Method for solving panel furniture energy consumption scheduling based on improved NSGA-II algorithm
Through the improved NSGA-II algorithm, the energy consumption scheduling model is constructed, and the energy use in panel furniture production is optimized, which solves the problem of energy waste in panel furniture production and achieves more efficient energy management.
Patent Information
- Application Number
- CN202410988291.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-23
- Publication Date
- 2024-11-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
There is a problem of energy waste in the production of existing panel furniture, resulting in unnecessary energy consumption during the production process.
The improved NSGA-II algorithm is used to construct an energy consumption scheduling mathematical model that weighs the total completion time and total energy consumption. Through double-layer coding rules and variable neighborhood search, the allocation and operation order of workpieces on different machines is optimized to reduce energy consumption.
By maximizing the trade-off time and total energy consumption, energy waste in the production process is significantly reduced, avoiding falling into local optimal solutions, and improving the efficiency of energy management.
Smart Images

Figure CN118940889A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of furniture production, and in particular to a method for solving energy consumption scheduling of panel furniture based on an improved NSGA-II algorithm. Background Art
[0002] Customized panel furniture adopts a small batch and multi-variety production mode to meet customers' customization needs. With the digital transformation of intelligent manufacturing and the integration of information systems, the production planning and scheduling process of panel furniture is also constantly changing. However, the current pull production ignores the energy management of the production process, resulting in unnecessary energy waste in the production process. Therefore, it is necessary to choose a suitable energy management method to meet the sustainable development of the furniture industry. Summary of the invention
[0003] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.
[0004] Therefore, the purpose of the present invention is to provide a method for solving the energy consumption scheduling of panel furniture based on an improved NSGA-Ⅱ algorithm, by constructing a mathematical model of energy consumption scheduling that weighs the total completion time and total energy consumption, and solving its optimal solution, maximizing the total completion time and total energy consumption, and reducing unnecessary energy waste in the production process.
[0005] To solve the above technical problems, according to one aspect of the present invention, the present invention provides the following technical solutions:
[0006] A method for solving the energy consumption scheduling of panel furniture based on an improved NSGA-Ⅱ algorithm includes:
[0007] S1. Establish a mathematical model for energy consumption scheduling that balances total completion time and total energy consumption;
[0008] S2. According to the energy consumption scheduling mathematical model, relevant parameters are set and the initialization population is generated according to the double-layer rule, which is denoted as P. The parameters include the population size N, the maximum number of iterations Maxgen, and the crossover probability p. c , mutation probability P m ;
[0009] S3, perform fast non-dominated sorting on the initial population P and calculate the crowding distance;
[0010] S4, select individuals from the current population through tournament selection to generate a new population N-Size;
[0011] S5. According to the crossover probability p c , perform crossover operations on the selected individuals to generate new solutions, and then according to the mutation probability P m , perform mutation operations on the selected individuals to increase the diversity of the population;
[0012] S6. Perform non-dominated sorting on the solution set obtained after the crossover and mutation operations, update the dominant solution set, and perform variable neighborhood search on the solutions in the non-dominated solution set to seek a better solution;
[0013] S7. Repeat steps S4-S6 in sequence, and after each generation, select the best N solutions from the current population and the offspring population as the next generation population through the elite retention strategy. When the number of iterations reaches the preset maximum number of iterations Maxgen, terminate the iteration and output the final non-dominated solution set.
[0014] As a preferred solution of the method for solving the energy consumption scheduling of panel furniture based on the improved NSGA-II algorithm described in the present invention, the energy consumption scheduling mathematical model is:
[0015]
[0016] minf2=E total =PE+SE+IE+TE+AE (2)
[0017]
[0018]
[0019] Among them, formulas (1) and (2) are objective functions. Formula (1) is the optimization goal of minimizing the maximum completion time, formula (2) is the optimization goal of minimizing the total energy consumption, and formulas (3)-(8) are constraints. Formula (3) ensures that each workpiece must be assigned to a machine at each stage. Formula (4) explains the priority constraints between each workpiece operation. Formula (5) indicates that under the transportation constraint, the same machine M f The previous workpiece of the last two consecutive workpieces The completion time of workpiece S ij The relationship between the start time and the start time of each job is guaranteed by formulas (6)-(8). Each job can only be processed by one machine at a time, and each machine cannot process multiple jobs at the same time.
[0020] As a preferred solution of the method for solving the energy consumption scheduling of panel furniture based on the improved NSGA-II algorithm described in the present invention, the two-layer encoding rule includes the machine selection information of each operation and the operation sequence information on each machine, and each chromosome is composed of two layers of strings of the same length, with a length of Each integer in the machine gene string represents the serial number of the optional machine set of the current process.
[0021] As a preferred solution of the method for solving the energy consumption scheduling of panel furniture based on the improved NSGA-II algorithm described in the present invention, in step S2, the GLR method is used to initialize the population, and the ratio of global selection GS, local selection LS and random selection RS is set to 6:3:1.
