Plantwide dynamic operability analysis from network perspective

Increasing raw material and energy costs have caused a shift in process design philosophy for better use of energy integration and recycling resulting in growing appearances of more complex chemical plants which consist of many process units interconnected in various configurations. The unit interactions caused by these interconnections often impose limitations on the plantwide operability of chemical processes, commonly in the form of control performance degradation and operational instability in the worst case. Despite the importance of plantwide operability analysis, it is often difficult to perform such assessment at the earlier stages of process design because of the degree of complexity, interactions, and nonlinearity of the processes.This thesis provides a framework of dynamic operability assessment for general nonlinear plantwide processes based on the concept of dissipative systems and network perspective. An entire large-scale system is represented as a network of smaller sub-systems and their interconnection topology can then be captured explicitly. A distinctive feature of the network perspective lies in the distinction between the physical interconnections between different sub-systems and the signal interconnections between the physical process and its control system. Using the dissipativity of each sub-system and the process topology, the plantwide stability, stabilizability, and achievable control performance can then be assessed. The most significant contribution of the proposed approach is on the explicit analysis of interaction effects on plantwide operability.Furthermore, the resulting operability analysis problems involve linear matrix inequalities (LMIs) which are easy to solve and potentially scalable.Improvements to the above framework are also proposed to reduce the level of conservativeness of the general nonlinear approach and provide a more detailed operability analysis. By decomposing the large-scale system down to the storage level, with sub-systems represented by single state equations, the dissipativity can be determined more easily. The dissipativity property of the entire nonlinear plantwide process can also be parametrized by the dissipativity of individual sub-systems, leading to a cluster of supply rates, which in turn are optimized to achieve much less conservative operability analysis results. In addition, linearized models can be used to obtain more detailed dynamic performance achievable, e.g., frequency domain performance indicators