UNS: Real-Time Economic Model Predictive Control of Nonlinear Processes

UNS：非线性过程的实时经济模型预测控制

• 批准号: 1506141
• 负责人: Panagiotis Christofides
• 金额: $ 30万
• 依托单位: University of California-Los Angeles
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2015
• 资助国家: 美国
• 起止时间: 2015-09-01 至 2021-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1506141&HistoricalAwards=false
• 关键词: UNS; Real; Time; Economic; Model

1506141 - ChristofidesThe development of next-generation advanced manufacturing (e.g., Smart Manufacturing, market-driven manufacturing, and real-time energy management) is of paramount importance to sustaining the future competitiveness of the U.S. chemical industry, a vital sector of the U.S. economy. The core of next-generation manufacturing objectives involves tightly integrating the components of the manufacturing processes to deliver increased safety, profitability, efficiency, variability, capacity and sustainability. Within the context of chemical process operations, process control systems should account for economic process considerations such as variable demand, changing energy prices, variable feedstock, and product transitions in the computation of the control actions and should be able to operate a process in a dynamic fashion to account for the volatile market conditions. Most of the existing control infrastructure has been designed to achieve the best possible performance with respect to steady-state (time-invariant) operation. Transitioning from steady-state operation to dynamic or time-varying operation represents a significant paradigm shift in chemical process operations and chemical process control. Economic optimization of chemical processes has traditionally been addressed through a two- layer architecture. In the upper layer, economic process optimization is completed by computing optimal process operation set-points using steady-state process models. These optimal set-points are used by the feedback control systems in the lower layer to force the process to operate on these steady-states. In the lower layer, model predictive control (MPC) has been widely adopted in the chemical process industry because of its ability to optimally control multivariable systems subject to input and state constraints. The conventional formulations of MPC use a quadratic performance index along a finite prediction horizon to steer the system to the optimal (economically) steady- state. While this strategy (steady-state optimization and operation) has been traditionally used in chemical process industries, steady-state operation may not necessarily be the economically best operation strategy. Recently, economic MPC (EMPC), an MPC scheme that uses a cost function that directly accounts for the process economics, has been introduced as an alternative approach to the two-layer economic process optimization and control. EMPC operates systems in a possibly time-varying fashion to optimize the process economics. However, the rigorous design of real-time EMPC systems, which address key practical considerations (time needed to compute control actions, guaranteed performance improvement over steady-state operation, time-varying cost functions, and monitoring and safety) poses significant fundamental and implementation challenges. Motivated by these considerations, the main objective of this research program is to develop the theory and methodology needed for the design and implementation of real-time economic model predictive control systems for chemical processes described by nonlinear dynamic models and to demonstrate the effectiveness of the proposed methods in the context of chemical processes of industrial importance. Specifically, this research will address: a) the development of real-time EMPC systems capable of handling real-time im- plementation issues and practical considerations including real-time calculation time and explicitly time-dependent cost functions accounting for variable energy price and demand, b) the development of monitoring schemes for evaluating the performance of EMPC systems accounting for time-varying operation and the design of EMPC systems that explicitly account for process safety constraints and deal with operational limitations due to possible malfunction of control system components, and c) applications of the real-time EMPC schemes to large-scale chemical plant simulators us- ing high-fidelity process models and a state-of-the-art experimental ultra-filtration/reserve osmosis water desalination system to demonstrate that EMPC can significantly reduce energy consumption. The development of real-time EMPC system methods is expected to signifi- cantly improve the operation and performance of nonlinear processes, thereby enhancing the com- petitiveness of the US economy. The integration of the research results into graduate and senior undergraduate courses and the writing of a new book on "Economic Model Predictive Control" will benefit students and researchers in the field.

1506141 - christofides下一代先进制造（如智能制造、市场驱动制造和实时能源管理）的发展对于维持美国化学工业的未来竞争力至关重要，化学工业是美国经济的重要部门。下一代制造目标的核心包括紧密集成制造过程的组件，以提供更高的安全性、盈利能力、效率、可变性、产能和可持续性。在化学过程操作的背景下，过程控制系统应该考虑到经济过程的考虑因素，如可变需求、不断变化的能源价格、可变的原料和控制动作计算中的产品转换，并且应该能够以动态方式操作过程，以考虑不稳定的市场条件。大多数现有的控制基础设施都是为了实现稳态（时不变）操作的最佳性能而设计的。从稳态操作到动态或时变操作的过渡代表了化学过程操作和化学过程控制的重大范式转变。传统上，化学过程的经济优化是通过两层结构来解决的。在上层，经济过程优化是通过使用稳态过程模型计算最优过程运行设定点来完成的。下层的反馈控制系统使用这些最优设定点来强制过程在这些稳态上运行。在较低层次，模型预测控制（MPC）由于能够最优控制受输入和状态约束的多变量系统而在化学过程工业中被广泛采用。传统的MPC公式使用沿有限预测范围的二次性能指标来引导系统达到最优（经济）稳态。虽然这种策略（稳态优化和运行）传统上用于化学过程工业，但稳态运行不一定是经济上最好的操作策略。最近，经济MPC （economic MPC， EMPC）作为两层经济过程优化和控制的替代方法被引入，它是一种使用成本函数直接说明过程经济的MPC方案。EMPC以可能的时变方式操作系统，以优化过程经济。然而，实时EMPC系统的严格设计，解决了关键的实际问题（计算控制动作所需的时间、保证稳态运行的性能改进、时变成本函数以及监控和安全），这给基础和实施带来了重大挑战。出于这些考虑，本研究计划的主要目标是发展设计和实施非线性动态模型描述的化学过程的实时经济模型预测控制系统所需的理论和方法，并证明所提出的方法在工业重要性的化学过程背景下的有效性。具体而言，本研究将涉及：a)开发能够处理实时实施问题和实际考虑因素的实时EMPC系统，包括实时计算时间和明确的与时间相关的成本函数，该函数考虑到可变的能源价格和需求；b)制定监测方案，以评估考虑时变运行的EMPC系统的性能，并设计明确考虑过程安全约束的EMPC系统，并处理由于控制系统组件可能发生故障而导致的运行限制；c)将实时EMPC方案应用于大型化工厂模拟器，包括高保真过程模型和最先进的超滤/储备渗透海水淡化系统，以证明EMPC可以显著降低能耗。实时EMPC系统方法的发展有望显著改善非线性过程的运行和性能，从而提高美国经济的竞争力。将研究成果整合到研究生和高级本科课程中，并撰写一本关于“经济模型预测控制”的新书，将使该领域的学生和研究人员受益。