Method and system for dynamic scheduling optimization of steel long process production based on time-of-use electricity price

Abstract

This invention provides a dynamic scheduling optimization method and system for long-process steel production based on time-of-use (TOU) pricing. Addressing the problems of high electricity costs and complex process constraints in steel production, it achieves effective reduction in electricity costs for enterprises by deeply integrating a full-process constraint model with TOU pricing strategies. This method comprehensively considers the continuity constraints, timing constraints, and by-product energy balance constraints of core processes such as coking, sintering, blast furnace ironmaking, converter steelmaking, refining, continuous casting, and rolling. It employs intelligent optimization algorithms to rationally arrange the production cycles of each process during peak, valley, and normal periods, precisely scheduling high-energy-consuming processes during off-peak electricity price periods and fully utilizing by-product gas power generation to replace expensive purchased electricity. This invention, while meeting steel production process requirements and quality compliance, can significantly reduce purchased electricity costs and substantially improve the economic benefits and energy efficiency of steel enterprises.

Description

Technical Field

[0001] This invention belongs to the field of dynamic scheduling and optimization technology for long-process steel production. Background Technology

[0002] The long-process steel production uses iron ore as raw material and coke as a reducing agent, and is a typical "continuous-discrete mixed" process, such as... Figure 1 As shown, the core processes include ironmaking processes (coking, sintering / pelletizing), blast furnace ironmaking, converter steelmaking, refining, continuous casting, and rolling, with each process subject to strict constraints.

[0003] Coking and sintering are continuous production processes that require a stable supply of raw materials. Interruptions will lead to equipment wear and capacity loss. Blast furnace ironmaking requires continuous operation to maintain thermal balance within the furnace. The supply and demand of raw materials (coke and sinter) must be matched in real time, and the load cannot be adjusted arbitrarily.

[0004] Continuous casting is based on "casting sessions" (combinations of several heats of molten steel). Each casting session cannot be interrupted and must be strictly synchronized with the output rhythm of the preceding converter steelmaking sessions to prevent the molten steel from cooling and solidifying. The rolling process must match the output sequence of the continuously cast billets, and the reciprocating / unidirectional rolling rhythm of roughing-finishing is fixed, resulting in large power surges.

[0005] Byproduct gas from each process (coke oven gas, blast furnace gas, converter gas) needs to be used in conjunction with gas holders through pipelines (e.g., blast furnace gas for sintering heating, coke oven gas for power generation). Imbalance between gas production and consumption will lead to venting losses or insufficient power generation, increasing reliance on purchased electricity.

[0006] To guide industrial users to use electricity during off-peak hours and improve the efficiency of power grid regulation, many regions in my country have implemented time-of-use (TOU) electricity pricing policies for large industrial users. A day is divided into peak periods (high electricity prices, such as 08:00-11:00 and 17:00-23:00), flat periods (medium electricity prices, such as 07:00-08:00 and 13:00-17:00), and valley periods (low electricity prices, such as 00:00-07:00 and 23:00-24:00). The price difference between peak and valley periods can reach 2-3 times (e.g., 0.83 yuan / kWh for peak periods, 0.55 yuan / kWh for flat periods, and 0.28 yuan / kWh for valley periods).

[0007] Long-process steel enterprises are typical high-energy-consuming users, with purchased electricity costs accounting for 15%-20% of total production costs. However, existing production scheduling suffers from significant shortcomings, including a single optimization dimension, insufficient constraint coordination, and poor algorithm adaptability. Specifically, traditional scheduling only focuses on "capacity targets" without considering time-of-use electricity pricing to adjust production cycles. For example, scheduling high-energy-consuming processes like refining and rolling during peak hours leads to persistently high electricity costs. Optimizing the timing of a single process (such as rolling) ignores the timing constraints of "blast furnace-converter-continuous casting" (such as the uninterrupted nature of continuous casting) or fails to consider the coordination between by-product gas power generation and purchased electricity, making optimization solutions unfeasible. Existing optimization methods often employ simple time-series shifting, failing to consider the "continuous-discrete hybrid" characteristics of long-process steelmaking and thus unable to handle global optimization under multiple constraints (such as raw material supply and demand and by-product gas storage), resulting in limited cost reductions (typically less than 5%). Summary of the Invention

[0008] This invention provides a dynamic scheduling optimization method for long-process steel production based on time-of-use electricity pricing. By constructing an integrated optimization model of "full-process constraints - time-of-use electricity pricing - by-product energy synergy", the timing and load intensity of production cycles of each process are reasonably arranged. Under the premise of meeting the constraints of production continuity, quality compliance and by-product energy balance, the cost of externally purchased electricity for enterprises is minimized.

