A demand response scheduling method for flexible manufacturing workshop based on hierarchical game

Abstract

The present invention relates to a flexible manufacturing workshop demand response scheduling method based on hierarchical game, which belongs to the technical field of production scheduling. First, a flexible manufacturing workshop production scheduling control model based on time-of-use electricity price is constructed, and then, based on the NSGA-II algorithm, the adversarial learning strategy and the adaptive crossover and mutation probability strategy are introduced for the convergence and diversity problems of the algorithm; for the machine load balancing problem, the balanced machine load strategy is adopted. The present invention is based on the production scheduling of flexible manufacturing workshops under time-of-use electricity price and demand response, which plays an important role in regulating the load of the power grid and improving the load rate and operation stability of the power system; through production scheduling, the energy consumption cost of flexible manufacturing production is also effectively reduced.

Description

Technical Field

[0001] The invention relates to a flexible manufacturing workshop demand response scheduling method based on hierarchical game, belonging to the technical field of production scheduling. Background Art

[0002] In recent years, most of the research work on demand response has focused on residential and commercial users. However, 42.6% of global electricity consumption is currently caused by industrial production, so demand response in the industrial sector is particularly important for reducing peak demand on the power grid. Currently, most research focuses on continuous production processes and lacks a general demand response model for discrete manufacturing. The main reason is that the complexity of discrete models makes it difficult to establish a specific general mathematical model.

[0003] Due to the flexibility of flexible manufacturing, flexible manufacturing workshops are considered as industrial users to participate in the dispatching process of the power grid, and at the same time, the production scheduling requirements of flexible manufacturing workshops need to be met. In addition, due to the complexity of discrete production, flexible manufacturing workshops need to consider the correspondence between orders, processes and machines, which is a typical NP-hard problem. Studying production scheduling based on electricity prices means studying the matching correspondence between processes, machines and time slots, making the problem more complicated. Therefore, how to choose a suitable algorithm to solve the production scheduling becomes crucial. Summary of the invention

[0004] The purpose of the present invention is to provide a demand response scheduling method for flexible manufacturing workshops based on hierarchical game, which plays an important role in regulating the load of the power grid, improving the load rate and operation stability of the power system, and can also effectively reduce the energy consumption cost of flexible manufacturing production.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] A flexible manufacturing workshop demand response scheduling method based on hierarchical game includes the following steps:

[0007] S1. Establish a flexible workshop production scheduling and control model based on time-of-use electricity prices, taking the maximum profit of the power company and the maximum benefit of industrial users as the objective function of the model;

[0008] S2. Determine the scheduling constraints, including process constraints, machine constraints, time constraints, and total power upper limit constraints;

[0009] S3. Obtain the electricity price announced by the power company;

[0010] S4, initialize to obtain the parent population;

[0011] S5, obtaining the offspring population according to the parent population;

[0012] S6, select a new parent population;

[0013] S7, if the termination condition is met, that is, the number of iterations is greater than the set value, then the process arrangement, machine selection, start time sequence of each process and total energy consumption value of each time period corresponding to the optimal individual are output and go to step S8, otherwise go to step S5;

[0014] S8. Calculate the optimal value of the utility function energy consumption of the power company according to the demand response scheduling strategy. If the energy consumption value of the industrial user at the current electricity price reaches the optimal value of the utility function energy consumption of the power company or the number of iterations reaches the set value, the maximum profit of the power company and the maximum benefit of the industrial user will be calculated, and the maximum profit of the power company and the maximum benefit of the industrial user will be output, and the program ends; if the energy consumption value of the industrial user at the current electricity price does not reach the optimal value of the utility function energy consumption of the power company and the number of iterations does not reach the set value, go to step S9;

[0015] S9. Iterate the electricity price according to the demand response scheduling strategy and go to step S3.

[0016] A further improvement of the technical solution of the present invention is that the maximum profit of the power company is:

[0017]

[0018] Where W represents the maximum profit of the power company, p(t) represents the electricity price announced by the power company to industrial users at each time, e(t) represents the energy consumption required for production of industrial users in each time slot, and a and b are the profit coefficients of the power company.