[0022] As a preferred solution of the method for solving the energy consumption scheduling of panel furniture based on the improved NSGA-II algorithm described in the present invention, in step S5, if the probability p of the new population N-Size is less than p c , then proceed to the subsequent crossover step, otherwise skip the crossover step;
[0023] The crossover operation steps are as follows:
[0024] The workpiece set U1, J2, ..., J n} Randomly divided into artifact set 1 / artifact set 2;
[0025] Copy the artifacts contained in artifact set 1 from P1 to C1, and copy the artifacts contained in artifact set 1 from P2 to C2, preserving their locations;
[0026] Copy the artifacts in artifact set 2 from P2 to C1, and copy the artifacts in artifact set 1 from P1 to C2, preserving their order;
[0027] In order not to destroy the good individual information and machine order, the machine layer uses the following UX operations:
[0028] Randomly generate an integer r in the interval [1, T0];
[0029] Randomly generate r mutually exclusive integers;
[0030] According to the generated integer, copy the genes at the corresponding positions in P1 and P2 to C1 and C2, and keep their positions unchanged;
[0031] The remaining genes in P1 and P2 were copied into C2 and C1, and their positions were kept unchanged.
[0032] As a preferred solution of the method for solving the energy consumption scheduling of panel furniture based on the improved NSGA-Ⅱ algorithm described in the present invention, in step S5, if the probability p of the new population N-Size is less than P m , then mutate, otherwise skip mutation;
[0033] The mutation operation steps are as follows:
[0034] Randomly select two non-adjacent positions in the gene string as intersection points r1 and r2, where r1 < r2;
[0035] Exchange the gene string portion between the crossover points r1 and r2;
[0036] For the machine selection part, in order to ensure that the good individual ground information and machine order are not destroyed, the following mutation operation is adopted:
[0037] Randomly select l positions in the machine layer gene;
[0038] Set the machine at each location in turn to any other candidate machine for the current operation.
[0039] As a preferred solution of the method for solving the energy consumption scheduling of panel furniture based on the improved NSGA-Ⅱ algorithm described in the present invention, in step S6, the following four neighborhood structures are used to perform variable neighborhood search on the solutions in the non-dominated solution set:
[0040] Swap two position structures: randomly select two different genes in the process layer gene scheme, and then swap these two genes;
[0041] Exchange a little element structure: perform domain transformation on the adjacent operations with the longest transmission time, and according to the decoding result, replace the machine corresponding to the next operation with the machine with the shorter transmission time for processing;
[0042] Exchange two position structures: randomly select a gene position r1 in the process layer gene, and exchange the elements of the two genes before and after r1.
[0043] Exchange element structure at the process level: Randomly select two positions r1 and r2 in the solution, where r1<r2, and exchange each pair of elements between the positions r1 and r2 in the process gene. Repeat the steps until all structures between r1 and r2 are exchanged or cannot be exchanged anymore.
[0044] Compared with the prior art, the present invention has the following beneficial effects:
[0045] (1) This paper constructs a mathematical model for energy scheduling that balances total completion time and total energy consumption, and solves its optimal solution to maximize the balance between total completion time and total energy consumption, thereby reducing unnecessary energy waste in the production process.
[0046] (2) An embedded hybrid structure is adopted to combine NSGA-II and variable neighborhood search. NSGA-II provides higher quality initial solutions for variable domain search and guides the algorithm to explore better solutions, while the neighborhood structure expands the search space by comparing the dominance of new solutions with the original solutions. It can quickly obtain a good solution to the energy consumption scheduling mathematical model and avoid falling into the local optimal solution. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the present invention will be described in detail below in combination with the accompanying drawings and detailed embodiments. 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 creative labor. Among them:
[0048] Figure 1 It is a flow chart of a method for solving the energy consumption scheduling of panel furniture based on the improved NSGA-Ⅱ algorithm of the present invention;
[0049] Figure 2 This is a pseudo code of a method for solving the energy consumption scheduling of panel furniture based on an improved NSGA-Ⅱ algorithm of the present invention;
[0050] Figure 3 An example diagram of coding provided by the present invention;
[0051] Figure 4 The decoding Gantt chart provided by the present invention, wherein: Figure 4 (a) is the decoding with idle period, Figure 4 (b) is the decoding without idle time period;
[0052] Figure 5 A schematic diagram of the process layer and machine layer gene crossover operation provided by the present invention;
[0053] Figure 6 A schematic diagram of the process and machine-level gene mutation operation provided by the present invention;
[0054] Figure 7 A schematic diagram of the structure of the switching element provided by the present invention;
[0055] Figure 8 A machine layout diagram of a panel furniture workshop provided by the present invention;
[0056] Fig. 9 Horizontal trend graphs of the three indicators (a) IGD, (b) HV and (c) GD provided by the present invention;
[0057] Fig.10 A schematic diagram of the Pareto frontier obtained by the VNS-NSGA-Ⅱ algorithm provided by the present invention;
[0058] Fig.11 The best energy-saving plan Gantt chart provided by the present invention, wherein: Fig.11 (a) is the optimal energy-saving schedule with transportation constraints, Fig.11 (b) the optimal energy-saving schedule without transportation constraints;
[0059] Fig.12 The proportion of each energy consumption module in the energy saving plan provided by the present invention;
[0060] Fig.13 Comparison of the Pareto front approximations of different algorithms provided by the present invention. DETAILED DESCRIPTION
[0061] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0062] Secondly, the present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention in detail, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.
