as follows: Figure 7 shown.

[0165] Figure 7 It can be seen that the optimal frontier surface is discontinuous, which shows that the three objectives of the scheduling scheme are nonlinear. max The peak-to-valley electricity ratio and processing energy cost are also different, and as the maximum completion time increases, the peak-to-valley electricity ratio and processing energy cost show a decreasing trend. As the peak-to-valley electricity ratio decreases, the processing energy cost shows an increasing trend.

[0166] During the iteration process, as the number of iterations increases, the population average and optimal value of each objective are gradually optimized, indicating that the algorithm has good convergence. The algorithm finally obtains a Paroto solution set with 15 individuals. By pruning redundant individuals, a Pareto solution set with 12 non-dominated solutions is obtained as shown in Table 2, which provides 12 relatively optimal scheduling schemes for the problem model in this embodiment. Taking into account the different actual needs of enterprises, three scheduling schemes with different focuses, namely, efficient, economical, and stable, are given in this embodiment. Each scheduling scheme is a scheme that minimizes the target value selected from the Pareto solution set. The optimal scheduling target values ​​under each scheme are shown in Table 1.

[0167] Table 1 Optimal target values ​​under different schemes

[0168]

[0169] From the data in Table 1, it can be seen that the maximum completion time of the high-efficiency scheme is the shortest, which is 14.75h, but its processing energy cost is 140.5418 yuan higher than that of the economical scheme, and its peak-valley electricity ratio is 0.1131 higher than that of the stable scheme; the processing energy cost of the economical scheme is the lowest, which is 2065.0256 yuan, and its maximum completion time is extended by 2h compared with the high-efficiency scheme, and its peak-valley electricity ratio is increased by 0.0283 compared with the stable scheme; the peak-valley electricity ratio of the stable scheme is the lowest, which is 0.1105, and its maximum completion time is extended by 2.5h compared with the high-efficiency scheme, and its processing energy cost is only increased by 63.8816 yuan compared with the energy consumption cost of the economical scheme. It can be seen that different scheduling schemes have different focuses. Overall, they can take into account the coordination of the three scheduling objectives of time, cost, and peak-valley electricity ratio to a certain extent. In the actual production process, the needs of the enterprise and the power grid can be comprehensively considered to select a balanced scheme. Figure 8-Figure 10 Gantt chart for optimal scheduling under different scenarios.

[0170] By comparing the best scheduling under different schemes, we can see that the efficient processing scheme has compact job arrangements, the least standby times, and balanced load on each machine. If high efficiency is pursued, the processing time required for workpieces on different machines will become the main influencing factor. Therefore, the efficient processing scheme is suitable for orders with high requirements for production efficiency and completion time. The standby period under the economical processing scheme is concentrated around 10 to 12 o'clock. The electricity price during this period is the highest. Therefore, it can be seen that the purpose of reducing energy costs can be achieved by arranging processing as little as possible during the peak electricity price period. In order to minimize the peak-to-valley electricity ratio, the stable processing scheme increases the number of standby machines and the standby time during the peak electricity price period compared to the economical processing scheme to maximize peak shaving and valley filling.

[0171] In order to explore the impact of the peak-valley electricity ratio on the maximum completion time and processing energy cost, the optimization target of the algorithm is changed to only the maximum completion time and processing energy cost, and the maximum completion time and processing energy cost are observed when the peak-valley electricity ratio is not used as one of the optimization targets. Through the evolution process of the maximum completion time, processing energy cost and peak-valley electricity ratio, it can be seen that when only the maximum completion time and processing energy cost are considered, their convergence speed is better than the previous three targets. Combined with Table 3, it can be seen that the average value and optimal value of the maximum completion time and processing energy cost of the final Paroto solution set of the two-objective scheme also optimize the three-objective scheme, but for the peak-valley electricity ratio indicator, the evolution process of the peak-valley electricity ratio optimal value of the two-objective scheme does not show a reasonable convergence trend, and its optimal value and average value are inferior to the three-objective scheme.

[0172] Table 2 The Pareto optimal solution set obtained in the final generation

[0173]

[0174] Table 3 Comparison of the Paroto solutions of the two-objective scheme and the three-objective scheme

[0175]

[0176] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A self-organizing optimization method for lean production in a flexible workshop with a peak-shaving and valley-filling strategy, characterized in that: The following steps are involved: S1. Based on the flexible job shop scheduling problem under peak shaving and valley filling, establish the problem model and determine the optimization objectives and constraints; S2. Encode and decode the chromosomes using a genetic algorithm based on improved global sorting, and initialize the population, which is used as the first-generation parent population; S3. Sort the individuals in the initial parent population based on the improved global sorting method, perform selection, crossover and mutation according to the sorting results, and use the equipment efficiency synergy operator and the energy efficiency synergy operator to form the initial child population; S4, merge the parent and child populations, perform sorting, selection, crossover and mutation again, and use the equipment efficiency synergy operator and the energy efficiency synergy operator to form a new child population, with the previous child population as the new parent population; S5. Repeat the iterative steps of S4 and output the optimal scheduling plan after reaching the termination condition.

