發(fā)布時間：2018-01-09 14:07

  本文關(guān)鍵詞：分時電價下煉鋼連鑄生產(chǎn)調(diào)度優(yōu)化方法　出處：《山東大學(xué)》2017年博士論文　論文類型：學(xué)位論文

  更多相關(guān)文章： 煉鋼連鑄 車間調(diào)度 智能優(yōu)化 分時電價 交叉熵算法

【摘要】：煉鋼連鑄生產(chǎn)過程是鋼鐵企業(yè)生產(chǎn)中的一個重要環(huán)節(jié),具有多階段、多機器、以及離散與連續(xù)生產(chǎn)相混合的特點。其生產(chǎn)調(diào)度優(yōu)化可以使鋼鐵企業(yè)各工序間有效銜接,并加快生產(chǎn)節(jié)奏、提高生產(chǎn)效率、降低生產(chǎn)成本,歷來是企業(yè)界和研究領(lǐng)域關(guān)注的熱點問題。煉鋼連鑄生產(chǎn)過程消耗的能源介質(zhì)種類多,其中電能消耗量大而且電力成本與分時電價密切相關(guān)。優(yōu)化分時電價下的電能消耗和電力成本可以降低生產(chǎn)過程的總能耗及能源成本,提高企業(yè)經(jīng)濟效益,具有重要的理論研究意義和實際應(yīng)用價值。本文針對分時電價下電能消耗及電力成本最小的煉鋼連鑄生產(chǎn)調(diào)度優(yōu)化這一復(fù)雜問題,從工藝路徑相同、復(fù)雜工藝路徑、加工時間不確定、多目標(biāo)等情況進行深入研究。針對工藝路徑相同的煉鋼連鑄生產(chǎn)過程電能消耗及電力成本最小化問題,建立了分時電價下的調(diào)度優(yōu)化模型。針對該模型引入分時電價后0-1變量急劇增加、目標(biāo)計算復(fù)雜、求解速度慢等問題提出了一種基于局部搜索的混合啟發(fā)式交叉熵算法。該算法采用矩陣編碼策略和基于階段順序的倒推解碼方法、基于FIFO啟發(fā)式規(guī)則的混合樣本生成和基于行列交換的局部搜索等策略,能在較短的時間內(nèi)求得高質(zhì)量的調(diào)度方案,具有很好的穩(wěn)定性和收斂性。仿真結(jié)果表明,與優(yōu)化爐次駐留時間以間接降低能耗相比,該優(yōu)化模型對降低煉鋼連鑄生產(chǎn)過程的電能消耗效果更好,尤其是分時電價下的電力成本優(yōu)化效果更明顯。針對復(fù)雜工藝路徑的煉鋼連鑄生產(chǎn)過程電能消耗及電力成本最小化問題,建立了分時電價下的調(diào)度優(yōu)化模型。與工藝路徑相同的情況比較,復(fù)雜路徑約束導(dǎo)致交叉熵算法的編碼與解碼更困難;分時電價的引入使優(yōu)化模型的決策變量規(guī)模擴大至少三倍,導(dǎo)致目標(biāo)計算更復(fù)雜、模型更難以求解。因此提出了一種基于動態(tài)參數(shù)的混合自適應(yīng)交叉熵算法。該算法采用基于操作順序的倒推解碼方法,以及基于全局選擇和隨機置換啟發(fā)式規(guī)則的混合樣本生成、基于矩陣分割與行列交換的局部搜索和參數(shù)動態(tài)調(diào)整等策略,求解質(zhì)量高、求解速度快、自適應(yīng)能力強。仿真結(jié)果表明,該優(yōu)化模型能有效地描述更復(fù)雜的大規(guī)模煉鋼連鑄生產(chǎn)過程,在優(yōu)化分時電價下電能消耗及電力成本方面比只考慮爐次駐留時間時的效果更好。針對煉鋼連鑄生產(chǎn)過程中爐次LF精煉時間和爐次基本加工時間等不確定的情況,建立了分時電價下電能消耗及電力成本最小化問題的調(diào)度優(yōu)化模型。該模型整數(shù)變量及其約束條件增加,LF精煉時間需要調(diào)整,從而決策變量更多、規(guī)模更大、求解更困難。因此,提出了一種離散與連續(xù)交叉熵算法相混合的串級交叉熵算法。該算法將不確定加工時間的求解與爐次機器分配狀態(tài)的求解分別進行,簡化了問題的編碼及解碼過程,減少了不可行解的數(shù)量,避免了遺傳算法染色體太長、交叉變異復(fù)雜、解碼困難的問題,從而縮短了求解時間。提出了基于關(guān)鍵爐次的混合調(diào)整方法,對爐次LF精煉時間進行調(diào)整以補償溫度損失,降低了分時電價下增加的電能消耗及電力成本。仿真結(jié)果表明,與加工時間確定的隨機實例和特殊實例的求解結(jié)果相比,該模型優(yōu)化了不確定加工時間的組合,在減少電能消耗及電力成本方面合理且有效。最后,針對煉鋼連鑄生產(chǎn)過程考慮分時電價后電能消耗、電力成本以及完工時間等多個目標(biāo)相互矛盾而難以抉擇的問題,建立了多目標(biāo)調(diào)度優(yōu)化模型。針對該模型目標(biāo)種類多、引入分時電價后計算更復(fù)雜、個體排序及評價困難、求解結(jié)果多樣性差、分布較集中等問題,提出了一種基于Pareto最優(yōu)的混合多目標(biāo)交叉熵算法。該算法采用混合多樣本生成、基于快速非支配排序的個體評價、基于擁擠距離和精英樣本聚類的多樣性保持等策略,達到了很好的求解效果。尤其是聚類算法的引入,有效地避免了非支配解向Pareto前沿中間部分聚集,提高了非支配解的多樣性和分布廣泛性。仿真結(jié)果表明,該模型的求解既能為調(diào)度決策者提供各目標(biāo)相對均衡的折中方案,又能提供在基本不惡化其它目標(biāo)的情況下偏向某個目標(biāo)的調(diào)度方案。
[Abstract]:Steelmaking and continuous casting production process is an important link in the production of iron and steel enterprises, with multiple stages, multiple machines, and the characteristics of discrete and continuous production mix. The production scheduling optimization of iron and steel enterprises can make effective connection between each process, and accelerate the pace of production, improve production efficiency, reduce production costs, is always a hot topic business and research in the field of energy consumption. The medium type of steelmaking and continuous casting production process, the electricity consumption and electricity cost and TOU price are closely related. Optimization of TOU electricity under the total energy consumption and cost of energy consumption and electricity costs can be reduced in the production process, improve the economic efficiency of enterprises, has the important theoretical significance and practical value. Based on TOU power and power consumption under the minimum cost of steelmaking and continuous casting production scheduling optimization of this complex problem, from The same process path, complex process, uncertain processing time, in-depth research on multi objective and so on. In the same process of steelmaking and continuous casting production process of electricity consumption and electricity cost minimization problem, a scheduling optimization model. The price for 0-1 variables a sharp increase in the model introduced tou, target complex calculation, problem solving speed presents a hybrid heuristic cross entropy algorithm based on local search. The algorithm uses the matrix encoding and decoding method based on the strategy of pushing down the stage order, based on the mixed sample to generate FIFO heuristic rules and based on the ranks of exchange local search strategies such as scheduling scheme can be obtained in high quality a short period of time, has good stability and convergence. The