[0063] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.
[0064] The present invention provides a method for solving the energy consumption scheduling of panel furniture based on an improved NSGA-Ⅱ algorithm. By constructing an energy consumption scheduling mathematical model that weighs total completion time and total energy consumption, and solving its optimal solution, the total completion time and total energy consumption are maximized, and unnecessary energy waste in the production process is reduced.
[0065] like Figure 1 and Figure 2 As shown in the figure, the specific steps of the method for solving the energy consumption scheduling of panel furniture based on the improved NSGA-Ⅱ algorithm are as follows:
[0066] S1. Establish a mathematical model for energy consumption scheduling that balances total completion time and total energy consumption;
[0067] S2. According to the energy consumption scheduling mathematical model, relevant parameters are set and the initialization population is generated according to the double-layer rule, which is denoted as P. The parameters include the population size N, the maximum number of iterations Maxgen, and the crossover probability p. c , mutation probability P m ;
[0068] S3, perform fast non-dominated sorting and crowding distance calculation on the initial population P;
[0069] S4, select individuals from the current population through tournament selection to generate a new population N-Size;
[0070] S5. According to the crossover probability p c , perform crossover operations on the selected individuals to generate new solutions, and then according to the mutation probability Pm , perform mutation operations on the selected individuals to increase the diversity of the population;
[0071] S6. Perform non-dominated sorting on the solution set obtained after crossover and mutation operations, update the dominant solution set, and perform variable neighborhood search on the solutions in the Pareto dominated solution set to seek better solutions;
[0072] S7. Repeat steps S4-S6 in sequence, and after each generation, select the best N solutions from the current population and the offspring population as the next generation population through the elite retention strategy. When the number of iterations reaches the preset maximum number of iterations Maxgen, terminate the iteration and output the final Pareto solution set.
[0073] In step S1, a mathematical model for energy consumption scheduling that balances total completion time and total energy consumption is established as follows:
[0074] In order to balance the total energy consumption and the total completion time, a mixed integer linear programming (MILP) model is established. In order to establish the objective function, this paper gives the energy consumption module of the panel furniture workshop, and the total energy consumption E considers the following important indicators:
[0075] (1) Processing energy (PE). PE refers to the power consumed by the machine to complete the current processing task. In the actual application of panel furniture manufacturing, parallel machines in the same stage will have different processing speed settings, resulting in different completion times and processing powers. The higher the speed, the shorter the processing time and the corresponding power is greater, and vice versa. Machine M f The corresponding power pe at speed s fs , its energy consumption can be calculated by formula (1):
[0076]
[0077] (2) Setup Energy (SE). SE is the energy required for machine preparation activities, such as loading and unloading, positioning, clamping workpieces, and switching operations (such as changing tools and materials). It is related to the basic power SE required to maintain the normal operation of the machine. f and the preparation time t for each workpiece ijf related.
[0078]
[0079] (3) Idle Energy (IE). IE refers to the energy consumed when the machine is idle and waiting for the next operation to be processed. Idle time refers to the duration between two consecutive operations on the same machine, which depends on the choice of machine, the sequence of operations and the associated transportation time. IE can be obtained by formula (3):
[0080]
[0081] (4) Transport energy consumption (TE). TE refers to the energy consumption of transport resources. The transport equipment in the panel furniture workshop is mainly an electric conveyor belt used for continuous and batch transport operations (transmission operations are performed through rollers), and its motor drive system will generate energy consumption. This mainly depends on the transport time between two continuously operating workstations selected by the workpiece and the energy consumption per unit time. The calculation formula is as follows:
[0082]
[0083] (5) Auxiliary energy consumption (AE). The auxiliary energy consumption of panel furniture manufacturing mainly comes from dust removal, lighting, ventilation and air conditioning equipment. Usually, the auxiliary power is relatively stable and only depends on the completion time of the entire manufacturing process. The calculation formula is as follows
[0084]
[0085] In this paper, we build a mathematical model for HFSP to trade off total completion time and total energy consumption:
[0086]
[0087] minf2=E total =PE+SE+IE+TE+AE (7)
[0088]
[0089] Formula (6) gives the optimization goal of minimizing the maximum completion time, and formula (7) gives the optimization goal of minimizing the total energy consumption. Constraint (8) ensures that each workpiece must be assigned to a machine at each stage. Constraint (9) explains the priority constraints between each workpiece operation. Constraint (10) indicates that under the transportation constraint, the same machine M f The previous workpiece of the last two consecutive workpieces The completion time of workpiece S ij Constraints (11)-(13) ensure that each job can only be processed by one machine at a time, and each machine cannot process multiple jobs at the same time.