[0019] A further improvement of the technical solution of the present invention is that the maximum benefit of the industrial user is:

[0020] π=max(log(α t +β t ·e(t))-p(t)·e(t)) (2)

[0021]

[0022] Among them, π is the maximum total revenue of producing the order; p(t) represents the electricity price announced by the power company to industrial users at each time; e(t) represents the energy consumption required for production of industrial users in each time slot; α t , β t is the industrial user profit coefficient; j is the workpiece number; D is the total number of orders; h is the process number; H j For order d j The number of processes; i is the machine number; m jh Process d jhThe number of machines that can be selected for processing; ijh Process d jh On machine m i Energy consumption required for processing; when process d jh On machine m i Processing time x ijh The value is 1, otherwise x ijh The value is 0; when process d jh At time t, the processing time z jht The value is 1, otherwise z jht The value is 0.

[0023] A further improvement of the technical solution of the present invention is that the specific steps of step S4 are:

[0024] S4.1. Initialize the population and use a two-segment encoding method to represent the chromosome information. The first segment is the process code OS, which is used to determine the processing order of the workpiece; the second segment is the machine code string MS, which is used to indicate the machine assigned to the process. A population with a population size of N is randomly generated.

[0025] S4.2, using the adversarial learning strategy to generate an opposite population;

[0026] S4.3, using the balanced machine load strategy to ensure that the loads of all machines in the two populations are as balanced as possible;

[0027] S4.4, merge the initial population with the population after its adversarial learning, and the population size is 2N;

[0028] S4.5, perform fast non-dominated sorting and crowding calculation on the merged population;

[0029] S4.6. The parent population is selected using tournament and elite selection strategies, and the population size is N.

[0030] A further improvement of the technical solution of the present invention is that the specific steps of step S5 are:

[0031] S5.1, calculate the adaptive crossover,mutation probability;

[0032] S5.2, perform selection, crossover, and mutation operations to obtain a progeny population with a population size of N;

[0033] S5.3, using the adversarial learning strategy, conduct adversarial learning on the offspring population to produce an opposite population;

[0034] S5.4, implement a balanced machine load strategy in the offspring population and its opposite population to ensure that the loads of all machines in the two populations are as balanced as possible;

[0035] S5.5, merge the offspring population with its opposite population after oppositional learning, and the population size is 2N;

[0036] S5.6, perform fast non-dominated sorting and crowding calculation on the merged population;

[0037] S5.7. Use tournament and elite selection strategies to select a new offspring population with a population size of N.

[0038] A further improvement of the technical solution of the present invention is that the specific steps of step S6 are:

[0039] S6.1. Merge the newly obtained parent population and offspring population, and the population size is 2N;

[0040] S6.2, perform fast non-dominated sorting and crowding calculation on the population after the merger of the parent and child generations;

[0041] S6.3. Use the tournament and elite selection strategies to select a new parent population with a population size of N.

[0042] A further improvement of the technical solution of the present invention is that the specific steps of step S5.1 are:

[0043] First, select the maximum value (CP) of the crossover probability (CP) and the mutation probability (MP) max and MP max ) and minimum value (CP min and MP min ); Set Fitness i is the relative fitness value of the i-th individual, Fitness max and Fitness min are the maximum relative fitness value and the minimum relative fitness value in the population respectively; then, the crossover probability of the i-th individual is calculated as:

[0044]

[0045] Among them, ε is a very small real number to prevent the denominator from being zero; the mutation probability calculation method of the i-th individual is:

[0046]

[0047] Among them, set CP max =0.9, CP min =0.6, MP max =0.25, MP min = 0.06, suppose that after fast non-dominated sorting, the number of assigned levels is R, the population size is Size_pop, and the number of individuals in each level is f i(where i = 0, 1, ..., R-1); let the relative fitness value be Ra_Fitness = Size_pop, then the relative fitness value of each individual is Fitness i The calculation is as follows:

[0048] Ra_Fitness=Ra_Fitness-Size_pop / R / f i +ε (14)

[0049] Fitness i =Ra_Fitness (15)

[0050] According to the order of i=0, 1, ..., R-1, the relative fitness value of each individual at each level is calculated in turn. i .

