Optimization method for multi-process energy consumption in iron and steel enterprises under the coordinated use of comprehensive energy

Sustainable development goals, such as energy conservation, emission reduction, cost reduction, and efficiency enhancement, present pressing challenges for the iron and steel industry. Process optimization technology for efficient production, energy conservation, and consumption reduction in the steel manufacturing process is based on the coordination of material and energy flow. Therefore, this article takes the production process of a converter steelmaking plant as the object, considers the regulatory requirements of multi-process and multi constraint operation optimization under the cooperation of production organization and energy conservation and consumption reduction, and it explores a production scheduling optimization method for steelmaking plants oriented towards material flow and energy flow synergy. Based on the characteristics of the problem space, this study suggests a hybrid genetic algorithm with local search (MOHGALS) that employs both standard and advanced evolution techniques. Tests on concrete problems demonstrate that the MOHGALS approach yields a superior set of multi-objective Pareto schemes. Additionally, random-generated test cases were used to evaluate the algorithm’s performance. The outcomes show that MOHGALS performs better than competing algorithms on multi-objective evaluation indicators, and that the suggested improved evolution approach may effectively boost the algorithm’s performance.


Introduction
The steel industry is a cornerstone of China's economy and a key component of the country's military strategy.Steel products have several applications in building, machinery, and shipbuilding, among others, and have a substantial impact on those sectors.Overcapacity and a lack of premium goods are two major structural imbalances affecting the steel sector right now [1].Companies in the steel industry are under scrutiny for their high pollution emissions, excessive use of resources and energy, and high manufacturing costs as a result of supply side reform efforts.There is an immediate need for intelligent manufacturing transformation and green manufacturing practices in the steel industry.Achieving efficient production, energy conservation, and consumption reduction in process operation optimization will be made possible by technology that optimizes production operations with a focus on the synergy of material and energy flow in the steel manufacturing process [2].To achieve multiobjective optimization of product quality, low cost, low energy consumption, and low emissions, the steel manufacturing process relies on the dynamic and ordered flow of ferrite material, driven and acted upon by energy flow, along a specific process network according to a set operating program [3].Business and academic communities have paid a lot of attention to the optimization of material flow operation and energy flow operation for steel manufacturing processes in recent years, thanks to the improvement of digitalization and informatization infrastructure of steel enterprises.The majority of the current literature examines the flow of either materials or energy, defines the features of the operational optimization issues involving these flows, and then develops mathematical models to explain them.The optimization of material flow operations has been the subject of much related research, but there is a lack of specificity in the results, both theoretically and in terms of their practical applications [4].Most of the prior work on this topic has focused on production scheduling and planning in order to connect and match the material flow time parameters.However, the material flow operation's temperature drop control goals and the material flow's temperature parameters are typically overlooked in the present study.The material flow and energy flow in the manufacturing units of the main processes are intricately related and influenced by one another in real production.There has been a lack of focus on the interdependence and shared objectives of material and energy flows on production units in the available literature, which treats these systems as separate entities [5].
Static and deterministic modeling approaches form the backbone of most prior work on optimizing material flow operations and energy flow operations, but these approaches fail to adequately account for the unpredictable aspects of production processes, making the model methods less applicable to real-world scenarios.Intelligent optimization and control of production material flow operations is currently challenging for Chinese steel businesses due to their reliance on manual scheduling expertise for real production organization and regulation.Complicating matters further, real manufacturing often encounters issues with energy dissipation or delayed energy supply, which disrupts smooth operations, increases energy consumption costs, and decreases the effectiveness of the synergy between material and energy flows.For this reason, steel companies place a premium on studies examining algorithms and models for optimizing production schedules that account for the interaction between material and energy flows.Improved material flow and energy flow synergy efficiency may lead to lower production costs, more efficient energy utilization, and overall improved production efficiency [6].The production system of a converter steelmaking plant experiences energy loss and heat conversion as a result of the material flow process.Plus, the main manufacturing unit of the process has very linked flows of materials and energy.The interaction between the material and energy flows has a significant impact on the overall process system's efficiency and energy consumption in a converter steelmaking facility.
Consequently, this article delves into the topic of material flow temperature drop control as it pertains to the production scheduling optimization problem.It then goes on to develop a model for optimizing scheduling in steelmaking facilities and an algorithm to solve it.By delving into the details of high-temperature operation in steelmaking plants and incorporating the conventional scheduling model with the minimum casting superheat and target tapping temperature constraints, this paper develops a multi-objective mathematical programming model specific to the problem at hand.In this research, we provide a multi-objective genetic algorithm that combines neighborhood search technology with an enhanced evolution method.