[0090] The specific representation of each parameter is shown in Table 1
[0091]
[0092]
[0093]
[0094] In step S2, the two-layer encoding rule contains the machine selection information of each operation and the operation sequence information on each machine. Each chromosome consists of two layers of strings of the same length. Each integer in the machine gene string represents the serial number of the optional machine set of the current process, and each integer in the operation gene string represents the workpiece number. The order of their appearance indicates the processing order between workpiece processes. Figure 3 As shown in the figure, the operation and processing machine sequence corresponding to the chromosome is O 21 (M 13 ) → O 11 (M3) → O 12 (M5) → O 22 (M4) → O 13 (M8) → O 23 (M7).
[0095] Decoding is the process of converting chromosome encoding into a scheduling solution. The present invention applies a plug-in "greedy" decoding algorithm to HFSP with transportation time to ensure that active scheduling can be generated after decoding. Assume that workpiece J i The jth operation O ij Yu S ijf Always on Machine M f The processing starts at S ijf Can be decoded as follows:
[0096] Read the genes in the operation gene string from left to right and determine the artifact J i The operation sequence is then used to find the processing machine M from the machine gene string. f .
[0097] Computation Operation O ij On machine M f The earliest start time S in ijf , which can be divided into the following three situations:
[0098] a. If process O ij It's Machine M f and workpiece J i The first operation is to set S ijf Set to 0.
[0099] b. If process O ij It's Machine M f The first operation, but not working J i The first step is to use S ijf Set to J i The sum of the completion time and the transfer time of the previous process (C i(j-1)w +T i(j-1)jwf ).
[0100] c. If process O ij Not a machine f The first operation is not workpiece J i The first process, then find machine M f All interval idle time on [IS i , IE i ], and determine the workpiece J i Is it possible to insert an idle state, if there is idle time, such as Figure 4 (a), S ijf =max(C i(j-1)w +T i(j-1)jwf , I.S. i ).otherwise, like Figure 4 (b) as shown.
[0101] In step S2, considering the quality of the initial population P0 for multi-objective scheduling, P0 is generated in the following way: Genes in the process layer gene string are randomly generated, and the genes in the machine layer are initialized using the GL method. The ratio of global selection GS, local selection LS, and random selection RS is set to 6:3:1. The initialization rules of each part are as follows:
[0102] Global Selection GS: 1. Randomly select a job from the set of jobs that have never been assigned to a machine, and assign machines to the job in order. 2. Select the machine with the smallest cumulative load for each job. 3. Repeat the above steps until all jobs are assigned.
[0103] Local selection LS: 1. Select the workpieces to be processed in order and arrange the processing machines for the selected workpieces in order. 2. Select the candidate machine with the shortest processing time for each workpiece. 3. Repeat the above steps until all the workpieces are arranged.
[0104] Random selection of RS: 1. Select the workpieces to be processed in order and arrange the processing machines for the selected workpieces in turn. 2. Randomly select a machine for each operation. 3. Repeat the operation until all jobs are arranged.
[0105] Among them, the GS and LS methods mainly consider the load of machine selection, aiming to balance the workload of the machine and improve the efficiency of the machine. The RS method ensures the diversity of the initial population. By combining these three methods, the quality of the initial solution can be improved.
[0106] In step S5, if the probability of the new population N-Size p<p c , then proceed to the subsequent crossover step, otherwise skip the crossover step;
[0107] like Figure 5 As shown in (a), the cross-operation steps at the process level are as follows:
[0108] The workpiece set {J1,J2,…,J n} Randomly divided into artifact set 1 / artifact set 2;
[0109] Copy the artifacts contained in artifact set 1 from P1 to C1, and copy the artifacts contained in artifact set 1 from P2 to C2, preserving their locations;
[0110] Copy the artifacts in artifact set 2 from P2 to C1, and copy the artifacts in artifact set 1 from P1 to C2, preserving their order;
[0111] In order not to destroy the good individual information and machine order, the machine layer uses the following uniform crossover operation, such as Figure 5 (b) as shown:
[0112] Randomly generate an integer r in the interval [1, T0];
[0113] Randomly generate r mutually exclusive integers;
[0114] According to the generated integer, copy the genes at the corresponding positions in P1 and P2 to C1 and C2, and keep their positions unchanged;
[0115] The remaining genes in P1 and P2 were copied into C2 and C1, and their positions were kept unchanged.
