Hybrid Analytical and Data-Driven Modeling Techniques for Digital Twin Applications
Andrew Wunderlich, Kristen Booth, Enrico Santi
Abstract
This paper summarizes, compares, and contrasts various modeling techniques which can be used in the notional digital twins which are under development for naval shipboard power and energy systems. The models can range from fully predetermined analytical models to fully data-driven empirical models, along with many hybrid approaches which fall between the two extremes. Hybrid approaches are of particular interest for digital twin applications as they can share benefits of both rigid analytical models and empirical models. One hybrid approach is proposed and discussed in detail as a particularly effective modeling solution. Drawbacks of the proposed approach are presented, along with further improvements which can mitigate those shortcomings.