Related work
Reference [7] proposes a method for predicting and scheduling oxygen in steel mills.Firstly, an energy consumption prediction model based on granular computing was established, and a mixed integer programming model for optimal scheduling of oxygen and nitrogen systems was established.In terms of energy consumption analysis and system energy-saving in the steel manufacturing process, reference [8] proposes energy-saving theories and technologies for the steel industry system based on concepts such as energy carriers and systems, and it proposes an energy consumption analysis method for steel production logistics based on benchmark logistics maps, known as the "e-p analysis method".Reference [9] constructed an integrated process model for minimizing energy consumption and waste emissions by analyzing the relationship between the overall and unit consumption and emissions in steel mills, providing decision-making assistance for promoting energy conservation and environmental protection.Reference [10] proposed a algorithm to solve complex production scheduling problems in steel mills.
Reference [11] proposes a combination of improved PSO algorithm and greedy algorithm for solving steelmaking production scheduling problems.The complete solution is constructed through greedy algorithm and iterative optimization is achieved based on PSO algorithm.The production scheduling problem in steel mills has attracted considerable attention in the literature.In this regard, reference [12] introduced an efficient artificial bee colony algorithm specifically tailored for addressing this problem.Moreover, a heuristic algorithm was devised, taking into account the problem characteristics, to expedite the attainment of viable solutions.Alternatively, reference [13] proposed a two-stage solution approach for the production scheduling problem in steelmaking plants by merging decentralized search techniques with mixed integer programming, with a particular emphasis on ladle allocation considerations.Additionally, reference [14] contributed to this area by presenting a hybrid solution method that combines genetic algorithm and linear programming techniques.This approach was designed to tackle the production scheduling problem in steelmaking plants while considering process jumping and adjustable processing time factors.This method involves using GA to quickly optimize problems, and then further improving solution quality through linear programming algorithms.

Optimization method for energy consumption
It is assumed that the process operations, ladle temperature conditions, and insulation conditions for furnace transit and waiting stay constant, which simplifies issue modeling and solution.Therefore, the process's temperature decline may be represented as a linear function that depends solely on time.
where  is the total time of transportation and waiting, and  0 is the tapping process.The temperature drop statistic caused by various factors such as ladle, where  1 represents the statistical unit time temperature drop coefficient.Since the pouring schedule has been determined and the planned start time of each pouring has been formulated in advance, the primary goal of energy consumption optimization is to optimize the deviation between the actual start time of each pouring and the planned start time, while ensuring the continuous pouring of each pouring.The second goal of this problem is to minimize the total temperature drop of all material flows entering and leaving the furnaces while in transit and waiting, thereby maximizing energy efficiency and minimizing energy waste. (2) where  1 represents the penalty for minimizing the deviation of the start time of the pouring process, and  2 represents the total temperature drop of the material flow in the minimized furnace during the transfer process. 1 represents the penalty coefficient for early start pouring time, and  2 represents the penalty coefficient for delayed start pouring time.
Production organization constraints encompass various aspects.First, according to the task operation sequence constraint, the total of a furnace's end processing time in one process plus its transit time between processes can't be less than the starting processing time of the succeeding process.Secondly, the continuous casting constraint ensures the continuity of casting by maintaining adjacent heats within the same casting process.Thirdly, machine preparation time is required between consecutive pours produced on the same casting machine.Fourthly, each furnace is limited to processing only once in each process.Fifthly, a processing priority relationship exists between different heats within each process.Sixthly, machine capability constraints indicate that each machine can process only one furnace task simultaneously.Seventhly, the release time constraint of molten iron dictates that the starting processing time of the first process for a furnace should be later than its iron feeding time.Eighthly, constraints on decision variable values are imposed.The liquidus temperature of the steel grade plus the minimum superheat of the process dictates the minimum casting temperature limitation, which is the beginning processing temperature of the furnace in the continuous casting process.
Tenth, process tapping temperature constraint, which means that the furnace is tapping at the target tapping temperature in non-continuous casting processes.
Eleventh, temperature related variable value constraints.
Considering the complexity of constraints, which increases the difficulty of solving the model, an optimization model for steelmaking plant production scheduling considering material flow temperature drop can be obtained by converting the constraints into objectives.

