Algorithmic Advances for Large-Scale Dynamic Process Optimization

大规模动态过程优化的算法进步

• 批准号: 0314647
• 负责人: Lorenz Biegler
• 金额: $ 30.39万
• 依托单位: Carnegie-Mellon University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2003
• 资助国家: 美国
• 起止时间: 2003-07-01 至 2007-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0314647&HistoricalAwards=false
• 关键词: Algorithmic; Advances; Large; Scale; Dynamic

Research:Dynamic optimization is an essential component of many process design activities. In previous work the PI developed an efficient simultaneous approach for dynamic optimization that handles both unstable and path-constrained differential-algebraic systems. He developed an improved nonlinear programming strategy based on interior point or barrier methods, which leads to significant gains in performance. For example, on his dynamic process optimization case, nonlinear programs with over 1.2 million variables can be solved in approximately two hours of CPU time on a 1.0 GHz computer. In this project he will extend his simultaneous optimization strategies to large systems. To do this, he will consider several improvements in the formulation and algorithm for dynamic optimization problems as well as extending the scope of the dynamic optimization problems to include discrete decisions that involve phase equilibriums and other complementary relationships. Because these features are exploited in a novel way by the interior point algorithm, difficult optimization problems that include a class of discrete decisions can be treated with little additional computational cost.He plans to incorporate these into a modeling framework that will combine well-developed user interfaces and object-oriented software as well as demonstrate the application of the algorithms within a state-of-the-art parallel computing environment on a wide variety of process optimization problems drawn from process control, operations and design.Broader Impact:The Chemical Process Industries (CPI) could greatly benefit from the development of more efficient dynamic optimization methodologies.

研究：动态优化是许多流程设计活动的重要组成部分。 在之前的工作中，PI 开发了一种高效的动态优化同步方法，可以处理不稳定和路径受限的微分代数系统。 他开发了一种基于内点或障碍方法的改进的非线性编程策略，从而显着提高了性能。 例如，在他的动态过程优化案例中，具有超过 120 万个变量的非线性程序可以在 1.0 GHz 计算机上用大约两个小时的 CPU 时间来求解。 在这个项目中，他将把他的同步优化策略扩展到大型系统。 为此，他将考虑对动态优化问题的公式和算法进行多项改进，并将动态优化问题的范围扩展到包括涉及相平衡和其他互补关系的离散决策。 由于内点算法以一种新颖的方式利用了这些特征，因此可以用很少的额外计算成本来处理包括一类离散决策的困难优化问题。他计划将这些特征合并到一个建模框架中，该框架将结合成熟的用户接口和面向对象的软件，并演示算法在最先进的并行计算环境中对来自过程控制、操作和设计的各种过程优化问题的应用。更广泛的影响：化学过程工业（CPI）可以从更高效的发展中受益匪浅动态优化方法。