基于自适应协同优化算法的流程工业生产调度研究

发布时间：2018-11-10 19:39

【摘要】：生产调度作为流程企业生产管理的核心,对提升企业综合竞争力和经济效益具有重要作用。流程工业生产调度是一个典型的NP-hard优化问题,具有复杂性、多约束性与多目标性,因此需要一种高效可行的优化算法用于问题的求解。而协同优化(Collaborative Optimization,CO)是一种新兴的多学科优化设计算法,它将复杂的模型分解为若干部分,降低了系统的复杂程度,减小了问题的求解难度,在流程工业生产调度领域具有很高的应用价值。本文的主要研究内容如下:(1)针对协同优化算法在学科级缺乏对目标函数的寻优能力,提出了一种自适应的协同优化算法(Self-adaptive Collaborative Optimization,SCO)。首先,在系统级引入协同不一致性,改进动态松弛因子使优化设计点快速收敛于极值点。其次,在学科级以动态权重将一致性目标函数和子学科最优目标函数相加作为子学科目标函数,考虑一致性同时又兼顾子学科独立性。最后,采用二阶段优化过程,在迭代后期去除动态松弛因子与子学科最优目标函数,防止收敛过程震荡。使用经典算例进行仿真,优化结果证明SCO算法对初始点不敏感,优化效率得到显著提升且具有较强的鲁棒性。(2)针对复杂的流程工业生产调度问题,建立了基于离散时间的流程工业MILP(Mixed Integer Linear Programming)模型,并将其应用于啤酒企业的糖化酿造车间七日生产调度实例。使用SCO算法将模型分解为七个糖化车间单日生产调度子学科与一个酿造车间七日生产调度子学科,同时使用遗传算法对SCO算法的学科级与系统级进行求解。通过对该案例的仿真与分析,验证了模型的合理性以及SCO算法用于求解流程工业生产调度问题的可行性与高效性。
[Abstract]:Production scheduling, as the core of production management in process enterprises, plays an important role in enhancing the comprehensive competitiveness and economic benefits of enterprises. Production scheduling in process industry is a typical NP-hard optimization problem with complexity, multi-constraint and multi-objective. Therefore, an efficient and feasible optimization algorithm is needed to solve the problem. Collaborative optimization (Collaborative Optimization,CO) is a new multidisciplinary optimization design algorithm, which decomposes the complex model into several parts, reduces the complexity of the system and reduces the difficulty of solving the problem. It has high application value in production scheduling field of process industry. The main contents of this paper are as follows: (1) an adaptive cooperative optimization algorithm (Self-adaptive Collaborative Optimization,SCO) is proposed to solve the problem of the lack of the ability to optimize the objective function at the subject level. Firstly, the cooperative inconsistency is introduced at the system level, and the dynamic relaxation factor is improved to make the optimal design point converge rapidly to the extremum point. Secondly, the consistency objective function and the subdiscipline optimal objective function are added as the subdiscipline objective function with dynamic weight at the subject level, and the consistency is considered and the subdiscipline independence is taken into account. Finally, the two-stage optimization process is used to eliminate the dynamic relaxation factor and the subdiscipline optimal objective function in the late iteration to prevent the convergence process from oscillating. The simulation results show that the SCO algorithm is insensitive to the initial point, and the optimization efficiency is improved significantly. (2) aiming at the complex production scheduling problem in the process industry, the simulation results show that the algorithm is not sensitive to the initial point, and the optimization efficiency is improved significantly. A discrete time based MILP (Mixed Integer Linear Programming) model for process industry was established and applied to the seven days production scheduling of saccharified brewing workshop in beer enterprises. SCO algorithm is used to decompose the model into seven sub-disciplines of single-day production scheduling in saccharification workshop and one sub-discipline of production scheduling in brewing workshop. At the same time, genetic algorithm is used to solve the SCO algorithm at the subject and system levels. Through the simulation and analysis of this case, the rationality of the model and the feasibility and efficiency of SCO algorithm used to solve the production scheduling problem in the process industry are verified.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TB497

【参考文献】