Steel production equipment start-stop scheduling optimization method and device based on particle swarm algorithm

Abstract

This invention provides a method and apparatus for optimizing the start-up and shutdown scheduling of steel production equipment based on particle swarm optimization (PSO), belonging to the field of production scheduling technology. The method includes: receiving production data to perform discrete-time modeling of a short-process steelmaking production flow, and introducing virtual batches after the actual production batches during the modeling stage; constructing an objective function with minimizing peak power as the main objective, iteratively solving it using an improved PSO algorithm, and obtaining a preliminary solution under the condition of reaching the maximum number of iterations or a preset convergence condition; projecting and decoding the preliminary solution, and eliminating virtual batches during the decoding process to obtain the actual production scheduling scheme for the short-process steelmaking. This invention, by introducing virtual batches, employing a decoding placement strategy, and utilizing the improved PSO algorithm in a collaborative manner, fully leverages additional degrees of freedom in the extended space to explore a better peak-shaving structure, outputting a start-up and shutdown scheduling scheme for steel production equipment that significantly reduces the peak power of the entire plant while meeting process constraints.

Description

Technical Field

[0001] This invention relates to the field of production scheduling technology, and in particular to a method and apparatus for optimizing the start-up and shutdown scheduling of steel production equipment based on particle swarm optimization algorithm. Background Technology

[0002] Short-process steelmaking typically involves three main stages: electric arc furnace (EAF) smelting, refining furnace (LF) refining, and continuous casting machine (CCM) billet casting. Each stage differs significantly in equipment configuration, processing time, power requirements, and upstream / downstream connections, with strict sequential constraints and maximum waiting time limits between stages. Exceeding these limits can lead to temperature drops or process risks, thus requiring stringent time control strategies. In the continuous casting stage, continuity constraints such as "two furnaces per casting" and "casting intervals" must also be met, further increasing scheduling complexity.

[0003] In actual production, there are also similar equipment working in pairs in parallel, which further affects the instantaneous power level of the entire plant. In addition, the parallel processing of high-power equipment in different processes can also easily form peak overlap, which further affects the peak power.

[0004] Current production scheduling relies heavily on fixed sequences or manual experience, making the results highly sensitive to operator experience and potentially leading to excessively high peak power. While some studies have attempted to solve scheduling problems using heuristic or metaheuristic methods, most employ static search strategies or fixed solution parameters. Under conditions of tight process coupling and limited time windows, the resulting scheduling is still prone to high peak power or violations of other constraints. Summary of the Invention

[0005] This invention provides a method and apparatus for optimizing the start-up and shutdown scheduling of steel production equipment based on particle swarm optimization (PSO) algorithm. It addresses the shortcomings of existing short-process steelmaking scheduling methods that struggle to simultaneously achieve peak shaving and production scheduling feasibility. The invention implements an improved PSO algorithm-based method for optimizing the start-up and shutdown scheduling of steel production equipment, enabling feasible production scheduling schemes for large-batch production within a limited time window.

[0006] This invention provides a method for optimizing the start-up and shutdown scheduling of steel production equipment based on particle swarm optimization, comprising:

[0007] Receive production data to perform discrete-time modeling of the short-process steelmaking production process, and introduce virtual batches after the actual production batches during the modeling stage to expand the actual production batches to obtain the batches to be optimized.

[0008] The objective function is constructed with the minimum peak power as the main objective. Under the constraints set in advance based on the short-process steelmaking production process, an improved particle swarm optimization algorithm is used to solve iteratively. A preliminary solution is obtained when the maximum number of iterations is reached or the preset convergence condition is met.

[0009] The preliminary solution is projected and decoded, and virtual batches are eliminated during the decoding process to obtain the actual production scheduling scheme for short-process steelmaking;

[0010] In each iteration, the improved particle swarm optimization algorithm, based on the processing priority order of each production process generated by decoding the particle position vector, arranges processing machines and start times for each batch in accordance with the principle of minimizing the peak power of the already scheduled parts and the batch coding order, and ensures that all batches are scheduled through a preset forced placement strategy; the inertia weight and learning factor of the improved particle swarm optimization algorithm are updated based on the scheduling scheme of each iteration to enhance the global exploration capability when the power peak pressure is high and enhance the local repair capability when the cross-process waiting pressure is high.

[0011] According to the present invention, a method for optimizing the start-up and shutdown scheduling of steel production equipment based on particle swarm optimization is provided. The preset forced placement strategy includes: if there is no feasible position within the time window of the current batch, searching for the last available gap from back to front within the global time window and forcibly placing it; and if there is no available gap within the global time window, placing it directly at the end.

[0012] According to the present invention, a method for optimizing the start-up and shutdown scheduling of steel production equipment based on particle swarm optimization (PSO) algorithm, the step of updating the inertia weights and learning factors of the improved PSO algorithm based on the production scheduling scheme of each iteration specifically includes:

[0013] Calculate the peak power pressure and cross-process waiting pressure of the production scheduling plan in each iteration;

[0014] The inertial weight is updated based on the peak power pressure, such that when the peak power pressure is greater than a first preset value, the inertial weight increases accordingly.

[0015] Based on the cross-process waiting pressure, the individual learning factor and the group learning factor are updated so that when the cross-process waiting pressure is greater than the second preset value, the individual learning factor increases accordingly, while the group learning factor decreases accordingly.

[0016] The velocity and position of the next generation of particles are determined based on the updated inertial weights and learning factors.

[0017] According to the present invention, a method for optimizing the start-up and shutdown scheduling of steel production equipment based on particle swarm optimization algorithm, the step of calculating the peak power pressure of the production scheduling scheme in each iteration specifically includes:

[0018] Determine the peak power of the production plan for each particle, and calculate the ratio of the peak power of the production plan for each particle to the theoretical maximum power, which is used as the peak power pressure index of the corresponding particle.

[0019] Calculate the average peak power pressure index of all particles to obtain the peak power pressure of the production scheduling scheme in each iteration.

[0020] According to the present invention, a method for optimizing the start-up and shutdown scheduling of steel production equipment based on particle swarm optimization algorithm, the step of calculating the cross-process waiting pressure of the production scheduling scheme in each iteration specifically includes:

[0021] Calculate the ratio of the number of waiting constraint violations to the total number of batches to be optimized in the production scheduling plan for each particle, and obtain the average waiting constraint pressure per batch;

[0022] Take the baseline value and add it to the product of the average waiting constraint pressure per batch and the adjustment coefficient to obtain the waiting pressure index of the corresponding particle.

[0023] Calculate the average waiting pressure index of all particles to obtain the cross-process waiting pressure of the production scheduling scheme in each iteration.

[0024] According to the present invention, a method for optimizing the start-up and shutdown scheduling of steel production equipment based on particle swarm optimization algorithm, the step of receiving production data to perform discrete-time modeling of the short-process steelmaking production process specifically includes:

[0025] The production data includes shift duration, time granularity, batch data, equipment data used in each process, and maximum waiting time between processes in the short-process steelmaking production process.

[0026] Based on the production data, a resource task network method is used to define a set of batches, equipment, and time slots to achieve discrete-time modeling of the short-process steelmaking production process.