[0051] Due to the adoption of the above technical solution, the technical effects achieved by the present invention are as follows:

[0052] The present invention is based on flexible manufacturing workshop production scheduling under time-of-use electricity prices and demand response, which plays an important role in regulating the power grid load and improving the load rate and operation stability of the power system; through production scheduling, the energy consumption cost of flexible manufacturing production is also effectively reduced.

[0053] In order to improve the convergence and diversity of the NSGA-II algorithm, the present invention introduces an adversarial learning strategy and an adaptive crossover and mutation probability strategy. Aiming at the machine load balancing problem, a balanced machine load strategy is adopted to make the loads of all machines as balanced as possible, effectively reducing the energy consumption cost of flexible manufacturing production. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 is a flow chart of the present invention;

[0055] Figure 2 is a flow chart of the application of the machine load balancing strategy of the present invention;

[0056] Figure 3 It is a schematic diagram of the distance between individuals of the same level in the present invention;

[0057] Figure 4 This is a second flow chart of the application of the machine load balancing strategy of the present invention. DETAILED DESCRIPTION

[0058] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0059] The present invention provides a flexible manufacturing workshop demand response scheduling method based on hierarchical game, and builds a flexible manufacturing workshop production scheduling control model based on time-of-use electricity price on the traditional job shop scheduling model; based on the NSGA-II algorithm, the adversarial learning strategy and adaptive crossover and mutation probability strategy are introduced to address the convergence and diversity issues of the algorithm, and the balanced machine load strategy is adopted to address the machine load balancing problem. The specific process is as follows Figure 1 As shown, the following steps are included:

[0060] S1. Establish a flexible workshop production scheduling control model (FJSP model) based on time-of-use electricity prices, and take the maximum profit of the power company and the maximum benefit of industrial users as the objective function of the model

[0061] Assume that there is a set of n independent jobs and a set of m independent machines. Each job has a continuous sequence of operations that are processed in sequence according to the operation priority constraints. Each process can be processed on different machines, and the processing time on different machines is different. The maximum profit of the power company and the maximum benefit of the industrial user are taken as the objective function of the model and described as follows by mathematical functions. Table 1 defines and describes the main variable symbols involved in the FJSP model and the constraints and evaluation indicators listed below:

[0062] Table 1 Variable description

[0063]

[0064]

[0065] The maximum profit of the power company is expressed as:

[0066]

[0067] Where W represents the maximum profit of the power company, p(t) represents the electricity price announced by the power company to industrial users at each time. Since the power company considers the problem of maximizing its revenue in each time slot {t, t+1}, the value of electricity in each time slot is not necessarily equal; e(t) represents the energy consumption required for production of industrial users in each time slot, and a and b are the revenue coefficients of the power company.

[0068] The maximum benefit for industrial users is expressed as:

[0069] π=max(log(α t +β t ·e(t))-p(t)·e(t)) (2)

[0070] in:

[0071]

[0072] After establishing the dynamic multi-objective flexible job shop scheduling model, appropriate initial parameters are selected, the population size sizepop is set to 100, and the maximum number of algorithm iterations c is set to 100.

[0073] S2. Determine the scheduling constraints, including process constraints, machine constraints, time constraints, and total power upper limit constraints

[0074] The factory production scheduling problem is a constrained optimization problem with the following constraints:

[0075] (1) At the same time, only one machine can be selected for each process:

[0076]

[0077] Given a 0-1 variable to determine which machine each process is performed on: when x = 0, it indicates that the process of the workpiece is not performed on machine h, and when x = 1, it indicates that the process of the workpiece is performed on machine h. At a point in time, each process of each workpiece can only select one machine for processing until this process is completed, and when this process is completed, it will not be processed a second time.

[0078] (2) Given a 0-1 variable to determine whether a process of a workpiece is in the processing state at a certain moment, the sum of the values ​​of this variable at each time point in the total processing time is equal to the processing time of the process.

[0079]

[0080] (3) The start time of the subsequent process cannot be earlier than the end time of the previous process.

[0081] b jh +x ijh ·t ijh <=f jh (6)

[0082] f jh <=b j(h+1) (7)

[0083] The production process of each workpiece is strictly carried out in a certain order, and the subsequent process cannot be produced in advance when the previous process is not completed. Therefore, the production time of the process must be restricted in order, and equations (6) and (7) together constitute this constraint.

