Abstract: This presentation develops the theory and demonstrates an efficient implementation algorithm to extract locally linear state-space models from large-scale physical models.  State-space models are systems of low-order, linear, transient, ordinary differential equations that predict the system response at particular operating states.  Such state-space models may be used to predict and interpret the system's electrochemical impedance spectra (EIS) and to develop model-predictive-control (MPC) algorithms.  The present approach applies system-identification algorithms that exercise physical models with pseudo-random binary sequences (PRBS) that accurately capture frequency responses.  The approach is illustrated using a particular 18650-format cell for which a high-fidelity physical model is available.  Gain scheduling is used to blend locally linear state-space models in ways that accurately represent nonlinear dynamic responses.  The composite state-space models are validated by direct comparison with the large physical models.  Examples are used to illustrate MPC scenarios for applications such following load-demand cycles and optimal charging strategies.

Bio: Dr. Kee's research efforts are primarily in the modeling and simulation of thermal and chemically reacting flow processes, with applications to combustion, chemical processing, electrochemistry, and materials processing.  His fuel-cell research concentrates on elementary chemistry and electrochemistry formulations and their coupling with reactive fluid flow.  Primary applications are to solid-oxide fuel cells operating on hydrocarbon fuels.  Research on hydrocarbon reforming and microchannel reactors is closely tied to the fuel-cell research.  In addition to research on the microstructural electrode behavior in Li-ion batteries, recent research is also concerned with large-scale sodium-based batteries.  Other electrochemistry research is concerned with mixed- and proton-conducting ceramics for application in fuel cells and membrane reactors.  The combustion research emphasizes the use of elementary chemical kinetics to understand fundamental flame structure.  Recent efforts concentrate on flame-droplet interactions in strained flames with multicomponent hydrocarbon fuels.  The materials-processing efforts focus on the development of chemical-vapor-deposition processes, with applications in thin-film photovoltaics and microelectronics.  All the research includes development of computational methods and software to solve systems of stiff differential equations.  Prof. Kee has published over 200 archival papers documenting his research.