[0027] This invention also provides a start-up and shutdown scheduling optimization device for steel production equipment based on particle swarm optimization, comprising:

[0028] The receiving module is used to receive production data to perform discrete-time modeling of the short-process steelmaking production process, and to introduce a virtual batch after the actual production batch during the modeling stage to expand the actual production batch to obtain the batch to be optimized.

[0029] The optimization module is used to construct an objective function with the minimum peak power as the main objective. Under the constraints set in the short-process steelmaking production process, an improved particle swarm optimization algorithm is used to iteratively solve the problem and obtain a preliminary solution when the maximum number of iterations is reached or the preset convergence condition is met.

[0030] The output module is used to project and decode the pre-decode solution, and to remove virtual batches during the decoding process to obtain the actual production schedule of short-process steelmaking.

[0031] In each iteration, the improved particle swarm optimization algorithm, based on the processing priority order of each production process generated by decoding the particle position vector, arranges processing machines and start times for each batch in accordance with the principle of minimizing the peak power of the already scheduled parts and the batch coding order, and ensures that all batches are scheduled through a preset forced placement strategy; the inertia weight and learning factor of the improved particle swarm optimization algorithm are updated based on the scheduling scheme of each iteration to enhance the global exploration capability when the power peak pressure is high and enhance the local repair capability when the cross-process waiting pressure is high.

[0032] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the start-up and shutdown scheduling optimization method for steel production equipment based on particle swarm optimization as described above.

[0033] The present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the start-up and shutdown scheduling optimization method for steel production equipment based on particle swarm optimization as described above.

[0034] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described particle swarm optimization method for the start-up and shutdown scheduling of steel production equipment.

[0035] The present invention provides a method and apparatus for optimizing the start-up and shutdown scheduling of steel production equipment based on particle swarm optimization. By introducing virtual batches to expand the search space, and using a decoding placement strategy to make the virtual batches bear the pressure of constraint violation, the improved particle swarm optimization algorithm controls the direction and rhythm of particle movement. Finally, through the collaboration of the three, the additional degrees of freedom are fully utilized in the expanded space to explore a better peak-shaving structure. At the same time, the process feasibility between real batches is always maintained as the basis for the search. After eliminating virtual batches, the output is a start-up and shutdown scheduling scheme for steel production equipment that can significantly reduce the peak power of the entire plant and meet the process constraints. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0037] Figure 1 This is one of the flowcharts illustrating the start-up and shutdown scheduling optimization method for steel production equipment based on particle swarm optimization provided by this invention.

[0038] Figure 2 This is a Gantt chart of the baseline scheme in the steel production equipment start-up and shutdown scheduling optimization method based on particle swarm optimization provided by this invention;

[0039] Figure 3 This is a Gantt chart of the ordinary particle swarm optimization algorithm solution in the steel production equipment start-up and shutdown scheduling optimization method based on particle swarm optimization algorithm provided by the present invention;

[0040] Figure 4 This is a Gantt chart of the solution scheme of the algorithm of the present invention in the optimization method for start-up and shutdown scheduling of steel production equipment based on particle swarm optimization provided by the present invention;

[0041] Figure 5 This is a comparison of the power curves of the ordinary particle swarm optimization algorithm solution and the baseline solution in the steel production equipment start-up and shutdown scheduling optimization method based on particle swarm optimization algorithm provided by this invention.

[0042] Figure 6 This is a comparison chart of the power curves of the solution scheme of the algorithm of the present invention and the baseline scheme in the optimization method for start-up and shutdown scheduling of steel production equipment based on particle swarm optimization provided by the present invention.

[0043] Figure 7 This is a schematic diagram of the start-stop scheduling optimization device for steel production equipment based on particle swarm optimization algorithm provided by the present invention.

[0044] Figure 8 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0046] Specifically, in the actual production of short-process steelmaking, similar equipment often works in pairs in parallel. For example, a production line may include two electric arc furnaces, two refining furnaces, and two continuous casting machines. Each piece of equipment is only allowed to process one furnace at a time. At the same time, both electric arc furnaces and refining furnaces are high-power processes. If the production schedule is not properly arranged, power peaks can easily form. If the start time of a certain process is delayed to avoid peak transition, it can easily cause the waiting time across processes to exceed the requirements of the production standard, making production infeasible.

[0047] Although some studies have attempted to solve the scheduling problem of short-process steelmaking using heuristic or metaheuristic methods, most of them adopt static search strategies. Under conditions of tight process coupling and limited time windows, the scheduling results may still result in high peak power or infeasible solutions that violate waiting constraints, leading to repeated adjustments to the production scheduling plan and low efficiency.

[0048] In scenarios where the required production batches are large or the shift schedules are tight, the above contradictions become more prominent, making it difficult to simultaneously achieve power peak shaving and production scheduling feasibility. To address this, this invention proposes an optimized start-stop scheduling method for steel production equipment based on an improved particle swarm optimization algorithm and a virtual batch projection scheduling mechanism.

[0049] The following is combined Figures 1 to 6 This invention introduces a method for optimizing the start-up and shutdown scheduling of steel production equipment based on particle swarm optimization, such as... Figure 1 As shown, it includes:

[0050] Step 101: Receive production data to perform discrete-time modeling of the short-process steelmaking production process, and introduce virtual batches after the actual production batches during the modeling stage to expand the actual production batches to obtain the batches to be optimized.

[0051] Optionally, production data includes equipment data and batch data to be produced on the short-process steelmaking production line. Equipment data includes equipment type, quantity, theoretical peak power, duration of each operation, maximum waiting time, etc., determined based on the actual equipment conditions of the production line. Batch data includes the number of batches to be produced and shift duration, i.e., the production task of this batch number needs to be completed within the specified shift duration.

[0052] Based on this, a discrete-time model of this round of production process is performed.

[0053] Optionally, the shift duration is discretized into multiple time slots according to a preset time granularity to form a time slot set; the batch set and the equipment set are defined, and based on the idea that work starts and occupies, the processing status of each batch on each equipment in each time slot is represented by binary decision variables, thereby transforming the continuous equipment start-up and shutdown scheduling problem into a combinatorial optimization problem on discrete time slots.

[0054] In the modeling phase, virtual batches are introduced. That is, when defining the batch set, virtual batches are introduced after the last actual production batch, based on the actual production batches {1, ..., B}, to obtain the set of batches to be optimized I={1, ..., N} for modeling, and N=B+V, where N represents the number of batches to be optimized, B represents the number of actual production batches, and V represents the number of virtual batches.

[0055] Optionally, the number of virtual batches is preferably set to 1. If no feasible solution is found within the search space expanded by a virtual batch, the search space can be further expanded by increasing the number of virtual batches.

[0056] Step 102: Construct an objective function with the minimum peak power as the main objective. Under the constraints set in advance based on the short-process steelmaking production process, use an improved particle swarm optimization algorithm to iteratively solve the problem. Obtain a preliminary solution when the maximum number of iterations is reached or the preset convergence condition is met.