[0116] In step S5, if the probability p of the new population N-Size is less than P m , then mutate, otherwise skip mutation;
[0117] The process layer of the present invention adopts two-point insertion mutation operation, which can effectively improve the diversity of offspring, such as Figure 6 As shown in (a), the mutation operation steps are as follows:
[0118] Randomly select two non-adjacent positions in the gene string as intersection points r1 and r2, where r1 < r2;
[0119] Exchange the gene string portion between the crossover points r1 and r2;
[0120] For the machine selection part, in order to ensure good individual ground information and machine sequence is not destroyed, e.g. Figure 6 As shown in (b), the following mutation operation is adopted:
[0121] Randomly select l positions in the machine layer gene;
[0122] Set the machine at each location in turn to any other candidate machine for the current operation.
[0123] In step S6, the following four neighborhood structures are used to perform variable neighborhood search on the solutions in the non-dominated solution set:
[0124] Swap two position structures: randomly select two different genes in the process layer gene scheme, and then swap these two genes;
[0125] Exchange a little element structure: perform domain transformation on the adjacent operations with the longest transmission time, and according to the decoding result, replace the machine corresponding to the next operation with the machine with the shorter transmission time for processing;
[0126] Exchange two position structures: randomly select a gene position r1 in the process layer gene, and exchange the elements of the two genes before and after r1.
[0127] Exchange element structure: such as Figure 7 As shown, two positions r1 and r2 in the solution are randomly selected at the process level, where r1 < r2, and each pair of elements between the positions r1 and r2 in the process gene are exchanged, and the steps are repeated until all structures between r1 and r2 are exchanged or cannot be exchanged anymore.
[0128] The mathematical model and VNS-NSGA-II algorithm proposed in this paper were tested in HFSP of customized panel furniture. First, the algorithm parameters were calibrated through orthogonal experiments. Second, in order to help decision makers screen the optimal scheduling rules, this paper combined entropy weight TOPSIS to screen the optimal scheduling solution from the Pareto optimal solution set. Third, in order to prove the superiority of the algorithm, the VNS-NSGA-II algorithm proposed in this paper was compared with the results of MOEA / D and SPEA2 under different test instances.
[0129] 1. Panel furniture workshop environment
[0130] like Figure 8 As shown in Figure 2, it is a machine layout diagram of a panel furniture workshop. Each job has three processing stages, namely sawing, edge banding and drilling. There are parallel machines with different speeds in each stage. Usually, a large number of jobs with similar processes and colors are organized into a batch (a task) for processing. Therefore, the machine processing time in this article refers to the execution time of a task. Table 2 shows the processing time, setup time and power under each machine tool state, and all times are evenly distributed in a given interval. The transportation time between adjacent machine tools is shown in Table 3. In addition, the workpiece is transported between machine tools by conveyor belts so that it can reach the next processing machine tool as soon as possible after the previous process is completed. The unit transportation power between machines is set to 1.5kW, and the auxiliary power required to support the production environment is set to 1.2kW.
[0131] Table 2 - Basic information of each machine processing
[0132]
[0133] Table 3-Transportation time between different machines (unit: min)
[0134]
[0135]
[0136] 2. Parameter calibration
[0137] The performance of VNS-NSGA-II may be affected by related parameter settings, including: population size P, maximum number of iterations Maxgen, crossover probability P c and mutation probability P m . To ensure the rationality of parameter settings, we used the Taguchi method to measure the optimal parameter combination. Table 4 lists the parameters and the values of each level. According to the number of parameters and levels, the L16 (45) orthogonal table was selected for the experiment. In addition, the inverse generation distance (IGD), hyper volume (HV) and generation distance (GD) were used as comprehensive indicators to evaluate the convergence, diversity and distribution of the algorithm. All experimental tests were run independently 30 times. The experimental results expressed as the mean and standard deviation of IGD, HV and GD are shown in Table 5.
[0138] According to the parameter combination results in Table 5, the horizontal trend diagram of each parameter can be obtained. Fig. 9 The main effect diagram of the four parameters on the three performance indicators is also shown. According to comprehensive judgment, the best parameter combination is P = 200, Maxgen = 500, P c =0.8, P m =0.15.
[0139] Table 4 - Parameter Level Table
[0140]
[0141] Table 5 - Orthogonal experiment table
[0142]
[0143]
[0144] 3. Experimentation and Evaluation
[0145] To evaluate the energy-saving scheduling model designed in this paper, the enhanced NSGA-II is used to find the optimal or near-optimal solution. Fig.10The obtained Pareto frontier is shown in Figure 6, where the blue circle represents each feasible solution on the frontier. Table 6 shows the corresponding effective time and energy consumption values in the experimental results. It can be seen that the maximum completion time of the solution set is between 677min and 720min, and the corresponding total energy consumption values are 700.29kWh and 749.84kWh, respectively. In addition, there is a negative correlation between the two objective functions. When the time span is the smallest, the corresponding energy consumption is the largest. In order to evaluate the quality of the obtained solution set, we compare it with the results obtained in actual production (first come first served, according to the order of job number production): makespan=712min,E total =773.42kWh. Compared with the actual production plan, the proposed VNS-NSGA-II algorithm can increase the production time span by 5.17% and reduce the total energy consumption by 10.44%. Therefore, the proposed algorithm can solve this problem well.