[0009] Firstly, this invention proposes a dynamic scheduling optimization method for long-process steel production based on time-of-use (TOU) pricing, comprising: collecting and inputting TOU parameters, all-process equipment parameters, production order information, and by-product gas parameters; establishing a basic information database and establishing data interfaces with the manufacturing execution system, energy management system, and distributed control system; based on the data in the basic information database and the technical characteristics of each process in the long process, constructing a full-process constraint model including continuity constraints, timing constraints, by-product energy constraints, and quality and equipment constraints; and, based on the full-process constraint model, taking the minimization of electricity purchase costs within and outside the scheduling cycle as the optimization objective, integrating the cost deduction effect of by-product gas self-generation, and constructing a cost objective function under TOU pricing. A hybrid algorithm combining mixed-integer linear programming and particle swarm optimization is employed, using the cost objective function as the fitness function, to globally optimize the start time, load intensity, and gas allocation ratio of each production cycle. Under the premise of satisfying all constraints of the full-process constraint model, an optimal production cycle plan is generated. This optimal production cycle plan is then issued to the distributed control system, and production data of various categories contained in the basic information database are collected in real time from the distributed control system. When the actual value of any production data deviates from the expected value corresponding to the optimal production cycle plan by more than a set threshold, the optimization solution steps are re-executed based on the updated basic information database to generate an adjusted production cycle plan.

[0010] In some examples, the continuity constraints include: minimum continuous operating time window constraints for coking and sintering processes; raw material input rate fluctuation range constraints for blast furnace ironmaking; uninterrupted constraints for continuous casting processes on a per-casting-time basis; and matching constraints between the start time of a casting and the output time of molten steel from the preceding converter.

[0011] In some examples, the timing constraints include: establishing a timing correlation matrix for ironmaking-blast furnace-converter-refining-continuous casting, constraining the matching deviation range between blast furnace molten iron production and converter molten iron demand, and the lag interval between the refining furnace receiving time and the converter tapping time.

[0012] In some examples, the by-product energy constraints include: constructing a production-consumption balance model for coke oven gas, blast furnace gas, and converter gas; setting upper and lower limits for gas holder storage capacity; constraining the upper limit for gas venting rate; and establishing a scheduling strategy that prioritizes gas-fired power generation for high-power-consuming processes during peak periods.

[0013] In some examples, the cost objective function is expressed as:

[0014] in, For the scheduling period, for Time-of-use electricity pricing for different time periods for Electricity consumption purchased by enterprises during the period for Electricity generated from by-product coal gas during certain periods.

[0015] In some examples, the cost objective function incorporates penalty mechanisms, including: penalties for interruption of continuous casting, insufficient continuous operation time of coking furnace, excessive gas venting rate, excessive deviation between blast furnace molten iron and converter demand, and deviation in refining furnace receiving time.

[0016] In some examples, the hybrid integer linear programming (HIM) and particle swarm optimization (PSO) fusion algorithm includes: an initialization phase where the PSO algorithm randomly generates production round schemes that satisfy the process sequence and equipment capacity limits, serving as the initial dataset for HIM; a local optimization phase where HIM uses the continuous variables output by the PSO algorithm as boundary conditions to solve for the integer variables in each production round scheme, with the uninterrupted continuous casting and the upper limit of gas emission rate as mandatory constraints; and a global iteration phase where the PSO algorithm updates the particle positions and velocities based on the locally optimal production round scheme optimized by HIM, using the minimization of purchased electricity cost as the fitness function. If the HIM solution finds that a production round scheme violates the mandatory constraints, it is fed back to the PSO algorithm to penalize the production round scheme. The two iterate collaboratively until global convergence.

[0017] In some examples, the optimization solution steps also include time-segmented load scheduling strategies: prioritizing high-power-consuming processes and increasing purchased electricity load during off-peak electricity price periods, reducing gas-fired power generation, and storing by-product gas in gas holders; reducing purchased electricity load during peak electricity price periods, prioritizing the use of gas in gas holders for power generation to replace purchased electricity; and coordinating the load of each process during flat electricity price periods to achieve a smooth transition between purchased electricity and gas-fired power generation, and prioritizing continuous casting in flat periods.