[0057] In each iteration, the improved particle swarm optimization algorithm, based on the processing priority order of each production process generated by decoding the particle position vector, arranges processing machines and start times for each batch in accordance with the principle of minimizing the peak power of the already scheduled parts and the batch coding order, and ensures that all batches are scheduled through a preset forced placement strategy; the inertia weight and learning factor of the improved particle swarm optimization algorithm are updated based on the scheduling scheme of each iteration to enhance the global exploration capability when the power peak pressure is high and enhance the local repair capability when the cross-process waiting pressure is high.

[0058] Based on the constructed discrete-time model, an objective function is constructed with the minimum peak power as the main objective, and constraints are pre-set based on the short-process steelmaking production process to form a constraint system for solving the objective function.

[0059] In one feasible implementation, a system is constructed with the primary objective of minimizing peak power and ensuring that the power per hour of the slot does not exceed a certain limit. As a hard constraint, the search is guided by a composite objective of mean power weight and violation, resulting in the objective function shown below:

[0060] ;

[0061] ;

[0062] In the formula, This is the upper limit of peak power. for t Total power over the time period T A set of time slots; For a composite objective, Weighted by mean power Penalty points are incurred, at least in cases of breach of the rules.

[0063] in, The operating power of the electric arc furnace. Indicates batch i In the t Time slot in electric arc furnace m The processing status is indicated by 1, where 1 indicates processing is in progress and 0 indicates no processing. The operating power of the refining furnace, Indicates batch i In the t Time slot in refining furnace m The processing status on it; The working power of the continuous casting machine. Indicates batch i In the t Time slot in continuous casting machine m The processing status.

[0064] Optionally, the constraint system includes process sequence constraints, maximum waiting time constraints, and continuous casting changeover constraints. The process sequence constraints ensure that each batch is produced sequentially in the order of electric arc furnace, refining furnace, and continuous casting machine. The maximum waiting time constraints ensure that the waiting time between the electric arc furnace and refining furnace stages does not exceed the maximum time threshold required by the process (60 minutes in this embodiment), and also ensure that the waiting time between the refining furnace and continuous casting machine does not exceed the maximum time threshold required by the process (45 minutes in this embodiment). The equipment capacity constraints limit the maximum number of heats that can be processed in the same time slot using similar equipment (1 heat in this embodiment). The continuous casting changeover constraints limit the changeover time that needs to be inserted between every two heats (60 minutes in this embodiment).

[0065] Based on this, the present invention employs an improved particle swarm optimization algorithm to solve the objective function under the constructed constraint system.

[0066] Specifically, the solution process uses particle position vectors as encoding, with each particle carrying a position vector. The position vector is divided into three priority keys, which correspond to the three production processes of electric arc furnace, refining furnace and continuous casting machine. Each segment is a sub-vector containing N (total number of batches to be optimized) real number components. The sub-vector takes values ​​in the range of [0,1] and is used to represent the processing priority of the same batch in the corresponding process.

[0067] During decoding, the real number components within each segment are sorted in ascending order of value. The resulting sequence number determines the processing order of all batches in the corresponding process of that segment.

[0068] Based on this, each batch is placed sequentially according to its batch code, and the specific processing machine and start time for each batch in each process are assigned accordingly. Specifically, for the currently being arranged batch and process, within the feasible window that satisfies the process constraints defined by the constraint system and the processing sequence determined by decoding, all available machines and possible start time slots are traversed; for each candidate placement point, the peak value of the real-time power curve of the entire plant after placement is simulated and calculated, and the machine and time slot that minimizes this peak power are selected as the actual placement of the current batch in this process.

[0069] It should be noted that for this process of continuous casting machine, the rule of two heats per pour must be followed, that is, after two heats are arranged in succession, a pouring interval must be forcibly inserted before the start time slot of the next heat.

[0070] It is understandable that after the batches are placed in the above manner, the feasible window of each batch gradually narrows, and there may be no feasible window when the later batches are placed. Therefore, a forced placement strategy is pre-set to force the placement of batches with no feasible window, so as to ensure that there are no batches that have not been scheduled for production. When forced placement occurs, a penalty score is triggered so that the particle swarm algorithm of this invention can iteratively repair under the drive of the penalty score.

[0071] After forced placement, the production scheduling plan corresponding to the particle can be obtained, and specific equipment start-up and shutdown timing and power curves can be formed, resulting in a complete scheduling solution.

[0072] Furthermore, the peak power, waiting time between processes, and constraint violations are calculated for the obtained complete scheduling solution. These are then used to construct a stress index to provide feedback for adjusting the inertia weights and two learning factors of the particle swarm, thereby guiding the updated particles to change their speed and position in the next iteration.

[0073] The pressure indicators constructed in this embodiment include peak power pressure and peak power pressure. Peak power pressure represents the degree to which the highest value of the plant's power curve under the current scheduling solution is close to the theoretical maximum power of all equipment running at full load. The higher the peak power pressure, the more urgent the peak shaving demand. Cross-process waiting pressure represents the actual waiting time of each batch between the electric arc furnace and the refining furnace, and between the refining furnace and the continuous casting machine, under the current scheduling solution, which is close to or exceeds its maximum allowable waiting limit. The higher the value, the higher the risk of violating the waiting constraint.

[0074] When updating the inertia weight and two learning factors of the particle swarm algorithm using pressure indices, the inertia weight is increased when the power peak pressure is greater, in order to enhance the global exploration capability of the particle swarm and enable it to jump out of the current region that may lead to a high power peak. When the cross-process waiting pressure is greater, the individual learning factor is increased and the group learning factor is decreased, in order to enhance the tendency of particles to move toward their own optimal position and prioritize the repair of process feasibility.

[0075] Optionally, based on the above principles, the two pressure indicators can be linearly mapped to the offsets of the inertia weight and the two learning factors, and constrained within a preset reasonable range, so as to achieve adaptive adjustment of the inertia weight and the two learning factors, forming an improved particle swarm algorithm.

[0076] By iteratively solving the problem using the above method, a preliminary solution can be obtained when the maximum number of iterations is reached or the preset convergence condition is met.

[0077] Optionally, the preset convergence condition can be the emergence of a feasible solution that does not violate the constraints of the system and whose power peak value is at a preset high value.

[0078] Step 103: Project and decode the prepared solution, and remove virtual batches during the decoding process to obtain the actual production schedule of short-process steelmaking;

[0079] Using the same method to decode and place the pre-decode, after obtaining the production schedule corresponding to the pre-decode, the placeholders corresponding to the virtual batches are removed to obtain the actual production schedule.

[0080] In one feasible implementation, the scheduling table, Gantt chart, and power curve of the pre-determined solution are decoded, and virtual batches are removed from the scheduling table and Gantt chart to obtain the actual production schedule. Then, the power curve is updated.

[0081] It is understandable that the introduction of virtual batches will inevitably bring a larger search space, thereby increasing the possibility of finding a feasible solution. However, the introduction of virtual batches will also lead to the destruction of the feasibility of the original solution when virtual batches are removed and a real production schedule is formed. For example, the real production schedule may have a waiting time between batches that violates the constraint.