2. The method for self-organizing optimization of lean production in a flexible workshop with a peak-shaving and valley-filling strategy according to claim 1 is characterized in that: The optimization goal of the problem model is to smooth the load curve of production activities, reduce the peak-valley electricity ratio and reduce electricity costs; In the problem model, the objective function established is: f=min(C max ,TEC,PVER) TEC=WEC+REC PVER=PEC / VEC The constraints are: In the formula, C max represents the maximum completion time; TEC stands for total energy cost; PVER means peak-to-valley electricity ratio; C i represents the completion time of workpiece i; n and m represent the number of processed workpieces and the number of processing machines, respectively; WEC, REC, and TEC represent total processing energy cost, total standby energy cost, and total energy cost, respectively; i, j, and k represent the workpiece number, the process number of the workpiece, and the machine number, respectively; n i represents the number of processes included in workpiece i; m i,j represents the total number of available machines for the jth process of job i; L represents the total number of electricity price intervals; t i，j，k，l represents the processing time of the jth process of workpiece i on machine k in the lth electricity price interval; W i,j,k represents the processing power of the jth process of machine k processing workpiece i; EP l represents the electricity price in the lth electricity price interval; t k，l represents the idle time of machine k in the lth electricity price interval; W k represents the no-load power of machine k; PEC and VEC represent peak energy consumption and valley energy consumption, respectively; p l Indicates whether the lth electricity price interval is a peak electricity price interval; v l Indicates whether the lth electricity price interval is a valley electricity price interval; S i,j represents the start processing time of the jth process of workpiece i; p i,j,k represents the processing time of the jth process of workpiece i on machine k; x i,j,k Indicates whether the jth process of workpiece i is processed on machine k; J represents the set of artifacts; It indicates the start time of the rth process of machine k; y i，j，k，r Indicates whether the j-th process of workpiece i is the r-th process of machine k; r represents the processing priority of the machine; u k represents the total number of processes completed by machine k; Q represents an infinite number; c i represents the completion time of workpiece i; The nth i The start time of the process; The nth i The processing time of the process on machine k; The nth i Whether the process is processed on machine k; T l Indicates the duration of the lth electricity price interval.

3. The method for self-organizing optimization of lean production in a flexible workshop with a peak-shaving and valley-filling strategy according to claim 1 is characterized in that: The method of encoding and decoding chromosomes by using a genetic algorithm based on improved global sorting and initializing a population, wherein the initialized population is used as a first-generation parent population, comprises the following steps: S21, using a three-layer coding method of process sorting, machine selection and the type of electricity price in the first process of the workpiece to encode the chromosome; S22, actively decoding the encoded chromosome using an insertion method combining a genetic algorithm with improved global sorting and an insertion greedy decoding algorithm; S23, using a hybrid heuristic method to form a preset number of populations, the number of individuals in each population is 2N, and N is the expected number of individuals in the population; S24, traverse each population, traverse each individual in each population, and calculate the crowding degree of the individual; S25. Arrange the individuals in each population in descending order of crowding degree, and select the first N individuals to form several new populations; S26. Calculate the entropy of each new population, and arrange the new populations in order from large to small, select the new population with the largest entropy, and use the new population as the initial population for evolution.

4. The method for self-organizing optimization of lean production in a flexible workshop with a peak-shaving and valley-filling strategy according to claim 3 is characterized in that: The three-layer encoding method of using process sorting, machine selection and the type of electricity price of the first process interval of the workpiece to encode the chromosome includes: For the coding of the process chain, the workpiece number is used as the gene for identification, and the total number of genes corresponds to the accumulation of the number of all workpiece processes; for the coding of the machine chain, the serial number of the processing machine specified by the process in the set of optional machines for the current process is used; instead of the direct machine code, the third-level coding method is adopted: the interval electricity price type of the first process of the workpiece is judged, if it is the peak electricity price, the first digit of the code is 0, if it is the valley electricity price, the first digit of the code is 1, and if it is the flat electricity price, the first digit of the code is 2.

5. The method for self-organizing optimization of lean production in a flexible workshop with a peak-shaving and valley-filling strategy according to claim 3 is characterized in that: The plug-in method combining the improved global sorting genetic algorithm with the plug-in greedy decoding algorithm to actively decode the encoded chromosome includes: When the completion time of the j-th process of workpiece i is earlier than the end time of the idle state of machine k, and the time difference between the two times is greater than the processing time of the j+1-th process of workpiece i, the j+1-th process of workpiece i is scheduled to the idle time period, while ensuring that the start time of the j+1-th process is later than the completion time of the j-th process. When there are multiple such idle time periods, the idle time period that minimizes the energy consumption cost of the processing machine is selected for insertion. If the conditions are not met, the processes will be processed on the corresponding machines in a pre-set order.