[0146] After obtaining the optimal solution, the decision maker is still faced with the problem of how to select the most satisfactory solution from Pareto. This paper uses the TOPSIS method to select the scheduling scheme. This method can make a comprehensive decision based on the management experience of the decision maker and the information entropy of each goal, and obtain the most satisfactory low-energy and high-efficiency scheduling from Pareto.
[0147] First, according to the workshop scheduling objectives, a scheduling target set A = {A1, A2, ..., A b , ..., A n Then, according to the Pareto optimal solution obtained by VNS-NSGA-II, a scheduling solution set X = {X1, X2, ..., X f , ..., X m}. Therefore, the optimal scheduling decision matrix E = (a fb ) m×n It can be written as:
[0148] A1 … A b … A n
[0149]
[0150] Among them, X f is the fth scheduling scheme, A b is the bth scheduling object, a fb Is the scheduling plan X f The fth dispatch object value.
[0151] Scheduling optimization objectives are usually divided into benefit type (the smaller the value, the better the result) and cost type (the larger the value, the better the result). Let B1 and B2 be the indexes of the benefit target set and the cost target set respectively, then there exists B1∪B2={1,2,...,n}.
[0152] The steps of the EW-TOPSIS method are as follows:
[0153] Step 1: Establish a standardized decision matrix. The purpose of establishing a standardized decision matrix is to convert various attribute values into dimensionless attributes so that the attributes can be compared with each other. Since both construction period and energy consumption are benefit targets, they can be normalized using Formula 15, and the decision matrix E = (a fb ) m×n Converted to the standard decision matrix R = (r fb ) m×n .
[0154]
[0155] Step 2 uses the entropy weight method (EW) to calculate the weight of each target. Information entropy is used to measure disordered information. The greater the information entropy, the higher the degree of information disorder. The weight vector of the evaluation attribute is solved based on the entropy weight method. The entropy weight method eliminates the interference of human factors in the calculation of the weight of each indicator to a certain extent, making the evaluation result more realistic. The calculation steps are as follows:
[0156] Step 2.1 Normalize each target. Calculate the dimensionless attribute value pfb of the fth scheduling solution under the bth target:
[0157]
[0158] Step 2.2 Calculate the bth target E b The entropy value is calculated as (17):
[0159]
[0160] Where δ represents a constant, δ = 1 / ln m (1n is the natural logarithm), thus ensuring 0 ≤ E b ≤1.
[0161] Step 2.3 Calculate the information deviation d of the bth target b :
[0162] d b =1-E b (18)
[0163] The weight is calculated by normalizing the degree of information bias. If the decision maker has no preference between targets, the weight of target b can be obtained by formula (19):
[0164]
[0165] If the decision maker has a preference for the target set, let the subjective weight be γ b , then the above formula is further modified to obtain the comprehensive weight of attribute b:
[0166]
[0167] Step 3: Establish a weighted standardized decision matrix. Use the entropy weight method to obtain the target weight vector W = {ω1, ω2, ..., ω b , ..., ω n}, And construct the standardized decision matrix Q as:
[0168] Q=(q fb ) m×n =(ω b r fb ) m×n (twenty one)
[0169] Step 4: Determine the positive and negative ideal solutions. The formula is as follows:
[0170]
[0171]
[0172] Among them, X + is a positive ideal solution, and its various target values all reach the optimal value among the evaluation targets; X- is a negative ideal solution, and its various target values all reach the worst value among the evaluation targets.
[0173] Step 5 uses the n-dimensional Euclidean distance to calculate the distance between each target and the positive ideal solution, which can be obtained by formula (24):
[0174]
[0175] Similarly, the distance from each scheduling plan to the negative ideal solution is:
[0176]
[0177] Step 6: Calculate the closeness of each scheduling schedule to the positive ideal solution. f With the ideal solution + The proximity of is defined as follows:
[0178]
[0179] If X f =X + , then the relative closeness C f=1; if X f =X + , then the relative closeness C f = 0. When the relative proximity C f When it converges to 1, the properties of the scheduling schedule are close to the positive ideal solution.
[0180] Step 7 Determine the best scheduling plan. Press C f The values of are sorted, and the solution closest to the ideal solution is defined as the Pareto optimal solution.