[0082] Therefore, in solving the objective function, this invention first uses the principle of sequential placement to ensure that the decoding and placement strictly follow the batch encoding order, prioritizing the scheduling of all real batches. This ensures that the temporal relationship between real batches is optimized and fixed in a relatively feasible resource environment. Furthermore, it allows for greater application of forced placement to virtual batches, which bear the constraint violation pressure caused by the search. Finally, through an improved particle swarm optimization algorithm, the position of the next round of particles is updated based on the pressure index of the solution in each iteration. As a result, the waiting time of the process corresponding to the preliminary solution obtained after multiple iterations is usually far from the upper limit of the constraint, providing a safety margin. Ultimately, the real production scheduling scheme after removing virtual batches can still not violate the waiting time constraint, resulting in a production-feasible real production scheduling scheme.

[0083] This invention expands the search space by introducing virtual batches, uses a decoding placement strategy to make virtual batches bear the pressure of constraint violations, controls the direction and rhythm of particle movement through an improved particle swarm algorithm, and finally explores a better peak-shaving structure by fully utilizing the additional degrees of freedom in the expanded space through the collaboration of the three. At the same time, the process feasibility between real batches is always taken as the basis of the search, so that after eliminating virtual batches, the output is a steel production equipment start-up and shutdown scheduling scheme that can significantly reduce the peak power of the whole plant and meet the process constraints.

[0084] In the steel production equipment start-up and shutdown scheduling optimization method based on particle swarm optimization algorithm of the present invention, the preset forced placement strategy includes: if there is no feasible position within the time window of the current batch, searching for the last available gap from back to front within the global time window to force placement, and if there is no available gap within the global time window, placing it directly at the end.

[0085] In this embodiment, the following preset forced placement strategy is adopted: when there is no feasible position for the current batch in the remaining time window of the current stage, for example, if the particle decoding order of the current batch requires it to be placed after a certain batch A in the electric arc furnace stage, but there is no time slot for the electric arc furnace to be placed after batch A, then the last available gap is searched from back to front within the global time window for forced placement.

[0086] If there is still no gap, for example, if there is no time slot to be placed before or after batch A of the electric arc furnace, then place it at the end and apply the penalty points corresponding to forced placement.

[0087] In this embodiment, a peak-sensing placer is set up to execute the above-mentioned forced placement strategy. Through the forced placement strategy, the penalty score of forced placement is incorporated into the objective function, guiding the particle swarm to adjust the batch order and interval in a directional manner in subsequent iterations, gradually reducing or even eliminating forced placement, so as to improve the feasible scheduling solution that fully meets the power and waiting time constraints.

[0088] In the steel production equipment start-up and shutdown scheduling optimization method based on particle swarm optimization (PSO) of this invention, the step of updating the inertia weights and learning factors of the improved PSO algorithm based on the production scheduling scheme of each iteration specifically includes:

[0089] Calculate the peak power pressure and cross-process waiting pressure of the production scheduling plan in each iteration;

[0090] Decoding each generation of particles yields the corresponding production schedule, which is used to calculate peak power pressure. and waiting pressure across processes .

[0091] The inertial weight is updated based on the peak power pressure, such that when the peak power pressure is greater than a first preset value, the inertial weight increases accordingly.

[0092] In this embodiment, peak power pressure is used. Update inertia weights:

[0093] ;

[0094] In the formula, For the updated inertia weights, To preset the initial inertia weight value, For adjustment coefficients, The inertial weights are pre-defined minimum and maximum boundaries.

[0095] That is, when the peak power pressure is greater than a first preset threshold (set to 1 in this embodiment), the peak power pressure is mapped to a parameter offset. The first preset threshold is determined based on the calculation method and experience of the peak power pressure.

[0096] Alternatively, the inertia weight can also be updated through other feasible calculation methods, as long as the adjustment effect on the initial inertia weight value described above can be achieved.

[0097] By using the above method, when the peak power pressure of this generation of particles exceeds the first preset threshold, the inertial weight increases accordingly. The next generation of particles tends to maintain the current speed and expand the search step size, thereby enhancing global exploration and helping to break out of the batch arrangement pattern that forms peak loads.

[0098] When the peak power pressure of this generation of particles is less than the first preset threshold, the inertial weight decreases accordingly, and the search becomes more focused on local refinement.

[0099] Based on the cross-process waiting pressure, the individual learning factor and the group learning factor are updated so that when the cross-process waiting pressure is greater than the second preset value, the individual learning factor increases accordingly, while the group learning factor decreases accordingly.

[0100] In this embodiment, cross-process waiting pressure is used. The learning factors in the particle swarm optimization algorithm include individual learning factors. and group learning factor Based on this, a feasible method for updating learning factors is as follows:

[0101] ;

[0102] ;

[0103] In the formula, and These are the updated individual learning factor and the group learning factor, respectively. The clip function restricts the parameters to preset upper and lower bounds. and These are the preset initial individual learning factor values ​​and the preset initial group learning factor values, respectively. and For adjustment coefficients, and These are the preset upper and lower bounds for individual learning factors and group learning factors, respectively.

[0104] That is, when the cross-process waiting pressure is greater than the second preset threshold (set to 1 in this embodiment), the cross-process waiting pressure is mapped to the parameter offset, wherein the second preset threshold is determined based on the calculation method and experience of the cross-process waiting pressure.

[0105] Using the above method, when the cross-process waiting pressure exceeds the second preset threshold, indicating that waiting constraints are generally tightening or even being violated, the individual learning factor... This will increase accordingly, and particles will place greater emphasis on their own historically feasible individual optimality. pbest To strengthen localized feasibility repair capabilities, and at the same time, appropriately reduce the group learning factor. Reduce the tendency toward the global optimum ( gbest The strong convergence of the population prevents the entire group from being pulled into a structure with a low peak value but infeasible construction. Conversely, when the waiting pressure is low, Close to the baseline value, Approaching or slightly exceeding the benchmark value makes the particle swarm converge more closely toward the global excellent solution.

[0106] The velocity and position of the next generation of particles are determined based on the updated inertial weights and learning factors.

[0107] The updated inertia weights and two learning factors are directly incorporated into the next generation's particle swarm velocity update formula to update the velocity and position of the next generation of particles:

[0108] ;

[0109] ;

[0110] In the formula, and The first k The velocity and position of the +1 generation particles and The first k The velocity and position of the particles; and Let be an independent uniform random vector on [0,1]. This represents element-wise multiplication. For particles p The optimal position of an individual This represents the optimal position reached by all particles in the swarm throughout the search history. Indicates inertia weight, and These are the updated individual learning factors and group learning factors, respectively.

[0111] The state feedback mechanism constructed in the above manner, without changing the basic structure of the particle swarm, adjusts the balance between the three driving forces of "inertia preservation, convergence towards individual experience, and convergence towards the optimal group" in each iteration by updating the inertia weight, individual learning factor, and group learning factor of the particle swarm in real time. This allows the global search and local repair to be dynamically balanced as the scheduling state (i.e., the two constructed pressure indicators) changes.

[0112] In the steel production equipment start-up and shutdown scheduling optimization method based on particle swarm optimization of this invention, the step of calculating the peak power pressure of the production scheduling scheme in each iteration specifically includes:

[0113] Determine the peak power of the production plan for each particle, and calculate the ratio of the peak power of the production plan for each particle to the theoretical maximum power, which is used as the peak power pressure index of the corresponding particle.

