Control Co-Design of Partitioned Energy Systems Using Graph-Based Models

Abstract Energy systems, systems that create, use, or transfer energy, are traditionally designed using one of two methods. They can be designed by optimizing each component in isolation; however, this may lead to a design with components that are incompatible with the entire system. On the other hand, they can be designed by optimizing the entire system together, which can lead to computational challenges due to problem size and complexity. There is a need to optimize single subsystems while ensuring consistency with the rest of the system in a computationally efficient manner. Two partitioned control co-design approaches are proposed to optimize single subsystems while ensuring continuity with the complete system. One approach is based on analytical models and the other approach is data based. Both approaches utilize a versatile graph-based modeling framework to capture the dynamics of multidomain energy systems. These two approaches can be used to determine the best subsystem to optimize when only a subset of subsystems can be optimized due to resource or time limitations. The approaches are applied to a notional aircraft fuel thermal management system, and they provide a design within 0.5% of the optimal design found by optimizing the entire system at once. The results demonstrate that the proposed methods provide a solution near the optimal point with limited information on the system.