Solution to energy consumption optimization problem
Due to the complexity and multi-objective nature of this optimization problem, it is necessary to design effective multi-objective optimization algorithms to achieve the solution of the problem.Therefore, this chapter designs a MOHGALS based on the NSGA-II algorithm framework to solve the problem, and it improves the algorithm's solving performance by designing an improved genetic evolution mechanism.The overall process of the MOHGALS algorithm is as follows.
Step 1: In the initialization stage of the MOHGALS algorithm, the non-dominated solution of the algorithm evolution process is saved by setting the elite solution file and initializing it.This article suggests two evolutionary ways for strengthening the MOHGALS algorithm: first, by enhancing the genetic operator to speed up the algorithm's convergence, and second, by enhancing the algorithm's deep search capacity via local search.During the evolution process, the enhanced genetic operator offers saving and updating non dominated solutions by using an external elite solution file K. Furthermore, in order to prevent the loss of exceptional individuals and speed up the convergence of the algorithm evolution process, the non-dominant solution in K is crossed with the people in the matching pool.The following enhanced genetic procedures are carried out when two parent individuals have been picked from the mating pool.To begin, pick any Gi from of set K at random.If both parent people are dominant, then go to Step 2. The second step is to undertake a crossover operation to produce new offspring by replacing the individuals.Third, create a new offspring by performing a crossover on the parent.Incorporating local search (LS) into multi-objective optimization algorithms has been proven in previous research on evolutionary algorithms to effectively increase the search capabilities of the algorithm.To further improve the algorithm's deep search capabilities, this chapter employs a local search algorithm for non-dominated solutions, which is generated by genetic evolution and non-dominated sorting.

Experimental preparation
To test and analyze the effectiveness of the models and algorithms proposed in this chapter, this section takes the production and manufacturing process of a certain steel plant as the object, and obtains a multi-objective optimization scheduling solution set by considering the production scheduling optimization model and algorithm of material flow temperature drop, for flexible selection by production scheduling personnel.

Experimental results and analysis
By implementing the algorithm on a computer, the multi-objective non-dominated solution set of this case, namely the Pareto solution set, was ultimately obtained, as shown in figure 2. It can be seen that the solution set has a wide distribution on the target F1 or F2, providing production scheduling personnel with multiple scheduling options.Further from figure 3, it can be seen that the optimization scheduling scheme S2 selected based on the principle of optimal F2 function value can further reduce the process temperature drop of furnace material flow operation compared to S1.The average total temperature drop of material flow can be reduced by 18.85%, thereby reducing thermal energy loss and improving energy efficiency.When analyzing the efficacy of a multi-objective algorithm quantitatively, it is common practice to consider both the distribution of the Pareto solution set obtained by the algorithm and the distance between that set and the optimal Pareto solution set.
To determine the best Pareto solution set and the average HV index of the MOHGALS method under this parameter combination, we will run the MOHGALS, NSGA-II, and SPEA2 algorithms 10 times each, separately and iteratively, for case 1.The chapter employs the same maximum CPU time limit as the method completion criteria for all parameter combinations, as indicated in table 1, despite the fact that the calculation time varies significantly depending on the parameters.The contour map of marginal mean estimation for each parameter is drawn based on SPSS software, as shown in figure 4, and used to select the optimal algorithm parameter combination.Due to the significant impact of population size PS and neighborhood solution reception rule AR on the performance of the MOHGALS algorithm, the optimal PS value of 200 is first determined based on the marginal mean estimation contour map of PS, and the AR value is the corresponding algorithm performance indicator HV optimal value.Due to the fact that other parameters have no significant impact on the MOHGALS algorithm, it is possible to flexibly set their values.In this chapter, the performance of MOHGALS is analyzed through a series of comparison tests after the best parameter combination has been determined.Numerical experiments were used to fine-tune the algorithm parameters of NSGA-II and SPEA2 before to the comparison tests.
Figure 4.Estimated marginal mean of HV. Figure 5 displays the final outcomes of the experiments.Most of the time, MOHGALS has the best IGD indicator, and its average IGD values are lower than those of NSGA-II and SPEA2.Because of this, we can conclude that MOHGALS is better to competing algorithms in terms of its ability to create a multi-objective solution set that is quite near to the optimal Pareto border.In most circumstances, the optimal solution is found using MOHGALS when the RD indication is used.Figure 6 displays the experimental findings, which reveal that the average RD value is less than that of NSGA-II and SPEA2.This shows that the MOHGALS algorithm yields better results in terms of distance between solutions compared to its competitors.