[0114] In this embodiment, the calculation of peak power pressure is based on the power curve of the scheduling scheme obtained from particle placement, assuming the first... p The power curve of the scheduling scheme for each particle is as follows: Then its peak power is: .

[0115] Redefining the theoretical maximum power This refers to the total power of all similar devices operating at full load, i.e.:

[0116] .

[0117] In the formula, , and These represent the quantities of electric arc furnaces, refining furnaces, and continuous casting machines, respectively. , and These represent the operating power of the electric arc furnace, refining furnace, and continuous casting machine, respectively.

[0118] Based on this, the first p The peak power pressure index of an individual particle is defined as:

[0119] .

[0120] Preferably, the calculated individual power peak pressure index is mapped to a preset range using a pinch operator to avoid parameter oscillations caused by extreme values. In this embodiment, let .

[0121] When the individual power peak pressure index approaches 1, it indicates that the current scheduling is approaching the limit state of "all equipment running at full power at the same time", and the peak pressure is relatively large; when the individual peak pressure is less than 1, it indicates that the peak still has room to decline.

[0122] Calculate the average peak power pressure index of all particles to obtain the peak power pressure of the production scheduling scheme in each iteration.

[0123] The peak power pressure of the particles in each generation is calculated by averaging the individual peak power pressure exponents of all particles in that generation.

[0124] ;

[0125] In the formula, This indicates the number of particles in a particle swarm.

[0126] Will It is used as a parameter to adjust the inertia weight.

[0127] In the particle swarm optimization method for starting and stopping steel production equipment, the step of calculating the cross-process waiting pressure of the production scheduling scheme in each iteration specifically includes:

[0128] Calculate the ratio of the number of waiting constraint violations to the total number of batches to be optimized in the production scheduling plan for each particle, and obtain the average waiting constraint pressure per batch;

[0129] Cross-process waiting pressure represents the degree of stress on the waiting constraints of the current particle scheduling scheme in the two stages from the electric arc furnace to the refining furnace and from the refining furnace to the continuous casting machine. Optionally, it can be defined based on the normalized deviation of each batch waiting time from the maximum waiting upper bound. If the definition of the first... i The waiting time from the electric arc furnace to the refining furnace for a batch is The maximum waiting upper bound is Then its normalized excess degree It can be written as:

[0130] ;

[0131] The waiting pressure from the refining furnace to the continuous casting machine can be defined in a similar way, and then the average waiting pressure of all batches can be used to define the cross-process waiting pressure.

[0132] However, to reduce computational complexity, this implementation method counts the number of wait constraint violations for each batch to be optimized within the current particle scheduling scheme. Construction waiting pressure:

[0133] ;

[0134] In the formula, N represents the average waiting constraint pressure per batch, and N is the number of batches to be optimized.

[0135] Take the baseline value and add it to the product of the average waiting constraint pressure per batch and the adjustment coefficient to obtain the waiting pressure index of the corresponding particle.

[0136] Based on this, and taking into account the waiting constraint pressure of all batches, the waiting pressure index of the particles is calculated:

[0137] ;

[0138] In the formula, Represents particles p The waiting pressure index is an adjustment coefficient used to control the degree of influence of waiting violations on parameter adjustment; where 1 is the baseline value selected in this embodiment.

[0139] Preferably, the To avoid drastic parameter fluctuations, the clamp is positioned within a preset range. In this embodiment, Grab to [0.5, 2.0].

[0140] Calculate the average waiting pressure index of all particles to obtain the cross-process waiting pressure of the production scheduling scheme in each iteration.

[0141] The average waiting pressure exponent for all particles in a generation yields the inter-process waiting pressure for that generation of particles.

[0142] ;

[0143] Received Used to achieve adaptive feedback of learning factors.

[0144] In the particle swarm optimization method for starting and stopping steel production equipment, the step of receiving production data to perform discrete-time modeling of the short-process steelmaking production process specifically includes:

[0145] The production data includes shift duration, time granularity, batch data, equipment data used in each process, and maximum waiting time between processes in the short-process steelmaking production process.

[0146] In this embodiment, the production data includes shift duration, time granularity, and maximum waiting time between each process.

[0147] It also includes equipment data used in each process, specifically the duration of each heat for each piece of equipment, the continuous casting changeover time of the continuous casting machine, the quantity of each piece of equipment, the power parameters of each piece of equipment, and the number of batches to be scheduled for production.

[0148] In one specific implementation, the shift duration H = 700 minutes, the duration of one heat cycle for the electric arc furnace, refining furnace, and continuous casting machine is 80 minutes, 30 minutes, and 60 minutes, respectively, and the continuous casting changeover time for the continuous casting machine is 60 minutes. There are two electric arc furnaces, two refining furnaces, and two continuous casting machines. The maximum waiting time between the electric arc furnace and the refining furnace is no more than 60 minutes, and the maximum waiting time between the refining furnace and the continuous casting machine is no more than 40 minutes. The power parameters (unit: MW) of the electric arc furnace, refining furnace, and continuous casting machine are as follows: The number of batches B to be scheduled for production is 9. After that, 2 virtual batches are introduced, resulting in a total of 11 batches to be optimized.

[0149] Based on the production data, a resource task network method is used to define a set of batches, equipment, and time slots to achieve discrete-time modeling of the short-process steelmaking production process.

[0150] The Resource Task Network (RTN) method is used to build discrete-time models based on production data.

[0151] Specifically, let the batch set I = {1, ..., ...} N}, equipment collection Time slot set T ={0, ..., T -1}; Duration (slot) d E , d L , d C Replacement pouring (grooving) Waiting for the upper bound (slot) , .

[0152] Based on the concept of "occupation upon commencement of construction," the definition is... Indicates whether batch i starts in slot t of EAF machine m; This indicates whether processing is required in that slot; the same applies to LF and CCM. In implementation, this is achieved through an occupancy matrix. This indicates that it is updated directly by the placer.

[0153] The discrete-time modeling of production data is completed using the above method, and a constraint system is constructed based on this model as shown below:

[0154] The number of operations constraint is used to ensure that each operation is executed only once. ;

[0155] Single-machine capacity constraint is used to limit the number of batches that can be executed per heat per machine. Taking electric arc furnace as an example, the same settings are used for refining furnace and continuous casting machine: ;

[0156] Process sequence and duration constraints: Start and end slots are set. Taking an electric arc furnace as an example, the same settings apply to refining furnaces and continuous casting machines. ,satisfy ;

[0157] Maximum waiting upper bound constraint: ;

[0158] Schedule boundary constraints: ;

[0159] Continuous casting changeover constraint: This constraint must be satisfied for adjacent groups on the same continuous casting machine. .