[0018] Secondly, this invention proposes a dynamic scheduling optimization system for long-process steel production based on time-of-use electricity pricing, comprising: a data acquisition and initialization module, configured to collect and input time-of-use electricity pricing parameters, all-process equipment parameters, production order information, and by-product gas parameters, establish a basic information database, and establish data interfaces with the manufacturing execution system, energy management system, and distributed control system, while receiving real-time production data feedback from the scheme output and dynamic adjustment module; a full-process constraint model module, configured to construct a full-process constraint model including continuity constraints, timing constraints, by-product energy constraints, and quality and equipment constraints based on the data in the basic information database and the technical characteristics of each process in the long process; and an objective function and optimization algorithm module, configured to construct a cost objective function under time-of-use electricity pricing based on the full-process constraint model, with the optimization objective being the minimization of electricity purchase costs within and outside the scheduling cycle, integrating the cost deduction effect of by-product gas self-generation, and employing mixed integer linear programming and particle swarm optimization. The algorithm fusion method performs global optimization on the start time, load intensity, and gas allocation ratio of production cycles to generate an optimal production cycle scheme that satisfies all constraints. The scheme output and dynamic adjustment module is configured to receive the optimal production cycle scheme and issue it to the distributed control system. Simultaneously, it collects production data of various categories contained in the basic information database from the distributed control system in real time and feeds the data back to the data acquisition and initialization module. When the actual value of any production data deviates from the expected value corresponding to the optimal production cycle scheme by more than a set threshold, the objective function and optimization algorithm module is triggered to re-execute the optimization solution based on the updated basic information database to generate an adjusted production cycle scheme. The effect evaluation and self-learning module is configured to periodically evaluate the implementation effect of the system and analyze historical operating data through machine learning methods to continuously correct the parameters of the full-process constraint model module and optimize the algorithm configuration of the objective function and optimization algorithm module.

[0019] Thirdly, the present invention proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the aforementioned method for dynamic scheduling optimization of long-process steel production based on time-of-use electricity pricing.

[0020] This invention comprehensively considers the continuity constraints, timing constraints, and by-product energy balance constraints of core processes such as coking, sintering, blast furnace ironmaking, converter steelmaking, refining, continuous casting, and rolling. It employs an intelligent optimization algorithm to rationally schedule the production cycles of each process during peak, valley, and normal periods, precisely arranging high-power-consuming processes during off-peak electricity prices and fully utilizing by-product coal gas power generation to replace expensive purchased electricity. Implementation results show that, while meeting production process requirements and quality compliance, this invention can reduce purchased electricity costs by 12%-18%, significantly improving the economic benefits and energy efficiency of steel enterprises. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the steel production process.

[0022] Figure 2 This is a flowchart of an embodiment of the present invention of a dynamic scheduling optimization method for long-process steel production based on time-of-use electricity pricing.

[0023] Figure 3 This is a time-of-use electricity price chart set according to an embodiment of the present invention.

[0024] Figure 4 This is an optimized by-product gas generation curve according to an embodiment of the present invention.

[0025] Figure 5 This is an optimized balance diagram of by-product gas consumption and storage according to an embodiment of the present invention.

[0026] Figure 6 This is a Gantt chart of the production arrangement of each process after optimization according to an embodiment of the present invention. Detailed Implementation

[0027] Current electricity price optimization technologies for steel companies mostly focus on short-process steelmaking (with electric arc furnace as the core) or a single process, which cannot adapt to the constraints of the entire process in long processes. Although some technologies mention time-of-use electricity pricing, they do not integrate the cost deduction effect of self-generated electricity from by-product gas, nor do they solve rigid constraints such as the continuity of continuous casting and the thermal balance of blast furnaces. As a result, optimization schemes are prone to causing quality problems or capacity losses in actual production, making it difficult to meet the dual needs of enterprises for "cost reduction and compliance".

[0028] The core of this invention lies in the deep integration and optimization of time-of-use electricity pricing policies, complex production process constraints, and the utilization of by-product energy, thereby minimizing the electricity costs for steel enterprises.

[0029] This invention is applicable to long-process steel enterprises with "coking-sintering-blast furnace ironmaking-converter steelmaking-refining-continuous casting-rolling" as the core process. It aims to minimize the enterprise's electricity costs by optimizing the timing of production cycles in each process, while meeting constraints such as production continuity, quality compliance, and synergistic utilization of by-product energy.

[0030] Example 1: A dynamic scheduling optimization method for long-process steel production based on time-of-use electricity pricing. The following section combines... Figure 2 Please explain this method in detail.

[0031] Step 1: System initialization and data preparation.

[0032] 1.1 Construct a basic information database, including inputting electricity price parameters, equipment parameters, production plans, and energy parameters.

[0033] Clarify the time-of-use electricity pricing policy in the enterprise's location, precisely divide the day into three periods: peak, flat, and valley, and record the electricity price for each period (e.g., peak period 0.83 yuan / kWh, flat period 0.55 yuan / kWh, valley period 0.28 yuan / kWh).

[0034] Enter the technical parameters of core equipment for each process (coking, sintering, blast furnace, converter, refining, continuous casting, rolling, etc.), including rated power, processing capacity, minimum / maximum operating load, start-up and shutdown characteristics, etc.