Conclusion
As a type of manufacturing system operating at high temperatures, the material flow temperature drop caused by transportation, waiting for processing, and other processes during the production process of steel mills can directly cause a large amount of thermal energy loss, thereby increasing process energy consumption and having a significant impact on the total production energy consumption.Therefore, this chapter focuses on the production scheduling optimization problem of material flow and temperature drop control, and studies the multi-objective optimization problem of steelmaking plant production scheduling considering high temperature operation constraints and temperature drop objectives.This study offers a multi-objective hybrid genetic algorithm MOHGALS paired with neighborhood search to solve the SPSPPCTD issue and creates an improved evolutionary strategy, taking into account the original problem's multi-objective and complicated constraint features.The suggested technique in this chapter is proven to be able to achieve an optimum multi-objective Pareto solution set through case testing.The average total temperature drop of material flow operations may be reduced by 18.85 percent using a scheduling system that takes into account the principle of minimal penalty for material flow temperature drop as opposed to the notion of minimum penalty for opening time deviation.The MOHGALS method outperforms competing algorithms on the SPSPCTD problem in terms of the IGD, RD, and HV multi-objective evaluation indicators.

Step 2 :
Generate a parent population with an individual size of N through a population initialization algorithm based on casting order.Step 3: Perform a decoding heuristic algorithm on the chromosome encoding of each individual in the parent population to obtain its corresponding complete solution.Step 4: Perform a fast non dominated sorting on the parent population, dividing the corresponding solutions of individuals in Pt into multiple levels and calculating the crowding degree between solutions.Step 5: Select N/2 individuals based on the fast non dominated sorting results to enter the matching pool.Afterwards, improved genetic operators such as crossover and mutation were performed on individuals in the matching pool and elite solution file to obtain a new offspring population.Step 6: Merge the parent population and the offspring population to obtain a larger population with a size of 2N individuals and perform fast non dominated sorting and crowding calculation.Step 7: Select N individuals from R based on the sorting results to form a new parent population.Step 8: Determine whether the algorithm iteration termination condition has been met.If the termination condition is met, the algorithm ends and outputs the non-dominated solution in the population.Otherwise, go to Step9.Step 9: Perform a local search on the non-dominated solutions, and obtain an improved parent population to update the t+1 elite solution file, and set t+1→t to go to Step 3. The flowchart of the algorithm is shown in figure 1.

Figure 2 .
Figure 2. Pareto solution distribution obtained by case solution.Further from figure3, it can be seen that the optimization scheduling scheme S2 selected based on the principle of optimal F2 function value can further reduce the process temperature drop of furnace material flow operation compared to S1.The average total temperature drop of material flow can be reduced by 18.85%, thereby reducing thermal energy loss and improving energy efficiency.

Figure 3 .
Figure 3.The comparison of total temperature drops of each charge.When analyzing the efficacy of a multi-objective algorithm quantitatively, it is common practice to consider both the distribution of the Pareto solution set obtained by the algorithm and the distance between that set and the optimal Pareto solution set.To determine the best Pareto solution set and the average HV index of the MOHGALS method under this parameter combination, we will run the MOHGALS, NSGA-II, and SPEA2 algorithms 10 times each, separately and iteratively, for case 1.The chapter employs the same maximum CPU time limit as the method completion criteria for all parameter combinations, as indicated in table 1, despite the fact that the calculation time varies significantly depending on the parameters.Table1.The eight combinations of algorithm parameter.
on Applied Physics and Mathematics Journal of Physics: Conference Series 2729 (2024) 012017

Table 1 .
The eight combinations of algorithm parameter.