[0160] in i Indicates batch index; m Indicates the device index; k This indicates the furnace sequence number for continuous processing on the same continuous casting machine. , , , , , Each represents a batch i Start and end time slots on electric arc furnaces, refining furnaces, and continuous casting machines; d E , d L , d C These represent the duration (in time slots) of a single heat cycle for the electric arc furnace, refining furnace, and continuous casting machine, respectively. W EL , W LC These represent the maximum allowable waiting time between the electric arc furnace and the refining furnace, and between the refining furnace and the continuous casting machine, respectively. T This indicates the upper bound of the time slot corresponding to the total time of each shift. γ This indicates the time interval required for continuous casting and changing of the pouring vessel; i k and i k+1 These represent the first and second consecutive processes performed on the same continuous casting machine. k Furnace and the first k +1 batch.

[0161] The aforementioned constraint system is executed in two phases: the placement phase and the verification phase of the solution for each iteration of the particle swarm.

[0162] Furthermore, regarding the production data in the above embodiments, this invention compares the baseline of the fixed line, the production scheduling scheme obtained iteratively using a common particle swarm optimization algorithm, and the production scheduling scheme determined using the improved particle swarm optimization algorithm of this invention, such as... Figures 2 to 6 As shown.

[0163] in, Figure 2 , Figure 3 and Figure 4 These are Gantt charts representing the baseline, the iterative solution of the ordinary particle swarm optimization algorithm, and the production scheduling scheme of the algorithm of this invention. Figures 5 to 6 These are power curves comparing the algorithm of this invention with those of the ordinary particle swarm optimization algorithm and the baseline.

[0164] As can be seen, optimization comparisons were performed using the same production time window and production batches. The algorithm of this invention exhibits improved power output stability compared to ordinary particle swarm optimization (PSO) algorithms. In practical applications, this helps reduce the impact on the power grid, improve power supply stability, and reduce costs while increasing efficiency. Furthermore, ordinary PSO algorithms may exhibit forced placement and constraint violations in the output, such as... Figure 2 As can be seen, in order to ensure that the peak power does not exceed the constraint, the ordinary particle swarm algorithm does not schedule the continuous casting machine process of batch 8 (B8). However, the algorithm of the present invention ensures that all batches and all processes are scheduled while the peak power does not exceed the power limit. It can be seen that the particle swarm algorithm of the present invention is highly adaptable to steel production scheduling.

[0165] The following describes the steel production equipment start-up and shutdown scheduling optimization device based on particle swarm optimization algorithm provided by the present invention. The steel production equipment start-up and shutdown scheduling optimization device based on particle swarm optimization algorithm described below can be referred to in correspondence with the steel production equipment start-up and shutdown scheduling optimization method based on particle swarm optimization algorithm described above.

[0166] like Figure 7 As shown, the steel production equipment start-up and shutdown scheduling optimization device based on particle swarm optimization algorithm includes a receiving module 701, an optimization module 702 and an output module 703.

[0167] The receiving module 701 is used to receive production data to perform discrete-time modeling of the short-process steelmaking production process, and to introduce a virtual batch after the actual production batch in the modeling stage to expand the actual production batch to obtain the batch to be optimized.

[0168] Optionally, production data includes equipment data and batch data to be produced on the short-process steelmaking production line. Equipment data includes equipment type, quantity, theoretical peak power, duration of each operation, maximum waiting time, etc., determined based on the actual equipment conditions of the production line. Batch data includes the number of batches to be produced and shift duration, i.e., the production task of this batch number needs to be completed within the specified shift duration.

[0169] Based on this, a discrete-time model of this round of production process is performed.

[0170] Optionally, the shift duration is discretized into multiple time slots according to a preset time granularity to form a time slot set; the batch set and the equipment set are defined, and based on the idea that work starts and occupies, the processing status of each batch on each equipment in each time slot is represented by binary decision variables, thereby transforming the continuous equipment start-up and shutdown scheduling problem into a combinatorial optimization problem on discrete time slots.

[0171] In the modeling phase, virtual batches are introduced. That is, when defining the batch set, virtual batches are introduced after the last actual production batch, based on the actual production batches {1, ..., B}, to obtain the set of batches to be optimized I={1, ..., N} for modeling, and N=B+V, where V is the number of virtual batches.

[0172] The optimization module 702 is used to construct an objective function with the minimum peak power as the main objective. Under the constraints set in advance based on the short-process steelmaking production process, the improved particle swarm algorithm is used to solve iteratively and obtain a preliminary solution when the maximum number of iterations is reached or the preset convergence condition is met.

[0173] In each iteration, the improved particle swarm optimization algorithm, based on the processing priority order of each production process generated by decoding the particle position vector, arranges processing machines and start times for each batch in accordance with the principle of minimizing the peak power of the already scheduled parts and the batch coding order, and ensures that all batches are scheduled through a preset forced placement strategy; the inertia weight and learning factor of the improved particle swarm optimization algorithm are updated based on the scheduling scheme of each iteration to enhance the global exploration capability when the power peak pressure is high and enhance the local repair capability when the cross-process waiting pressure is high.

[0174] Based on the constructed discrete-time model, an objective function is constructed with the minimum peak power as the main objective, and constraints are pre-set based on the short-process steelmaking production process to form a constraint system for solving the objective function.

[0175] Optionally, the constraint system includes process sequence constraints, maximum waiting time constraints, and continuous casting changeover constraints. The process sequence constraints ensure that each batch is produced sequentially in the order of electric arc furnace, refining furnace, and continuous casting machine. The maximum waiting time constraints ensure that the waiting time between the electric arc furnace and refining furnace stages does not exceed the maximum time threshold required by the process (60 minutes in this embodiment), and also ensure that the waiting time between the refining furnace and continuous casting machine does not exceed the maximum time threshold required by the process (45 minutes in this embodiment). The equipment capacity constraints limit the maximum number of heats that can be processed in the same time slot using similar equipment (1 heat in this embodiment). The continuous casting changeover constraints limit the changeover time that needs to be inserted between every two heats (60 minutes in this embodiment).

[0176] Based on this, the present invention employs an improved particle swarm optimization algorithm to solve the objective function under the constructed constraint system.

[0177] Specifically, the solution process uses particle position vectors as encoding, with each particle carrying a position vector. The position vector is divided into three priority keys, which correspond to the three production processes of electric arc furnace, refining furnace and continuous casting machine. Each segment is a sub-vector containing N (total number of batches to be optimized) real number components. The sub-vector takes values ​​in the range of [0,1] and is used to represent the processing priority of the same batch in the corresponding process.

[0178] During decoding, the real number components within each segment are sorted in ascending order of value. The resulting sequence number determines the processing order of all batches in the corresponding process of that segment.

[0179] Based on this, each batch is placed sequentially according to its batch code, and the specific processing machine and start time for each batch in each process are assigned accordingly. Specifically, for the currently being arranged batch and process, within the feasible window that satisfies the process constraints defined by the constraint system and the processing sequence determined by decoding, all available machines and possible start time slots are traversed; for each candidate placement point, the peak value of the real-time power curve of the entire plant after placement is simulated and calculated, and the machine and time slot that minimizes this peak power are selected as the actual placement of the current batch in this process.

[0180] It should be noted that for this process of continuous casting machine, the rule of two heats per pour must be followed, that is, after two heats are arranged in succession, a pouring interval must be forcibly inserted before the start time slot of the next heat.