[0035] Import production orders within the planned cycle (e.g., the next 24 hours), including product specifications, output requirements, processing time for each process, and composition of continuous casting cycles.

[0036] Set the production coefficient, calorific value, upper and lower limits of gas holder capacity, generator set efficiency, and ramp rate of by-product gas (coke oven gas, blast furnace gas, converter gas).

[0037] 1.2 Establish communication and data interfaces.

[0038] Establish data interfaces with the enterprise's existing Manufacturing Execution System (MES), Energy Management System (EMS), and Distributed Control System (DCS) to ensure that production plans, real-time equipment status, and energy data can flow into the optimization system automatically and accurately.

[0039] Step 2: Construct a full-process constraint model for long-process steelmaking.

[0040] Based on the technical characteristics of each process in a long process, clear constraint boundaries are defined, including: continuity constraints, timing constraints, by-product energy constraints, and quality and equipment constraints.

[0041] 2.1 Continuity Constraints. Define continuous production time windows for coking and sintering, and constrain the raw material supply rate for blast furnace ironmaking to avoid heat loss due to interruptions. For example, a single continuous operation of the coking furnace should be ≥72 hours, and the fluctuation in coke / sinter input should be ≤5% / hour. The continuous casting process is based on "casting sessions," setting the duration of a single casting session and the interval between sessions. For example, the duration of a single casting session is 2-4 hours, and the interval between sessions should be ≤1 hour. Constraints are placed on the fact that heats within the same casting session cannot be split, and the start time of the casting session must match the steel output time of the preceding converter steelmaking process; for example, the steel waiting time should be ≤30 minutes.

[0042] 2.2 Timing Constraints. Establish a timing correlation matrix for "ironmaking-blast furnace-converter-refining-continuous casting". For example, the blast furnace molten iron output must match the molten iron demand of converter steelmaking (deviation ≤10%), and the molten steel receiving time of the refining furnace must lag the converter tapping time by ≥10 minutes and ≤20 minutes to avoid a drop in molten steel temperature.

[0043] 2.3 Constraints on secondary production capacity. Construct a production-consumption balance model for coke oven gas, blast furnace gas, and converter gas, and set upper and lower limits for gas holder storage, such as blast furnace gas holder capacity ≥ 20% of daily output; constrain gas-fired power generation to be prioritized for high-power-consuming processes during peak periods (such as steel rolling), and ensure that the gas venting rate is ≤ 2%.

[0044] 2.4 Quality and Equipment Constraints. Set quality indicators such as carbon content of blast furnace molten iron (4.2%-4.5%) and decarburization rate of converter molten steel (≥90%), and set heating temperature (hot-rolled billet heating temperature 1100-1250℃) and rolling force constraints for the rolling process to avoid product scrap due to excessive parameters.

[0045] Step 3: Construct the cost objective function under time-of-use pricing.

[0046] With the core objective of minimizing the cost of purchased electricity within a company's cycle (e.g., 1 day), and integrating the cost-deductible effect of self-generated electricity from by-product coal gas, the objective function is as follows:

[0047] in: The scheduling period is (e.g., 1440 minutes, discretized into 96 time periods at 15-minute intervals). for Time-of-use electricity pricing for different time periods (peak / short / valley periods are charged at corresponding prices, such as 0.83 / 0.55 / 0.28 yuan / kWh); for The electricity consumption (kW) purchased by enterprises during the specified period must meet the grid's incoming capacity constraints (e.g., ≤200MW). for The power generation (kW) from by-product gas during a given period is calculated based on the output and power generation efficiency (35%) of by-product gas from each process, and is prioritized for use during peak hours. The 35% figure represents the industry benchmark average for mixed power generation from by-product gas in long-process steel enterprises and will vary depending on the type of gas and the load on the generator set.

[0048] Meanwhile, the following penalty mechanisms and corresponding setting principles are introduced: (1) The penalty cost for interruption of continuous casting is ≥ 50,000 yuan / time, to ensure the continuity of the continuous casting process and avoid the scrapping of molten steel due to cooling and solidification and equipment wear; (2) The penalty cost for continuous operation of coking furnace < 72 hours is ≥ 70,000 yuan / time, to ensure the stability of the coking process and avoid coke quality fluctuations and capacity loss; (3) When the gas release rate is > 2%, the penalty cost = release amount × unit energy value of gas × 1.2, to ensure full utilization of by-product gas and reduce energy waste; (4) The penalty cost for deviation between blast furnace molten iron and converter demand > 10% is 50,000 yuan / time, to ensure the matching of molten iron supply and demand and avoid converter waiting for materials or molten iron waste; (5) The penalty cost for the time of receiving molten steel in refining furnace deviating from the range of 10-20 minutes is 20,000 yuan / time, to ensure that the temperature of molten steel meets the refining requirements and ensure refining quality.

