Stable Dynamic Optimization Strategies for Large-Scale Chemical Processes

点击翻译按钮获取中文摘要

Biegler CTS-9729075 Strategies for process optimization have seen widespread applications over the past decade. These have been applied in the design of new chemical processes, real-time optimization of process units and plants in petrochemical processes, and for sophisticated analyses related to the operability and flexibility of chemical processes. The vast majority of these applications have been for steady state process models, described by systems of algebraic equations. Simultaneously, development of powerful, large-scale dynamic simulation tool are also becoming widespread and equation based dynamic process models have been constructed with up to 100,000 differential-algebraic equations (DAE). The success of dynamic process simulation has led to the demand for optimization tools that deal readily with process models described by differential algebraic equations (DAEs). However, development of dynamic optimization strategies has lagged that of simulation for a number of reasons. Among these are conceptual limitations of current optimization strategies. These become important for DAE models that are highly constrained or have unstable dynamic modes. The latter are often true in reactive and reactive-separation systems including exothermic reactors and reactive distillation. The PI is planning on developing a simultaneous formulation for dynamic optimization that includes a stable large-scale decomposition based on boundary value formulations. Novel nonlinear programming, strategies that detect unstable modes, exploit problem structure for large scale decomposition and adapt leading, edge methods for the treatment of large sets of inequality constraints will be considered. The development of these strategies will be aided by research collaborations with applied mathematicians that specialize in nonlinear programming and DAE solution algorithms. The combination of these strategies will lead to large-scale dynamic optimization strategies that can be applied to dynamic simulation models currently considered in industrial applications. The methods developed will be validated by large scale industrial applications in reaction engineering and reactive distillation.