The objective of this research is to develop a new class of modeling and analysis tools for energy processing systems. The approach is based on a two-level structure in which lower-level (component) models in the form of computer code, lookup data or equations are used to directly answer questions about the higher system level without requiring explicit models in the form of equations.
Intellectual Merit: The project develops an overall framework and specific tools for multi-scale robust modeling, analysis and simulation of energy processing systems. Specifically, the framework allows for models specified as computer code, enables multi-level analysis and simulation that is "equation-assisted" at the system level for numerical performance, and enables hybrid simulators by allowing for run-time measurement inputs.
Broader Impact: The project contributes to unification of knowledge in the areas of system and energy engineering, controls, and applied mathematics. The long-term goal is to improve the understanding and containment of complex system-wide events such as blackouts. The underlying ideas are to be presented at the undergraduate level to illuminate the complementary roles of simulations and measurements in understanding complex systems and to prepare students for leadership roles in technical fields and in public policy. Educational aspects include training of two graduate students, who will attend professional meetings and participate in undergraduate and graduate teaching. New teaching modules for undergraduate, graduate, and professional development courses in energy engineering and a new graduate class on multi-scale simulation and analysis of large-scale energy systems are to be developed.
Historically, one can classify dynamical system models in three groups. The first one isthe Newtonian that calls for modeling a system using a differential equation, that must be solve explicitly. Although very successful in examples it tackled (mostly from physics), it also proved too restrictive, as there is still no general procedure for finding closed-form solutions when they exist. Moreover, there are elementary functions of practical importance whose integrals are not elementary. The second group would also model the system using a differential equation, but rather than solving it, will provide only qualitative information on the system directly. The third groupwould be the algorithmic modeling, which basically uses software and algorithms to reconstruct models using existingdata, usually resulting in proprietary and legacy code. Equation-free approach states that rather than aiming for a complete model (that may, most of the time, leave us with enormous amounts of unneeded data and with an imense computational effort), we might be able to construct an approximate system-level model that may still answer key questions regarding system behavior by acting directly on a hierarchical multi-scale model.In addition, this will address two major problems: one is the possibility to model energy processing components,whose models are, unfortunately, only available in rigid, inflexible formats as proprietary and legacy computer code. Second, is taking advantage of the fact that there exists many good models for the component level, while in many cases the system level is the one of interest, for which a physical model would be difficult or practically impossibleto obtain. In this project we have developed a set of general tools for equation-free system modeling and analysis. Our examples came from power electronics, and have included switched-mode and resonant converters, and were analyzed using the dyamic phasor modeling approach. Our procedure allows for increased model exchange among different industrial entities, as models are used as simulation "black boxes' without requiring possibly sensitive analytical model information.