[0049] Step 4: Production round optimization algorithm.

[0050] A hybrid algorithm of "mixed integer linear programming-particle swarm optimization" (MILP-PSO) is adopted to achieve global optimization under multiple constraints.

[0051] 4.1 Variable Definition. Decision variables are defined as: the start time of each production cycle (e.g., the start time of the i-th batch of molten iron in the blast furnace). ), load intensity (such as the power of the j-th cycle of the sintering machine) ), gas allocation ratio (e.g., the proportion of blast furnace gas used for power generation during time period t) ).

[0052] 4.2 Constraint Embedding. The continuity, temporal coherence, by-product energy constraints, and quality and equipment constraints from step 1 are transformed into mathematical inequalities (e.g., minute, (minutes), embedding algorithm constraint matrix.

[0053] 4.3 Iterative Optimization. Initialize the particle swarm (each particle corresponds to a set of production round schemes), and iteratively update the particle positions using the cost function from step 2 as the fitness function.

[0054] Initialization phase (PSO-led): The particle swarm algorithm randomly generates a set of production round plans that satisfy basic constraints (such as the order of processes and the upper limit of equipment capacity). Each particle corresponds to a complete plan, including time variables, load variables, and allocation variables. This set serves as the initial input dataset for MILP, quickly covering the global feasible region.

[0055] In the local optimization stage (MILP-led), for each particle's corresponding scheme, local fine optimization is carried out through MILP: (1) The continuous variables output by PSO (such as rolling load intensity) are used as boundary conditions to accurately solve integer variables (such as the start time of continuous casting and the arrangement of converter smelting cycles); (2) Rigid constraints are strictly satisfied (such as continuous casting cannot be interrupted and gas release rate ≤2%), and the constraints are transformed into linear inequalities and embedded in the MILP model to ensure the feasibility of the scheme.

[0056] Global Iteration Phase (both work together): (1) PSO receives the local optimal solution after MILP optimization, updates the particle position and velocity with "minimum external power cost" as the fitness function, and guides the population to converge to the global optimal region; (2) If MILP finds that a particle's solution violates rigid constraints (such as timing connection deviation exceeding 10%), it is fed back to PSO, and a penalty term is applied to the particle (such as increasing the fitness function value) to avoid repeated invalid searches in subsequent iterations.

[0057] During off-peak hours (00:00-07:00), high-power-consuming processes (such as refining and rough rolling) are prioritized, and the load of purchased electricity is increased (e.g., rolling power is increased to 90% of the rated value), while reducing gas-fired power generation (gas is stored in gas holders). During peak hours (08:00-11:00, 17:00-23:00), the load of purchased electricity is reduced; processes such as fine rolling are scheduled at the beginning / end of the cycle, and gas-fired power generation is prioritized (e.g., gas-fired power generation is increased to 80% of the rated value), while blast furnaces and converters maintain basic loads (to avoid disrupting the heat balance). During flat hours (07:00-08:00, 13:00-17:00), the loads of various processes are coordinated to achieve a smooth transition from purchased electricity to gas-fired power generation. For example, continuous casting is prioritized during flat hours to avoid waiting for molten steel during peak hours.

[0058] Figure 3 The time-of-use electricity price curve chart clearly defines the time intervals and corresponding electricity prices for peak periods (08:00-11:00, 17:00-23:00, 0.83 / kWh), flat periods (07:00-08:00, 13:00-17:00, 0.55 yuan / kWh), and valley periods (00:00-07:00, 23:00-24:00, 0.28 yuan / kWh). This not only matches the parameter description of peak-valley electricity price difference in the background technology, but also provides an intuitive basis for the logic of "prioritizing high-energy-consuming processes during valley periods, prioritizing the use of gas-fired power generation during peak periods, and coordinating load transition during flat periods" in the iterative optimization strategy. At the same time, it provides a specific calculation value for the "time-of-use electricity price for period t" in the cost objective function.

[0059] 4.4 Convergence Judgment. When the cost difference of 10 consecutive iterations is ≤0.1%, the optimal production cycle plan (including the start time of each process cycle, load intensity, and gas allocation ratio) is output.

[0060] Step 5: Optimize the output and dynamically adjust the solution.

[0061] Generate a visual production cycle plan, clearly defining the production arrangements for each process during peak, valley, and flat periods. Simultaneously, output a time-slot allocation table for gas-fired power generation and purchased electricity. Specific production instructions are then transmitted to the MES and DCS systems via interfaces to guide on-site production.