[0181] It is understandable that after the batches are placed in the above manner, the feasible window of each batch gradually narrows, and there may be no feasible window when the later batches are placed. Therefore, a forced placement strategy is pre-set to force the placement of batches with no feasible window, so as to ensure that there are no batches that have not been scheduled for production. When forced placement occurs, a penalty score is triggered so that the particle swarm algorithm of this invention can iteratively repair under the drive of the penalty score.

[0182] After forced placement, the production scheduling plan corresponding to the particle can be obtained, and specific equipment start-up and shutdown timing and power curves can be formed, resulting in a complete scheduling solution.

[0183] Furthermore, the peak power, waiting time between processes, and constraint violations are calculated for the obtained complete scheduling solution. These are then used to construct a stress index to provide feedback for adjusting the inertia weights and two learning factors of the particle swarm, thereby guiding the updated particles to change their speed and position in the next iteration.

[0184] The pressure indicators constructed in this embodiment include peak power pressure and peak power pressure. Peak power pressure represents the degree to which the highest value of the plant's power curve under the current scheduling solution is close to the theoretical maximum power of all equipment running at full load. The higher the peak power pressure, the more urgent the peak shaving demand. Cross-process waiting pressure represents the actual waiting time of each batch between the electric arc furnace and the refining furnace, and between the refining furnace and the continuous casting machine, under the current scheduling solution, which is close to or exceeds its maximum allowable waiting limit. The higher the value, the higher the risk of violating the waiting constraint.

[0185] When updating the inertia weight and two learning factors of the particle swarm algorithm using pressure indices, the inertia weight is increased when the power peak pressure is greater, in order to enhance the global exploration capability of the particle swarm and enable it to jump out of the current region that may lead to a high power peak. When the cross-process waiting pressure is greater, the individual learning factor is increased and the group learning factor is decreased, in order to enhance the tendency of particles to move toward their own optimal position and prioritize the repair of process feasibility.

[0186] Optionally, based on the above principles, the two pressure indicators can be linearly mapped to the offsets of the inertia weight and the two learning factors, and constrained within a preset reasonable range, so as to achieve adaptive adjustment of the inertia weight and the two learning factors, forming an improved particle swarm algorithm.

[0187] By iteratively solving the problem using the above method, a preliminary solution can be obtained when the maximum number of iterations is reached or the preset convergence condition is met.

[0188] Optionally, the preset convergence condition can be the emergence of a feasible solution that does not violate the constraints of the system and whose power peak value is at a preset high value.

[0189] Output module 703 is used to project and decode the pre-decode solution, and remove virtual batches during the decoding process to obtain the actual production schedule of short-process steelmaking.

[0190] Using the same method to decode and place the pre-decode, after obtaining the production schedule corresponding to the pre-decode, the placeholders corresponding to the virtual batches are removed to obtain the actual production schedule.

[0191] In one feasible implementation, the scheduling table, Gantt chart, and power curve of the pre-determined solution are decoded, and virtual batches are removed from the scheduling table and Gantt chart to obtain the actual production schedule. Then, the power curve is updated.

[0192] It is understandable that the introduction of virtual batches will inevitably bring a larger search space, thereby increasing the possibility of finding a feasible solution. However, the introduction of virtual batches will also lead to the destruction of the feasibility of the original solution when virtual batches are removed and a real production schedule is formed. For example, the real production schedule may have a waiting time between batches that violates the constraint.

[0193] Therefore, in solving the objective function, this invention first uses the principle of sequential placement to ensure that the decoding and placement strictly follow the batch encoding order, prioritizing the scheduling of all real batches. This ensures that the temporal relationship between real batches is optimized and fixed in a relatively feasible resource environment. Furthermore, it allows for greater application of forced placement to virtual batches, which bear the constraint violation pressure caused by the search. Finally, through an improved particle swarm optimization algorithm, the position of the next round of particles is updated based on the pressure index of the solution in each iteration. As a result, the waiting time of the process corresponding to the preliminary solution obtained after multiple iterations is usually far from the upper limit of the constraint, providing a safety margin. Ultimately, the real production scheduling scheme after removing virtual batches can still not violate the waiting time constraint, resulting in a production-feasible real production scheduling scheme.

[0194] This invention expands the search space by introducing virtual batches, uses a decoding placement strategy to make virtual batches bear the pressure of constraint violations, controls the direction and rhythm of particle movement through an improved particle swarm algorithm, and finally explores a better peak-shaving structure by fully utilizing the additional degrees of freedom in the expanded space through the collaboration of the three. At the same time, the process feasibility between real batches is always taken as the basis of the search, so that after eliminating virtual batches, the output is a steel production equipment start-up and shutdown scheduling scheme that can significantly reduce the peak power of the whole plant and meet the process constraints.

[0195] Figure 8 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 8As shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830, and a communication bus 840. The processor 810, communication interface 820, and memory 830 communicate with each other via the communication bus 840. The processor 810 can call logical instructions in the memory 830 to execute a particle swarm optimization method for the start-up and shutdown scheduling of steel production equipment. This method includes: receiving production data to perform discrete-time modeling of the short-process steelmaking production process, and introducing a virtual batch after the actual production batch during the modeling stage to expand the actual production batch to obtain the batch to be optimized; constructing an objective function with minimum peak power as the main objective, and iteratively solving the problem using an improved particle swarm optimization algorithm under a pre-defined constraint system based on the short-process steelmaking production process, obtaining a preliminary solution upon reaching the maximum number of iterations or under a preset convergence condition; and projecting and decoding the preliminary solution. The algorithm eliminates virtual batches during the decoding process to obtain the actual production schedule for short-process steelmaking. In each iteration, the improved particle swarm optimization (PSO) algorithm, based on the processing priority order of each production process generated by decoding the particle position vector, assigns processing machines and start times to each batch according to the principle of minimizing the peak power of the already scheduled portion and the batch coding order. A preset forced placement strategy ensures that all batches are scheduled. The algorithm updates its inertia weights and learning factors based on the production schedule in each iteration to enhance global exploration capabilities when power peak pressure is high and local repair capabilities when cross-process waiting pressure is high.

[0196] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0197] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the steel production equipment start-up and shutdown scheduling optimization method based on particle swarm optimization provided by the above methods. This method includes: receiving production data to perform discrete-time modeling of the short-process steelmaking production process, and introducing a virtual batch after the actual production batch during the modeling stage to expand the actual production batch to obtain the batch to be optimized; constructing an objective function with minimum peak power as the main objective, and iteratively solving it using an improved particle swarm optimization algorithm under a pre-set constraint system based on the short-process steelmaking production process. A preliminary solution is obtained by reaching the maximum number of iterations or under a preset convergence condition; the preliminary solution is projected and decoded, and virtual batches are eliminated during the decoding process to obtain the actual production scheduling scheme for short-process steelmaking; wherein, in each iteration, the improved particle swarm algorithm is based on the processing priority order of each production process generated by decoding the particle position vector, and arranges processing machines and start times for each batch in accordance with the principle of minimizing the peak power of the already scheduled part and the batch coding order, and ensures that all batches are scheduled through a preset forced placement strategy; the inertia weight and learning factor of the improved particle swarm algorithm are updated based on the production scheduling scheme of each iteration to enhance the global exploration capability when the power peak pressure is high and enhance the local repair capability when the cross-process waiting pressure is high.

