Computationally efficient models enabling behaviour exploration of large scale cyber physical systems by simulations

Abstract

Cyber physical systems (CPS) enable applications not previously possible by networking components incorporating tightly connected physical and computational (cyber) functions. We propose a method for studying and optimizing such systems at an early development stage by using simulation. Models, which mimic the cyber physical components behaviour, combine analytical models for the physical part with neural network models for the cyber part. Such models are computationally efficient, can be easily upgraded and can be networked to simulate even large scale cyber physical systems. Applications for studying architectural issues, overall reliability and commercial aspects of cyber physical energy systems will be discussed.