[0084] (4) A machine can only process one process at a time.

[0085] b jh +t ijh <=b kl+L·(1-y ijhkl ) (8)

[0086] f kl <=b j(h+1) +L·(1-y iklj(h+1) ) (9)

[0087] When each machine starts processing a certain process, it cannot stop processing or switch to other processes before the process is completed. Therefore, a 0-1 variable is given to determine the order between two different processes. When y=1, it means that process d jh Before d kl On machine m i Processing, otherwise, y=0.

[0088] (5) Upper limit constraint on the total power of each process.

[0089]

[0090] The amount of electricity purchased from the power company is subject to physical limitations such as distribution lines and transformer ratings, so the total power produced by each process is less than or equal to the maximum power purchased from the grid.

[0091] S3. Obtain the (initial) electricity price announced by the power company

[0092] The time window T (considering T = 24h) is divided into three different time periods, namely 0 to 8, 9 to 16, and 17 to 24, namely P1, P2, and P3 (the initial electricity price is the same for the three time periods). Among them, the values ​​of the power company's profit coefficients a and b in each time period are shown in Table 2:

[0093] Table 2a, b values ​​for each time period

[0094]

[0095] S4. Initialize and obtain the parent population

[0096] S4.1. Initialize the population and use a two-segment encoding method to represent chromosome information. The first segment is the process code OS, which is used to determine the processing order of the workpiece; the second segment is the machine code string MS, which is used to indicate the machine to which the process is assigned. The population size is N.

[0097] Initially, a population with a fixed value is randomly generated. In this population, each individual is a chromosome. Each chromosome contains two parts of information: one is a randomly generated process selection sequence, which must ensure that each process is in the sequence; the other is a randomly generated machine selection sequence, which must ensure that each process corresponds to a machine.

[0098] S4.2. Use the adversarial learning strategy to generate an opposite population

[0099] Using the adversarial learning strategy, the greedy opposite path of each individual process sequence is calculated, an opposite process sequence is created based on the path, and the machine sequence is adjusted accordingly to generate an opposite population.

[0100] The opposition method for discrete domain problems adopts open path opposition. Open paths refer to paths that complete them when we reach the last vertex on the path. The goal of open path opposition is to maximize the total proximity in the path by dispersing adjacent nodes. Note that calculating the exact opposite situation is itself a combinatorial problem, so we define a greedy opposite path with the open path greedy reverse algorithm. The pseudo code of the algorithm solution process is shown in Table 3, and an opposite population can be created based on the path to save processing time. For example, a four-node example (1 2 34), using the open path greedy reverse algorithm, calculates its greedy opposite path as (1 4 2 3).

[0101] Table 3 Pseudocode of the solution process of the open path greedy reverse algorithm

[0102]

[0103] S4.3. Use the balanced machine load strategy to ensure that the loads of all machines in the two populations are as balanced as possible.

[0104] With the balanced machine load strategy, the maximum loaded machine is first found, and the operations on the maximum loaded machine are redistributed to reduce the load on the maximum loaded machine, ensuring that the loads of all machines are as balanced as possible.

[0105] The machine load balancing strategy used includes two practical application scenarios. Here is the first practical application scenario. When the search is in the initialization stage, after the initialization is completed to the initial position of each particle, the maximum load machine can be found by decoding the original position vector of the initialized particle, and then these operations on the maximum load machine are randomly redistributed once (including the current maximum load machine) to reduce its load. The flowchart of the application process is shown in Figure 2 shown.

[0106] S4.4. Merge the initial population with the population after its adversarial learning, and the population size is 2N.

