Embedded Reduced Models in Flowsheet Optimization

Advanced energy systems demand powerful and systematic optimization strategies for analysis, high performance design and efficient operation. Such processes are modeled through a heterogeneous collection of device-scale and process scale models, which contain distributed and lumped parameter models of varying complexity. This work addresses the integration and optimization of advanced energy models through multi-scale optimization strategies. In particular, we consider the optimal design of advanced energy processes by merging device-scale (e.g., CFD) models with flowsheet simulation models through sophisticated model reduction strategies. Recent developments in surrogate-based optimization have led to a general decomposition framework with multiple scales and convergence guarantees to the overall multi-scale optimum. Here, we develop two trust region-based algorithms where gradients are not required from the original detailed model (ODM). These algorithms borrow from derivative-free optimization (DFO) methods for unconstrained optimization and we extend them to the constrained case, as well as to process flowsheet optimization with different scale models that allow only limited recourse to the ODM. Both methods demonstrate multi-scale optimization of advanced energy processes. The resulting theoretical derivations and developed algorithms are interesting and justify our previous work, including methodologies of reduced model (RM) development and flowsheet optimization, with reduced models based on their CFD counterparts.