A Massively Parallel Population-Based Approach to Operations Scheduling

9505668拉梅什操作调度问题涉及以最经济的方式调度受几个约束的发电机组。问题是确定在从一天到一周的一段时间内每个OU中应该打开(提交)哪些单元，并确定每个提交单元在每个小时中的生成级别。求解该问题的主要困难在于它是一个大型混合整数非线性规划(MINLP)问题，计算量大。正因为如此，目前采用的算法技术没有对许多相关的约束进行建模，并且使用了几个导致次优解的近似。这项研究的主要目标是：a)研究和开发一种新的基于种群的优化技术来解决MINLP问题，以及b)使用一个全面的模型，该模型包含了作业调度问题的大部分相关约束。我们将在这项研究中探索的新的组合优化技术是一种分解方法，它并行地为总体问题的每个子问题评估一组解。该方法将被应用于求解作业调度问题的综合模型。我们将研究所提出的技术的复杂性和最优性。***

9505668 Ramesh The operations scheduling problem concerns the scheduling of generating units in the most economic manner subject to several constraints. The problem is to identify which units should be ON (committed) in each our over a period of time ranging from a day to a week, and to determine the generation level of each committed unit in each hour. The main difficulty in solving the problem is that it is a large mixed-integer nonlinear programming (MINLP) problem, and is compute-intensive. Because of this, the algorithmic techniques currently employed do not model many relevant constraints, and use several approximations which lead to suboptimal solutions. The main objectives of this research are: a) to investigate and develop a new population-based optimization technique for MINLP problems, and b) to use a comprehensive model encompassing most, if not all, relevant constraints for the operations scheduling problem. The new combinatorial optimization technique that we will explore in this research is a decomposition methodology that evaluates a population of solutions, in parallel, for each subproblem of the over-all problem. The approach will be applied to solve a comprehensive model of the operations scheduling problem. We will investigate the complexity and optimality of the proposed technique. ***