[0107] S4.5. Fast non-dominated sorting and crowding calculation of the merged population

[0108] The initial population is merged with the population after its adversarial learning to obtain a population R of size 2N t At this time, you need to keep Rt The best N members in the next generation are first sorted using a fast non-dominated sorting algorithm. t Divide into several non-dominated layers such as F1, F2, F3, and then merge all individuals in each non-dominated layer from top to bottom until a population S with a size just exceeding or exactly equal to N is generated t , assuming S t =F1∪F2∪...∪F v , where F v is the last acceptable critical layer. At this time, if S t If the scale is exactly N, proceed to the next step, otherwise discard F v Some individuals in the t The size is N. Then calculate the crowding degree of each individual, in F v In the process, the individuals with small crowding degree are discarded. Figure 3 This is a schematic diagram of the distance between individuals of the same level. The crowding calculation formula is:

[0109]

[0110] S4.6. Use tournament and elite selection strategies to select the parent population, the population size is N

[0111] The roulette strategy and the elite strategy are combined to select the next generation of excellent individuals. The selection of excellent individuals depends on the fitness value under each strategy. The elite strategy can keep the direction of population evolution optimal, while the roulette strategy can prevent the optimization process from falling into local optimization. The combination of the two can effectively accelerate the iteration of the population and obtain the optimal solution.

[0112] S5. Obtain the offspring population based on the parent population

[0113] S5.1. Calculate adaptive crossover and mutation probabilities

[0114] First, select the maximum value (CP) of the crossover probability (CP) and the mutation probability (MP) max and MP max ) and minimum value (CP min and MP min ); Set Fitness i is the relative fitness value of the i-th individual, Fitness max and Fitness min are the maximum relative fitness value and the minimum relative fitness value in the population respectively; then, the crossover probability of the i-th individual is calculated as:

[0115]

[0116] Among them, ε is a very small real number that can prevent the denominator from being zero; the mutation probability calculation method of the i-th individual is:

[0117]

[0118] Among them, set CP max =0.9, CP min =0.6, MP max =0.25, MP min = 0.06, suppose that after fast non-dominated sorting, the number of assigned levels is R, the population size is Size_pop, and the number of individuals in each level is f i (where i = 0, 1, ..., R-1); first, set the relative fitness value to Ra_Fitness = Size_pop, then the relative fitness value Fitness of each individual i The calculation is as follows:

[0119] Ra_Fitness=Ra_Fitness-Size_pop / R / f i +ε (14)

[0120] Fitness i =Ra_Fitness (15)

[0121] The relative fitness value of each individual at each level is calculated in the order of i=0, 1, ..., R-1, as shown in equations (12) and (13). If an individual has a higher fitness than other individuals in the group, then it will have a smaller CP i and MP i Value, that is, it is difficult for crossover and mutation to occur, and such individuals are directly retained to the next generation.

[0122] S5.2. Perform selection, crossover, and mutation operations to obtain a progeny population with a population size of N.

[0123] Crossover operation is the main means of generating new individuals and determines the global search capability of the algorithm. Before crossover, the individuals in the population must be randomly paired, and then the two paired chromosomes are allowed to exchange some genes to generate two new individuals.

[0124] Assume that P1 and P2 are two paired parent chromosomes, and two daughter chromosomes C1 and C2 are generated through crossover. The specific operation process is as follows:

[0125] 1) All processes of an individual are randomly divided into two parts, J1 and J2;

[0126] 2) Copy the process P1 contained in J1 to C1, and copy the process P2 contained in J2 to C2, and keep the positions unchanged;

[0127] 3) Copy the process P1 included in J1 to C2, and copy the process P2 included in J2 to C1, and keep the order unchanged;

[0128] Mutation operation is an auxiliary means to generate new individuals. It only operates on genes at individual positions of chromosomes, thus determining the local search capability of the algorithm. For the process sorting part, the exchange mutation operator is used, that is, genes at two positions are randomly selected for exchange.

[0129] S5.3. Use the adversarial learning strategy to conduct adversarial learning on the offspring population to produce an opposite population.

[0130] S5.4. Implement a balanced machine load strategy in the offspring population and its opposite population to ensure that the loads of all machines in the two populations are as balanced as possible.

[0131] The balanced machine load strategy is implemented in the child population and its opposite population. First, the machine with the minimum load is found, and the process of the candidate machine set containing the current minimum load machine is temporarily assigned to the current minimum load machine to help further balance the load of other machines. This is the second practical application of the balanced machine load strategy. During the iteration process, once the candidate machine set contains the current minimum load machine, the operation is temporarily reallocated to the minimum load machine to help further balance the load of other machines. The flowchart of the application process is as follows Figure 4 shown.

[0132] S5.5. Merge the offspring population with its opposite population after oppositional learning, and the population size is 2N.