[0062] The system collects production data for each process in real time (such as blast furnace gas output and molten steel temperature). If the actual value deviates from the optimized value by ≥10% (e.g., a sudden drop in gas output), the algorithm is triggered to iterate again and adjust subsequent production cycles (e.g., shifting the peak-slot rolling process to the flat-slot). For deviations, the system can initiate local or global re-optimization to quickly calculate and generate adjusted production plans (e.g., shifting the originally planned peak-slot rolling process to the flat-slot), thereby ensuring production operates on a cost-optimal path.

[0063] Figure 4 The optimized by-product gas production curves show the trends in the production and total output of coke oven gas, blast furnace gas, and converter gas during the scheduling cycle. This reflects the differences in the timing of the three types of gas production, such as the continuous and stable output of blast furnace gas and the fluctuation of converter gas output with each smelting cycle. This aligns with the modeling object of "constructing a production-consumption balance model for the three types of gas" in the by-product energy constraint, verifying the feasibility of "prioritizing the use of gas for power generation during peak periods" in the iterative optimization. It also provides data feature support for the scenario setting of "sudden drop in gas production triggering re-optimization" in the dynamic adjustment mechanism.

[0064] Figure 5 To optimize the balance diagram of by-product gas consumption and storage, the dynamic balance relationship of total gas production, consumption, storage and release is presented. The constraint boundaries of "release rate ≤ 2%" and "gas holder capacity > daily production 20%" are marked. This intuitively verifies the goal of reducing gas release and making full use of by-product energy in the by-product energy constraint. It is completely consistent with the operational logic of "reducing gas power generation and storing it in gas holders during off-peak hours, and using gas power generation to replace purchased electricity during peak hours" in the iterative optimization. It clearly reflects the core path of gas self-generation to offset the cost of purchased electricity.

[0065] Figure 6To optimize the production schedule of each process, a Gantt chart is used to display the operating status and corresponding power of processes such as blast furnace, coking, sintering, converter, refining, continuous casting, and rolling at different times in the form of a time axis. For example, rolling runs from 03:00 to 08:00 (power 50MW) and continuous casting runs from 13:00 to 16:00 (power 20MW). This intuitively presents the optimization logic of "prioritizing high power-consuming processes during off-peak hours, arranging continuous casting during flat periods, and controlling external power load during peak periods" in the iterative optimization. It is a specific expression of the "visualized production cycle plan table" in the output of the optimization scheme, and at the same time, it verifies the implementation effect of the time sequence connection constraints, such as the time sequence matching between converter and blast furnace, and between continuous casting and converter.

[0066] Step 6: Effectiveness evaluation and continuous system optimization.

[0067] The system's implementation effectiveness will be evaluated regularly, with key indicators including: the reduction in purchased electricity costs (expected to reach 12%-18%), the rate of by-product gas release, the rate of continuous casting interruption, and the steel qualification rate. The system accumulates historical operational data, and through machine learning methods, continuously refines the parameters of the constraint model and optimizes the algorithm configuration to make future optimization decisions more accurate, forming a closed loop of continuous improvement.

[0068] Example 2: A dynamic scheduling and optimization system for long-process steel production based on time-of-use electricity pricing. The system includes: a data acquisition and initialization module, a full-process constraint model, an objective function and optimization algorithm module, a scheme output and dynamic adjustment module, and an effect evaluation and self-learning module.

[0069] The data acquisition and initialization module is configured to build a basic information database, including inputting electricity price parameters, equipment parameters, production plans, and energy parameters. It clarifies the time-of-use electricity pricing policy for the company's location, precisely dividing the day into peak, flat, and valley periods, and inputs the electricity price for each period. It inputs the technical parameters of equipment for each process step, including rated power, processing capacity, minimum / maximum operating load, and start-up / shutdown characteristics. It imports production orders within the planned cycle, including product specifications, output requirements, processing time for each step, and continuous casting batch composition. It sets the by-product gas generation coefficient, calorific value, upper and lower limits of gas holder capacity, generator set efficiency, and ramp-up rate.

[0070] The data acquisition and initialization module is also responsible for establishing communication and data interfaces, and establishing data interfaces with the enterprise's existing Manufacturing Execution System (MES), Energy Management System (EMS) and Distributed Control System (DCS) to ensure that production plans, real-time equipment status and energy data can automatically and accurately flow into the optimization system.

[0071] The full-process constraint model is constructed based on the technical characteristics of each process in a long process, including continuity constraints, timing constraints, by-product energy constraints, and quality and equipment constraints. The specific constraint definitions are consistent with step 2 in Example 1.