[0181] Assuming that the decision maker does not set preferences, the comprehensive scores of the Pareto solutions are shown in Table 6. It can be seen that maxC f The value of is 0.6620. Therefore, Scheme 16 is the most satisfactory energy-saving scheme. In order to evaluate the impact of transportation on the objective function, this paper analyzes the Gantt chart of the optimal energy-saving scheme, such as Fig.11 (a) is shown. It is not difficult to find that when the job meets the processing conditions, it will be processed on the idle machine first. At the same time, the transportation time will also affect the maximum time and energy consumption to a certain extent. In addition, from the perspective of energy saving, we can reasonably adjust the earliest start time of the third-level machines (M6 and M7) to reduce idle energy consumption and ensure that the total completion time remains unchanged.
[0182] Table 6 - Pareto solution set and total score
[0183]
[0184] In order to observe the distribution of each energy consumption module more intuitively, Fig.12 The proportion of each energy consumption module in the energy saving plan is shown. It can be seen that PE (95.46%) is the main source of energy consumption in the workshop, which plays a vital role in reducing total energy consumption and achieving green manufacturing. AE (1.92%) and TE (1.73%) also have a great impact on total energy consumption, requiring managers to optimize and control total processing time and transportation scheduling time.
[0185] 4. Comparison between VNS-NSGA-Ⅱ and other algorithms
[0186] Two well-known multi-objective evolutionary algorithms (including MOEA / D and SPEA2) are used for comparison to demonstrate the effectiveness of VNS-NSGA-Ⅱ. For fair comparison, P and Maxgen of MOEA / D and SPEA2 are set to 200 and 500 respectively, and P c =0.8 and P m=0.15. The neighborhood size of MOEA / D is 40. The archive size of SPEA2 is 100. At the same time, all algorithms use the same encoding and decoding as well as crossover and mutation methods. In order to verify the effectiveness of the VNS-NSGA-II algorithm, we considered a larger scale problem. As shown in Table 7, the three algorithms were tested at six scales. Each instance was run independently 30 times. The comprehensive indicators IGD, HV, and GD were also used. Since the values of the two objectives vary greatly at different scales under large-scale problems, the coefficient of variation (CV) and the average value are used to measure these three indicators.
[0187] Table 7 - Test question parameters
[0188]
[0189] As can be seen from Table 8, under different problem scales, VNS-NSGA-II is superior to MOEA / D and SPEA2 in the three comprehensive indicators of IGD, HV and GD, which shows that VNS-NSGA-II can obtain Pareto solutions with better convergence and wider distribution of solutions in this case. On problem 20×7, the results of the three algorithms are closest, but when the scale becomes larger, VNS-NSGA-II shows more advantages in achieving compromise solutions and has good diffusion and convergence performance. In problem 320×14, VNS-NSGA-II covers almost all Pareto optimal boundaries and obtains the lowest IGD and GD values. Fig.13 The Pareto optimal frontiers of all algorithms in 6 instances are shown. Obviously, the proposed VNS-NSGA-II produces more Pareto optimal solutions compared to the other algorithms. This means that the decision maker has more candidate Pareto solutions to choose from. Therefore, it can be concluded that VNS-NSGA-II is better than the compared algorithms.
[0190] Table 8 - Statistical results of the three main indicators of all algorithms
[0191]
[0192]
[0193] Although the present invention has been described above with reference to the embodiments, various modifications may be made thereto and parts thereof may be replaced by equivalents without departing from the scope of the present invention. In particular, as long as there is no structural conflict, the various features in the embodiments disclosed in the present invention may be used in combination with each other in any manner, and the fact that these combinations are not exhaustively described in this specification is only for the sake of omitting space and saving resources. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.
Claims
1. A method for solving the energy consumption scheduling of panel furniture based on the improved NSGA-Ⅱ algorithm. It is characterized in that include: S1. Establish a mathematical model for energy consumption scheduling that balances total completion time and total energy consumption; S2. According to the energy consumption scheduling mathematical model, relevant parameters are set, and the initialization population is generated according to the double-layer rule, denoted as P. The parameters include the population size N, the maximum number of iterations Maxgen, and the crossover probability p. c , mutation probability P m ; S3, perform fast non-dominated sorting on the initial population P and calculate the crowding distance; S4, select individuals from the current population through tournament selection to generate a new population N-Size; S5. According to the crossover probability p c , perform a crossover operation on the selected individuals to generate a new solution, and then according to the mutation probability P m , perform mutation operations on the selected individuals to increase the diversity of the population; S6. Perform non-dominated sorting on the solution set obtained after the crossover and mutation operations, update the Pareto solution set, and perform variable neighborhood search on the solutions in the non-dominated solution set to find a better solution; S7. Repeat steps S4-S6 in sequence, and after each generation, select the best N solutions from the current population and the offspring population as the next generation population through the elite retention strategy. When the number of iterations reaches the preset maximum number of iterations Maxgen, terminate the iteration and output the final non-dominated solution set.