[0133] S5.6. Perform fast non-dominated sorting and crowding calculation on the merged population.

[0134] S5.7. Use tournament and elite selection strategies to select a new offspring population with a population size of N.

[0135] S6. Select a new parent population

[0136] S6.1. Merge the newly obtained parent population and offspring population, and the population size is 2N.

[0137] S6.2. Perform fast non-dominated sorting and crowding calculation on the population after the merger of the parent and child generations.

[0138] S6.3. Use the tournament and elite selection strategies to select a new parent population (the parent population for the next iteration) with a population size of N.

[0139] S7. If the termination condition is met, that is, the number of iterations is greater than the set value, the process arrangement, machine selection, start time sequence of each process and total energy consumption value of each time period corresponding to the optimal individual are output and go to step S8, otherwise go to step S5.

[0140] S8. According to the demand response scheduling strategy, the current electricity price of each time period is substituted into formula (16) to calculate the optimal value of the utility function energy consumption of the power company. If the energy consumption value of the industrial user at the current electricity price reaches the optimal value of the utility function energy consumption of the power company or the number of iterations reaches the set value, the maximum profit of the power company is calculated according to formula (1), and the maximum benefit of the industrial user is calculated according to formula (2). The maximum benefit of the industrial user and the maximum profit of the power company are output, and the program ends. If the energy consumption value of the industrial user at the current electricity price does not reach the optimal value of the utility function energy consumption of the power company and the number of iterations does not reach the set value (100), go to step S9.

[0141] The optimal value of energy consumption of the utility function of the power company is calculated as follows:

[0142]

[0143] Among them, x=1,2,3 represent three time periods, b1=0.001, c1=0.00015.

[0144] S9. Iterate the electricity price according to the demand response scheduling strategy and go to step S3.

[0145] The demand response scheduling strategy is as follows: First, the power company announces the initial electricity price, and the industrial users schedule production according to this initial electricity price. The energy consumption value of each time period is obtained by scheduling production. The current electricity price of each time period is substituted into formula (16) to calculate the optimal value of the utility function energy consumption of the power company. If the energy consumption value of the industrial user at the current electricity price reaches the optimal value of the utility function energy consumption of the power company or the number of iterations reaches the set value, the maximum profit of the power company will be calculated according to formula (1), and the maximum profit of the industrial user will be calculated according to formula (2). The maximum profit of the industrial user and the maximum profit of the power company are output, and the program ends. If the energy consumption value of the industrial user at the current electricity price does not reach the optimal value of the utility function energy consumption of the power company and the number of iterations does not reach the set value (100), the price iteration is used to adjust its own price. The price iteration equation is:

[0146]

[0147] In the formula, δ>0 represents the price adjustment step of the power company, E represents the total energy consumption of industrial users in each iteration, It represents the optimal energy consumption value of the utility function of the power company under the current electricity price. Industrial users get updated electricity prices, and then schedule production, iterating in sequence until both industrial users and power companies reach Nash equilibrium and output the optimal result.

Claims

1. A flexible manufacturing workshop demand response scheduling method based on hierarchical game, characterized in that The following steps are involved: S1. Establish a flexible workshop production scheduling and control model based on time-of-use electricity prices, taking the maximum profit of the power company and the maximum benefit of industrial users as the objective function of the model; S2. Determine the scheduling constraints, including process constraints, machine constraints, time constraints, and total power upper limit constraints; S3. Obtain the electricity price announced by the power company; S4, initialize to obtain the parent population; S5, obtaining the offspring population according to the parent population; S6, select a new parent population; S7, if the termination condition is met, that is, the number of iterations is greater than the set value, then the process arrangement, machine selection, start time sequence of each process and total energy consumption value of each time period corresponding to the optimal individual are output and go to step S8, otherwise go to step S5; S8. According to the demand response dispatch strategy, the optimal value of the utility function energy consumption of the power company is calculated. The optimal value of the utility function energy consumption of the power company is calculated as follows: Among them, x=1,2,3 represent three time periods, b1=0.001, c1=0.00015, represents the optimal value of energy consumption of the utility function of the power company; If the energy consumption value of the industrial user at the current electricity price reaches the optimal energy consumption value of the utility function of the power company or the number of iterations reaches the set value, the maximum profit of the power company and the maximum benefit of the industrial user will be calculated, and the maximum profit of the power company and the maximum benefit of the industrial user will be output, and the program ends; if the energy consumption value of the industrial user at the current electricity price does not reach the optimal energy consumption value of the utility function of the power company and the number of iterations does not reach the set value, go to step S9; S9. Iterate the electricity price according to the demand response scheduling strategy and go to step S3.