[0072] The objective function and optimization algorithm module are configured with "minimizing the enterprise's internal and external electricity purchase costs" as the core objective, integrating the cost deduction effect of self-generated electricity from by-product coal gas to construct a cost objective function. A "mixed-integer linear programming-particle swarm optimization" fusion algorithm is used for global optimization solution, and the specific algorithm flow is consistent with steps 3 and 4 in Example 1.

[0073] The output and dynamic adjustment module is configured to generate a visual production cycle plan, clearly defining the production arrangements for each process during peak, valley, and flat periods. It also outputs a time-slot allocation table for gas-fired power generation and purchased electricity. Specific production instructions are transmitted to the MES and DCS systems via interfaces to guide on-site production. Real-time collection of actual production data for each process is conducted; if deviations exceed set thresholds, the algorithm is triggered to iterate again, adjusting subsequent production cycles.

[0074] The effectiveness evaluation and self-learning module is configured to periodically evaluate the system's implementation effectiveness. Key indicators include the reduction in purchased electricity costs, the rate of by-product gas release, the interruption rate of continuous casting, and the steel qualification rate. The system accumulates historical operating data and uses machine learning methods to continuously correct the parameters of the constraint model and optimize the algorithm configuration, forming a closed loop of continuous improvement.

[0075] Example 3: A computer system.

[0076] A computer system includes a processor and memory. The memory and processor are interconnected via a bus system and / or other forms of connection. The processor may include any one or more processors such as a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP). The memory may include volatile memory, such as random access memory (RAM). The memory may also include non-volatile memory. volatile memory), such as read-only memory (read-only memory). The memory can be a ROM (Mobile On Demand), flash memory, hard disk drive (HDD), or solid-state drive (SSD). The memory stores executable program code, which the processor executes to implement the time-of-use electricity pricing-based dynamic scheduling optimization method for long-process steel production described in Example 1. In other words, the memory stores instructions for executing the time-of-use electricity pricing-based dynamic scheduling optimization method for long-process steel production.

[0077] Example 4: A computer-readable storage medium.

[0078] A computer-readable storage medium, for example, a non-transitory computer-readable storage medium, such as a read-only memory (ROM). Read-only memory (ROM), flash memory, and compact disc (CD-ROM) Only memory, CD ROM, magnetic tape, floppy disk, and optical data storage devices, etc. This computer-readable storage medium is used to store non-transitory computer-readable instructions, which, when executed by a computer, can implement one or more steps in the dynamic scheduling optimization method for long-process steel production based on time-of-use electricity pricing described in Example 1.

Claims

1. A dynamic scheduling optimization method for long-process steel production based on time-of-use electricity pricing, characterized in that, include: Collect and input time-of-use electricity price parameters, full-process equipment parameters, production order information and by-product gas parameters, establish a basic information database, and establish data interfaces with the manufacturing execution system, energy management system and distributed control system; Based on the data in the aforementioned basic information database and the technical characteristics of each process in the long process, a full-process constraint model is constructed, which includes continuity constraints, timing constraints, by-product energy constraints, and quality and equipment constraints. Based on the full-process constraint model, with the optimization objective of minimizing the electricity purchase cost inside and outside the scheduling cycle, the cost deduction effect of self-generated electricity from by-product gas is integrated to construct a cost objective function under time-of-use pricing. A hybrid algorithm combining mixed integer linear programming and particle swarm optimization is adopted. Using the cost objective function as the fitness function, the start time, load intensity, and gas allocation ratio of the production cycle are globally optimized. Under the premise of satisfying all constraints of the full-process constraint model, the optimal production cycle scheme is generated. The optimal production cycle plan is issued to the distributed control system, and production data of the categories contained in the basic information database are collected from the distributed control system in real time. When the actual value of any production data deviates from the expected value corresponding to the optimal production cycle plan by more than a set threshold, the optimization solution step is re-executed based on the updated basic information database to generate an adjusted production cycle plan.

2. The method for dynamic scheduling optimization of long-process steel production based on time-of-use electricity pricing as described in claim 1, characterized in that, The continuity constraints include: minimum continuous operating time window constraints for coking and sintering processes; raw material input rate fluctuation range constraints for blast furnace ironmaking; uninterrupted constraints for continuous casting processes in terms of casting cycles; and matching constraints between casting cycle start time and the output time of molten steel from the preceding converter.

3. The method for dynamic scheduling optimization of long-process steel production based on time-of-use electricity pricing according to claim 1, characterized in that, The time-series connection constraints include: establishing a time-series correlation matrix for ironmaking-blast furnace-converter-refining-continuous casting, constraining the matching deviation range between blast furnace molten iron production and converter molten iron demand, and the lag interval between the refining furnace receiving time and the converter tapping time.