simulation results show that compared with the optimal furnace dwell time to indirectly reduce energy consumption, the optimization The model can better reduce the consumption of electric steelmaking and continuous casting production process, especially the TOU power cost optimization effect is more obvious. In view of the complex process path of steelmaking and continuous casting production process of electricity consumption and electricity cost minimization problem, a scheduling optimization model for electricity price under time. Compared with the same process conditions that complex path constraints lead to encoding and decoding cross entropy algorithm is more difficult; tou was introduced to make decision variable scale optimization model is enlarged at least three times, the target calculation is more complex, more difficult to solve the model. This paper presents a method based on the dynamic parameters of the hybrid adaptive cross entropy algorithm. This algorithm is used to push down decoding method based on operation sequence, and based on the mixed sample generation global selection and random permutation heuristic rules, local search matrix segmentation and exchange based on ranks The cable and the parameters of dynamic adjustment strategy, high quality solution, solution speed, strong adaptive ability. The simulation results show that this model can effectively describe large-scale steelmaking and continuous casting production process more complicated, in optimizing the TOU price under the power consumption and the cost of electricity is better than only considering the effect of furnace residence time for steelmaking and continuous casting production process of furnace LF refining furnace time and basic processing time is uncertain, a scheduling optimization model of TOU power consumption and electricity cost minimization problem. The model increases the integer variables and constraints, the refining time of LF need to be adjusted, thus more decision variables, scale more and more difficult to solve. Therefore, put forward the cascade cross entropy algorithm combined a discrete and continuous cross entropy algorithm. This algorithm will not determine the processing time of the solution and furnace machine To solve the allocation state respectively, simplifying the encoding and decoding process, reduce the number of infeasible solutions, avoid the genetic algorithm chromosome crossover and mutation is too long, complicated, difficult decoding problems, shorten the solving time. Put forward the mixed adjustment method based on key furnace, furnace refining time of LF adjusted to compensate for temperature loss, reduce the TOU price increase power consumption and electricity cost. The simulation results show that, compared with the processing time of the solution to determine the random instances and special examples, the model optimizes the combination of processing time uncertainty in reducing power consumption and electricity cost is reasonable and effective finally, for steelmaking and continuous casting production process considering tou after power consumption, power cost and the completion time of a number of conflicting goals and difficult choices, establish a multi-objective Scheduling optimization model. According to the model of the target species, the introduction of TOU price calculation is more complex, individual ranking and evaluation results difficult, diversity, centralized distribution, proposes a cross entropy hybrid multi-objective algorithm based on Pareto optimal. The algorithm generated by mixed samples, non fast individual evaluation dominated sorting based on diversity crowding distance and clustering and elite strategy based on reaching a good solution. Especially the introduction of clustering algorithm, can effectively avoid the non dominated solution to the Pareto front in the part of aggregation, improve the wide diversity and distribution of non dominated solutions. The simulation results show that the model can provide decision makers for scheduling compromise each target is relatively balanced, but also provide a target bias scheduling scheme basically does not worsen the effects of other target case.

【學(xué)位授予單位】：山東大學(xué)
【學(xué)位級別】：博士
【學(xué)位授予年份】：2017
【分類號】：TF758
，

本文編號：1401681