2. According to claim 1, a flexible manufacturing workshop demand response scheduling method based on hierarchical game is characterized in that: The maximum profit of the power company is: Where W represents the maximum profit of the power company, p(t) represents the electricity price announced by the power company to industrial users at each time, e(t) represents the energy consumption required for production of industrial users in each time slot, and a and b are the profit coefficients of the power company.

3. According to claim 1, a flexible manufacturing workshop demand response scheduling method based on hierarchical game is characterized in that: The specific steps of step S4 are: S4.

1. Initialize the population and use a two-segment encoding method to represent chromosome information. The first segment is the process code OS, which is used to determine the processing order of the workpiece; the second segment is the machine code string MS, which is used to indicate the machine assigned to the process. A population with a population size of N is randomly generated; S4.2, using the adversarial learning strategy to generate an opposite population; S4.3, using the balanced machine load strategy to ensure that the loads of all machines in the two populations are as balanced as possible; S4.4, merge the initial population with the population after its adversarial learning, and the population size is 2N; S4.5, perform fast non-dominated sorting and crowding calculation on the merged population; S4.

6. The parent population is selected using tournament and elite selection strategies, and the population size is N.

4. According to claim 1, a flexible manufacturing workshop demand response scheduling method based on hierarchical game is characterized in that: The specific steps of step S5 are: S5.1, calculate the adaptive crossover,mutation probability; S5.2, perform selection, crossover, and mutation operations to obtain a progeny population with a population size of N; S5.3, using the adversarial learning strategy, conduct adversarial learning on the offspring population to produce an opposite population; S5.4, implement a balanced machine load strategy in the offspring population and its opposite population to ensure that the loads of all machines in the two populations are as balanced as possible; S5.5, merge the offspring population with its opposite population after oppositional learning, and the population size is 2N; S5.6, perform fast non-dominated sorting and crowding calculation on the merged population; S5.

7. Use tournament and elite selection strategies to select a new offspring population with a population size of N.

5. The method for demand response scheduling of a flexible manufacturing workshop based on hierarchical game according to claim 1 is characterized in that: The specific steps of step S6 are: S6.

1. Merge the newly obtained parent population and offspring population, and the population size is 2N; S6.2, perform fast non-dominated sorting and crowding calculation on the population after the merger of the parent and child generations; S6.

3. Use the tournament and elite selection strategies to select a new parent population with a population size of N.

6. The method for demand response scheduling of a flexible manufacturing workshop based on hierarchical game according to claim 4 is characterized in that: The specific steps of step S5.1 are: First, choose the maximum value CP of the crossover probability CP and the mutation probability MP max and MP max and minimum CP min and MP min ; Set Fitness i is the relative fitness value of the i-th individual, Fitness max and Fitness min are the maximum relative fitness value and the minimum relative fitness value in the population respectively; then, the crossover probability of the i-th individual is calculated as: In the formula, CP i represents the crossover probability of the i-th individual; Among them, ε is a very small real number to prevent the denominator from being zero; the mutation probability calculation method of the i-th individual is: In the formula, MP i represents the mutation probability of the i-th individual; Among them, set CP max =0.9, CP min =0.6, MP max =0.25, MP min = 0.06, suppose that after fast non-dominated sorting, the number of assigned levels is R, the population size is Size_pop, and the number of individuals in each level is f i (where i = 0, 1, ..., R-1); let the relative fitness value be Ra_Fitness = Size_pop, then the relative fitness value of each individual is Fitness i The calculation is as follows: Ra_Fitness=Ra_Fitness-Size_pop / R / f i +ε (14) Fitness i =Ra_Fitness(15) According to the order of i=0, 1, ..., R-1, the relative fitness value of each individual at each level is calculated in turn. i .

Images