An integrated approach to discovery and optimal operation in process and reaction engineering

本提案的重点是在过程和反应工程中发现和优化操作的综合方法。它解决了化学过程复杂性所带来的各种问题：大量的物种/反应，空间异质性，紧密耦合和网络相互作用的子系统。第一个原因来自复杂进料的使用，第二个原因来自过程的空间异质性，第三个原因是紧密耦合的传热和传质操作，这是过程强化和试图提高操作效率的自然结果。复杂性对最优动态运行提出了许多挑战。在生物质转化和沥青升级等系统中，无法识别所有物种对确定最佳操作和了解潜在化学成分构成了重大挑战。我们提出了一种基于统计、机器学习和图论方法的方法来假设和发现这些系统中的主导化学。提出了一种结合多光谱源信息的方法，以提高对化学成分的识别。化学和反应机理的识别使精确的过程模型和先进的过程控制的发展成为可能。通过将过程模型转换为网络模型或基于图的模型来解决油气储层空间非均质性问题。利用图中节点的可达性概念来确定传感器和执行器的位置，从而实现最佳的操作和控制。在化学过程及其单元的紧密耦合网络的情况下，过程模型和控制器的基于热力学的框架与网络模型相结合，以开发全厂控制系统的优化重构框架。该方案所描述的集成网络或基于图的方法有潜力解决两个主要问题：不确定性下过程系统的智能和自适应优化操作，以及可持续和优化的过程操作。应用于与能源相关的过程也确保了研究结果实施时的重大影响。

The focus of this proposal is on an integrated approach to discovery and optimal operation in process and reaction engineering. It addresses the various issues presented by complexity in chemical processes: a large number of species/reactions, spatial heterogeneity, and tightly coupled and networked interacting subsystems. The first cause arises from the use of complex feeds, the second from spatial heterogeneity of the processes, and the third because of tightly coupled heat and mass transfer operations that are a natural consequence of process intensification and attempts to improve the efficiency of operation. Complexity presents many challenges to optimal dynamic operation. In systems like biomass conversion and bitumen upgrading, the inability to even identify all species presents a significant challenge to determining optimal operation and understanding the underlying chemistry. We propose a method to hypothesize and discover the dominant chemistry in such systems based on statistical, machine learning and graph-theoretic methods. A method is developed to incorporate information from multiple spectroscopic sources to improve identification of the chemistry. The identification of chemistry and reaction mechanism enable the development of accurate process models and advanced process control. The issues presented by spatial heterogeneity of petroleum reservoirs are addressed by transforming the process models to network or graph-based models. Concepts of reachability of nodes in graphs are invoked to identify the sensor and actuator locations that will result in optimal operation and control. In the case of tightly coupled networks of chemical processes and their units, a thermodynamics-based framework for the process models and controllers is combined with the network models to develop a framework for optimally reconfiguration of the plantwide control system. The integrated network or graph-based approach described in the proposal has the potential to address two major issues: smart and adaptive optimal operation of process systems under uncertainty, and sustainable and optimal process operation. The application to energy-related processes also ensures significant impact when the results of the research are implemented.