4. The method for dynamic scheduling optimization of long-process steel production based on time-of-use electricity pricing according to claim 1, characterized in that, The by-product energy constraints include: constructing a production-consumption balance model for coke oven gas, blast furnace gas, and converter gas; setting upper and lower limits for gas holder storage capacity; constraining the upper limit for gas release rate; and establishing a scheduling strategy that prioritizes the use of gas-fired power generation for high-power-consuming processes during peak periods.

5. The method for dynamic scheduling optimization of long-process steel production based on time-of-use electricity pricing according to claim 1, characterized in that, The cost objective function is expressed as follows: in, For the scheduling period, for Time-of-use electricity pricing for different time periods for Electricity consumption purchased by enterprises during the period for Electricity generated from by-product coal gas during certain periods.

6. The method for dynamic scheduling optimization of long-process steel production based on time-of-use electricity pricing according to claim 1, characterized in that, The cost objective function introduces a penalty mechanism, including: penalty for interruption of continuous casting, penalty for insufficient continuous operation time of coking furnace, penalty for excessive gas emission rate, penalty for excessive deviation between blast furnace molten iron and converter demand, and penalty for deviation in the receiving time of molten steel in refining furnace.

7. The method for dynamic scheduling optimization of long-process steel production based on time-of-use electricity pricing according to claim 1, characterized in that, The hybrid integer linear programming and particle swarm optimization algorithm includes: Initialization phase: The particle swarm optimization algorithm randomly generates production round schemes that satisfy the process sequence and equipment capacity limit, which serve as the initial dataset for mixed-integer linear programming; Local optimization stage: The mixed integer linear programming uses the continuous variables output by the particle swarm algorithm as boundary conditions to solve the integer variables in each production round scheme, and in the solution process, the continuous casting cycle cannot be interrupted and the upper limit of the gas emission rate are used as mandatory constraints. Global Iteration Phase: The particle swarm optimization algorithm is based on the locally optimal production round scheme optimized by mixed-integer linear programming. It updates the particle position and velocity with the fitness function of minimizing the cost of purchased electricity. If the mixed-integer linear programming solution finds that the production round scheme violates the mandatory constraints, it is fed back to the particle swarm optimization algorithm to impose a penalty on the production round scheme. The two work together to iterate until global convergence.

8. The method for dynamic scheduling optimization of long-process steel production based on time-of-use electricity pricing according to claim 1, characterized in that, The optimization solution steps also include a time-based load scheduling strategy: High-energy-consuming processes are scheduled during off-peak electricity prices, and the load of purchased electricity is increased to reduce coal gas power generation and store by-product coal gas in gas holders. During peak electricity price periods, reduce the load on purchased electricity and use gas stored in gas holders to generate electricity instead of purchased electricity; During the flat electricity price period, the load of each process is coordinated to achieve a smooth transition between purchased electricity and gas-fired power generation, and the continuous casting process is given priority in the flat electricity price period.

9. A dynamic scheduling optimization system for long-process steel production based on time-of-use electricity pricing, characterized in that, include: The data acquisition and initialization module is configured to collect and input time-of-use electricity price parameters, full-process equipment parameters, production order information and by-product gas parameters, establish a basic information database, and establish data interfaces with the manufacturing execution system, energy management system and distributed control system, while receiving real-time production data feedback from the scheme output and dynamic adjustment module; The full-process constraint model module is configured to construct a full-process constraint model based on the data in the basic information database and the technical characteristics of each process in the long process, including continuity constraints, timing constraints, by-product energy constraints, and quality and equipment constraints. The objective function and optimization algorithm module is configured to, based on the full-process constraint model, take minimizing the electricity purchase cost inside and outside the scheduling cycle as the optimization objective, integrate the cost deduction effect of self-generated electricity from by-product gas, construct a cost objective function under time-of-use pricing, and use a hybrid integer linear programming and particle swarm optimization algorithm to globally optimize the start time, load intensity and gas allocation ratio of the production cycle, generating the optimal production cycle scheme that satisfies all constraints. The scheme output and dynamic adjustment module is configured to receive the optimal production round scheme and issue it to the distributed control system. At the same time, it collects production data of the categories contained in the basic information database from the distributed control system in real time and feeds the data back to the data acquisition and initialization module. When the actual value of any production data deviates from the expected value corresponding to the optimal production round scheme by more than a set threshold, the objective function and optimization algorithm module is triggered to re-execute the optimization solution based on the updated basic information database to generate an adjusted production round scheme. The effect evaluation and self-learning module is configured to periodically evaluate the implementation effect of the system, analyze historical operating data through machine learning methods, continuously correct the parameters of the full-process constraint model module, and optimize the algorithm configuration of the objective function and optimization algorithm module.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the dynamic scheduling optimization method for long-process steel production based on time-of-use electricity pricing as described in any one of claims 1 to 8.

Images