[0198] Furthermore, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the above-described particle swarm optimization method for the start-up and shutdown scheduling optimization of steel production equipment. This method includes: receiving production data to perform discrete-time modeling of the short-process steelmaking production flow, and introducing a virtual batch after the actual production batch during the modeling stage to expand the actual production batch to obtain the batch to be optimized; constructing an objective function with minimum peak power as the main objective, and iteratively solving the function using an improved particle swarm optimization algorithm under a pre-defined constraint system based on the short-process steelmaking production flow, until the maximum number of iterations or a preset convergence threshold is reached. A preliminary solution is obtained under the given conditions; the preliminary solution is projected and decoded, and virtual batches are eliminated during the decoding process to obtain the actual production scheduling scheme for short-process steelmaking; wherein, in each iteration, the improved particle swarm algorithm is based on the processing priority order of each production process generated by decoding the particle position vector, and arranges processing machines and start times for each batch in accordance with the principle of minimizing the peak power of the already scheduled part and the batch coding order, and ensures that all batches are scheduled through a preset forced placement strategy; the inertia weight and learning factor of the improved particle swarm algorithm are updated based on the production scheduling scheme of each iteration to enhance the global exploration capability when the power peak pressure is high and enhance the local repair capability when the cross-process waiting pressure is high.

[0199] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0200] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0201] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing start-stop scheduling of a steel production plant based on a particle swarm algorithm, characterized in that, include: Receive production data to perform discrete-time modeling of the short-process steelmaking production process, and introduce virtual batches after the actual production batches during the modeling stage to expand the actual production batches to obtain the batches to be optimized. The objective function is constructed with the minimum peak power as the main objective. Under the constraints set in advance based on the short-process steelmaking production process, an improved particle swarm optimization algorithm is used to solve iteratively. A preliminary solution is obtained when the maximum number of iterations is reached or the preset convergence condition is met. The preliminary solution is projected and decoded, and virtual batches are eliminated during the decoding process to obtain the actual production scheduling scheme for short-process steelmaking; In each iteration, the improved particle swarm optimization algorithm decodes the processing priority order of each production process based on the particle position vector. It then arranges processing machines and start times for each batch according to the principle of minimizing the peak power of the already scheduled parts and the batch coding order. A preset forced placement strategy ensures that all batches are scheduled. The algorithm updates the inertia weight and learning factor of the improved particle swarm optimization algorithm based on the scheduling scheme of each iteration to enhance the global exploration capability when the power peak pressure is high and enhance the local repair capability when the cross-process waiting pressure is high. The preset forced placement strategy includes: if there is no feasible position within the time window of the current batch, searching for the last available gap from back to front within the global time window and forcibly placing it there; and if there is no available gap within the global time window, placing it directly at the end.

2. The method for optimizing the start-up and shutdown scheduling of steel production equipment based on particle swarm optimization as described in claim 1, characterized in that, The step of updating the inertia weights and learning factors of the improved particle swarm optimization algorithm based on the production scheduling scheme of each iteration specifically includes: Calculate the peak power pressure and cross-process waiting pressure of the production scheduling plan in each iteration; The inertial weight is updated based on the peak power pressure, such that when the peak power pressure is greater than a first preset value, the inertial weight increases accordingly. Based on the cross-process waiting pressure, the individual learning factor and the group learning factor are updated so that when the cross-process waiting pressure is greater than the second preset value, the individual learning factor increases accordingly, while the group learning factor decreases accordingly. The velocity and position of the next generation of particles are determined based on the updated inertial weights and learning factors.

3. The steel production equipment start-stop scheduling optimization method based on a particle swarm algorithm according to claim 2, characterized in that, The step of calculating the peak power pressure of the production scheduling scheme in each iteration specifically includes: Determine the peak power of the production plan for each particle, and calculate the ratio of the peak power of the production plan for each particle to the theoretical maximum power, which is used as the peak power pressure index of the corresponding particle. Calculate the average peak power pressure index of all particles to obtain the peak power pressure of the production scheduling scheme in each iteration.

4. The steel production equipment start-stop scheduling optimization method based on a particle swarm algorithm according to claim 2, characterized in that, The step of calculating the cross-process waiting pressure of the production scheduling plan in each iteration specifically includes: Calculate the ratio of the number of waiting constraint violations to the total number of batches to be optimized in the production scheduling plan for each particle, and obtain the average waiting constraint pressure per batch; Take the baseline value and add it to the product of the average waiting constraint pressure per batch and the adjustment coefficient to obtain the waiting pressure index of the corresponding particle. Calculate the average waiting pressure index of all particles to obtain the cross-process waiting pressure of the production scheduling scheme in each iteration.

5. The steel production equipment start-stop scheduling optimization method based on a particle swarm algorithm according to claim 1, characterized in that, The step of receiving production data to perform discrete-time modeling of the short-process steelmaking production process specifically includes: The production data includes shift duration, time granularity, batch data, equipment data used in each process, and maximum waiting time between processes in the short-process steelmaking production process. Based on the production data, a resource task network method is used to define a set of batches, equipment, and time slots to achieve discrete-time modeling of the short-process steelmaking production process.

6. A steel production equipment start-stop scheduling optimization device based on a particle swarm algorithm, characterized in that, include: The receiving module is used to receive production data to perform discrete-time modeling of the short-process steelmaking production process, and to introduce a virtual batch after the actual production batch during the modeling stage to expand the actual production batch to obtain the batch to be optimized. The optimization module is used to construct an objective function with the minimum peak power as the main objective. Under the constraints set in the short-process steelmaking production process, an improved particle swarm optimization algorithm is used to iteratively solve the problem and obtain a preliminary solution when the maximum number of iterations is reached or the preset convergence condition is met. The output module is used to project and decode the pre-decode solution, and to remove virtual batches during the decoding process to obtain the actual production schedule of short-process steelmaking. In each iteration, the improved particle swarm optimization algorithm decodes the processing priority order of each production process based on the particle position vector. It then arranges processing machines and start times for each batch according to the principle of minimizing the peak power of the already scheduled parts and the batch coding order. A preset forced placement strategy ensures that all batches are scheduled. The algorithm updates the inertia weight and learning factor of the improved particle swarm optimization algorithm based on the scheduling scheme of each iteration to enhance the global exploration capability when the power peak pressure is high and enhance the local repair capability when the cross-process waiting pressure is high. The preset forced placement strategy includes: if there is no feasible position within the time window of the current batch, searching for the last available gap from back to front within the global time window and forcibly placing it there; and if there is no available gap within the global time window, placing it directly at the end.

7. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steel production equipment start-up and shutdown scheduling optimization method based on particle swarm optimization as described in any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by the processor, it implements the steel production equipment start-up and shutdown scheduling optimization method based on particle swarm optimization as described in any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steel production equipment start-up and shutdown scheduling optimization method based on particle swarm optimization as described in any one of claims 1 to 5.

Images