Data-Driven Process Systems Optimization under Uncertain Environment

对于今天的流程工业来说，基于已建立的过程控制和信息技术，已经在数据采集和存储方面奠定了坚实的基础。虽然海量的数据是通过日常运营产生的，但流程工业正在向数据驱动的决策方向发展。运营和业务数据分析用于支持流程、工厂和企业范围的优化计划。要实现这种过渡，需要将数据分析和流程控制方面与计划和调度的运营优化无缝结合。其目标是为工业流程创造附加值，提高流程的敏捷性，以同时对能源效率和可持续性做出反应。 在过去的几年里，我们利用稳健优化和随机规划技术，在不确定条件下的过程操作优化方面取得了一定的成果。有了大量的运行数据，就有可能改进过程和不确定性模型，做出更实际的系统优化决策。在这项研究计划中，我们计划通过实施不确定情况下的数据驱动优化的系统方法，将过程优化和过程数据分析领域结合起来。具体地说，我们将：1)研究数据驱动的过程和借助数据分析的不确定性建模；2)开发数据驱动的稳健和自适应复杂过程系统优化的方法；3)开发软件工具，使数据驱动的优化问题的建模和求解自动化。 拟议的研究将被用来优化系统系统，从单个工艺单元到整个工厂现场。这将提供重要的理论和计算支持，使流程行业的经理、操作员和工程师能够在信息驱动的环境中使用实时数据和分析进行协作和工作。

For today's process industries, a solid foundation in data acquisition and storage has been developed based on the established process control and information technologies. While huge volumes of data are generated through daily operations, process industries are moving towards data-driven decision making. The analysis of operations and business data is used to support process, plant, and enterprise-wide optimization initiatives. To achieve this kind of transition, it requires seamlessly merging data analytics and process control aspects with operational optimization of planning and scheduling. The target is to create added value to the industrial process, increase the agility of processes to react to changes simultaneously focusing on energy efficiency and sustainability. In the past few years, we have made achievements in process operations optimization under uncertainty using the robust optimization and stochastic programming techniques. With a large amount of operational data available, it is possible to improve the process and uncertainty model and make more practical decisions towards systems optimization. In this research program, we plan to combine the fields of process optimization and process data analytics by implementing systematic methods for data-driven optimization under uncertainty. Specifically, we will: 1) investigate the data-driven process and uncertainty modeling with the aid of data analytics; 2) develop approaches for data-driven robust and adaptive optimization of complex process systems; 3) develop software tools to enable the automation of modeling and solution of the data-driven optimization problem. The proposed research will be utilized to optimize systems of systems, from a single process unit to an entire plant site. This will provide significant theoretic and computational support that enables managers, operators, and engineers in the process industry to collaborate and work together using real-time data and analysis in an information-driven environment.