6. The method for self-organizing optimization of lean production in a flexible workshop with a peak-shaving and valley-filling strategy according to claim 3 is characterized in that: The method of forming a preset number of populations using a hybrid heuristic method includes: Generate the first level code: 40% of the population is generated using the maximum remaining process strategy; 30% of the population is generated using the minimum average processing time strategy; the rest is generated randomly; Generate the second layer of code: 40% of the population is generated using the global selection strategy; 30% of the population is generated using the local selection strategy; the rest is generated randomly; Generate the third layer code: randomly generate.

7. The method for self-organizing optimization of lean production in a flexible workshop with a peak-shaving and valley-filling strategy according to claim 3 is characterized in that: The calculation formula of the congestion degree is: The formula for calculating the entropy of a population is: Where D ci represents the crowding degree of individual ci; md represents the target dimension; M represents the total number of target dimensions; f md (ci) represents the function value of individual ci on target md; f md (ci+1) represents the function value of individual ci+1 on the target md; f md (ci-1) represents the function value of individual ci-1 on the target md; H represents the entropy of the population; LC indicates chromosome length; s represents the size of the symbol set D; P gk represents the probability that the kth symbol in the symbol set D appears at the gene locus g.

8. The method for self-organizing optimization of lean production in a flexible workshop with a peak-shaving and valley-filling strategy according to claim 7 is characterized in that: The method of sorting individuals in the current population based on the improved global sorting method, performing selection, crossover and mutation according to the sorting results, and forming an initial generation of offspring population using the equipment efficiency synergy operator and the energy efficiency synergy operator includes the following steps: S31, normalize the function value data of the individual multidimensional target, and based on the normalized results, calculate the global ranking value GR(i) and the individual aggregation degree A of each individual in the population P. i And global fitness fitness(i), and arrange the individuals in the population in descending order according to the global fitness; S32, calculating the entropy of the parent population and the global fitness of the individuals in the population, and judging whether the entropy of the population satisfies the requirement of continuous G-generation decrease, if so, applying Boltzmann selection, if not or the entropy of the population does not reach G-generation, applying binary tournament selection; S33, randomly select k solutions from the parent population as the centroids of each cluster Where l = 1, 2, ..., k; S34. Divide the population into k subclusters C l , traverse each individual, calculate the similarity between the individual and each centroid, and batch the individual to the cluster where the centroid is located. If the similarity with different centroids is equal, the individual is assigned to the cluster with less than the preset number of individuals in the cluster; S35. Calculate the Euclidean distance between each individual and other individuals respectively, and take the individual with the smallest Euclidean distance variance as the new centroid of each sub-cluster S36: Determine whether the conditions are met Or the maximum number of cycles is reached, where ε represents the acceptable similarity between the new and old centroids. If yes, the algorithm ends. If no, And return to S33; S37, randomly pairing individuals from two different clusters for crossover, and selecting individuals for mutation at a preset mutation rate; S38, calculating the number of times each machine in the machine selection segment is selected for the chromosome after crossover mutation, and determining the machine with the largest number of processes to be processed and the machine with the smallest number of processes to be processed; traversing each process of the machine with the largest number of processes to be processed, and if the current process can be processed by the machine with the smallest number of processes to be processed, replacing the processing machine of the process; S39. Calculate and sort the power consumption of each process of each workpiece of the chromosome of the application equipment efficiency coordination operator, determine the process with the largest power consumption, and replace the selected machine with the currently unloaded minimum power machine from the optional machine set of the process to form an initial generation population.

9. The method for self-organizing optimization of lean production in a flexible workshop with a peak-shaving and valley-filling strategy according to claim 8 is characterized in that: The expression of the normalization process is: The calculation formula of the global ranking value GR(i) is: Individual aggregation degree A i The calculation formula is: The calculation formula of global fitness fitness(i) is: In the formula, f m (i) represents the function value of individual i at target m; f m ′(i) represents the normalized function value of individual i at target m; and Respectively represent the maximum and minimum values ​​of individual i in the current target; S ij Represents the similarity between individual i and individual j; γ represents the similarity threshold; represents the number of identical genes between individuals i and j; N represents the number of individuals in the population; ω1 and ω2 represent the weight coefficients of the global ranking value and individual aggregation degree, respectively.

10. The method for self-organizing optimization of lean production in a flexible workshop with a peak-shaving and valley-filling strategy according to claim 8, characterized in that: The individual j and each centroid The calculation formula for the similarity between is: The formula for calculating variance is: In the formula, Represents individual j and each centroid The similarity between Represents individual j and the centroid The number of identical genes; μ i represents the mean of the Euclidean distance; g indicates the gene locus in the chromosome; x ig represents the g-th gene value in individual i; x jg represents the g-th gene value in individual j; (C l ) represents the subcluster C l The number of individuals in ; s i It represents the variance of the Euclidean distance between individual i and other individuals.

Images