2. According to the method for solving the energy consumption scheduling of panel furniture based on the improved NSGA-Ⅱ algorithm described in claim 1, It is characterized in that The energy consumption scheduling mathematical model is: MINFI 2 =E total =PE+SE+IE+TE+AE (2) Among them, formulas (1) and (2) are objective functions. Formula (1) is the optimization goal of minimizing the maximum completion time, formula (2) is the optimization goal of minimizing the total energy consumption, and formulas (3)-(8) are constraints. Formula (3) ensures that each workpiece must be assigned to a machine at each stage. Formula (4) explains the priority constraints between each workpiece operation. Formula (5) indicates that under the transportation constraint, the same machine M f The previous workpiece C of the last two consecutive workpieces i-j- The completion time of workpiece S ij The relationship between the start time and the start time of each job is guaranteed by formulas (6)-(8). Each job can only be processed by one machine at a time, and each machine cannot process multiple jobs at the same time.
3. According to the method for solving the energy consumption scheduling of panel furniture based on the improved NSGA-Ⅱ algorithm according to claim 1, It is characterized in that The two-layer encoding rule contains the machine selection information of each operation and the operation sequence information on each machine. Each chromosome consists of two layers of strings of the same length. Each integer in the machine gene string represents the serial number of the optional machine set of the current process.
4. According to the method for solving the energy consumption scheduling of panel furniture based on the improved NSGA-Ⅱ algorithm according to claim 1, It is characterized in that In step S2, the GLR method is used to initialize the population, and the ratio of global selection GS, local selection LS and random selection RS is set to 6:3:
1.
5. According to the method for solving the energy consumption scheduling of panel furniture based on the improved NSGA-Ⅱ algorithm according to claim 1, It is characterized in that In step S5, if the probability of the new population N-Size p<p c , then proceed to the subsequent crossover step, otherwise skip the crossover step; The crossover operation steps are as follows: The artifact set {J 1 ,J 2 ,…,J n } Randomly divided into artifact set 1 / artifact set 2; Remove the artifacts contained in artifact set 1 from P 1 Copy to C 1 , and remove the artifacts contained in artifact set 1 from P 2 Copy to C 2 , retaining their position; Remove the artifacts in artifact set 2 from P 2 Copy to C 1 , and move the artifacts in artifact set 1 from P 1 Copy to C 2 , preserving their order; In order not to destroy the good individual information and machine order, the machine layer uses the following uniform crossover operation: In the interval [1,T 0 ] randomly generates an integer r; Randomly generate r mutually exclusive integers; According to the generated integer, P 1 and P 2 The gene at the corresponding position in C 1 and C 2 and keep its position unchanged; P 1 and P 2 The remaining genes in the 2 and C 1 and keep their positions unchanged.
6. According to the method for solving the energy consumption scheduling of panel furniture based on the improved NSGA-Ⅱ algorithm according to claim 1, It is characterized in that In step S5, if the probability p of the new population N-Size is less than P m , then mutate, otherwise skip mutation; For the process part, the variation operation steps are as follows: Randomly select two non-adjacent positions in the gene string as the intersection point r 1 and r 2 , where r 1 <r 2 ; Exchange intersection r 1 and r 2 The part of the gene string between; For the machine selection part, in order to ensure that the good individual ground information and machine order are not destroyed, the following mutation operation is adopted: Randomly select l positions in the machine layer gene; Set the machine at each location in turn to any other candidate machine for the current operation.
7. A method for solving the energy consumption scheduling of panel furniture based on an improved NSGA-II algorithm according to claim 1, It is characterized in that In step S6, the following four neighborhood structures are used to perform variable neighborhood search on the solutions in the non-dominated solution set: Swap two position structures: randomly select two different genes in the process layer gene scheme, and then swap these two genes; Exchange a little element structure: transform the adjacent operations with the longest transmission time, and according to the decoding result, replace the machine corresponding to the next operation with the machine with the shorter transmission time for processing; Exchange two position structures: randomly select a gene position r in the process layer gene 1 , exchange r 1 Elements of two genes before and after. Swap element structure: Randomly select two positions r in the process layer solution 1 and r 2 , where r 1 <r 2 , r in the exchange process gene 1 and r 2 Repeat steps for each pair of elements between positions until r 1 and r 2 All structures between them have been exchanged or can no longer be exchanged.
Citation Information
Patent Citations
Method for solving shop energy-saving scheduling problem based on improved NSGA-II algorithm
CN110543151A
Double-resource flexible job shop scheduling method for aerospace complex components
CN116757411A
Flexible job shop scheduling method and device based on double-population hybrid genetic algorithm
CN116859869A
Cited By
Intelligent production unit layout and energy consumption optimization method and system
CN119575901A
Intelligent production unit layout and energy consumption optimization method and system
CN119575901B
Local power grid pumped storage capacity configuration method considering cost dredging
CN119834371A
