Abstract:

Digital twins are a promising technology for vapor compression cycles because their capabilities can enable new approaches to system analysis, design, and control. As the computational models utilized by these digital twins must capture the observed dynamics of these physical systems, we describe important characteristics of these systems that impact the structure and implementation of these models. Specific attributes of these systems that govern model development include large-scale structures with tens or hundreds of thousands of equations, time constants that range over 10 orders of magnitude, and derivative discontinuities that affect the performance of solvers. We then describe candidate modeling, calibration, and estimation techniques that leverage modern mathematical and computational methods to meet these requirements, and present use cases demonstrating their efficacy.